EPA 600/R-12/032 | May 2012 | www.epa.gov/ord
United States
Environmental Protection
Agency
Removing Biological and
Chemical Contamination from
a Building's Plumbing System:
Method Development and Testing
Office of Research and Development
National Homeland Security Research Center
-------
&EPA
EPA/600/R-12/032
United States
Environmental Protection
Agency
Removing Biological and Chemical Contamination
from a Building's Plumbing System: Method
Development and Testing
United States Environmental Protection Agency
Cincinnati, Ohio 45268
National Institute of Standards and Technology
Gaithersburg, Maryland 20899
-------
Disclaimer
The U.S. Environmental Protection Agency (EPA) through its Office of Research and
Development partially funded and collaborated in the research described herein under
Interagency Agreement DW13-921677 to the Department of Commerce, National
Institute of Standards and Technology. It has been reviewed by the Agency but does not
necessarily reflect the Agency's views. No official endorsement should be inferred. EPA
does not endorse the purchase or sale of any commercial products or services.
This report was prepared by the National Institute of Standards and Technology (NIST)
through an interagency agreement with the U.S. Environmental Protection Agency.
The research in this document has been funded wholly by EPA under IA identification
number DW-13-92167701-6.
-------
Foreword
Following events of September 11, 2001, U.S. Environmental Protection Agency's
mission was expanded to cover critical needs related to homeland security. Presidential
directives identified EPA as the primary federal agency responsible for the country's
water supplies and for decontamination following a chemical, biological, or radiological
attack. To provide scientific and technical support to meet this expanded role, EPA's
National Homeland Security Research Center (NHSRC) was established. The NHSRC
research program is focused on conducting research and delivering products that improve
the capability of the Agency to carry out its homeland security responsibilities.
As part of this mission, NHSRC conducts research on contaminants that could be
intentionally injected into a community's water supply and water distribution system.
The possibility of such intentional contamination raised a new set of questions: What is
the fate of the contaminant in the water system? How can the contaminant be removed?
The approach to such questions depends on where the contaminant is located, in the
water treatment system, in the water distribution system, or in a building's plumbing
system. For example, following intentional contamination, the remediation strategy for a
large underground water main would probably be different than a remediation strategy
for a typical house.
The objective of this project was to study and model the adherence and subsequent
removal of chemical and biological contaminants in building plumbing systems. General
recommendations are given for plumbing decontamination, which can serve as a starting
point for more specific action plans.
Jonathan Herrmann
Director, National Homeland Security Research Center
11
-------
Abstract
This report presents an overview of measurements and analysis of the accumulation of
chemical and biological contaminants and their subsequent removal in building plumbing
systems. In addition, methods for decontaminating building plumbing systems and
restoring their operation, based on both specific and general contaminant characteristics,
are presented. Measurements consistently showed that most of the contaminants did stick
to the plumbing material substrates after the initial exposure. Some (diesel fuel, toluene)
showed a substantial reduction from flushing with clean tap water, while others required
the addition of high levels of chlorine to effect removal (phorate, gasoline, biologicals).
Measurements were used to develop fundamental models to predict maximum
contamination levels and required flushing times. These models were, in turn, used to
initiate development of computer software that could eventually be used as part of a
response to a contamination event.
in
-------
Table of Contents
Disclaimer i
Foreword ii
Abstract iii
Table of Contents iv
List of Tables v
Table of Figures vi
Acronyms and Abbreviations ix
Acknowledgements xii
Executive Summary xiii
Chapter 1 INTRODUCTION 1
1.1 Background 1
1.2 Potential Contamination Scenarios for Building Plumbing Systems 4
Chapter 2 PROCEDURES AND METHODS 7
2.1 Bench Scale Tests 7
Chapter 3 NEW METHOD DEVELOPMENT 16
3.1 Bench Scale Tests 16
3.2 Dynamic Fluid/Surface Interface Measurements 24
3.3 Screening Tests 30
3.4 Full-Scale Dynamic Tests 31
Chapter 4 MEASUREMENT RESULTS 34
4.1 Bench Scale Tests 34
4.2 Dynamic Fluid/Surface Interface Measurements 44
4.3 Screening Tests 49
4.4 Hot Water Heater Tests 51
4.5 Full-Scale Tests 55
Chapter 5 MODEL DEVELOPMENT 57
5.1 Semi-empirical 57
5.2 Flow in idealized pipe geometries 66
Chapter 6 DISCUSSION OF RESULTS AND RECOMMENDATIONS FOR
CONTAMINANT REMOVAL 72
6.1 Inactivation and Flushing of Spores and Bacteria 72
6.2 Dynamic Flushing of Diesel 73
iv
-------
6.3 General Considerations for Real World Scenarios 74
6.4 Recommendations for Contaminant Removal from Plumbing Surfaces 77
7.0 REFERENCES 81
Appendix A. Chemlmage: NIST Pipe Contamination Study A-l
Appendix B. Determination of Soluble Copper Cyanide Complexes in Tap
Water B-l
Appendix C. Tables Cited in the Text C-l
List of Tables
Table ES 1 Summary of Contaminant and Substrate Tests xiv
Table ES 2 Summary of Chemical Contaminant Interaction with Pipe Materials and
Decontamination Results xvi
Table ES 3 Results from Screening Tests xviii
Table ES 4 Results from Hot Water Heater Tests xix
Table ES 5 Results from Full Scale Plumbing SystemTests xx
Table ES 6 Comparison of Flushing Model Predictions to Measurements xxi
Table ES 7 Comparison of Contaminant Removal Times in Various Pipe Geometries xxii
Table ES 8 General Decontamination Procedures Based on Contaminant Type xxiv
Table 1 Summary of Contaminant and Substrate Tests 3
Table 2 Contaminant Mass Fractions for Screening Tests 31
Table 3 Summary of Chemical Contaminant Interaction with Pipe Materials and
Decontamination Results 35
Table 4 Results from Screening Tests 50
Table 5. Results from Hot Water Heater Tests 52
Table 6 Results from Full Scale Plumbing System Tests 56
Table 7 Comparison of Flushing Model Predictions to Measurements 61
Table 8 Comparison of Contaminant Removal Corrected for Flow Rates 70
Table 9 General Decontamination Procedures Based on Contaminant Type 79
-------
Table of Figures
Figure A 1. Raman spectra of pure components and precipitate samples A-4
Figure A2. Visible reflectance image of samples A-6
Figure A3. Overlay of area masks with samples A-7
Figure A 5. Spectra extracted from image of samples A-10
Figure A 4. Frame from spectral image data for copper samples A-10
Figure A 6. Histogram of score image for PC 1 (copper) A-ll
Figure A 7. Mask 1 based on PCI score image from copper A-12
Figure A 8. Mask 2 based on PCI score image of copper A-12
Figure A 9. Results table A-14
Figure B 1. Shows the calibration curve for the [Cu(CN)3]2" standard B-2
Figure B 2. Shows a chromatogram of 26.33 ppm [Cu(CN)3]2-standard B-3
Figure B 3. Shows a chromatogram of sample #4-50 ppm KCN + Cu pipe (old) mean
concentration = 0.963 ppm, (zoomed in to enhance peak) B-3
Figure B 4. Shows a chromatogram of sample #11-50 ppm KCN + Cu pipe (fresh) mean
concentration = 59.49 ppm B-4
Figure C 1. Flowchart depicting process of research C-5
Figure C 2. Potential contamination scenarios for building plumbing systems C-5
Figure C 3. Contaminant/substrate interactions and the exposure conditions C-5
Figure C 4. Schematic of dynamic flow test loop C-5
Figure C 5. Schematic of spectrofluorometer, test section, and linear positioning device... C-5
Figure C 6. Schematic of screening test setup C-5
Figure C 7. Schematic view of the full-scale plumbing system test facility C-5
Figure C 8. Schematic of full-scale test loop C-5
Figure C 9. Schematic of hot water heater testing apparatus C-5
Figure C 10. IR micro spectroscopy for Cu biofilm pipes exposed to 12.4 mg/1 phorate... C-5
Figure C 11. Thirty g/1 toluene in water exposed to PVC pipe C-5
Figure C 12. Gas chromatograms for gasoline and diesel C-5
Figure C 13. Raman spectra of Cu flat pipe exposed to 2000 mg/L diesel in water C-5
Figure C 14. Optical image of control pipes and those exposed to strychnine C-5
Figure C 15. SEM image at 2600x magnification of mercury precipitate on Cu pipe C-5
Figure C 16. Mercury concentration profile of 500 mg/1 HgC12 solution with PVC and
Cu in-service pipe materials C-5
Figure C 17. Biofilm organism levels in the reactors with different conditions for growth
of the biofilm C-5
Figure CIS. Bacillus thuringiemis (BT) spore levels adhered to biofilm condition pipe
surfaces in the reactors with different spore contacting conditions C-5
vi
-------
Figure C 19. Effect of exposure time and flow rate on thickness of the diesel excess layer
fora 0.15 % bulk free stream mass fraction on aPVC disk C-5
Figure C 20. Effect of exposure time and flow rate on thickness of the diesel excess layer
fora 0.15 % bulk free-stream mass fraction on an iron disk C-5
Figure C 21. Diesel excess layer thickness as a function of Re (Reynolds number) for
PVC surface and water/diesel (99.95/0.15) C-5
Figure C 22. Diesel excess layer thickness as a function of Re (Reynolds number) for
iron surface and water/diesel (99.95/0.15) C-5
Figure C 23. Effect of exposure time and flow rate on thickness of the diesel excess layer
for a 0.2 % bulk free-stream mass fraction C-5
Figure C 24. Effect of exposure time and flow rate on thickness of the diesel excess layer
fora 0.3 % bulk free-stream mass fraction C-5
Figure C 25. Diesel excess layer thickness as a function of Re (Reynolds number) for
water/diesel (99.8/0.2) C-5
Figure C 26. Diesel excess layer thickness as a function of Rre (Reynolds number) for
water/diesel (99.7/0.3) C-5
Figure C 27. Raman intensity of coupons soaked in 100 % diesel fuel C-5
Figure C 28. Raman intensity of coupons soaked in 100 % gasoline C-5
Figure C 29. Raman intensity of coupons soaked in strychnine (0.5% mass fraction) C-5
Figure C 30. Raman intensity of coupons soaked in sodium cyanide (1 % mass fraction). C-5
Figure C 31. Raman intensity of coupons soaked in 100 % phorate C-5
Figure C 33. Measurement results for a hot water tank exposed to diesel fuel C-5
Figure C 34. Heated hot water heater test with diesel fuel, sampling from outlet C-5
Figure C 35. Heated hot water heater test with diesel fuel, sampling from drain C-5
Figure C 36. Hot water heater test with strychnine C-5
Figure C 38. Hot water heater measurements with BT spores sediment and anode
samples C-5
Figure C 39. Pipe loop measurements for copper pipe and diesel fuel C-5
Figure C 40. Pipe loop measurements for copper pipe and diesel fuel C-5
Figure C 41. Pipe loop measurements for % inch copper pipe and strychnine C-5
Figure C 42. Pipe loop measurements for /^ inch copper pipe and strychnine C-5
Figure C 43. Pipe loop measurements for % inch copper pipe and BT spores C-5
Figure C 44. Pipe loop measurements for /^ inch copper pipe and BT spores C-5
Figure C 45. Pipe loop measurements for /^ Inch CPVC pipe and BT spores C-5
Figure C 46. Flushing measurements used to fit coefficients of model C-5
Figure C 47. Ratio of the effective kinematic viscosity to the kinematic viscosity in the
bulk phase C-5
Figure C 48. Maximum contamination diesel layer on various pipe surfaces as predicted
by eq. (3.12) C-5
Figure C49. Preliminary validation of semi-empirical flushing model C-5
vii
-------
Figure C 50. Sample graphical users' interface for EPA/NIST building plumbing
contamination model C-5
Figure C 51. Effect of building area on diesel flushing time for 6 mm piping C-5
Figure C 51. Effect of building area on diesel flushing time for 6 mm piping C-5
Figure C 52. Velocity fields, as indicated by the arrows, for a Reynolds number of about
30 in the U-shaped pipe system C-5
Figure C 53. Velocity fields for a Reynolds number of about 3000 for the U-shaped pipe
system C-5
Figure C 54. Flow near a cavity C-5
Figure C 55. Normalized pipe contamination, C/Cinit C-5
Figure C 56. Low Re (Reynolds) number flow past a rectilinear obstruction C-5
Figure C 57. High re flow past a rectilinear obstruction C-5
Figure C 58. Normalized total concentrations for the rectilinear cases C-5
Figure C 59. Time sequence of ingress of contaminant C-5
Figure C 60. Re-scaled total concentration adjusted for actual flow rates C-5
Figure C 61. Schematic of injection of flush water through an exterior water spigot C-5
Figure C 62. Schematic showing injection of flush water through drain valve of a hot
water heater C-5
Vlll
-------
Acronyms and Abbreviations
1,2,3-(CH3)3C6H3
2,4-(CH3)2C6H3CH=CH2
1,2,3 trimethylbenzene
2,4 dimethylstyrene
A floor area of a building
AQ regression constant
AI regression constant
ASTM ASTM International (formerly the American Society for Testing and
Materials)
ATCC American Type Culture Collection
BI transition depth
BA Bacillus anthracis
BE binding energies
BFRL Building and Fire Research Laboratory
BI biological indicator
BT Bacillus thuringiensis
BSL-1 Biological Safety Level 1
C total contaminant in the pipe system
c concentration of fluorescent diesel
Cmit initial total contaminant in the pipe system
ฐC degrees Celsius
1,1' (methylenebisthio) bis-ethane
benzene
C6H5C2Hs ethylbenzene
CeHsCsH? propylbenzene
CyHg toluene
C7Hi7O2PS3 phorate
1,2,3,4 tetrahydronapthalene
1,2,3,4 tetrahydro-2 methylnapthalene
equilibrium concentration of contaminant on surface
calcium carbonate
Centers for Disease Control and Prevention
colony forming units
Cyanide
CC>2 carbon dioxide
CP corner patch (describes location of contaminant in a U shaped pipe)
CPVC chlorinated polyvinyl chloride
CT free chlorine concentration x time (mg-min/L)
Cu copper
CVAAS cold vapor atomic adsorption spectrometer
Ddw diffusion coefficient
DRIFTS Diffuse Reflectance Infrared Fourier Transform Spectroscopy
EDS energy dispersive X-ray analysis
cs
CaCO3
CDC
CPU
IX
-------
EDX energy dispersive X-ray analysis
EPA U.S. Environmental Protection Agency
eV Electron volt
F Fluorescence intensity
FTIR Fourier transform infrared spectroscopy
g/L grams/liter
GC gas chromatograph
GUI graphical user interface
gpm gallons per minute
h hour
HPLC high pressure liquid chromatography
/0 incident intensity
IR infrared
ISA ionic strength adjuster
ISE ion selective electrode
K Kelvin
k mass transfer coefficient
KD dimensionless constant: ratio of convective to diffusive influences
corrected dimensionless constant (convective/diffusive influences)
Kj entrainment constant
kg kilogram
kV kilovolt
L liters
le excess layer thickness
leo initial excess thickness layer
LB Luria-Bertani
LDH lactate dehydrogenase
LI total length of plumbing in a building
M moles/liter
mass flow rate
Mc molar mass of contaminant
m3/h cubic meters/hour
mA milliamp
mg/L milligrams/liter
min minutes
mL milliliters
mm millimeter
MP middle patch (describes location of contaminant in a U shaped pipe)
MS mass spectrometer
NDIR nondispersive infrared
NIST National Institute of Standards and Technology
nm nanometer
pw wetted perimeter of channel
Pa Pascal
PBS phosphate buffered saline
PE pass energy
-------
ppm
ppmv
PT
PVC
QC
Re
rpm
RO
SEM
SG
SPME
SS
STERIS
t
TOC
u
UV
V
Vj
v0
xm
xin
xout
XPS
XRD
8
5
s
F
ug/L
ug/m3
j, Q
PA
0
parts per million
parts per million by volume
purge and trap
polyvinyl chloride
Quality control
Reynolds number
revolutions per minute
reverse osmosis
scanning electron microscope
specific gravity
solid phase microextraction
Spic and Spanฎ
STERIS Corporation
time
total organic carbon
local axial velocity
friction velocity
ultraviolet
axial bulk velocity
entrainment velocity/contaminant velocity toward pipe surface
entrainment velocity/contaminant velocity away from pipe surface
maximum entrainment velocity
mass fraction of diesel in bulk liquid
mass fraction
mass fraction of contaminant entering building
mass fraction of contaminant leaving building
X-ray photoelectron spectroscopy
x-ray diffraction spectrometry
penetration depth
ratio of effective kinematic viscosity to kinematic viscosity in bulk phase
extinction coefficient
surface excess density
emi s si on/detecti on wave 1 ength
excitation wave length
viscosity of bulk liquid
micrograms per liter
micrograms per cubic meter
microohm
kinematic viscosity
kinematic viscosity
density of bulk mixture (diesel and water)
density of diesel
quantum efficiency of fluorescence
XI
-------
Acknowledgements
This document was prepared under an interagency agreement, DW 13-921677, between
the US EPA and the Department of Commerce, National Institute of Standards and
Technology (NIST). The contributing authors from NIST were Kenneth Cole, Stephen
Treado, Mark Kedzierski, Stephanie Watson, and Nicos Martys. The contributing author
from the U.S. EPA was Vicente Gallardo. The U.S. EPA project officer was Vicente
Gallardo. The EPA review of the report was conducted by Roy Haught, Matthew
Magnuson, and Marissa Lynch.
Contact Information
For questions on this report, contact Vicente Gallardo, National Homeland Security
Research Center, US EPA, 26 W. Martin Luther King Dr., Cincinnati, OH 45268
Phone: 513-569-7176, E-Mail: gallardo.vincente@epa.gov
xn
-------
Executive Summary
The objective of this project was to study and model the adherence and subsequent
removal of chemical and biological contaminants in building plumbing systems. The
project results would form a basis for effectively responding to an event where building
plumbing systems had been intentionally or accidentally contaminated. This report does
not detail procedures for the decontamination of affected plumbing systems since such
procedures would be highly site-specific. However, this report does give general
recommendations for decontamination which can serve as a starting point for more
specific action plans.
This report summarizes the results of adherence and decontamination experiments
conducted in bench, pilot and full scale systems and with a variety of contaminants and
plumbing materials. In addition, specific cases of contaminant adherence and removal
were modeled in order to predict required flushing times for a given set of conditions.
Contaminant exposure tests were conducted under different flow conditions and
configurations, including coupons (i.e., small cut out pieces of pipe), small intact pipe
sections, full-scale pipe loops, and hot water heater tanks. The range of test conditions
was intended to cover what could be encountered in an actual contamination event.
The contaminants studied were:
Diesel fuel
Gasoline
Toluene
Strychnine
Cyanide
Phorate
Mercuric Chloride
Escherichia coli
Bacillus anthracis (BA)
Bacillus thuringiensis (BT)
Ricin
The material substrates tested were:
Copper
Galvanized iron
Polyvinyl chloride (PVC )
Chlorinated polyvinyl chloride (CPVC)
Rubber
Brass
xiii
-------
Table ES-1 summarizes the contaminants and substrates studied in the bench, pilot, and
field tests conducted during this project. In addition, substrate tests were conducted using
previously used hot water heater tanks with diesel fuel, cyanide, strychnine, and BT
spores.
Table ES 1 Summary of Contaminant and Substrate Tests
Contaminants
Diesel fuel
Gasoline
Toluene
Strychnine
Cyanide
Phorate
Mercuric
Chloride
E. coli
BA or BT
Ricin
Substrates
Copper
A, C, D, E
A, C
A
A, C,E
A, C,E
A, C
A
B
B, E
B
Iron
A, C
A, C
A
A, C
A,C
A
A
PVC or CPVC
A, C,E
A, C
A
A, C,E
A, C,E
A
A
B
B,E
B
Rubber
A
A, C
A
A, C
A,C
A
A
Brass
A
A
A
A
A
A
A
Legend: A- Bench-top Static Tests
B- Bench-top Biological Tests
C- Dynamic Flow Fluorescence Measurements
D- Screening tests
E- Full-Scale Tests
The procedures for the various measurements performed in this project are provided in
Sections 2 and 3 of this report. Section 2 describes the more conventional analytical
methods used, while Section 3 describes analytical methods that were developed in order
to carry out challenging aspects of this project such as measuring the presence of a
contaminant on a wetted surface. Analysis in this wetted surface matrix was not always
amenable to conventional methods.
The methods that can be found in section 3 include the following:
Solid phase microextraction (SPME)
Cold vapor atomic absorption spectroscopy (CVAAS)
Diffuse Reflectance Infrared Spectroscopy (DRIFTS)
Infrared (IR) Micro Spectroscopy
Raman Microscopy
X-ray Photoelectron Spectroscopy (XPS)
Scanning Electron Microscopy
SPME, CVAAS, and DRIFTS proved to be less useful than anticipated. XPS required a
high vacuum to be applied on the sample which led to analyte loss and also proved to be
inadequate. IR micro spectroscopy was more effective but mainly for phorate on copper
pipe. PVC material absorbed in the IR spectrum and thus adsorbed species on plastic
pipe could not be analyzed using this method. Raman microscopy was not subject to
xiv
-------
interference from the presence of water and in this regard was superior to IR methods
since samples did not have to be dried prior to analysis. However, the laser used in
Raman spectroscopy did heat up the sample which could lead to analyte loss. Raman
microscopy was still a useful method, however, due to the laser issue, there was a limited
range of conditions for analysis.
Lastly Section 3 contains the following:
Description of the experiments with biological reactors for testing adherence and
disinfection of Bacillus spores
A summary of a fluorescence technique for measuring the activity of ricin (as
opposed to analyzing for ricin directly)
A description of small scale dynamic tests which measured, in real time, diesel
accumulation and removal on PVC, iron and copper pipe material
A description of screening tests with a wide variety of contaminants and plumbing
A description of full scale tests with a smaller number of contaminants and
materials
RESULTS
In Section 4, results of bench scale, pilot scale and full scale tests are given for
contaminant adherence and removal.
Bench Scale Chemical Tests
Bench scale static adsorption tests were performed on a variety of chemical contaminants
and plumbing materials. Typically the initial concentration of the contaminant in the
adsorption test solutions was very high in order to simulate a worst case scenario. All the
contaminants adhered to the plumbing materials. Using tap water to decontaminate the
materials was only partially successful. Table ES 2 (below) summarizes these results.
xv
-------
Table ES 2 Summary of Chemical Contaminant Interaction with Pipe Materials and
Decontamination Results
Contaminant
Diesel fuel
Gasoline
Toluene
Phorate
Strychnine
Cyanide Salts
Mercuric
Chloride
Sticks to Substrate*?
Yes, All materials
Yes, All materials
Yes (CuF, CuB, CuIS,
PVC, PVCB)
Yes, All materials
Yes, All materials
Yes (Reacts with all
metals; CuF, CuB, CuIS,
BR,Fe)
Yes, All materials
(reacts with all metals to
form elemental Hg and
can absorb into plastics)
Removed by Water Flush?
Yes, partial
Yes, partial
Yes, partial
Yes, partial
Yes, partial
No, reaction product formed on
metal surfaces.
Yes, partial (not advised for
metals as elemental Hg is
present)
Removed by Other
Additive (specified)?
Bleach solution, partial
Bleach solution, partial
Bleach solution, partial
* Substrates: Pipe materials represent Copper [clean flat (CuF), biofilm growth (CuB), used in hot water
heater (CuIS)], PVC [clean (PVC), biofilm growth (PVCB)], iron (Fe), brass (BR), and rubber (RB).
Biological Contaminants
Bacillus Spores
A number of studies were completed with Bacillus spores. Results have been reported
previously (EPA and NIST, 2011) and are also summarized in this report.
In one set of studies, three types of bioreactors were used to study Bacillus spore
adhesion onto biofilm and the subsequent disinfection of the adhered spores with chlorine
and chloramine. Higher flow conditions led to greater spore adhesion. In general, high
concentrations of chlorine were needed to achieve significant reduction of the spores
adhered to biofilm. In addition, the biofilm itself exhibited a high chlorine demand and
necessitated prolonged contact times and additional chlorine in order to achieve spore
inactivation.
In addition studies were completed using germinants to determine if germinating the
spores prior to disinfection would increase the effectiveness of the disinfection
procedures. Amino acids were contacted with the spores to encourage germination.
After germination, organisms were contacted with disinfection solution. A net effect of
the germinants was to significantly decrease the concentration of the spores on the
surface of the biofilm-conditioned coupons. The germination step also resulted in the
organisms (both in suspension and adhered to the biofilm) being more susceptible to
inactivation by chlorine, monochloramine, or heat (50 ฐC).
xvi
-------
Ricin
The biological activity of ricin, lactate dehydrogenase, and lysozyme, (the latter two
compounds being simulants for ricin), was measured after treatment with solutions of
chlorine and monochloramine. Chlorine efficiently inactivated all three proteins. A
bleach concentration of 5.6 mg hypochlorite/L resulted in no detectable biological
activity of ricin. The simulant proteins lactate dehydrogenase and lysozyme were also
readily inactivated by chlorine, but the reaction with monochloramine was much slower,
taking several hours
In addition fluorescence was evaluated as a technique to analyze for the presence of
proteins such as ricin. Analysis of ricin is typically a time consuming process that
requires culturing in mammalian cells. The two simulant proteins, lactate dehydrogenase
and lysozyme, had a good correlation between the inactivation of biological activity and
the loss of native fluorescence. Results suggest that fluorescence is a good indicator for
measuring the disappearance of ricin.
The results of these studies on ricin are included in this report and can also be found in
Cole et al., 2008.
Small Scale Dynamic Tests for Exposure
Mixtures of diesel and water (1,500 to 3,000 ppm diesel) were contacted with PVC, iron
and copper pipe material in a flow through a fluorescence device that allowed the
measurement of diesel accumulation in real time. For the exposure tests with PVC,
increasing the flow rate produced a thicker layer of adsorbed diesel, with a maximum
thickness ranging from approximately 3 to 4 |j,m. For iron, the thickness of diesel was
approximately constant with respect to flow rate and was on average about 0.7 |j,m. For
copper, two levels of diesel were used, 2,000 and 3,000 ppm. For the lower level, the
maximum thickness of adsorbed diesel was approximately 8 |j,m and occurred at an
intermediate flow rate, i.e., flow rates greater than and less than this intermediate rate
resulted in thinner layers of adsorbed diesel. The higher diesel level also led to a
maximum thickness (26 |j,m) at an intermediate flow rate; however this data had an
increased amount of variability.
Small Scale Dynamic Tests for Flushing
Flushing tests were carried out at a Reynolds number of approximately 5,000. These
tests were done on the pipe material sections that had previously been subjected to the
exposure tests mentioned above. At the end of these exposure tests, there was still an
excess layer that remained on the pipe material surfaces. The pipes with residual diesel
were then flushed, and the thicknesses of the layers were measured in real time using the
same apparatus.
For all three pipe materials, (PVC, iron, and copper), we hypothesized that flushing
removed the adsorbed layer of diesel. For most measurements with PVC the thickness of
the diesel layer was measured to be less than zero soon after flushing commenced. (The
nature of the fluorescence measurement device and subsequent calculations for layer
thickness can lead to negative values.) The negative thicknesses were interpreted as a
xvii
-------
clean surface. Similarly, interpretation of the flushing test results for iron pipe surfaces
led to the conclusion that diesel was removed from the surface soon after flushing
commenced. For copper pipe material, flushing also appeared to remove adsorbed diesel,
however, it required a longer flushing time. Calculations showed that the adsorbed diesel
layer was removed at a rate of approximately 0.1 |j,m/h. Thus for a layer 1 |j,m thick, a
flushing time of 10 hours would be needed to remove the diesel.
The above interpretation that flushing successfully removes diesel is limited by the
sensitivity of the fluorescence device. It is possible that improved sensitivity would have
shown that diesel was still present on the surface.
Screening Tests
Screening tests for both exposure and flushing were conducted to identify particular
contaminant/substrate combinations that merited further testing in full scale or hot water
heater testing. Raman spectroscopy was used to detect contaminant residual during the
screening tests and the results are summarized in the Table ES-3. All contaminants
adhered to the pipe materials tested and 24 hour flushing was not successful or was only
partially successful.
Table ES 3 Results from Screening Tests for Exposure and Flushing
Contaminant
Diesel fuel
Gasoline
Strychnine
Cyanide
Phorate
Substrate
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Sticks to
Substrate?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Removed by Water Flush?
Partially in 24 h
Partially in 24 h
Partially in 24 h
Partially in 24 h
Not in 20 h
Not in 20 h
Not in 20 h
Partially in 24 h
Not in 20 h
Not in 20 h
Not in 20 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
The results of the screening tests for diesel contradict the flushing results from the small
scale dynamic tests which showed that diesel was removed from copper, iron and PVC
xvin
-------
pipe material. This could be because different levels of sensitivity for the two different
methods: Raman spectroscopy for the screening tests and fluorescence for the small
scale dynamic tests. Alternatively, this could be because the coupons in the screening
tests were contacted with pure diesel, while for the small scale dynamic tests, the pipe
material was contacted with relatively dilute (2,000 and 3,000 ppm diesel) mixtures of
diesel in water.
Tests with Hot Water Heaters
We chose the hot water heater for study in this project because, of all water-using
appliances in buildings, it would likely be the most challenging to decontaminate. These
challenges include accumulation of contaminants on tank sediment, difficulty in
achieving high flushing velocities within the tank, and inaccessibility of the internal tank
surface. Table ES-4 summarizes the tests with hot water heaters. In general, sediment at
the tank bottom accumulated contaminants and proved difficult to clean. Contaminants
may have adhered to the tank walls as well, but wall samples were not taken and
analyzed.
Table ES 4 Results from Hot Water Heater Tests
Contaminant
Diesel fuel
Strychnine
Sodium Cyanide
Bacillus
thuringiensis
spores
Results
Diesel adhered to sediment.
Flushing with water (either hot or cold) was ineffective in
removing diesel from sediment.
Laboratory detergent appeared to be effective in removing
diesel from sediment.
Strychnine was still present in sediment after flushing.
Flushing results were inconclusive due to cyanide's tendency
to form complexes with metal substrates which hindered
analysis.
Flushing removed spores associated with bulk water.
Flushing ineffective in removing spores from sediment.
Full Scale Tests
Tests were conducted at NIST's Plumbing Test Facility, which emulates a four story full-
scale building water supply system in a controlled laboratory setting.
The results of the flushing studies are presented in Table ES-5. All contaminants adhered
to all pipe materials tested. Flushing was partially successful at best.
xix
-------
Table ES 5 Results from Full Scale Plumbing System Test
Contaminant
Diesel fuel
Strychnine
BT Spores
Substrate
Copper
CPVC
Copper
PVC
Copper
PVC
Sticks to
Substrate?
Yes
Yes
Yes
Yes
Yes
Yes
Removed by Water Flush?
Minimal removal in 12 h
Partially in 12 h
Not in 14 h
Not in 14 h
Partially in 13 h
Partially in 13 h
Model Development
In Section 5, two modeling efforts are described. The first effort entailed the modeling of
accumulation and flushing for the purpose of predicting the amount of plumbing material
impacted and the time required to flush the plumbing surfaces of contaminants. The
second effort focused on modeling flow in idealized pipe geometries. This latter effort
showed that flow disturbances, e.g., an elbow or valve in a pipe segment, can cause the
formation of vortices, which in turn can affect contaminant removal.
Model Development to Predict Contamination Amounts and Flushing Times
Under this effort two models were developed based on the theory and results of the
dynamic fluid/surface interface studies conducted in this project. The first model gives
the maximum contaminant excess layer thickness that can occur for a given contaminant
bulk mass fraction and surface affinity. The second model predicts the time estimated to
flush a surface clean with respect to a contaminant as a function of time, contaminant
transport properties, and flushing Reynolds number.
These models were developed, in part, because of the logistical difficulty in actually
quantifying the surface concentration of a contaminant on a length of plumbing. This
difficulty necessitated an alternative approach to aid in responding to an event that
impacted a building's plumbing system. The approach consisted of a two prong strategy
in which a model predicted the maximum amount of contaminant that could adhere to a
surface, and then a second model predicted the time required to flush that estimated
contaminant amount from the plumbing surface.
Results of the flushing models varied. Table ES 6 compares predictions using the
flushing models with observed values for diesel fuel. For copper and for one of the trials
with iron, the models underestimated flushing times. In all the other trials, predictions
were conservatively overestimated.
xx
-------
Table ES 6 Comparison of Flushing Model Predictions to Measurements
Surface
PVC
PVC
PVC
Iron
Iron
Iron
Copper
Copper
Re
(during exposure
testing)
0
3200
7000
0
3200
7000
4600
4600
Contaminant
layer
thickness
4o (Mm)
1.5
1.5
2.5
1.0
1.0
0.5
6.6
1.9
Predicted Flushing
time (h) for clean
6.4
6.4
9.7
4.3
4.3
2.3
20
7.5
Observed
flushing time
(h) for clean
~0
~0
~Q
~ O
~0
~0
~5
<60
~\5
Re, Reynolds number
In another set of comparisons, the flushing models were able to roughly approximate the
required flushing times at two different Reynolds numbers. At a Reynolds number of
30,000, the predicted flushing time was about 14.5 hours for copper, 15 hours for PVC
and 17 hours for iron. The experimental data showed that pipe surfaces were flushed
clean between 20 and 25 hours. At a Reynolds number of 5,000, the predicted flushing
time for PVC was 44 hours, while the experimental data showed that the pipe surface was
flushed clean somewhere between 30 and 93 hours. Limitations with respect to level of
detection precluded a more accurate measure of required flushing time.
Although for the majority of the modeling results in Table ES 6, there was not good
agreement between predicted and observed flushing times, the fact that some of the
flushing times were in good agreement suggests that the models could be improved with
additional development. In addition the poor agreement may have had more to do with
the observed measurements than the predictions. The observed flushing times required
the measurement of a contaminant on a wetted surface, which is a challenging matrix to
analyze. Improvements made in these types of measurements would also help
development of improved predictive models.
A software program with a graphics user interface was developed based on the two
aforementioned models. The user inputs include mass fraction of contaminant, Reynolds
number during flushing, and the program provides recommended flushing time.
Currently, the software tool has been validated for diesel using detailed measurements of
adherence and removal of the contaminant at the liquid/surface interface.
Model Development for Flow in Idealized Pipe Geometries
The aforementioned models did not take into account the effect that obstructions and
various pipe bends (e.g., elbows, U shaped turns) can have on contaminant transport and
deposition. Another modeling effort was undertaken that addressed the effect of such
flow disruptions. Unlike the previous models, this model gives a close up view of the
flow paths occurring in specific plumbing obstructions and bends, and shows how
xxi
-------
phenomena such as vortices formed in pipe geometries such as elbows can affect
contaminant removal.
At higher flow rates the model showed that vortices did form in structures such as U
shaped fittings and cavities. These vortices resulted in pockets of reverse flow in which
flow direction was opposite of the direction of the bulk flow and could thus hinder
contaminant removal. Patches of contaminant were modeled to reside in specific areas of
a pipe structure, and removal of these patches of contaminant was modeled for two
different flow rates. Results are summarized in Table ES-7.
Table ES 7 Comparison of Contaminant Removal Times (h) in Various Pipe Geometries.
Reynolds
Number
30
3000
Straight Pipe,
with
contaminant in
middle of a
straight pipe
section
1
0.009
U-Shape, with
contaminant in
the middle of
theU
1.5
0.012
U-Shape, with
contaminant in
the corner of
theU
2.5
0.03
Cavity, with
contaminant
at bottom of
the cavity
10
0.18
In general, the model showed that for flow in idealized pipe geometries, the following
can have a significant effect on the decontamination of pipe systems: pipe geometry,
location of contaminants, and flushing rates. The higher the flush rate, the faster the
removal of the contaminant; however, in some cases, the effect of the vortices could
reduce the efficiency of removal. That is, it would take a greater amount of water to reach
the same contamination level when using high flush rates because higher flows could
produce vortices. Vortices can temporarily trap a contaminant and prevent it from
moving forward at the same rate of the bulk fluid. However, if only hydrodynamics is
considered, parts of a pipe system that are hard to decontaminate are generally less likely
to be reached by contaminants in the first place. The above analysis is qualified by
pointing out that this analysis does not take into account diffusion and mechanisms of
wall interactions that are clearly important and could significantly affect removal times.
Recommendations for Contaminant Removal
In the last section of this report, general recommendations for decontamination are
detailed for a number of scenarios. These recommendations are based, in part, on the
studies and modeling described in this report. In addition, knowledge of plumbing
systems and sound engineering principles also played an integral part in developing
recommendations for decontamination.
For the contaminants studied, all of them adhered to plumbing surfaces. The differences
were in the strength of that adherence. Compounds such as diesel did not bond as
strongly to surfaces compared to cyanide and mercuric ions, both of which bonded to
copper pipe in an irreversible and complex manner.
xxn
-------
Specific recommendations that link decontamination procedures to particular
contaminants or groups of contaminants with similar characteristics are given in Table
ES-8. These recommendations should prove useful as a starting point for comprehensive
guidelines that support general response plans for effective recovery from water supply
system contamination events. Although some of the procedures recommended in ES 8
were not effective or only partially effective in some of the actual testing, the general
approach for specific scenarios is judged to be a reasonable approach from a practical
point of view. For example, continuous flushing is seen as a sound approach for water
soluble contaminants. Further testing using additives such as detergents or non-toxic
solvents may yield more effective procedures.
xxin
-------
Table ES 8 General Decontamination Procedures for Water System Components Based on
Contaminant Type
Contaminant
Category
Soluble chemicals
Immiscible
chemicals with
specific gravity less
than one
Immiscible
chemicals with
specific gravity
greater than one
Sediments or
particles
Bacteria
Bacterial spores
Toxins
Example
from the
Category
Strychnine,
Cyanide,
Salts
Diesel fuel,
Gasoline
Phorate
Foreign
particles
Escherichia
co//0157:H7
Bacillus
anthracis
Ricin
Key Methods
For Pipes
Continuous flushing with
water, water buffered with
chlorine, or water mixed
with cleaner
Continuous flushing with
water, water with chlorine,
or water mixed with
cleaner
Continuous flushing with
water, water with chlorine,
or water mixed with
cleaner
Continuous flushing with
water, drain from
cleanouts where available
Flood system with water
and disinfectant and let
stand, followed by short
flush. Repeat as needed
Flood system with
germinant solution and let
stand to allow spores to
germinate, followed by
bleach disinfection
Continuous flushing with
water, water with chlorine,
or water mixed with
cleaner
For Tanks
Continuous flushing with
water, water buffered with
chlorine, or water mixed
with cleaner
Flush through fitting at top
of tank
Drain through drain valve
at bottom of tank, and fill
with cleaning solution.
Repeat as needed
Drain and flush from
bottom
Flood system with water
and disinfectant and let
stand, followed by short
flush. Repeat as needed
Flood system with
germinant solution and let
stand to allow spores to
germinate, followed by
bleach disinfection
Continuous flushing with
water, water with chlorine,
or water mixed with
cleaner
XXIV
-------
Chapter 1 INTRODUCTION
All occupied commercial and residential buildings, structures, and facilities must be
equipped with adequate supplies of clean, safe drinking water. In the past, there have
been concerns about insufficient treatment of water supplies or undesirable migration of
contaminants from plumbing system materials into water distribution systems, usually
due to some change in water quality or operating conditions. After the terrorist attacks of
September 11, 2001, and the anthrax mailings in September and October 2011, concern
arose about the potential for intentional introduction of contaminants into water
distribution systems and the need for effective methods for dealing with such
contamination events.
The objective of this project was to study the adherence and subsequent removal of
chemical and biological contaminants in building plumbing systems. Project results
could then be used to form a basis for effectively responding to actual events in which
building plumbing systems had been intentionally or accidentally contaminated. This
report does not detail procedures for the decontamination of affected plumbing systems
since such procedures would be highly event-specific. However, this report does give
general recommendations for decontamination which can serve as a starting point for
more specific action plans. These recommendations are based on analysis of the results
of water contaminant measurement and modeling activities that investigated
contamination and decontamination issues related to building plumbing systems.
The general classes of contaminants that were considered include chemicals such as
fuels, solvents, pesticides and poisons, and biological materials, such as bacteria
(vegetative bacteria as well as endospores), and biological toxins. A range of typical
plumbing system materials were assessed, including; copper, polyvinyl chloride (PVC),
iron pipe, solder, rubber gaskets and sealants. The first phase of this project focused on
the decontamination of Bacillus spores from plumbing surfaces and can be found in EPA
and NIST, 2011. The presence of chemical deposits and biofilms typically found on the
interior of plumbing components present another consideration (Mays, 2000). In addition,
plumbing system designs can vary widely; and flow obstructions, water tanks and other
water-using appliances can significantly complicate the analysis (Wingender and
Flemming, 2004). As a result, many traditional measurement methods were not
sufficient for detecting the presence of accumulated contaminants, and new methods had
to be developed and used. Chapter 3 of this report describes this new methods'
development.
1.1 Background
To reach the objective outlined above, a combination of detailed static and dynamic
measurements were done along with a computer simulation. This was aimed at
identifying the tendency of various contaminants to accumulate in building plumbing
systems. The measurements pointed to effective methods for eliminating or rendering
innocuous any accumulated contaminants. The basic measurement methodology was to
expose a particular plumbing material substrate, component, or system to a
1
-------
water/contaminant mixture followed by a flushing or other decontamination activity.
This occurred while periodically monitoring or collecting samples of the substrate and/or
water to evaluate for the presence of the contaminant or any residual materials. Testing
conditions included both static and dynamic configurations, and ranged from small
samples of substrate materials to full-scale building plumbing systems.
Figured (Appendix C) presents a project flowchart showing the various project tasks
and their relationships, and general information flow. This helps illuminate how the
experiments were selected and conducted. Two key points should be noted. First, there is
the iterative process of the development of new measurement methods based on the
results from initial measurements. Second, there is the incorporation of detailed
simulations of contaminant transport and dispersal to generalize the measurement results.
All of the measurement and modeling results fed into the development of general
decontamination procedures and recommendations in order to make the
recommendations as robust as possible.
Contaminant exposure tests were conducted using various combinations of contaminants
and substrates under different flow conditions and configurations, including coupons,
small pipe sections, full-scale pipe loops, and hot water heater tanks. The range of test
conditions was intended to cover what might be encountered in an actual contamination
event, although conditions beyond the range might be possible in extreme or unusual
cases.
Different measurement methods and configurations were used because each task focused
on a different aspect of contaminant accumulation and removal mechanisms. For
example, the static bench-top tests concentrated on measuring the fundamental
interactions between waterborne contaminants and plumbing system materials, while the
bench-top biological tests introduced the effects of flow and biofilms in a controlled
manner. The dynamic flow fluorescent measurements allowed in situ measurement of
the thickness of the layer of adsorbed diesel fuel under controlled flow conditions.
Screening and full-scale tests provided measurement data for more realistic plumbing
system configurations but under less well-controlled testing conditions.
The contaminants tested included:
Diesel fuel
Gasoline
Toluene
Strychnine
Cyanide
Phorate
Mercuric Chloride
Escherichia coli
Bacillus anthracis or B. thuringiensis (BA, BT)
-------
Ricin
The material substrates tested were:
Copper
Galvanized iron
PVC (PVC)
Chlorinated polyvinyl chloride (CPVC)
Rubber
Brass
Table 1 summarizes which contaminants and substrates were tested for the various tasks
of the project. In addition, substrate tests were conducted in which previously used hot
water heater tanks were contacted with one of the following contaminants: diesel fuel,
cyanide, strychnine, and BT spores.
Table 1 Summary of Contaminant and Substrate Tests
Contaminant
Diesel fuel
Gasoline
Toluene
Strychnine
Cyanide
Phorate
Mercuric
Chloride
E. coli
BA or BT
Ricin
Substrate
Copper
A, C, D, E
A,C
A
A,C,E
A,C,E
A,C
A
B
B, E
B
Iron
A,C
A,C
A
A,C
A,C
A
A
PVC or
CPVC
A, C,E
A,C
A
A, C,E
A, C,E
A
A
B
B, E
B
Rubber
A
A,C
A
A,C
A,C
A
A
Brass
A
A
A
A
A
A
A
Legend: A- Bench-top Static Tests
B- Bench-top Biological Tests
C- Dynamic Flow Fluorescence Measurements
D- Screening tests
E- Full-Scale Tests
In addition to the measurements of contaminant behavior in building plumbing systems,
the impact of plumbing system design and operation on decontamination strategies was
investigated. Because building plumbing systems are typically composed of a
complicated pressurized water supply network of piping, fittings, valves, fixtures, and
water-using appliances, along with an unpressurized sanitary drain system the potential
-------
for contaminant accumulation and the associated strategies for removal require careful
consideration of real-world factors.
1.2 Potential Contamination Scenarios for Building Plumbing Systems
The contaminants could be introduced via devices such as tanker truck or mechanical
pumps. Contaminants injected into one building's plumbing system could migrate to
adjacent buildings if conditions (e.g., pressure differential) were favorable. In general,
however, contaminants may enter a building water supply anywhere upstream, as
represented generically by the four locations shown in Figure C2 (Appendix C) and
summarized below:
1. Contaminants that are introduced far upstream from the building, either before or
near the water treatment facility, and travel a significant distance through the
water distribution system to reach the building service line
2. Contaminants that are introduced into a water main supplying multiple branch
lines, including one that supplies the building
3. Contaminants that are introduced near to, but outside of, the building via the
building service line
4. Contaminants that are introduced to the building water supply from within the
building.
These different scenarios represent how the methods and the amount of contaminant
could vary. As the contaminant introduction scenario varies, the duration and level of
contaminant concentrations in the water supply system would also vary. The portions of
the water distribution system, including the building plumbing system, that are affected
would be different. Thus, the response a contamination event, including methods to
control the spread and the procedures to remove the contaminant, would differ depending
on where the contaminant was introduced.
In the first scenario, Figure C2 , Point 1, the effects of dilution and water treatment will
likely reduce, but not necessarily eliminate, the impact of most contaminants on
downstream building plumbing systems. The closer the point of contaminant
introduction is to the building, the higher the concentration and the greater the potential
for contaminant accumulation. However, longer lengths of affected water supply lines
will require longer flushing times.
Most contamination events will be detected based on such factors as: a consumer
complaint (odor, color, and/or taste), illness or reaction, a sensor reading (although
sensors maybe uncommon), or a verbal or written threat. After this initial detection, the
first goal would be to determine if there is an actual contaminant present in the water
supply, what it is, and its extent. The point of contaminant introduction may be deduced
by collecting water samples from a range of locations and then mapping the water lines
that are found to be contaminated, working upstream until unaffected water lines are
found. As a hypothetical, in Figure C2, Point 3, has been determined to be the location of
-------
contaminant injection. Once the affected water lines are identified, several concerns
follow:
Precisely determining how long the contaminant has been in the water supply
system or how much of the contaminant was introduced will be difficult.
Collected water samples can identify the contaminant, but the sample contaminant
concentration will likely be less than that at the point of injection.
Contaminant can accumulate on internal plumbing surfaces, so simply flushing
out the contaminated water may not remove all of the contaminant.
Additional flushing or special procedures might be needed to restore the water
supply.
The above concerns guided the research described in this report.
Based on these concerns, it is prudent to assume a worst case exposure condition. In a
worse case scenario, a highly concentrated contaminant would be introduced into the
water supply system and would then come into contact with internal surfaces of plumbing
systems. Those contaminants that accumulate on plumbing system surfaces would need
to be eliminated before the system can be restored. The accumulation might be due to the
combined effects of a number of mechanisms, including different types of adsorption,
chemical reactions, and sedimentation on/with pipe materials and pipe scale. It might be
possible to remove some small sections of piping or other components for laboratory
analysis, but that is not always practical. Thus alternative strategies are needed.
The magnitude and the location of contaminant accumulation will be a function of the
characteristics of the contaminant/substrate interactions and the exposure conditions. For
example, some substrates are more conducive to contaminant accumulation. Longer
exposure times and higher concentrations can lead to greater accumulations.
Contaminants may or may not be soluble in water, and either more or less dense than
water. Soluble contaminants will dissolve (although such dissolution can take time) and
mix with water and thus come into contact with most of the plumbing system surfaces.
Insoluble contaminants will either float (specific gravity, SG<1) or (sink, SG>1), thereby
preferentially coming into contact with the upper or lower inside surfaces of pipes and
tanks. See illustrations (a) and (b) in Figure C3. Contaminants with sedimentary
characteristics, such as bacteria and spores, will sink due to gravity when water is not
flowing, and collect at the bottom or top of plumbing system components. See
illustrations (c) and (d) in Figure C3.
The above considerations imply that the particular details of the contaminant
characteristics and the method of introduction to a plumbing system can strongly
influence the location and magnitude of contaminant accumulation within the water
system. In most cases, it will be difficult to reconstruct the precise details of such an
introduction. Therefore, it might be advisable to presume the worst-case assumption that
the contaminant has interacted with the plumbing surfaces, unless there is specific
5
-------
information otherwise. Such an assumption could result in shutting off the water supply
to the building in order to prevent further migration; however this would not always
possible.
The following section (Section 2) summarizes the conventional procedures and methods
used in the analysis. However, development of new methods was needed in order to help
reach project objectives, e.g., methods for the in situ measurement of a contaminant on a
plumbing surface. These new methods are described in Section 3. In keeping with the
assumption of a worst case scenario, the studies that are described in the proceeding
sections typically used high concentrations of contaminants to contact plumbing materials
and plumbing systems for extended periods of time.
-------
Chapter 2 PROCEDURES AND METHODS
2.1 Bench Scale Tests
This section provides the procedures and methods used to measure chemical and
biological contaminants on the bench scale.
2.1.1 Chemical Contaminants
2.1.1.1 Static Adsorption and Decontamination Experimental Procedures
Interaction of the waterborne chemical contaminants with pipe materials was investigated
by monitoring the changes in the concentration of the specific chemical contaminant in
the water over time. Experiments took place in a 600 mL beaker (500 mL of contaminant
solution) or for volatile organic compounds, a 500 mL capped glass jar (450 mL
contaminant solution). Solutions were stirred with a magnetic stir bar to accelerate
equilibration. Pipe materials consisted of copper (Cu) pipe, and PVC pipe (white
Schedule 40, % inch nominal diameter). (Note that throughout this report the
measurement for pipe interior diameter is given in inches, the industry standard measure
for pipe used in the United States.) Pipe scale was simulated using calcium carbonate
(CaCOs) powder. Clean (new) flat Cu pipe coupons were also prepared from new Cu
pipe that was sectioned and flattened. All freshly cut new pipe materials were cleaned in
methanol to remove any cutting oils. Samples of used Cu pipe from a local apartment
building were also obtained and are referred to as Cu in-service pipes. The used pipe
came from a pipe line connected to the outlet side of a hot water heater. Biofilms were
grown on clean pieces of Cu and PVC tubing that were sectioned then cut in half and
held together with shrink-wrap tubing. Biofilm was grown, as described in more detail
below, on the pipe sections by flowing synthetic water plus humic acids, an energy
source for bacteria, for a period of two weeks and then for three days with only synthetic
water. Pipe pieces with biofilm growth were carefully removed from the shrink-wrap
tubing and characterized using FT-IR (Fourier transform infrared) microscopy before
starting adsorption experiments. For each contaminant concentration, two replicates of
each pipe material were completed.
A range of chemical contaminant concentrations was chosen based on lethal dosage data
and chemical solubility (Budavari et al., 1996). Initial adsorption studies used deionized
water as the solvent, and later adsorption experiments used laboratory tap water for the
solvent for more realistic conditions. Due to the immiscibility and the greater density of
some chemical contaminants compared to water and to ensure a well-mixed solution,
mixing of chemicals in water took place in two increments: one-half the amount of water
with the entire chemical contaminant quantity for 20 min followed by mixing the
remaining half of water for 20 min. An aliquot of contaminant solution was taken to
determine the concentration of the initial chemical contaminant. After pipe material
addition, aliquots of the contaminant solution were taken over time periods to monitor the
change in chemical contaminant concentration with pipe exposure time. Aliquots of
contaminant solution were taken over a length of 5 days using a graduated syringe. With
pipe material addition, the adsorption process began and subsequent contaminant solution
7
-------
specimens were taken at 10 min, 20 min, 40 min, 60 min, 120 min, 240 min, and 360 min
after the initial pipe sample introduction. Contaminant solution specimens were then
taken every morning and afternoon for the following 4 days. Concentration profiles of
chemical contaminants in water with the pipe material were then produced. In addition,
exposed pipe materials were examined for residual chemical contaminant in an attempt to
directly determine chemical contaminant surface adsorption after the experiment was
complete.
Decontamination experiments consisted of mixing a series of pipe materials that were
previously exposed to a specific chemical contaminant with a decontaminating water
solution. Decontaminating water solutions consisted of fresh tap water, tap water with
various bleach concentrations, and tap water with a specific concentration of a detergent,
Spic and Spanฎ (SS) detergent (Sun Fresh liquid cleaner, Prestige Brands, Irvington,
NY)1. For 450 mL of tap water, one of the following amounts of bleach (active
ingredient: sodium hypochlorite 6%, other ingredients 94 %, contains no phosphorus)
were added: 21.1 mL, 10.6 mL, 5.3 mL, or 2.6 mL. The concentration of SS used in the
decontamination studies was 7.1 mL detergent in 450 mL tap water. During
decontamination, both decontaminant water solution and pipe material were examined for
evidence of decontamination, i.e., contaminant chemical presence in the decontaminating
water solution and a reduction or removal of contaminant on the pipe material.
Decontaminant water solution was monitored during mixing with contaminated pipe
material. Pipe material was examined before and after the decontamination experiment.
Decontamination of cyanide exposed pipe materials was performed in a beaker in which
500 mL of decontamination water solution was mixed with a stir bar. Cyanide levels
were continuously monitored using a cyanide ion selective electrode, described in section
2.1.1.8. Contaminated pipe samples were added to the decontamination water after initial
readings of pH and cyanide concentration. Cyanide (CN~) concentration and pH were
recorded on a 30 min interval for 3 h. When no change in CN" concentration after this
decontamination time was observed, the sample was removed from the decontamination
water and the decontamination process was considered complete. If a change (or
increase) was noted in the concentration, the test continued until the concentration
stabilized. If decontamination showed chemicals present in the solution, the
decontamination was repeated with another 500 mL of fresh decontamination water.
Decontamination of organic contaminant exposed pipe materials was performed in a
500 mL capped glass jar. Contaminated pipe samples were placed in 450 mL
decontamination water and mixed with a stir bar. On an hourly interval, a 10 mL or
40 mL aliquot of the water was collected for water analysis and the pH was also
Certain commercial equipment, instruments, or materials are identified in this report in
order to specify the experimental procedure adequately. Such identification is not
intended to imply recommendation or endorsement by the National Institute of Standards
and Technology, nor is it intended to imply that the materials or equipment identified are
necessarily the best available for the purpose.
-------
recorded. Deionized water was used to dilute to a final volume of 40 mL. This final
diluted volume was analyzed and two replicates were analyzed.
2.1.1.2 Conventional Water Analysis
The quality of the water used in the adsorption and decontamination experiments was
tested by using a series of standard tests to check the water quality parameters commonly
employed within the water supply industry. The tests included pH, conductivity, chlorine
content, alkalinity, turbidity, and total organic carbon content (TOC). The standard water
tests were used to monitor the water quality before and after both adsorption and
decontamination methods to determine changes.
The pH was measured using a combination pH electrode (Accumet, Thermo Fisher
[Cole-Parmer Instrument Company, Vernon Hills, IL]) on an electrode meter with ion
selective electrode (ISE), conductivity, and computer interface capabilities (Accumet
Research AR50 meter, Thermo Fisher) (AWWA, 1990). The pH electrode was
calibrated daily with two buffer solutions ranging from pH 4.00, 7.00, and 11.00
depending on the range of pH to be measured in the test solution.
Conductivity was measured on a conductivity meter (Accumet Basic AB30 conductivity
meter, Thermo Fisher) using a glass 2-cell, 1.0 cm"1 conductivity cell (Accumet glass,
Thermo Fisher). The conductivity cell was calibrated daily with a 0.01 mole/L potassium
chloride solution with a known conductivity of 1408.8 |iS/cm. The procedure used can be
found in ASTM, 1999, which during the time of work, was the current standard.
Chlorine content (free and total) was measured using a portable photometer (Pocket
Photometer, FTPฎ Scientific, Fort Myers, FL). The meter was zeroed daily according to
manufacturer's instructions and checked against chlorine solutions with a known chlorine
concentration, which was prepared according to ASTM D1253-86 (ASTM, 1996). In the
portable photometer procedure, solid reagents containing a chlorine-indicating dye were
added to a water sample in a cuvette. Proper mixing of the reagent with the water solution
and the use of the defined time for color development was necessary for consistent
results.
Alkalinity was measured using a portable water analysis system (Mini-Analyst Series
942, Orbeco-Hellige, Sarasota, FL). The meter was zeroed daily according to
manufacturer instructions and checked against a sodium carbonate solution having a 160
mg/L alkalinity value (AWWA, 199la). In the portable system, a solid reagent was
added to a water sample in the sample tube. Proper mixing of the reagent with the water
solution was necessary for consistent results.
Turbidity was measured using a turbidity meter (LaMotte 2020, LaMotte Company,
Chestertown, MD). The meter was calibrated daily using turbidity standards in the range
of 1.0 NTU to 10.0 NTU (AMCO Primary turbidity standard, LaMotte). In addition,
turbidity reference solutions were also prepared according to EPA Method 180.1 (EPA,
1978).
-------
Total organic carbon (TOC) was measured using an automated TOC analyzer (Phoenix
8000 UV-Persulfate TOC Analyzer, Teledyne Tekmarฎ, Mason, OH). TOC measurement
involves oxidizing organic carbon in the sample to produce carbon dioxide (CO2),
detecting and quantifying the oxidized carbon or CO2, and presenting the results in units
of mass of carbon per volume of sample. The TOC analyzer used a wet chemical method
with persulfate oxidation along with simultaneous UV irradiation and nitrogen to sweep
the resulting CO2 to the detector, a nondispersive infrared (NDIR) detector (AWWA,
1991b). Forty mL autosampler vials were used and reagents were 10 % volume fraction
persulfate in 5 % volume fraction phosphoric acid and 21 % volume fraction phosphoric
acid (from sodium persulfate and concentrated phosphoric acid, respectively). The
analyzer reagents lines were primed and cleaned daily. The analyzer was calibrated
weekly using a 1000 mg/L carbon standard solution (Teledyne Tekmarฎ) diluted between
0.1 mg/L to 20 mg/L. Two aliquots were analyzed from each autosampler vial.
2.1.1.3 Phorate Water Analysis
Phorate concentrations used in adsorption studies were 12.4 mg/L, 24.8 mg/L, 100 mg/L,
and 400 mg/L. Contaminated water was analyzed with an automated purge and trap-gas
chromatograph/mass spectral detector (PT-GC/MS) (Teledyne Tekmarฎ XPT and
Thermo Finnigan Trace GC-DSQ MS, Thermo Fisher Scientific, Waltham, MA) (EPA,
1995a). The gas chromatograph (GC) method used was as follows:
GC column: FIP-Ultra2ฎ (fused silica column cross-linked with 5 % phenyl methyl
silicone), column length: 25 m, column ID: 0.22 mm, and film thickness: 0.33 um.
Oven Parameters:
Initial Temperature: 40 ฐC for 2 min
Ramp 1, 14 ฐC/min to 85 ฐC and hold for 2 min
Ramp 2, 30 ฐC/min to 220 ฐC and hold 2 min
Ramp 3, 13 ฐC/min to 260 ฐC and hold for 1 min
GC Conditions:
Inlet Temperature: 250 ฐC
Constant Flow at 1.0 mL/min
Split Inlet with a Split Ratio of 10
The purge and trap (PT) conditions were based on default settings with some minor
adjustments for longer rinse and purge times between samples to avoid interference. In
general, the critical PT parameters used were:
Purge:
Valve Oven Temperature: 150 ฐC
Transfer Line Temperature: 150 ฐC
Purge Time: 11.0 min at ambient temperature
Purge Flow: 40 mL/min
Desorb:
GC begins at desorb start
Desorb Preheat Temperature: 245 ฐC
10
-------
Desorb Time: 2 min
Desorb Temperature: 250 ฐC
Desorb Flow: 300 mL/min
Bake:
Bake Temperature: 270 ฐC
Bake Flow: 400 mL/min
Bake Time: 2 min
A 10 mL aliquot was taken to ensure enough solution remained after the entire
experiment for further analysis. The contaminant solution was diluted with deionized
water to the 40 mL purge and trap autosampler vial capacity. After dilution, the
autosampler vials were inverted 5 times to allow for complete mixing. Two GC analyses
were completed with each autosampler vial.
The GC peaks at retention time of 12.63 min [CyHnC^PSs; phorate] and 8.69 min
(CsHi2S2; 1,1' (methyl enebisthio) bis-ethane] were monitored to follow the change in
phorate concentration.
2. 1. 1. 4 Toluene Water Analysis
Toluene concentrations used in adsorption studies were 500 mg/L, 1.9 g/L, 3.8 g/L,
7.5 g/L, 15.1 g/L, and 30 g/L. The PT-GC/MS method was used to monitor the change in
toluene concentration with pipe exposure (ASTM, 1995; EPA, 1989a, 1989b, 1995b).
The same GC column (HP-Ultra2ฎ, Agilent Technologies, Santa Clara CA) was used as
in the phorate experiments with some modifications to the GC method:
Oven Parameters:
Initial Temperature: 40 ฐC for 2 min
Ramp 1, 14 ฐC/min to 85 ฐC and hold for 2 min
Ramp 2, 30 ฐC/min to 220 ฐC and hold 1 min
GC Conditions:
Inlet Temperature: 250 ฐC
Constant Flow at 1.0 mL/min
Split Inlet with a Split Ratio of 10
The PT parameters were the same as that used in the phorate experiments. The GC peak
at retention time of 4.0 min (CyHg; toluene) was used to monitor the change in toluene
concentration.
2.1.1.5 Gasoline Water Analysis
Gasoline concentrations used in adsorption studies were 100 mg/L, 300 mg/L, 500 mg/L,
1000 mg/L, and 2000 mg/L. The PT-GC/MS method that was used for toluene was used
to monitor the change in gasoline with pipe exposure (ASTM, 1995; EPA, 1989a, 1989b,
1995b). The GC peaks at retention times of 2.52 min (CeHe; benzene), 3.82 min (CyHg;
toluene), 5.12 min (C6H5C2H5; ethylbenzene), 6.67 min (CeHsCsH?; propylbenzene), and
11
-------
7.52 min (1,2,3 (CHOsCeH?; 1,2,3 trimethylbenzene) were used to monitor the change in
gasoline concentration.
2.1.1.6 Diesel Fuel Water Analysis
Diesel fuel concentrations used in adsorption studies were 100 mg/L, 300 mg/L, 500
mg/L, 1000 mg/L, and 2000 mg/L. The PT-GC/MS method that was used for toluene
was used to monitor the change in diesel fuel with pipe exposure (ASTM, 1995; EPA,
1989a, 1989b, 1995b). The GC peaks at the following retention times were used to
monitor the change in diesel fuel concentration:
7.67 min (1,2,3-(CH3)3C6H3; 1,2,3 trimethylbenzene)
9.45 min ( 2,4-(CH3)2C6H3CH=CH2; 2,4 dimethylstyrene)
9.54min(Ci0Hi2; 1,2,3,4 tetrahydronapthalene)
9.93 (CnHi4; 1,2,3,4 tetrahydro-2 methylnapthalene)
2.1.1.7 Strychnine Water Analysis
Strychnine concentrations determined for adsorption studies were 0.24 mg/L, 0.47 mg/L,
0.97 mg/L, 1.91 mg/L, and 4.01 mg/L. A method using high pressure liquid
chromatography (HPLC) was designed and based on the work of Alliot et al., 1982,
which used a chloroform extraction procedure at basic pH (5 M NaOH) with a 10 |ig/mL
quinine internal standard, all reconstituted in methanol. The HPLC components in this
study consisted of a Waters Breeze HPLC system (Watersฎ 1525 Binary Pump, Watersฎ
2487 Dual A, Absorbance Detector, Watersฎ 717Plus Autosampler; Waters, Milford MA)
with a Cig column [5 jim, 2.1 mm x 150 mm (Sunfire , Waters )] and a concentrated
ammonium hydroxide and methanol (0.75 : 99.25 volume fraction) mobile phase using a
10 jiL sample injection.
2.1.1.8 Cyanide Salts Water Analysis
Cyanide salts used in this study consisted of sodium cyanide and potassium cyanide in
concentrations of 3 mg/L, 10 mg/L, 20 mg/L, and 50 mg/L. The cyanide ion (GST) was
detected using an ion selective electrode (ISE) (CN~ 9606 combination electrode, Thermo
Orion). The ISE sensor was cleaned daily with emery paper to remove deposits. The CN"
ISE could not be exposed to CN" concentrations greater than 25 mg/L due to severe
sensor erosion. The CN" ISE was calibrated daily using CN" standard solutions (2 mg/L
and 20 mg/L from sodium cyanide or potassium cyanide depending on the experiment)
prepared according to EPA Method 9213 (EPA, 1996) and the appropriate ionic strength
adjuster (ISA) (1 mL/100 mL solution of 10 M sodium hydroxide). An in-house
computer program interfaced to the electrode meter (Accumet Research AR50 meter,
Thermo Fisher; Cole-Parmer Instrument Company, Vernon Hills, IL) was used to
constantly monitor the pH, temperature, and CN" concentration over the entire pipe
exposure period. Prior to cyanide salt addition, the pH of the tap water ranged from 7.1
to 7.6, after addition, pH of the water solution after ranged from 12.6 to 14.0.
12
-------
2.1.1.9 Mercuric Chloride Water Analysis
Mercuric chloride concentrations used in adsorption studies were 20 mg/L, 50 mg/L,
100 mg/L, 300 mg/L, and 500 mg/L. In this study an automated cold vapor atomic
adsorption spectrometer (CVAAS) (Hydra Cฎ, Teledyne Leeman Labsฎ) was used to
measure mercury content in both experimental solutions and pipe materials. Calibration
standards were prepared from a 100 |ig/L Hg standard (Teledyne Leeman ) diluting with
5 % volume fraction nitric acid. The instrument was calibrated using Hg standard
solutions ranging from 0.1 |ig/L Hg to 1.0 |ig/L Hg. All liquid samples were diluted to
5 % volume fraction nitric acid to ensure Hg dissolution. Nickel boats, which hold all
specimens, were initially heated to 700 ฐC to remove trace Hg and then periodically
heated under the same conditions to clean sample contamination. Sample aliquots ranged
from 50 jiL to 300 jiL, noting that larger aliquots took longer to dry. Solid samples were
run as received. Quality control (QC) methods consisted of analyzing standard solutions
of Hg, as was used in calibration of the instrument.
2.1.2 Biological Simulants
2.1.2.1 Microbiological Measurements ofBiofilms, Bacteria, and Spores
An important first step in simulating a water system for testing was establishing a
realistic biofilm on the surface of the coupons or pipe sections. The approach was to use a
synthetic formulation of tap water supplemented with humic acids (Morrow et al., 2008).
The humic acids stimulated the naturally occurring water microorganisms to grow a
biofilm in a short period of time (approximately 3 weeks). This allowed a sufficient
number of experiments to be completed within a reasonable amount of time. The biofilm
bacteria were removed from pipe or coupon surfaces by vigorous scrapping with a sterile
plastic cell scraper. Rinsing the surface with phosphate buffered saline (PBS, 0.01 M
phosphate, 0.138 M NaCl, 0.0027 M KC1, pH 7.4 ) was completed, and the biofilm
organisms were dispersed by vortexing for 30 s (Morrow et al., 2008). The solutions were
diluted in PBS, plated on R2A nutrient media at ambient temperature, and monitored for
growth for a period of up to one week (Schwartz et al., 2003).
The bench scale testing on biological contaminants was done by using simulants for more
hazardous biological contaminants. The Bacillus thuringiensis (BT) spore solutions used
in the experiments were a commercial preparation (Thuricide, Bonide Insecticide,
Oriskany, NY) prepared as previously described by Morrow et al., 2008. The choice of
BT as a simulant for B. anthracis (BA) spores was based on the close genetic relationship
(Radnedge et al., 2003) and similar structure of their outer most layer (the exosporium)
(Matz et al., 2001). BT spores can be used at Biological Safety Level 1 (BSL-1) and are
suitable for large-scale experiments where it is difficult to maintain higher levels of
biological safety. BA (Sterne) strain spores, also known as the vaccine strain were used
because they lack the genetic element (pX02 plasmid) needed for virulence in humans
and animals. The BA (Sterne) strain spores require BSL- 2 laboratories and were used for
limited experiments in laboratories of the Biochemical Science Division. The properties
of the BA (Sterne) strain spore samples were previously described by Almeida et al.,
2006 and Almeida et al., 2008. The concentrations of the spore samples were determined
13
-------
by the spread plate method. The samples were diluted with PBS buffer containing 0.1%
(volume fraction) Triton X-100. It was essential to include the surfactant Triton X-100 to
avoid the spores sticking to the plastic tubes used for dilutions. The dilutions were first
plated on Luria-Bertani (LB) nutrient agar and then incubated at 35 ฐC overnight before
counting the colonies.
The source of BT spores used for flushing experiments in hot water tanks was also from
the commercial preparation mentioned in the previous paragraph. The Thuricide Bionide
solution (100 mL) was diluted into a final volume of 1 gal of distilled water containing
0.01% Triton X-100. The low amount of petroleum hydrocarbon solvent from the
Thuricide Bionide BT solution and the Triton X-100 were subsequently diluted in the 50
gal hot water tank. The low concentrations of solvent and surfactant were not expected to
have a significant effect on the adhesion of spores to surfaces.
It was not found to be necessary to heat treat the spore samples prior to spread plating
onto the LB agar since it was relatively straightforward to distinguish between BA or BT
spores and the native bacteria in the biofilm. The biofilm bacteria grew very slowly i.e.,
about one week before colonies were visible, whereas the spores germinated and grew
rapidly overnight on the plates and could be easily recognized by their colony
morphologies.
2.1.2.2 Measurement of Inactivation of Bacteria in Solution
The inactivation kinetics of the spores and bacteria were measured in solution by
exposing the samples to solutions of either active chlorine or monochloramine. Free
chlorine concentrations were measured using N, N-diethyl-p-phenylenediamine reagent
and chlorine standards (Hach Company , Loveland, CO). A calibration curve was
measured using chlorine standards (Hach Companyฎ, Loveland, CO). Free chlorine
solutions were made by dilution of commercial bleach (sodium hypochlorite). The
chlorine concentration was measured after 30 min of stirring. Monochloramine
formulations were prepared (Camper et al., 2003) and measured using the indophenol
method and standards (Hach Companyฎ, Loveland, CO).
The spores or bacteria were added to either a buffer or synthetic water formulation at a
concentration of approximately 106 colony forming units (CFU)/mL. A small Teflon
coated stir bar was added and the solution gently stirred to keep the bacteria or spores
from settling. The inactivation studies for BT and BA spores were conducted in small
sterile glass vials to prevent the loss of spores by minimizing adherence of the spores to
the walls of plastic containers. Inactivation studies used either chlorine bleach or
monochloramine as the inactivating agent. Samples of the solution were taken at time
zero and at time intervals. At the end of the inactivation experiment, sodium thiosulfate
was added to quench the remaining chlorine or monochloramine. The amount of sodium
thiosulfate added would have resulted in a 7.5 mM concentration had not the thiosulfate
reacted with the chlorine compounds. After quenching the samples were then diluted
with PBS containing 0.1 % Triton X-100 and the dilutions were plated on LB plates to
determine the viable spores remaining.
14
-------
2.1.2.3 Measurement ofRicin Activity
Ricin is a protein toxin that consists of two subunits, A and B. The B subunit is
responsible for binding ricin to mammalian cells and the A subunit is an N-glycosidase
enzyme that catalytically inactivates ribosomes. Inactivation of ribosomes inhibits protein
synthesis in the affected cell, and results in cell death (cytotoxicity). The measurement of
the biological activity of ricin therefore requires the use of mammalian cells grown in
culture. Vero cells (CCL 81ฎ, ATCCฎ, Manassas, VA) were seeded at 6 x 103 cells per
well of tissue culture-treated 96-well microtiter plates overnight at 37 ฐC in 5 % 62
(Cole et al., 2008). After this step the ricin samples were diluted in Alpha MEM cell
culture media (Invitrogen, Carlsbad, CA) containing 10% fetal bovine serum and added
to the Vero cells and incubated at 37 ฐC in 5 % CC>2. After 22 h, the cytotoxic effects of
ricin were measured using the yellow tetrazolium MTT assay (#30-1010K, ATCC,
Manassas, VA) (Cole et al., 2008).
15
-------
Chapter 3 NEW METHOD DEVELOPMENT
Plumbing system designs can vary widely, and flow obstructions, water tanks and other
water-using appliances can significantly complicate the analysis (Wingender and
Flemming, 2004). As a result, many traditional measurement methods were not
sufficient for detecting the presence of accumulated contaminants. Thus in order to reach
project goals, it was necessary to develop new methods, which are described in this
section. Some initially promising methods that were tried but found to be inadequate are
only briefly described.
3.1 Bench Scale Tests
3.1.1 Chemical Contaminants
To simplify conditions involved in pipe systems for accurate adsorption isotherms to be
measured, a series of static adsorption experiments were performed in beakers with a
watch glass placed on top or in screw-capped jars. Basically, the experiments consisted
of adsorption experiments of test aqueous solutions containing a contaminant with a
substrate phase representing the pipe material and/or a water deposit followed by a
desorption phase. The substrates and the solutions were periodically sampled to
determine contaminant accumulation and removal rates. The physical microstructure of
the substrate was characterized to aid in the modeling and interpretation of the sorption
processes.
Due to the large number of variables to be studied within a pipe system (different grades
of pipe, mixtures of deposits in the pipes, temperature, and pH of water, etc), initial
adsorption studies were limited to one temperature (ambient), pH of tap water (near
neutral), and varying concentrations of contaminants. The sorbents used were copper and
PVC pipes, and calcium carbonate (CaCOs), to simulate scale deposits. The solutions
were stirred with a magnetic stir bar to accelerate equilibration time, to prevent clumping
of the CaCOs and to ensure equal contact with volatile contaminants. Glass stir bars were
used to prevent wear observed in Teflon stir bars and avoid contamination (e.g.
adsorption of some organics onto Teflon). Non-buffered deionized water was also
used initially to simplify interactions and establish analytical methods. All adsorption
experiments used laboratory tap water (which was characterized daily) as the solvent.
Pieces of pipe were used as specimens or substrates because the interior of whole pipe
segments were difficult to analyze after adsorption experiments without cutting or
penetrating the pipe and risking the disturbance of the adsorbed species. Substrate
samples of approximately 100 mm2 surface area were placed on the bottom of the beaker.
Suspending pipe samples using a wire mesh basket did not work as the metal wire reacted
with contaminant.
For the scale simulant, CaCOs, experiments were done to determine the solubility of
CaCOs in water; the mass loss observed ranged from 2.2 % to 3.5 %. A known, constant
volume (500 mL or 450 mL) of contaminant/water was used for all measurements. The
16
-------
volume was large enough to allow for the removal of a series of aliquots for contaminant
concentration analysis over the adsorption time. The dimensions (particularly area)
and/or mass of pipe substrate were recorded. Four replicates for each contaminant/pipe
materials combination were performed. Analysis of the various pipe materials and
contaminant concentrations were completed within the project deadlines which
established an accurate deviation value. Both the inner and outer surface of the pipe
materials were exposed to the contaminant/water solution. Only one side (inner surface)
of the exposed pipe material was analyzed for adsorbed species and reaction products.
The contaminant/water solution was also monitored for changes in the concentration of
contaminant over the adsorption time. Automated methods of water analysis were used
to help facilitate the large number of specimens. For cyanide experiments using ion
selective electrodes, the addition of ionic strength adjuster reagent was omitted to prevent
interactions with the chemical contaminant and maintain realistic pH changes. Several
analytical methods were tested, but many of the methods, specifically solid phase
microextraction (SPME), were found to lack the quantification necessary to accurately
monitor the changes in concentration.
3.1.1.1 Mercuric Chloride Analysis in Pipe Sediment
An automated CVAAS, the Hydra C, was used to measure mercury (Hg) content in both
experimental solutions and pipe materials. In this method the sample is combusted at
high temperatures with oxygen. Gases are carried through a heated catalyst tube that
removes halogens, nitrogen oxides, and sulfur oxides, and the remaining products are
carried through a gold amalgamation tube. All Hg at this point is in the elemental form
and is captured, heated, and released as a gaseous bolus toward the CVAAS. The signal
is measured in series by a high sensitivity cell followed by a low sensitivity cell, and the
two peaks are integrated and reported against a calibration.
This instrument was the manufacturer's first generation and, several issues were revealed
upon method development. Samples are analyzed in sample boats which consisted of
nickel metal. The nickel metal was highly susceptible to corrosion from nitric acid
solutions and high temperature used during the analysis. Constant examination of the
nickel boats was necessary to remove any damaged boats (holes present). The
manufacturer also advertised the availability of ceramic boats, but a final production
method for such boats had not been perfected at the time of this purchase. A small set of
'trial' ceramic boats were tested and found to be defective: non-even ceramic coating on
nickel metal substrate and severe cracks in the ceramic coating before and after use in the
Hydra C (Teledyne Leeman Labs). It was found that small amounts of samples were
necessary to avoid saturating the instrument detector. Serial dilution of liquid samples
was required and if not carefully performed, added error to the results. Sample carryover
also occurred regularly, so two blank sample boats were run in between each
experimental sample to prevent such carryover. Although solid sample analysis is
possible on the Hydra C, it was not practical for Cu pipe samples with adsorbed
mercury. A precipitate formed in adsorption experiments using copper (Cu) pipe with
mercuric chloride and caused poisoning of the catalyst and furnace apparatus in the
Hydra C. This required replacement of several major parts of the instrument. Sample
17
-------
precipitates could not be analyzed because of these results, thus solid phase
concentrations of mercury species are not reported.
3.1.1.2 Diffuse Reflectance Infrared Spectroscopy for Substrate Analysis
It was anticipated that diffuse reflectance infrared Fourier transform spectroscopy
(DRIFTS) could be used to analyze adsorbed contaminants on pipe material; however, it
did not prove successful. The following paragraphs give background information and
preliminary experience with this method.
DRIFTS analysis provides the vibrational modes of chemically bound constituents much
like Fourier-Transform infrared spectroscopy (FTIR) for fine particles and powders in the
concentration range from 0.1 % (by mass) to neat. DRIFTS does not require any special
sample preparation; powders can be measured 'as is" or mixed with a diffusely scattering
matrix such as potassium bromide. Organic and inorganic materials with a particle size
between 2 jim to 5 jim can be analyzed. Solid samples could also be examined using
emery paper to remove amounts of the sample of interest; the emery paper can then be
analyzed for DRIFTS spectra. DRIFTS measures any changes that occur in the IR beam
when the beam interacts with a particulate sample. When the beam reflects off the
sample surface, true specular reflection is produced and considered as interference in IR
detection. When the IR beam penetrates a particle, it scatters or diffuses depending on
the angle of the beam; it can also penetrate other particles as it moves through the
specimen or reflect off their surfaces. Diffused light that travels through and is partially
absorbed by the particles of a specimen contain information about absorption
characteristics of the specimen and is called diffuse reflection. Specular reflections are
separated out by analyzing pure backgrounds of the diluting matrix.
Factors that affect diffuse reflection include: refractive index, particle size, homogeneity
and concentration of the specimen. Samples with higher refractive indices and higher
concentrations produce more specular reflection and require dilution. Larger particle
sizes increase specular reflection; grinding samples to reduce particle size before analysis
can eliminate problems. More uniform samples produce more linear relationships
between band intensity and sample concentration. So thorough mixing is necessary. A
sample holder or cup is often used in DRIFTS accessories. To properly prepare a powder
sample for maximum signal, the powder should overflow the cup. The powder is then
leveled with a spatula to reduce reflection from the sample surface. The sample should
not be tamped as particles with too close proximity reduce IR beam penetration. The rule
for diluting (by weight) samples is: for organic samples- 10 % of sample is mixed with
90 % of diluting matrix and for inorganic samples- 2 % to 5 % of sample is mixed with
95 % to 98 % of diluting matrix.
Since water produces broad peaks at high wave numbers in a FT-IR spectrum and can
obscure many peaks of interest produced from chemical contaminants, all powder
samples from adsorption experiments has to be filtered and dried. Experiments using
physical additions of organic contaminants to CaCCb showed new peaks that were
characteristic of the contaminant. Powder CaCOs samples used in adsorption
experiments were filtered using a house vacuum. This procedure appeared to remove
18
-------
much of the volatile organic contaminants as no discernable signal was observed. In
addition, DRIFTS did not provide any useful information for the cyanide experiment on
Cu pipe. A black precipitate formed during the CN" adsorption experiments, but DRIFTS
did not yield any information as to the identity of this precipitate. In addition DRIFTS
did not yield useful information for CN" adsorption experiments with CaCCb. For this
study, DRIFTS could not easily characterize the hard copper pipe substrates and was not
used beyond these preliminary experiments. .
3.1.1.3 Infrared Micro Spectroscopy for Substrate Analysis
Infrared (IR) micro spectroscopy in reflectance mode was used to examine Cu pipe
materials before and after phorate adsorption experiments and before and after each
decontamination step to determine the extent of phorate removal. The resulting
reflection-transmission spectra are equivalent to traditional absorption spectra after
converting them to absorbance without any mathematical correction. The IR micro
spectroscopy was performed with a Thermo Scientific /Nicolet Magna-IR550 FTIR
spectrophotometer interfaced with a Nic-Plan IR microscope. The microscope was
equipped with a video camera, a liquid nitrogen-cooled mercury cadmium telluride
detector (Thermo Scientificฎ, Inc.; Madison, WI), and a computer-controlled mapping
translation stage (Spectra-Techฎ, Inc.; Shelton, CT), which is programmable in the x and
y directions. IR micro spectroscopy testing consisted of individual FTIR measurements
of 3 to 5 spots on the pipe specimen and representative FTIR mapping if obtainable. The
individual spectra were taken from the middle and sides of the specimens. Usually, 5
spectra were taken if the first 3 spectra differed from each other. For most experiments,
representative, equally spaced spots on the Cu pipe samples were analyzed in the IR
spectroscopy mode. If the contaminant layer and pipe sample were sufficient for IR
mapping, an IR map was collected either based on a portion of the contaminant IR
spectrum or on a particular IR peak that represented the chemical contaminant. Spectral
point-by-point mapping of the pipe materials was done in a grid pattern with a computer
controlled microscope stage and Atlus software package. Spectra were collected from
4000 cm"1 to 650 cm"1 at the spectral resolution of 8 cm"1 with 32 scans and a beam spot
size of 400 jim x 400 jim. The spectra were normalized to the background of bare spots
on the pipe material and were successively measured during the mapping after ever 20
spectra to compensate for slight changes. The resulting images were displayed as color
contour maps in the desired region.
For all of the IR micro spectroscopy measurements, all spectra were collected at the same
exact spot on the pipe samples after the various treatments. Different morphologies (more
or less deposits on the Cu substrate) on the pipe samples were also purposefully chosen
for IR analysis. The IR spectrum for pure phorate was also analyzed for comparison. IR
spectra of typical Cu in-service pipe were collected and showed that all the IR spectra
were different for the several spots on the pipe. This was an indication of the
heterogeneity of the deposits on the surface of the pipe. Furthermore, broadness of the IR
peaks indicated the complex composition of the deposits on the pipe sample at several
spots. After phorate adsorption onto the Cu in-service pipe, peaks that were
characteristic of phorate appear as IR bands at approximately 1375 cm"1, 1100 cm"1,
1005 cm"1, 950 cm"1, and 795 cm"1 as shown in the IR spectrum. IR spectra were then
19
-------
examined for significant differences between phorate exposed pipe samples after water
decontamination. PVC pipe samples could not be successfully examined by IR micro
spectroscopy because the PVC material caused interfering IR peaks that obscured any
contaminant peaks.
3.1.1.4 Raman Microscopy for Pipe Substrate Analysis
Raman spectroscopy is used to determine molecular structures and compositions of
organic and inorganic materials in a similar manner to IR spectroscopy. When an intense
beam of monochromatic light from a laser impinges on a material, scattering can occur in
all directions with the frequency of the scattered light the same as the original light; this
effect is known as Rayleigh scattering. Another type of scattering that can occur
simultaneously with Rayleigh scattering is known as the Raman effect. It occurs at
frequencies both higher and lower than the original light with considerably diminished
intensities. The difference between the incident and scattered frequencies are equal to the
actual vibrational frequencies of the material and are characteristic to the chemical
functional groups in a material. Raman is particularly useful in examining aqueous
solutions of inorganic and organic compounds. Water is a poor Raman scatterer, so
observations of vibrational transitions normally obscured in intense water adsorptions in
IR spectroscopy are allowed.
For the adsorption and decontamination experiments, a Raman microscope (Bruker
Senterraฎ, Bruker Optics, Billerica, MA) was used to collect Raman spectra on 3 to 5
spots of Cu or PVC pipe. The locations were spaced out to achieve the best
representation of the sample. Analysis began with a 50x objective (longer working
distance) and then to a lOOx objective. The location of the x, y stage was recorded for
each analysis and optical images of the locations were also saved. The basic analysis
protocol was to determine if the spectra look the same for all locations on clean pipe
materials (rinsed in methanol) and to define the background for the pipe material. For
original clean pipe, all spectra were similar.
The same pipe sample was then subjected to a drop of fuel or other contaminant and
analyzed using the same protocol to determine the presence of new peaks from the fuel.
Both fuel evaporation in air and the power of the laser affected the stability of the new
Raman signals. This procedure was used to determine instrument parameters (type of
laser light, laser power, resolution [range of wave numbers to be examined], and laser
dwell time) for each pipe sample material and contaminant. The choice of laser power
was based on obtaining good Raman signal (low signal to noise) with no sample damage.
High laser power burned samples. For reflective samples lower laser power was required
to avoid saturating the detector. A lower dwell time and shorter range of wave numbers
decreased the overall analysis time, which reduced the probability of contaminant
evaporation.
These instrument parameters were then used as a starting point on pipe materials
subjected to adsorption and decontamination experiments. In general, Raman was used
to make qualitative identification of the chemical contaminant. For in-service Cu pipe,
very low laser power was necessary to avoid burning the original deposits; therefore, the
20
-------
resulting Raman signals were very weak. It was difficult to achieve quantitative
information due to the evaporation of the contaminant in air and by the instrument laser.
The data collected from Raman analysis provided optical images of the pipe, 3-D maps of
Raman spectra (particular peaks related to specific contaminants), and individual spectra
from the pipe materials (both the pipe and any deposits).
3.1.1.5 X-ray Photoelectron Spectroscopy and Scanning Electron Microscopy
After the adsorption experiments for residual contaminant, the pipe and powder
substrates were examined using X-ray photoelectron spectroscopy (XPS). XPS is a very
surface sensitive analytical technique that examines the top 5 nm of the surface and has a
detection limit of about 1 % atomic concentration. It provides quantitative elemental and
chemical species information. All XPS measurements were performed with a Kratos
Axis Ultra photoelectron spectrometer. Experiments were conducted at room
temperature with a base pressure in the 1.3x 10~6 Pa range. The monochromatic Al Ka x-
ray source was operated at 140 W (14 kV, 10 mA). The energy scale was calibrated with
reference to the Cu 2p3/2 and Ag 3ds/2 peaks at binding energies (BE) of 932.7 eV and
368.3 eV. A coaxial charge neutralization system provided charge compensation. Pipe
samples and pipe precipitates were attached to the XPS sample holder using double stick
tape. The analysis area for the high-resolution spectra was 2 mm x 1 mm.
A survey scan was performed on all samples to determine the elements of interest, In
general, O Is, N Is, C Is, S 2p, P 2p, Ca 2p and Cu 2p spectra were acquired at a pass
energy (PE) of 20 eV and a maximum acquisition time of 8 minutes per element. Peak
BEs were determined by referencing to the adventitious C Is photoelectron peak at
285.0 eV. Quantitative XPS analysis was performed with the Kratos VISIONฎ software
(version 2.1.2). The atomic concentrations were calculated from the photoelectron peak
areas by subtracting a linear-type background. The O Is, N Is, and C Is regions were
deconvoluted using mixed 70 % Gaussian/30 % Lorentzian components.
Challenges with XPS are the high vacuum requirements which limit all samples to dry
conditions. In addition, volatile contaminants that were not chemically reacted to the
surface also had a greater possibility of being pumped away before analysis. In general,
increases in carbon, phosphorus, and sulfur were observed. Due to limited instrument
availability and high vacuum requirements, XPS was not the best choice for most surface
analysis.
Scanning electron microscopy (SEM) was also used to examine pipe materials before and
after adsorption experiments. SEM is primarily used to study the surface topography of
solid samples and has a resolution of 1.5 nm to 3.0 nm. Electrically conductive materials
can be examined directly, but non-conductive materials require a thin conductive coating
(carbon and precious metals) to prevent electrical charging of the specimen. An electron
beam passing through an evacuated column is focused by electro-magnetic lenses onto
the specimen surface. The beam is then raster scanned over the specimen in synchrony
with the beam of a cathode ray display screen. The secondary electron emission
(inelastically scattered) from the sample is then used to modulate the brightness of the
cathode ray display screen, thereby forming the image. If back-scattered electrons
21
-------
(elastically scattered) are used to form the image, the image contrast is determined largely
by compositional differences in the sample surface rather than topographic features. In
this study, the pipe material had to be dried for SEM analysis. Depending on the
material, a carbon coating was used for non-conductive samples. Secondary electrons
were generally used to collect images and energy dispersive X-ray analysis (EDS or
EDX) was used for element identification.
3.1.2 Biological Contaminants
3.1.2.1 Biofilm Reactor Measurements
During the course of the project, three different configurations were used to establish
biofilms on plumbing materials. The first configuration used was the Centers for Disease
Control and Prevention (CDC) biofilm reactors (obtained from BioSurface
Technologiesฎ, Bozeman, MT). The second configuration was a pipe section reactor
constructed in a lab that was operated as a single flow through with creeping flow (1
mL/min, Reynolds number [Re] < 1). The third configuration was the pipe section reactor
operated in a loop mode with intermittent high flow rate (2.5 L/min, 2 h with flow and 2
h without flow in a continuous cycle). The first two configurations used synthetic water
supplemented with humic acids. The third used local (Gaithersburg, MD) tap water
supplemented with humic acids to establish the biofilm layers.
The CDC biofilm reactor has 24 coupon disks (13 mm diameter) made of PVC or copper
(coupons) suspended in a 1 L beaker. A Teflon baffle is suspended in the reactor and
the stirring rate of the baffle results in the fluid shear on the coupon surfaces. A separate
peristaltic pump was used to add the synthetic water solution to the CDC reactor. For
growing a biofilm layer on the coupons the CDC reactor was operated without stirring for
the first week and for the second and third week the baffle was maintained at 120 rpm
(Morrow et al., 2008). A variety of shear conditions was used for contacting the biofilm
conditioned surfaces with spores in the CDC reactor. The CDC reactor with the baffle
operated at 60 rpm was compared to the baffle removed (low shear) for contacting with
the Bacillus spores.
The pipe section reactor used a series of plumbing pipes (PVC and copper) with 19 mm
(3/4 inch) diameter and 51mm in length. The pipe sections were jointed together using
silicone tubing. Peristaltic pumps were used to control the flow of water in the pipe
section reactors. The third configuration used to grow biofilms on pipe surfaces was the
pipe section reactor configured in a loop and local tap water supplemented with humic
acids was used to establish the biofilms. This third configuration used an intermittent
high flow (2 h on and 2 h off) to better simulate the type of conditions found in a building
water system. Additional details are given in the EPA and NIST, 2011.
Experiments were also done using germinants to determine their effect on the
effectiveness of disinfectants on spores adhered to biofilm conditioned coupons.
22
-------
Biofilm conditioned coupons were prepared using the CDC reactor as previously
described. The coupons were contacted with solutions of BT and BA spores
(approximately 107 CFU/mL) for 24 h and then rinsed. A germinant solution consisting
of amino acids (1 mM inosine and 8 mM glycine) was then contacted with the spores
adhered to the coupons for 24 h with stirring in the CDC reactor vessel Additional details
are given in EPA and NIST, 2011.
3.1.2.2 Inactivation of Biofilm Spores on Pipe Materials
A Teflon baffle was used to establish the fluid shear in the CDC biofilm reactor.
Peristaltic pumps were used to control the fluid velocity in the pipe section reactor
(second and third configurations). When using the CDC biofilm reactor, the biofilm
conditioned coupons were contacted with spores by adding the spores to the reactor under
two shear conditions (one without baffle stirring and one with the baffle stirring at
60 rpm). After contacting with spores, the coupons were then removed from the reactor,
rinsed, placed in a tube, and then contacted with the disinfectant solutions without
stirring.
The spore solutions with a concentration of approximately 107 CFU/mL were contacted
with the biofilm conditioned pipe coupons or pipe sections. The contact time varied from
2 h to 24 h. The adhesion of the spores reached a plateau at approximately 24 h. After
contacting the spores, the coupons and pipe sections were rinsed with water. The
concentration of the spores adhered to the coupon or pipe section was determined by
sampling several coupons or pipe sections at this stage. The coupons and pipe sections
with spores adhered to the biofilm were then contacted with disinfectant solutions.
Different concentrations of active chlorine or monochloramine were contacted with the
coupons or pipe sections for varying periods of time. The coupons or pipe section were
removed from the disinfectant solution, rinsed with 7.5 mM sodium thiosulfate to stop
the reaction of chlorine or monochloramine, and then rinsed in water before sampling.
In the second configuration, the pipe section reactor was operated with a creeping flow
(1 mL/min) for biofilm growth (three weeks). After biofilm growth stage, the pipe
sections were disassembled and the individual pipe sections were contacted with spore
solutions, rinsed and then contacted with disinfection solutions. The contacting and
disinfection steps were done without flow.
A major difference in the conditions of spore contacting, flushing, and disinfection was
used in the third configuration compared to the first two configurations. In the third
configuration, the pipe sections operated in a loop mode with intermittent high flow for
the spore solution (2 x 106 CFU/mL in 2 L tap water) contacting phase for 24 h (cycling
2 h on and 2 h off). The water flushing and disinfection with chlorine solutions stages
were done at high flow (2.5 L/ h with continuous flow).
3.1.2.3 Inactivation ofRicin by Disinfectants
Ricin has native fluorescence due to the presence of aromatic amino acids in the subunits
(Gaigalas et al., 2007). Tryptophan has the highest fluorescence of the amino acids
23
-------
present. Chlorine and monochloramine are chemical oxidants known to react with many
amino acids, including tryptophan (Nightingale et al., 2000; Hawkins et al., 2003;
Alimova et al., 2005). The native fluorescence of ricin was monitored by exciting the
samples with light at 280 nm and measuring the fluorescence emission light at 340 nm.
Fluorescence is easily monitored property that can be done in real time and in a
noninvasive or destructive manner. The biological activity of ricin using cell culture
(Section 2.1.2.3) was measured. The biological activity of ricin and the fluorescence of
the same samples were also measured to determine if the loss of fluorescence correlated
to the loss of biological activity.
Since ricin is an enzyme, the inactivation of two model enzymes, lactate dehydrogenase
(LDH, from cow heart) and lysozyme (from chicken egg) were studied as simulants for
ricin. The native fluorescence of these two model proteins was monitored along with their
enzymatic activity. Cow heart LDH and chicken egg lysozyme were chosen as simulants
because these proteins could be obtained in a pure form for reasonable costs, and their
enzymatic activity can rapidly be measured. A through description and discussion of the
above can be found in Cole et al., 2008
3.2 Dynamic Fluid/Surface Interface Measurements
A fluorescence-based measurement technique was developed to measure diesel fuel
adsorption to PVC, iron, and copper during contaminated water flow and tap water
flushing (Kedzierski, 2006 and Kedzierski, 2008). The test apparatus was designed for
the purpose of studying adsorption of diesel fuel from a flowing dilute diesel/water
mixture. It was used to measure the mass of diesel fuel adsorbed per unit surface area
(the excess surface density) and the bulk concentration of the diesel fuel in the flow.
Both bulk composition and the excess surface density measurements were achieved via a
traverse of the fluorescent measurement probe perpendicular to the test surface. The
diesel adsorption to each test surface was examined for three different Reynolds
numbers : 0, 3,100 and 7,000. A Reynolds number of 3,100 is in the transition between
laminar and turbulent flow, and a Reynolds of 7,000 corresponds to fully turbulent flow.
Measurements for a given condition were made over a period of approximately 200 h for
a diesel mass fraction of approximately 0.15 % in tap water. The adsorbed diesel on the
surfaces was removed by flushing with tap water. Excess surface density measurements
were made during flushing.
3.2.1 Experimental Apparatus and Measurement Uncertainties
A test apparatus was designed and developed to use the fluorescent properties of diesel
fuel to study its adsorption and desorption to and from plumbing pipe materials. A
calibration technique was developed to measure both the mass of diesel adsorbed per unit
surface area (the excess surface density) and the bulk concentration of the diesel fuel in
the flow. The flow loop for measuring diesel fuel on pipe substrates and the development
2 The Reynolds number is a dimensionless number that characterizes the swiftness and
randomness of the flow. Larger Reynolds numbers indicate more random and swifter
flow, i.e., more turbulent flow.
24
-------
of the measurement technique with its uncertainties is thoroughly discussed in
Kedzierski, 2006 and Kedzierski, 2008).
Figure C4 schematically shows the flow loop for measuring diesel fuel on pipe substrates.
The primary components of the loop are the pump, the reservoir, and the test chamber
with the test section. The inside surfaces of the approximately 96 mm x 1.6 mm
rectangular flow cross-section of the aluminum test chamber, shown in Figure C5, were
black anodized to minimize stray light reflections. A centrifugal pump delivered the
contaminated water to the entrance of the rectangular test chamber at room temperature.
The flow rate was controlled and varied by varying the pump speed with a frequency
inverter. A heat exchanger immersed in the reservoir was supplied with brine from a
temperature-controlled bath to maintain the entrance temperature to the test chamber at
ambient temperature (293.8 K). This was done to ensure that the diesel fuel was at the
same temperature as it was during the fluorescence calibration to avoid the temperature
effect on fluorescence (Miller, 1981). An additional temperature-controlled bath was
used to maintain the fluorescence standards at the same ambient temperature. As
described in this section, the fluorescence standards were used to calibrate the range of
the fluorescence measurements.
Residential copper pipe was used to plumb together the various components of the loop.
Redundant volume flow rate measurements were made with an ultrasonic Doppler and a
turbine flowmeter with expanded uncertainties of ฑ 0.12 m3/h and ฑ 0.03 mVh,
respectively. As shown in Figure C4, three water pressure taps before and after the test
chamber permitted the measurement of the upstream absolute pressure and the pressure
drops along the test section with expanded uncertainties of ฑ 0.24 kPa and ฑ1.5 kPa,
respectively. Also, a sheathed thermocouple measured the water temperature at each end
of the test chamber to within an uncertainty of ฑ 0.25 K. The dissolved oxygen level, the
conductivity, and the pH, were monitored at the water reservoir with associated B-type
uncertainties of ฑ 0.5 %, ฑ 50 // Q/cm, and ฑ 0.3, respectively.
Figure C4 also shows the inlet tap used to flush the test section with fresh tap water. In
preparation for flushing, the test section was isolated from the rest of the test loop by
closing valves. Then the fluid was drained from the test chamber and returned to the
reservoir. Next, a tap water supply was connected to a test chamber port. The other test
chamber port was connected to a filter to absorb any diesel fuel before it was sent to a
drain.
Figure C5 shows a view of the spectrofluorometer that was used to make the fluorescence
measurements and the test chamber with the fluorescence probe perpendicular to the
flattened pipe test surface. The spectrofluorometer was modified by replacing the cuvette
holder with a special adapter with lenses and mirrors to remotely excite and measure
fluorescence via a bifurcated optical bundle. Two optical bundles consisting of 84 fibers
each originated from the spectrofluorometer. One of the bundles transmitted the
excitation light, i.e., the incident intensity (/0), to the test pipe surface. The other bundle
carried the emission, i.e., the fluorescence intensity (F), from the test surface to the
spectrofluorometer. The excitation wavelength (Ax) and the emission/detection
25
-------
wavelength (Am) were set to 434 nm and 485 nm, respectively, for all tests. Further
details on the fluorescence measurement technique are given in Kedzierski, 2006 and
Kedzierski, 2008.
Number 2 diesel fuel was used from a single batch throughout the experiment to avoid
property variations that might be caused by batch variations due to it being a complex
mixture of hydrocarbons. A nominally 1.5 % by mass diesel mixture was prepared with
local Gaithersburg, Maryland tap water for the exposure/flow rate tests. The measured
dissolved oxygen level, the conductivity, and the pH, of the water at 24 ฐC before mixing
with diesel fuel were 86.4 %, 358 |iQ/cm, and 7.04, respectively.
During exposure tests, diesel adsorbs to the test surface to an excess layer thickness (4).
Because the molar mass of the diesel is unknown, the surface excess density (F ) is
defined in this work on a mass basis as (Kedzierski, 2002):
r = ttPd ~ Pbxb) (3.2.1.1)
The density of liquid diesel fuel is/>d- The density of the flowing bulk mixture (pb) is
evaluated at the bulk mass fraction of the mixture (xb). The surface excess density is
roughly the mass of diesel attached per surface area. The F and 4 are the primary
measurements of this study. The 4 is measured perpendicular to the surface with the
origin at the fluid-surface interface.
Two different calibration methods had to be combined due to the additional complexity
caused by immiscible liquids. Both calibration techniques were used to quantify different
functional aspects of the Beer-Lamb ert-Bougher law (Amadeo et al., 1971), which forms
the basis of the calibration equation. The first method was used to obtain the relationship
between diesel composition and fluorescence intensity for a fixed light path length (fixed
probe height above the test surface). The first method would have been sufficient had the
bulk composition of the flow remained the same as it was charged in the reservoir. Due
to the immiscibility of the two fluids, the bulk composition of the flow differs from that
in the reservoir. As a result, a second method is necessary to determine both the
contaminant mass fraction and the excess layer thickness. The second method that was
developed in this study relies on a perpendicular traverse of the flow stream with the
measurement probe. To achieve this, a linear positioning device with a graduated knob
was adapted to the quartz tube as shown in Figure C5. The second method (traverse
method) is used to calibrate the effect of contaminant thickness (path length) and the
proximity of incident intensity. The traverse method is essential for splitting the total
measured fluorescent intensity into two components: that from the diesel fuel on the test
surface and that from the diesel in the bulk flow stream. In this way, the mass of the
diesel fuel on the test surface and the composition of the fluid stream are determined.
Two standard jars were used as reference standards to set the lower (0) and upper (100)
limits of the intensity signal on the spectrofluorometer for raw measurements made at the
test section (Fr). Ajar that contained only pure water was used to zero the intensity. A
second jar that contained pure diesel fuel was used to set the intensity on the
26
-------
spectrofluorometer to 100. All raw-measured intensities (Fr) were numerically
normalized by the intensity from the zero-jar (Fo) and the maximum-jar (Fioo). The
equation for this numerical normalization is:
F =
FT-F0
F - F
1 100 -1 0
(3.2.1.2)
where the intensity of the contamination data was adjusted by no more than 0.3 % to
account for the small temperature difference (typically within ฑ 1 K) between the test
section and the bath containing the maximum-jars and the zero-jars (Kedzierski, 2006).
The maximum correction for the flushing data was approximately 1.5 %, which was
larger than for the contamination measurements due to the colder temperature of the
house tap water.
The linear form of the Beer-Lambert-Bougher law (Amadeo et al., 1971) shows that the
measured fluorescence intensity is related to the incident light intensity (/0), the
extinction coefficient (s), the concentration of the fluorescent diesel (c), the path length
of light (/), and the quantum efficiency of the fluorescence (0) as:
(3.2.1.3)
The linear criteria for eq. (3.2.1.3) (scl < 0.05) is satisfied for 78 % of the calibration
data, and the absorbance (scl) did not exceed 0.063 for all of the data. As a result, the
calibration measurements gave:
1
2.37 O^AT =1.0473 5
C/ O
-209.231m-1]/
when fitted to eq. (3.2.1.3) (Kedzierski, 2006). Using the calibration and expressing the
concentration in terms of the bulk mass fraction and the bulk liquid density gives the
calibration of the fluorescence intensity in terms of the mass fraction and path length as:
_
F =
2.3/Qg
1.04735
m
kg
lxbpbe
-209.231m-1]/
(3.2.1.4)
Note that the concentration of the fluorescent diesel fuel has been replaced with the
product of the bulk contaminant (diesel) mass fraction (xb) and the density of the bulk
27
-------
mixture (#,) divided by the molar mass of the contaminant (Mc). The mixture densities
were calculated from a linear mass weighted basis of the pure fluid specific volumes.
Because the actual concentration of the diesel fuel entrained in the water flow stream is
unknown, eq. (3.2.1.4) cannot be directly used to obtain the excess layer of diesel on the
pipe surface. The /"and the 7e must be obtained from additional information that is
obtained from a perpendicular traverse of the flow stream. As shown in Figure C5, a
linear positioning device with a graduated knob was used to traverse and locate the quartz
tube relative to the test surface and thus measure the path length of the incident light
through the fluid. Measurements of the fluorescence intensity (F) for various path
lengths provided sufficient information for obtaining both the bulk mass fraction and the
excess layer thickness. The methodology for this is explained in the following.
The total fluorescence signal (F) can be separated into three components along the path
length while assuming a uniform bulk mass fraction. The total intensity is the sum of that
contributed by the bulk concentration for the entire path length and that in the diesel
excess layer minus the intensity that would have been due to the bulk concentration but
did not occur because it was displaced by the excess layer:
(3.2.1.5)
Where,
Fix -xh = the fluorescence contributed by the bulk concentration
ixmxb J
FI = fluorescence that would have occured without the excess layer and,
exm=xb
Fi = fluorescence due to diesel
eXm=X-L
Substitution of eq. (3.2.1.4) into the components of the above equation and grouping like
terms gives:
F = 2.3I0ฎeM;1 \hbpb -lexbpb +lepd] (32.
Here p& is the density of liquid diesel.
For a given probe traverse, the only variable in eq. (3.2.1.6) is the path length.
Consequently, eq. (3.2.1.6) can be arranged in terms of two regression constants for a
single traverse:
28
-------
?-209.23[nT1]/
s
(3.2.1.7)
Comparison of equations (3.2.1.6) and (3.2.1.7), yields the bulk mass fraction as:
-209.23m-1/
1.04735
i *' i i u
(3.2.1.8)
and the excess layer thickness as:
The average uncertainty of/e for the measurements with the iron, the PVC, and the
copper surfaces for the contamination measurements was approximately ฑ 0.06 |j,m,
ฑ0.1 |j,m, and ฑ 0.2 |j,m, respectively. The average uncertainty of x\> was approximately
ฑ 0.00008.
The diesel bulk mass fraction of the tap water used during the flushing tests is zero. For
flushing tests, eq. (3.2.1.8) produced a non-zero bulk mass fraction with a magnitude
3 The methodology presented here is a refinement of that given in Kedzierski (2006).
Equations (3.2.1.8) and (3.2.1.9) are more explicit than those presented in
Kedzierski (2006), but give the same results.
29
-------
close to the uncertainty of the measurement, i.e., typically 0.008 % (80 ppm). An
alternative approach for the flushing tests that forces the bulk mass fraction to be zero is
to start by setting x\, = 0 in eq. (3.2.1.6) and taking its derivative with respect to the path
length.
Rearranging the resulting differentiated equation and solving for the excess layer
thickness yields:
l = dl _ \ 209.231m-1]
209.23[m-1]pd2.3/0d>ffM;1 1.04735| ^ ป0 IAM 2 1
Equations (3.2.1.9) and (3.2.1.10) are equivalent for negligible A\, which is the case for
flushing. However, eq. (3.2.1.10) was used to obtain the 4 for all of the flushing
measurements because of its more explicit derivation. The average value of 4 was used
for a given measurement probe traverse. The average uncertainty in 4 for the iron and
the PVC flushing tests was approximately ฑ 0.02 |j,m, and ฑ 0.05 |j,m, respectively.
3.3 Screening Tests
A methodology for rapid identification of contaminant/substrate combinations that
merited more detailed analysis was developed and utilized for screening purposes. The
purpose of the screening tests was three fold. First, due to the large number of
contaminant/substrate combinations and the time required to conduct each full scale test,
it was deemed desirable to focus on the combinations that would be most useful in
providing information to support recommendations for decontamination. In other words,
rather than trying to exhaustively test every combination, which would not have been
possible in any case given the time constraints, the screening tests would identify the
contaminant/substrate combinations that showed the potential for accumulation, allowing
those to be tested in the full-scale test systems. The second purpose for the screening
tests was to identify contaminant/substrate combinations that might result in damage to
the full-scale plumbing loop. This could result in the need for time-consuming and costly
repairs or cause other troublesome measurement problems, such as odor or disposal
issues. The third purpose for the screening tests was to provide a broader range of
measurement results to help generalize the decontamination recommendations, albeit
with data from smaller scale measurements.
High concentrations (mass fractions) of contaminants were placed in small jars, and a
small piece (coupon) of each of the test materials was separately placed in a jar and
sealed. See details in Table 22. Following several days of exposure which varied based
on scheduling constraints, the test material was removed from each jar and evaluated for
the presence of contaminant. This was followed by flushing with clean tap water for one
hour and a recheck for contaminant, and then another clean tap water flush for 24 h and a
final measurement. The flushing methodology, shown in Figure C6, involved inserting
the exposed coupon into a piece of flexible hose through which cold tap water was
30
-------
directed at a flow rate of approximately 0.25 L/s (4 gallons per minute [gpm])
corresponding to a Reynolds number of approximately 30,000 (fully turbulent flow).
Table 2 Contaminant Mass Fractions for Screening Tests (Concentration)
Contaminant
Diesel
100%
Gasoline
100%
Strychnine
0.5 %
Cyanide
1%
Phorate
100%
These tests represent a somewhat severe exposure scenario, since the contaminant
remains in contact with the substrate for an extended period of time. This type of
scenario would be one that might occur near the point of contaminant introduction in an
actual event. Because the objective was to ensure the removal of any significant amount
of accumulated contaminants, this worst-case analysis is appropriate.
The presence of contaminant on the exposed coupons was evaluated using a Raman
spectrometer (microscope) to look for the characteristic Raman spectral signatures
associated with each of the contaminants. These signatures were determined by first
obtaining Raman spectra of each contaminant, along with Raman spectra of each
unexposed substrate material. Then, the Raman spectra of the exposed substrates were
compared to the baseline and contaminant spectra, and the presence of the contaminant
inferred by the observation of characteristic Raman peaks, and the magnitude of any
contaminant accumulation estimated by measuring the height of specific characteristic
peaks.
Since the purpose of the screening measurements was to identify the presence of the
contaminant rather than the quantity, the contaminant accumulation values are viewed as
relative values. However, a later section will attempt to estimate the thickness of the
layer of accumulated contaminant for some of the measurements. Note that acceptable
levels of residual contaminants within water supply systems are a public health matter
that is beyond the scope of this analysis. However, estimation of residual contamination
and the methods for effectively removing contaminants is the focus of this work.
3.4 Full-Scale Dynamic Tests
Tests were conducted at NIST's Building and Fire Research Laboratory (BFRL )
Plumbing Test Facility, which emulates a full-scale building water supply system in a
controlled laboratory setting. The test facility includes water supply piping, fittings, and
fixtures that represent a five-story building. It also incorporates a computer-based control
and data acquisition system which can be programmed to circulate water according to
any desired profile. The facility includes a measurement station for water tanks, such as
hot water heaters. Figure C7 presents a schematic profile of the full-scale plumbing
system test facility.
The tests introduced a water/contaminant mixture into the water distribution system, and
allowing it to circulate and/or stand for a set period of time, followed by a flushing or
cleaning operation. Pipe and water samples were collected at various stages of each test
31
-------
run and then analyzed for the presence of accumulated contaminant. In this manner, the
tendency of different contaminants to accumulate on exposed components of building
plumbing systems could be determined along with the effectiveness of flushing and
cleaning procedures.
3.4.1 Building Plumbing System Loops
The BFRL Plumbing Test Facility was used to measure accumulation and removal of
contaminants in full-scale building plumbing systems, including pipe loops and hot water
heater tanks. The testing methodology is briefly summarized as follows:
1. Water/contaminant mixtures were prepared in a large tank, and then circulated by
pumping them through a simulated three floor building plumbing system. The
simulated building plumbing system included measurement sections with small
removable pieces of copper (3/4 and 1A inch) and PVC (1/2 inch) pipe.
(Exception: For cyanide and strychnine, the entire test loop was not subjected to
contaminant exposure due to the toxic nature of these compounds. So, for these
tests, the pipe samples were exposed to the contaminant by suspension in a small
container, and then inserted into the pipe loop for flushing.
2. Pipe samples were removed after exposure for measurement using a Raman
spectrometer (microscope)
3. The plumbing system was then flushed with clean, cold tap water, at a flow rate
of 0.25 L/s (4 gallons per minute [gpm]) at the pipe samples, and additional pipe
samples were removed periodically and measured.
4. At the conclusion of the flushing tests, all components of the pipe loop were
thoroughly cleaned as needed with hot water and detergent.
Tests were conducted using diesel fuel, strychnine, sodium cyanide and Bacillus
thuringiensis (BT) spores, but the tests conducted using sodium cyanide did not provide
any useful results. Figure C8 shows a schematic of the test loop. While the figure
indicates only a single sink, in actuality there were three sinks on each of three levels,
served by different types and sizes of piping materials.
3.4.2 Hot Water Heaters
Tests were conducted with previously used nominally 189 L (50 gallon) electric hot
water heaters using the following contaminants:
Diesel fuel
Strychnine
Sodium cyanide
Bacillus thuringiensis (BT) spores
32
-------
As with the pipe loop tests, water/contaminant mixtures were made and added to empty
hot water heater tanks, then allowed to soak, followed by flushing and/or draining and
refilling with clean tap water. Periodic samples were taken of sediments from the bottom
of the tanks using a long thin tube connected to a vacuum source, with the opening of the
tube resting against the inner side of the tank bottom. Periodic samples were also taken of
tank water from the center of the tanks.
Except for the second diesel fuel test, the tank heating elements were not energized. In
some cases, multiple Raman measurements were made on each sample or multiple
samples, so data are shown as maximum, minimum and average readings. This occurs
because contaminant accumulation is not uniform, and the measurement area of the
Raman system is very small (less than a square mm). Multiple measurements were made
at various locations on the sample surface. The particular readings that showed maximum
values were viewed as the most significant since it was desired to ensure adequate
removal of accumulated contaminants. Figure C9 shows a schematic of the hot water
heater testing apparatus.
33
-------
Chapter 4 MEASUREMENT RESULTS
4.1 Bench Scale Tests
4.1.1 Chemical Contaminants
Detailed results of bench scale tests of water mixtures with phorate, toluene, gasoline,
diesel fuel, strychnine, cyanide salts, and mercuric chloride are provided in Subsections
4.1.1.1 through 4.1.1.6.
TableS below summarizes the results for static adsorption experiments and subsequent
decontamination procedures. These results can be used as a general guide for worst case
chemical contamination; however, the guide is limited because no realistic pipe-flow
conditions were used. In general, all contaminants physically stick to most of the tested
pipe substrates, but pipe composition affected the ability of some contaminants to
interact. For organic contaminants, one evidence for interaction was simply in the odor
of the pipe substrates which lingered after adsorption tests. Analytical methods were also
used to examine the pipe substrates for trace amounts of contaminants. Water
decontamination of exposed pipe substrates removed a fraction of most of the chemical
contaminants. This fractional removal is referred to as "partial" in Table 3. For these
decontamination experiments, trace amounts of the chemical were found analytically in
the water. In some instances, pipe characterization also revealed the presence of the
chemical in question. The qualitative nature of some of these above measurements
precluded efforts to quantify the amount of removal.
Decontamination using bleach solutions (2.6, 5.3,10.6 and 21.1 mL bleach/450 mL of tap
water) was also completed for a subset of chemical contaminants. In these cases, only
small amounts of the chemical were removed from the pipe and is indicated as "partial"
in the table. Trace amounts of the chemical were found analytically in the water. The
limitations of the data prevent the giving of a quantitative value for the amount partially
removed. In some instances, pipe characterization revealed the presence of the specific
chemical. Typically the chemical's odor was present both in the water and on the pipe
surface.
34
-------
Table 3 Summary of Chemical Contaminant Interaction with Pipe Materials and
Decontamination Results
Contaminant
Diesel fuel
Gasoline
Toluene
Phorate
Strychnine
Cyanide Salts
Mercuric
Chloride
Did It Stick?
Yes, All materials*
Yes, All materials*
Yes (CuF, CuB, CuIS,
PVC, PVCB)
Yes, All materials*
Yes, All materials*
Yes (Reacts with all
metals; CuF, CuB,
CuIS, BR, Fe)
Yes, All materials*
(reacts with all metals
to form elemental Hg
and can absorb into
plastics)
Did It Flush With
Water?
Yes, partial
Yes, partial
Yes, partial
Yes, partial
Yes, partial
No, reaction product
formed on metal
surfaces.
Yes, partial (not
advised for metals
as elemental Hg is
present)
Removed by Other
Additive (specified)
Bleach solution,
partial
Bleach solution,
partial
Bleach solution,
partial
*Pipe materials
heater (CuIS)],
(RB).
represent Copper [clean flat (CuF), biofilm growth (CuB), used in hot water
PVC [clean (PVC), biofilm growth (PVCB)], iron (Fe), brass (BR), and rubber
4.1.1.1 Phorate
Phorate adsorption experiments showed an interaction of phorate with all pipe materials
[Cu, PVC, calcium carbonate (CaCCb), brass, iron, and rubber]. A simple test of odor
detection further indicated the presence of phorate. It was assumed that phorate, an
organic compound, would absorb more on plastic and rubber pipes than with metal pipes.
Phorate in water was analyzed with PT-GC/MS. Concentration profiles from water
analysis showed a general decreasing trend in phorate concentration from initial phorate
concentration with exposure time to some equilibrium phorate concentration for all pipe
materials. This equilibrium phorate value depended on the initial phorate concentration
and type of pipe material. GC/MS results showed similar phorate species for control
solutions and all pipe materials with no evidence of secondary products. The extent of
phorate interaction was determined by comparing the change in concentration between
the initial and the equilibrium values. A ranking of decreasing interaction was CaCCb,
PVC, Cu pipe with deposits, and clean flat Cu. Factors influencing the ranking included
high surface area of CaCCb, polymeric material (i.e., PVC) interaction with the organic
compound, and composition and roughness of the Cu deposit.
35
-------
The measurement of water quality parameters also yielded insight into contaminant
interaction with pipe material. An increase in TOC concentration was observed, but it
never reached a level that was the stoichiometric equivalent to the initial phorate
concentrations, indicating the lack of dissolution of the organic material in water. There
were no clear trends in TOC values with pipe material, but phorate/PVC pipe solutions
showed the greatest increase in TOC values, perhaps indicating a partial dissolution of
PVC by phorate. Turbidity values fluctuated over phorate concentration ranges, but an
increase was observed with Cu in-service pipes in which visible pipe deposits were
observed in the water solution.
Pipe material characterization was essential in determining the fate of phorate
contamination on pipes because of low solubility of phorate in water. In the preparation
of the phorate solutions, a separate layer of phorate was observed even though the
solutions were made at the solubility limit or below. This additional phase complicated
fate determination and necessitated a more detailed characterization of pipe material. By
analyzing the pipe material, deposition of the contaminant can be measured directly to
establish a correlation with data from water analysis. Fourier Transform infrared (FTIR)
micro spectroscopy data showed that a film of contaminant was present on all Cu pipe
samples. IR bands in the 1000 to 800 cm-1, characteristic of phorate dominate the
phorate exposed Cu pipe. Figure CIO shows IR mapping of the contaminant on pipe
materials. The light blue plot represents pure phorate, the other plots represent the IR
map of the pipe surfaces under different conditions. For the latter spectra some of the
peaks in the spectrum were due to deposits on the pipe surface which might have
obscured phorate peaks. The dark blue plot indicates that there was a heterogeneous layer
of phorate spread on the surface of the pipe. Decontamination of phorate from the pipe
required a bleach solution which likely oxidized the phorate molecule. Ku and Lin, 2002
and Hong and Pehkonen, 1998 describe the environmental fate of phorate, and oxidation
was given as one mechanism for phorate degradation. Decontamination of pipe materials
with fresh tap water did remove small amounts of phorate as observed in
decontamination solutions using PT-GC/MS for initial phorate concentrations > 24.8
mg/L phorate. FTIR microscopy showed that water rinsing alone did not remove phorate
on Cu in-service pipe, but solutions of bleach (as low as 5.3 mL/L) removed the phorate
from the surface of the Cu pipe to below the detection limit of the FTIR microscopy
method (Figure CIO).
4.1.1.2 Toluene
Since toluene is a component of gasoline, it was chosen as a reference material for
gasoline. Toluene interacts more strongly with polymeric pipe materials than with metal
pipe. PT-GC/MS was used to follow the change in toluene concentration over exposure
time for Cu, PVC, and CaCOs. GC/MS analysis showed no evidence of secondary
products, and species were the same for control solutions and all pipe materials. Most of
the toluene concentrations (500 mg/L, 1.9 g/L, 3.8 g/L, 7.5 g/L, 15.1 g/L, and 30.0 g/L)
used in these adsorption experiments were much greater than its solubility in water (500
mg/L). Because it is less dense than water, more toluene could have been lost to the
atmosphere upon collecting water samples for analysis. Concentration profiles from PT-
36
-------
GC/MS water analysis showed a general decreasing trend in toluene concentration from
the initial toluene concentration with exposure time to some equilibrium value for all pipe
materials. This equilibrium toluene value depended on initial toluene concentration and
type of pipe material. There were no clear trends between different pipe materials. This
could be due to the fact that the initial toluene concentration was high, which would have
made it difficult to see a relatively small decrease in concentration due to adsorption.
Pipe material characterization was used to determine the fate of toluene water mixture on
the pipe material. Raman micro-spectroscopy was used to examine PVC and Cu pipe
materials. Although there was a detectable toluene smell from the pipes, for all Cu pipe
materials (clean flat, biofilm growth, and in-service), there was no significant difference
in the Raman spectra for initial Cu pipes and those exposed to various concentrations of
toluene. This could be explained by the volatility of toluene and power intensity of the
laser used to obtain Raman signals. However, modifying Raman instrument parameters
to decrease laser intensity did not change results for Cu pipe samples. PVC pipe material
showed visible swelling of the pipe material induced from toluene in water, and toluene
was observed in the Raman spectrum for 15.1 g/L and 30.0 g/L pipe samples (Figure
Cll). No toluene was observed on exposed CaCOs powder using diffuse reflectance
Fourier-Transform spectroscopy (DRIFTS). This lack of signal could be due to toluene
evaporation from the filtration method used to remove contaminated water for DRIFTS
analysis.
Water quality parameters were measured to gain insight into contaminant interaction with
pipe material. TOC concentration did not always increase linearly with toluene
concentration and never reached a level that was the stoichiometric equivalent to the
initial toluene concentration, an indication of the insolubility of toluene in water. PVC
exposed pipe materials showed the greatest increase in TOC value. Turbidity values
exhibited no clear trend with toluene concentration, and no trends were observed in the
conductivity, chlorine, and alkalinity methods.
Decontamination of toluene pipes with fresh tap water appeared to remove a fraction of
the toluene from the exposed pipe. Analysis of decontamination water with PT-GC/MS
showed that amounts of toluene were present in the decontamination water for all pipe
materials. The toluene concentration detected was greatest for water that had been used
to decontaminate PVC pipe samples. The measured toluene concentration in the
decontamination water was about half of the measured initial toluene concentration in the
adsorption experiments, regardless of initial toluene contamination concentration. Water
characterization of decontamination water showed a slight increase in conductivity and
increase in TOC value. Decontaminated pipe material analysis with Raman microscopy
did not show any change in the toluene spectrum compared with the initial contaminated
spectrum, thus indicating toluene was still present after decontamination. However,
based on the analysis for toluene in the water samples, decontamination with water did
remove a fraction of the adsorbed toluene.
4.1.1.3 Gasoline and Diesel Fuel
37
-------
Gasoline and diesel fuels were tested separately on Cu (clean flat, with biofilm, and in-
service from a hot water heater line), PVC (clean and with biofilm), and CaCOs pipe
materials. Fuel concentrations (500 mg/L, 1000 mg/L, and 2000 mg/L) in these
adsorption experiments were much lower than that used in toluene adsorption
experiments. PT-GC/MS was used to follow the change in fuel concentration in the fuel
water mixture with exposure time to the pipe materials. Each fuel water mixture had a
unique gas chromatogram (Figured 2). The chromatogram for fuels consisted of many
components, each of which was identified using GC/MS and a MS library search. There
was no evidence of secondary products and the species were the same for control
solutions and all pipe materials with exposure time. For gasoline experiments, toluene
and four components consisting of benzene and benzene derivatives were followed over
exposure time. For diesel fuel experiments, benzene and naphthalene components were
monitored. For both fuels, a slight decrease in peak concentration with exposure time
was observed in concentration profiles for all monitored components. Furthermore, there
was no significant difference in concentration profiles for control or blank experiments
(no pipe materials). These results might indicate that the initial fuel concentrations were
too high to measure any adsorption onto the pipe material.
Raman micro-spectroscopy was used to determine the fate of fuel water mixture on PVC
and Cu pipe materials. Although there was a detectable fuel odor from all exposed pipe
materials, for most pipe samples no Raman signal was detected for the fuel. However,
the Raman spectrometer did detect the presence of fluorescence for pipe materials
exposed to diesel fuel. It should be noted that the biofilm growth on the PVC and Cu
pipe materials also caused fluorescence, a potential interference for those pipe materials
with biofilm. In general, there was no significant difference in the Raman spectra for the
initial Cu pipes and those exposed to the various concentrations of fuel. However, diesel
peaks were observed on Cu flat pipe that was exposed to 2000 mg/L diesel (Figured 3).
The general lack of Raman signal of the fuel on most pipe samples could be explained by
the volatility of the fuel and the power intensity of the laser used to measure Raman
signals. However, modifying the instrument parameters to decrease laser intensity did
not change the results for any of the pipe samples.
Water quality parameters were measured to gain insight into contaminant interaction with
pipe material. An increase in TOC concentration was observed, but did not increase
linearly with increasing fuel concentration. Also, it never reached a level that was the
stoichiometric equivalent to the initial fuel concentration, an indication of the insolubility
of the fuel in water. In general, the pH increased with fuel addition, but was not
dependent on fuel concentration. A decrease in chlorine content was observed for all
adsorption experiments. There was also an increase in alkalinity, but no significant
change in conductivity. Turbidity values increased with fuel addition, but no clear trend
was observed with pipe material. This increase in turbidity could be due to release of
scale that had formed on pipe material or the insolubility of the fuel in water.
Decontamination of fuel-exposed Cu and PVC pipes with fresh tap water appeared to
remove a fraction of the fuel from the contaminated pipe materials. Bleach solutions
(2.6, 5.3, 10.6 and 21.1 mL bleach/450 mL of tap water) were also used to decontaminate
38
-------
both fuels individually from clean flat Cu pipe material. Analysis of the decontamination
tap water with PT-GC/MS showed that fuel was present in the decontamination water for
all pipe materials. The amount of fuel present depended on the time elapsed between
contamination (adsorption experiments) and decontamination procedures. All exposed
pipe samples were removed from the contaminated water solution and placed in sealed
jars without water. For samples immediately decontaminated, more fuel was present in
the decontaminated water as shown by PT-GC/MS analysis. The measured
concentration of fuel in the decontamination water ranged from 350 times to 1000 times
lower than that measured for the initial contamination fuel concentration in the adsorption
experiments. No clear trend was observed for the various initial fuel concentrations and
pipe materials because the time elapsed between decontamination varied. PT-GC/MS
analysis of bleach solutions used in decontamination did not show any peaks for the fuel
components.
Measurement of water quality parameters in the decontamination water further confirmed
the removal of fuel. TOC values increased with increasing initial contamination fuel
concentration, but were on average 7 times lower than the TOC value observed in
adsorption experiments. The TOC value also depended on the time elapsed between
adsorption and decontamination, so a greater TOC value was observed for more
immediate decontaminations. The pH also increased slightly and the chlorine content
decreased to near zero. In general, decontaminated pipe material analysis with Raman
microscopy did not show any change in the spectra compared to the fuel exposed
coupons prior to decontamination. Only the 2000 mg/L diesel exposed clean flat Cu pipe
showed evidence of fuel removal via the disappearance of the diesel Raman peaks and
the fluorescence after water decontamination.
4.1.1.4 Strychnine
The entire set of adsorption and decontamination experiments were not carried out for
strychnine. Preliminary results of pipe materials analysis using fluorescence imaging was
obtained for a "worse case" scenario experiment using 1 g strychnine hydrochloride
dihydrate in 65 mL of tap water. Pipe samples consisting of Cu in-service from a hot
water heater, brass, iron, rubber and PVC were exposed to the strychnine solution for 2
weeks and were rinsed in 450 mL water before fluorescence imaging analysis
(Appendix A). No visible evidence of reaction or interaction was observed (Figure
C14a). However, a distinctive fluorescence signal from the strychnine exposed samples
was observed. Fluorescence imaging was used to determine percent coverage of the
fluorescing signal on the pipe material (Figure C14b) using principal component analysis.
Principal component analysis showed that brass and iron were 100 % covered, PVC was
78 % covered, and Cu in-service was 66 % covered. These results show that strychnine,
an organic nitrogenous base, does interact or stick to pipe surfaces even after tap water
decontaminati on.
4.1.1.5 Cyanide Salts
Adsorption experiments showed a reaction of CN" with metal pipe materials. This is not
surprising as cyanide solutions are used in electroplating applications for brass and zinc.
39
-------
The reaction was especially strong with Cu pipe, a focus of these adsorption experiments.
Metal/ CN" reactions are a potential complication. They may leave CN" and metal CN
complexes in the contaminated water, which could produce toxic HCN during
decontamination (Ismail et al., 2009; Hamid and Aal, 2009). The extent of reaction
depended on the pipe material and initial CN" concentration. At CN" concentrations of
310 g/L in tap water, a Ig piece of Cu pipe dissolved. Iron and brass pipe samples with
similar masses also showed a reaction with this high cyanide concentration and a
passivation layer resulted on the surface of the metal, leaving both metal pipes with a
silver surface. The rate of change in CN" concentration was measured by a CN" ion
selective electrode (ISE), and attempts were made to correlate CN" concentration with the
amount deposited on the surface of the Cu pipe. The CN" concentration did decrease
slightly with increasing amounts on the surface, but given the heterogeneous nature of the
deposit, no correlation was developed. For the lowest initial CN" concentration, 3 mg/L,
an equilibrium CN" concentration value approached zero. There were also visible
differences in the appearance of Cu materials from a shiny orange color to a dull black
deposit. The black deposit also precipitated into the water solution. Analytical
techniques, X-ray photoelectron spectroscopy (XPS), x-ray diffraction spectrometry
(XRD), and scanning electron microscopy (SEM), showed no cyanide species present on
the surface of Cu pipe materials. The techniques showed that the black deposit was made
up of copper oxide and a carbonaceous substance.
There was no apparent interaction between CN" and PVC pipe. No apparent reaction
occurred with PVC pipe or rubber plumbing materials. An adsorption study using 310
g/L CN" resulted in no physical change to the PVC or rubber material, but the CN"
solution changed from clear and colorless to clear amber. With CN" concentrations up to
50 mg/L, the CN" ISE response was that of a slight increase in CN" concentration over
time, but the signal was very noisy, an indication of the presence of some interfering ion
with the CN" ISE probe, perhaps Cl".
Water quality parameters were measured to gain insight into contaminant interaction with
pipe material. A large increase in conductivity values was observed for all CN" solutions.
The conductivity value leveled off to about 16 mS regardless of initial CN" concentration.
Typical tap water conductivity was 250 jiS. Turbidity for CN" solutions increased at the
completion of the experiment. The free and total chlorine did not always approach zero
values as was observed for the organic contaminants. The total organic carbon (TOC)
concentration increased with increasing CN" concentration, but never approached a level
that was the stoichiometric equivalent to the initial CN" concentrations. A research group
at University of Maryland, Baltimore County, has developed an ion chromatography
method based on an application note to detect metal/CN complexes in water for this
research project. Results showed that metal/CN complexes are present in CN" solutions
exposed to Cu substrates. This was expected since no CN compounds were detected on
the pipe surface. In addition the metal/CN complex concentration depended on the initial
CN" concentration. Furthermore, the concentration of the metal/CN complex changed
over time; it was found that the metal/CN complex decreased with storage age of the
experimental solution after the adsorption experiment (Appendix B).
40
-------
Decontamination by rinsing the exposed Cu and PVC pipe materials with fresh water did
not produce any CN" species in the decontamination water. In light of these results,
decontamination studies with bleach or the other detergent additive were not performed.
4.1.1.6 Mercuric Chloride
Adsorption experiments showed that mercuric chloride (HgCb) reacted with metal pipe
materials. Metal oxides and chlorides were formed for all metals. For experiments using
36.9 g/L HgCb, elemental mercury was formed for the brass and iron pipe samples and
found in contaminant solution. Furthermore, a mercury mirror formed on the surface of
the brass and iron pipe samples. The Cu substrate showed a light blue precipitate rather
than evidence of elemental mercury. However, SEM analysis of the Cu substrate showed
droplets of elemental mercury intermixed with copper oxide (Figured 5). Other pipe
substrate analysis methods are reported later (Appendix A) and showed no evidence of
mercury attaching to Cu substrates, however this method (Raman) was not as sensitive as
SEM.
Mercury levels in water were also monitored for a portion of the adsorption experiments.
For PVC, mercury was monitored for the experiments which had starting concentrations
of 100 and 500 mgl/L HgCb, and for copper, mercury was monitored for the experiment
which had a starting concentration of 1000 mg/L HgCb. The mercury concentration
steadily decreased over exposure time for Cu pipe, but no change was observed for PVC
pipe ( Figure C16). These results further indicate a reaction with Cu, but not for PVC.
The results from water quality measurements for the HgCb /Cu experiments showed an
increase in conductivity and turbidity, which increased with increasing initial HgCb
concentration. The TOC concentration for HgCb /Cu experiments only slightly increased
compared to the initial tap water value. Conductivity and turbidity did not change
significantly for HgCb/PVC experiments, but the TOC concentration was twice the value
of tap water.
Given that mercury metal was formed with metal pipe materials, it was determined that
decontamination would not be possible due to safety restrictions in our laboratory. PVC
pipe material should be analyzed for the presence of mercury before determining if
decontamination would be appropriate. A flushing of tap water with 5 % nitric acid can
work to solubilize any mercury compounds, although this was not attempted due to the
safety concerns.
4.1.2 Biological Contaminants
Results of bench scale tests of water mixtures with biofilms, spores, bacteria, and ricin
are provided in Subsections 4.1.2.1 through 4.1.2.4. The results of the work with
Bacillus spores are reported in more detail (EPA and NIST, 2011).
4.1.2.1 Biofilm Reactor Measurements
The biofilm organisms for these three biofilm reactor configurations reached similar
levels during the three week growth period (Figure C17). These levels are comparable to
41
-------
the levels measured in a water distribution system that had measured values of 2.3 x 10
CFU/m2 and 1.4 x 106 CFU/m2 for PVC and copper, respectively (Schwartz et al., 2003).
4.1.2.2 Inactivation of Spores with Biofilm Conditioned Pipe Materials
As mentioned in Section 3, three reactor configurations for testing spore adhesion to
biofilms and subsequent disinfection. Summarized below are the reactor configurations
and the ways in which spore contact was accomplished:
CDC bioreactor. After biofilm had developed, spore contact was accomplished in
one of two ways, either with stirring by stirring blade in place (60 rpm) or without
stirring (only gentle mixing to prevent the settling of the spores). Both methods
were carried out for 24 hours.
Pipe section reactor with creeping flow. After biofilm development, pipe sections
were placed in beaker and contacted with BT spores for 24 h (with gentle mixing
to prevent the spores from settling).
Pipe section reactor with high flow. After biofilm development, the pipe section
reactor was left intact and a spore suspension was recirculated for 24 h with
intermittent high flow (2.5 L/min cycling at 2 h of flow and 2 h no flow).
The concentrations of BT spores adhered to the different configurations are shown in
Figure CIS. The pipe section reactor contacted with spores with creeping flow and the
CDC reactor contacted with spores without the baffle (both low shear conditions) had
low levels of spore adhesion. The CDC reactor with baffle stirring (60 rpm) during spore
contacting and the pipe section reactor with high flow (2.5 L/min) during contacting had
much higher levels of adhered BT spores (Figure CIS). This data suggests that both
greater mixing and higher contact flow rates resulted in a greater amount of spores
adhering to the biofilm-condition coupons and pipe sections.
In the disinfection experiments using the CDC reactor and the pipe section with creeping
flow, the coupons and pipe sections were contacted with chlorine and monochloramine
solutions in containers without shear. In these experiments, the spores adhered to the
biofilms were very resistant to disinfection by chlorine or monochloramine. Extremely
high concentrations of chlorine (100 mg/L) were required to achieve any significant
reduction in the levels of adhered spores even after prolonged contact times, i.e., 20 to 60
minutes. This is discussed in more detail in Morrow et al., 2008.
Disinfection results were also obtained with the pipe section reactor operated with high
flow rate (2.5 L/min). The pipe section reactor was flushed first with water, drained, and
then flushed with chlorine solutions (10 mg/L); both flushing steps were at high flow
rate. In these experiments a flow rate of 2.5 L/min in 19 mm diameter pipe sections
resulted in a Reynolds number of approximately 2800, indicating the flow is transitioning
to turbulent conditions. The effect of a 2 h flush with tap water on the removal of spores
was reduction by approximately 0.5 logic.
42
-------
The pipe section reactor operated at high flow rate required chlorine disinfection with
two solutions of chlorine. This was because the large surface area of the pipe sections and
tubing with the biofilm resulted in a large chlorine demand. The first chorine solution (2
L with a concentration of 10 mg/L) was depleted to approximately half of the initial value
after 30 min. At this time, a new chlorine solution was used to complete the disinfection
process. The second disinfection solution (2 L with concentration of 10 mg/L) was also
reduced to approximately half of the initial value by the end of the second disinfection
step (150 min). The first 30 min disinfection step resulted in a reduction of approximately
1.5 logic of BT spores associated with the surface and the second disinfection step
resulted in a further reduction of approximately 1.5 logic. The overall total reduction of
BT spores was approximately 3 to 4 logic in the pipe section loop for both the PVC and
copper pipe sections. Results are reported in more detail in the previous report on
Bacillus spore disinfection (EPA and NIST, 2011). The biofilm organisms were reduced
by a similar magnitude during the disinfection process using the high flow rate
conditions. This is in contrast to the minor reduction of biofilm organisms seen in the
disinfection steps using the first two configurations. These results point to the value of a
high flow rate to achieve disinfection of BT spores adhered to biofilm conditioned pipes,
when compared to previous studies using low flow rates.
Studies using germinants were also completed using germinants to determine if
germinating the spores prior to disinfection would increase the effectiveness of the
disinfection procedures. Amino acids were contacted with the spores to encourage
germination. After germination commenced, the organisms were then contacted with
disinfection solution. Results are reported in more detail (Morrow et al., 2009; EPA and
NIST, 2011). The net effect of the germinants was to significantly decrease the
concentration of the spores on the surface of the biofilm-conditioned coupons (reduction
of 1.5 to 3 logic). The germination step resulted in the organisms (both in suspension and
adhered to the biofilm) being more susceptible to inactivation by chlorine,
monochloramine, or heat (50 ฐC).
4.1.2.3 Measurement of Inactivation of Bacteria in Solution
The inactivation of spores and bacteria in solution was measured in order to have data to
compare the inactivation results when the spores or bacteria are adhered to biofilms on
the surface of pipes. A common method to express inactivation of bacteria is to multiply
the concentration of the disinfectant by the time required to achieve a specified log
reduction in the number of viable bacteria. This CT value (concentration x time) was
used to compare disinfection efficiencies. For a given logic reduction a number of
important solution variables will determine the CT values obtained. Temperature, pH,
the and the ionic composition can have a major effect. With an initial chlorine
concentration of 10 mg/L chlorine in synthetic water (pH 8.2, ambient temperature), CT
values, along with one standard deviation of the mean (in parentheses), were calculated to
be 615 (13) mg-min/L for BT and 294 (12) mg-min/L for BA, for a 2 logic reduction. In
addition the CT value of BT spores was calculated from inactivation tests using a 0.1 M
sodium phosphate buffer (pH 7.8) and an initial test solution concentration of 10 mg/L
chlorine. A CT of 150 (60) mg-min/L was calculated for 2 logic reduction The
differences in these two values could be due to the phosphate buffer which results in a
43
-------
lower pH and an accompanying increase in the ratio between hypochlorous acid (the
more germicidal species) to hypochlorite ion. The above results are reported in more
detail in Morrow et al., 2008 and EPA and NIST, 2011.
4.1.2.4 Inactivation ofRicin by Disinfectants
The biological activity of ricin, lactate dehydrogenase, and lysozyme, (the latter two
compounds being simulants for ricin) was measured after treatment with solutions of
chlorine and monochloramine. Chlorine efficiently inactivated all three proteins. Low
concentrations of bleach (2.8 and 3.2 mg hypochlorite/L) added to samples containing
67 ug/mL ricin inactivated over 93 % of the biological activity in 10 min, and higher
concentrations of bleach (5.6 mg/L) resulted in no detectable biological activity of ricin.
The simulant proteins lactate dehydrogenase and lysozyme were also readily inactivated
by chlorine, but the reaction with monochloramine was much slower, taking several
hours (Cole et al., 2008). Lactate dehydrogenase and lysozyme had a good correlation
between the inactivation of biological activity and the loss of native fluorescence.
Analysis of the proteins by gel electrophoresis and size exclusion chromatography
indicated that the treatment with bleach resulted in extensive modification of the structure
observed in a change of the protein charge and in an increased size due to aggregation.
The native fluorescence of proteins due mainly to tryptophan, and to a lesser extent
tyrosine and phenylalanine, was reduced by the treatment with chlorine. A reduction of
biological activity corresponded with a decrease in native fluorescence in all three
proteins, indicating the utility of monitoring of fluorescence as a way to measure the
inactivation of toxins such as ricin. Monochloramine solutions reacted more slowly with
the proteins and was not effective at inactivating ricin even at levels 10-fold higher than
chlorine. The slower kinetics of monochloramine (compared to chlorine) was observed
for all three proteins. The correspondence between inactivation and loss of native
fluorescence in the model enzymes (lactate dehydrogenase and lysozyme) makes LDH
and lysozyme useful simulants to study decontamination processes. The above results
have been reported in more detail (Cole et al., 2008).
4.2 Dynamic Fluid/Surface Interface Measurements
The dynamic fluid/surface contamination and flushing measurements that were made
following the procedure described in Subsection 3.2 are presented here. The
contamination measurements over an approximate 200 h time period were made for three
different Reynolds numbers varying from 0 to 7000. Reynolds number for pipe and other
channels with a non-circular cross-section is defined in the following equation:
(4.2.1)
where the wetted perimeter of the channel (pw) was 195 mm, the viscosity of the mixed
bulk flow (jUb) was calculated using a nonlinear mixture equation, and the mass flow rate
( m ) was obtained from the turbine meter. Flushing measurements were done for a fixed
44
-------
Reynolds number of approximately 5000. This Reynolds number was estimated to be
typical for flow rates through half-inch diameter pipe found in buildings. After each
contamination/flushing test, the test surface was cleaned with acetone and clean tap
water.
4.2.1 Contamination Excess Layer Thickness
Figure C19 provides the measured diesel layer thickness on a PVC surface as caused by
an exposure to a flowing water/diesel (99.85/0.15) mixture, i.e., diesel fuel at
approximately 0.15 % bulk mass fraction (1500 ppm). The exposure time is the duration
of the exposure test; it is the time that the test surface is exposed to the contaminated flow
starting with a clean surface. The open circle, square, and triangle symbols represent
contamination measurements for the Reynolds number = 0, 3200, and 7000 conditions,
respectively.
Figure C19 shows that the Reynolds number = 3200 and the soak (Reynolds number = 0)
exposure tests gave similar results. Specifically, the excess layer thickness was
established immediately upon exposure of the PVC surface to the water/diesel mixture
and remained nearly constant for the 200 h test duration. Only a marginal increase in the
time-averaged 4 was observed from approximately 1.32 |j,m to 1.48 |j,m when the
Reynolds number was increased from 0 to 3200, respectively. However, an increase in
the Reynolds number to 7000 resulted in roughly a 142 % increase in the time-averaged 4
over the soak condition to an average value of approximately 3.2 |j,m. In addition, a
Reynolds number of 7000 condition did not produce a nearly constant 4 with respect to
exposure time as did the conditions at Reynolds numbers of 0 and 3000. Rather, the
conditions at a Reynolds number of 7000 gave a maximum diesel thickness of
approximately 4.4 |j,m at an exposure time of 20 h. With further exposure to 140 h, the
diesel thickness decreased from this maximum to nearly the thickness at the initial
contamination, which was approximately 3 |j,m. For the PVC surface, the approximate
average 4 for the Reynolds numbers of 0, 3200, and 7000 conditions was 1.32 |j,m,
1.48 |j,m, and 3.16 |j,m, respectively. Averaging over all contaminating flow rates and
exposure times, the average 4 for Xb = 0.15 % on the PVC surface was approximately
2.0 urn.
Figure C20 gives the measured diesel layer thickness4 on an iron surface due to exposure
to a flowing water/diesel (99.85/0.15) mixture, i.e., the same composition as for the PVC
surface. In general, the flow rate had little effect on the diesel excess layer thickness.
For the iron surface, the approximate average 4 for Reynolds numbers of 0, 3200, and
7000, was 0.87 |j,m, 0.66 |j,m, and 0.60 |j,m, respectively. Averaging over all
contaminating flow rates and exposure times, the average 4 for x\> = 0.15 % on the iron
surface was approximately 0.71 |j,m. Consequently, the PVC surface adsorbs
approximately 180 % more diesel fuel than the iron surface. For copper, which will be
4 The soak measurements on the iron surface were corrected as outlined in
Kedzierski (2008) to account for additional rust resulting when the surface was exposed
to air during repair of the apparatus.
45
-------
discussed below, the average accumulation of diesel fuel for Xb = 0.2 % was
approximately 2.3 |j,m, which was comparable to the diesel accumulation on PVC
(Kedzierski, 2006).
Figure C21 cross-plots all of the contamination excess layer measurements on PVC of
Figure C19 as a function of Reynolds number. Figure C21 shows that the maximum
diesel excess layer thickness on the PVC surface of approximately 4.4 |j,m occurred
between Reynolds number of 5700 and 6300. The peak 4 for Reynolds number near
3200 was approximately 2 |j,m, which is approximately 55 % less than the maximum 4
for tests at a Reynolds number of 6000. Another 20 % reduction in the peak 4 on the
PVC surface was observed when the nominal Reynolds number was reduced from 3200
to 0. The peak 4 for the soak tests (Reynolds number = 0) was approximately 1.6 |j,m for
the PVC surface. The variation in Reynolds number for a given set of tests for "fixed"
Reynolds number was caused by an approximate 1 % variation in the water temperature
during start-up and an approximate 15 % variation in the water flow during the nearly
200 h test duration. Figure C21 and Figure C19 contain the same number of
measurements, however in Figure C19 multiple measurements of the same 4 at a "fixed"
Reynolds number appeared as a single symbol when plotted against time despite the
variation in Reynolds number. In Figure C21 the small differences in Reynolds number
are apparent, and result in the individual data points being more obvious.
Figure C22 cross-plots all of the excess layer measurements on the iron surface of Figure
C20 as a function of Reynolds number. Figure C22 shows that the maximum film
thickness of approximately 1.3 |j,m occurred for the soak tests on the iron surface. The
peak 4 on the iron surface decreases slightly from the soak condition for increasing
Reynolds number. The peak 4 on the iron surface for Reynolds number near 3200 and
6000 was approximately 1.1 (im and 1.0 |j,m, respectively. The dashed lines given in
Figure C21 and Figure C22 indicate the maximum measured excess layer for the
exposure.
Figure C23 provides the measured diesel layer thickness on oxidized copper as caused by
an exposure to a flowing water/diesel (99.8/0.2) mixture, i.e., diesel fuel at approximately
0.2 % bulk mass fraction (2000 ppm). The exposure time is the duration of exposure of
the test surface to the flow starting from when the clean surface was first exposed to a
particular flow condition. For all flow rates and exposure times, the average 4 for
Xb = 0.2 % obtained from the eq. (6) methodology was approximately 2.3 |j,m.
Figure C24 provides the measured diesel layer thickness on oxidized copper as caused by
an exposure to a flowing water/diesel (99.7/0.3) mixture, i.e., diesel fuel at approximately
0.3 % bulk mass fraction (3000 ppm). A much larger variability in the measurements is
evident for the 0.3 % mass fraction than for the 0.2 % mass fraction condition. For all
exposure times and Reynolds number, the average 4 for x\> = 0.3 % was approximately
7.4 |j,m, which is 5.1 |j,m (222 %) thicker than the average thickness observed for the
0.2 % mass fraction tests.
46
-------
Figure C25 cross-plots all of the excess layer measurements on oxidized copper of Figure
C23 as a function of Reynolds number. Figure C25 shows that the maximum diesel
excess layer thickness of approximately 8 |j,m occurred at a Reynolds number near 4800.
For Reynolds number larger and smaller than 4800, the diesel excess layer was thinner.
For example, the 4 for Reynolds number near 1900 and 3800 was approximately 1 |j,m,
which is nearly eight times less than the maximum 4- For Reynolds number greater than
6000, the 4 was approximately 3 |j,m. Figure C26 cross-plots all of the excess layer
measurements of Figure C24 (the 0.3 % mass fraction tests) as a function of Reynolds
number. Figure C26 shows that the maximum film thickness of approximately 26 |j,m
occurred at a Reynolds number of approximately 4000. Consequently, a maximum for
the diesel adsorption exists near a Reynolds number of 4000 for both free stream
concentrations. The dashed lines given in Figure C25 and Figure C26 represent the
maximum measured excess layer for each range of Reynolds number tests. The variation
in Reynolds number for a given set of tests for "fixed" Reynolds number was caused by
an approximate 1 % variation in the water temperature during start-up and an
approximate 15 % variation in the water flow during the nearly 200 h test duration.
4.2.2 Flushing Excess Layer Thickness
The flushing tests were carried out at a Reynolds number of approximately 5,000. These
tests were done on pipe material sections that had previously been subjected to the
exposure tests mentioned earlier in 4.2.1. At the end of these exposure tests, there was
still an excess layer that remained on the test pieces. For the flushing tests these exposed
test pieces were flushed with tap water and the thickness of the excess layer measured in
real time using the same apparatus as was used for the exposure tests.
Figure C19 and Figure C20 show the results of the flushing tests for PVC and iron pipe
respectively. Measurements of 4 during flushing of the surfaces that had been previously
exposed at Reynolds numbers of 0, 3200, and 7000 are represented by the filled circle,
square, and triangle symbols, respectively. For the PVC surfaces, most of the flushing
measurements are close to but less than zero. The average of all the flushing
measurements on the PVC surface is approximately -0.2 |j,m. It is likely that an
unknown bias error has caused the measurement to be less than zero because 0.2 |j,m is
larger than the uncertainty of the 4 measurement. The negative thicknesses are
interpreted as a clean surface. Consequently, the surface is clean nearly immediately
after the inception of flushing.
Negative thicknesses during flushing were observed with the exception of the flushing
test on the PVC surface that had been previously exposed at a Reynolds number of 7000.
For the first 3.6 hours of the flushing test, the excess layer thickness was measured as
positive. After this time period negative thicknesses were observed. For the
aforementioned case the initial 4 was about 0.6 |j,m and decreased to approximately
0.13 |j,m at 3.6 hours. This corresponds roughly to a 0.13 |j,m/h removal rate, which is
similar in magnitude to the flushing diesel removal rate, 0.10 |j,m/h, found for a copper
surface (discussed below).
47
-------
For the flushing of iron pipe material, the iron surface that had been previously exposed
at a Reynolds number of 7,000 was the most successfully cleaned. The average le for this
flushing test was -0.2 |j,m as a result of flushing. On average, the flushing after the
Reynolds numbers of 0 and of 3200, exposure tests left similar quantities of diesel fuel on
the surface, 0.6 |j,m and 0.8 |j,m, respectively. These numbers indicated that flushing after
the test at a Reynolds number of 3200 did not show any diesel fuel removal, while the
flushing after the soak tests (Reynolds number = 0) gave only roughly a 0.3 |j,m diesel
fuel removal. However, both flushing tests of the pipe material that had been exposed at
the two lower Reynolds numbers (0 and 3200), were corrected as outlined in Kedzierski
et al., 2008. The corrected measurements have an estimated uncertainty of ฑ 0.8 |j,m.
Consequently, flushing likely cleaned the iron surfaces that had been exposed at
Reynolds numbers of 0 and 3200 given that the measurements are within the
measurement uncertainty. This is consistent with the non-corrected flushing
measurements after the exposure test at a Reynolds number of 7000. These last results
indicated that the iron surface was clean after flushing.
Flushing tests were performed on the oxidized Cu disk that had been exposed to a
water/diesel mixture of 99.8/0.2 (i.e., 2,000 ppm diesel) at a Reynolds number of 4,600.
Flushing results are shown in Figure C23. The flushing measurements start at an 4 near
6.5 |j,m, which agrees with the value of/e at the end of the exposure test at a similar
Reynolds number, thus, confirming the repeatability of the measurement technique. The
4 decreased from approximately 6.5 |j,m to approximately 1.5 |j,m after 55 h of flushing.
This corresponds approximately to a 0.09 |j,m/h removal rate and a 77 % reduction of the
total diesel thickness over 55 h.
Flushing tests were also performed on the oxidized Cu disks that had been exposed at a
Reynolds number of 4600 and at a more concentrated water/diesel mixture of 99.7/0.3
(i.e., 3,000 ppm diesel). Flushing results are shown in Figure C24. The flushing
measurements start at an 7e of 1.5 |j,m, which agrees with the value of 4 at the end of the
exposure tests at a similar Reynolds number, again, confirming the repeatability of the
measurement technique. After approximately 20 h of flushing, the excess layer thickness
was reduced to approximately -0.5 |j,m. Given the uncertainty of the measurement, most
all of the diesel fuel was likely removed from oxidized copper by flushing with clean tap
water. The removal rate achieved was approximately 0.1 |j,m/h and agrees closely with
that achieved for the flushing tests of the pipe material that had been exposed to 2000
ppm diesel at a Reynolds number of 4600. This suggests a constant removal rate of
approximately 0.1 |j,m/h of diesel fuel from a copper surface for a flushing Reynolds
number of 5,000 and is independent of initial thickness and original contamination
concentration. No removal rate could be calculated for the flushing tests
of the pipe material that had been exposed to 3000 ppm diesel at a Reynolds number of
7000 because the tests produced a 4 near -0.5 |j,m for nearly all measurement times.
48
-------
4.3 Screening Tests
Screening tests were conducted to help identify particular contaminant/substrate
combinations that merited further testing. Raman spectroscopy was used to detect
contaminant residual during the screening tests, and are presented in this section. As
mentioned earlier, the Raman measurements do not provide absolute quantitative
information about the magnitude of contaminant accumulation. Thus the Raman
intensities at different flush times should be seen as relative values.
As mentioned in Chapter 3, the coupons were contacted with the contaminant of interest
prior to measurement for accumulation. The coupons were exposed to one of the
following contaminants at the respective mass fractions:
diesel, 100%, gasoline 100%, strychnine 0.5%, cyanide 1%, phorate 100%.
Flushing was accomplished by inserting the exposed coupon in a flexible hose and
flushing water through the hose at a Reynolds number of 30,000.
4.3.1 Diesel Fuel
Figure C27 presents measurement results for the diesel fuel contaminant. Diesel fuel is
clearly seen to stick to each of the substrates, particularly the cast iron, PVC, and copper
coupons. The initial one hour clean water flush was successful at reducing the diesel
accumulation by about one order of magnitude. Continued flushing removed additional
diesel, however the removal rate slowed noticeably. Residual diesel fuel levels following
24 h of clean water flushing leveled off at about 4 % of the initial values. Later in this
section comparison of these results with the diesel results presented in 4.2 (Dynamic
Fluid/Surface Interface Measurements) will be discussed.
4.3.2 Gasoline
Figure C28 presents the results of measurements for gasoline exposure. Gasoline
accumulation after initial exposure was greatest for copper, and least for iron, with PVC
and rubber falling in the middle. Flushing only was effective for the rubber substrate,
which showed a reduction to approximately 3 % of the initial level.
4.3.3 Strychnine
Figure C29 presents the measurement results for the strychnine exposure tests. These
results show that strychnine accumulated on all of the substrates, and flushing with tap
water was not very effective in removing the residual contaminant.
4.3.4 Cyanide
Figure C30 presents measurement results for the cyanide exposure tests. The greatest
accumulation was on the copper sample, which also showed some reduction from
flushing. The other samples accumulated less cyanide, but the levels were not
substantially reduced by flushing.
49
-------
4.3.5 Phorate
Figure C31 presents the measurement results for the phorate exposure tests. Due to its
low solubility in water, phorate accumulation was observed to be nonuniform, a probable
explanation for the low initial measured accumulations on copper and PVC (i.e.,
measurement area at a spot where a lower amount of phorate was present.) Flushing with
water was not effective at removing the accumulated phorate from the PVC and copper,
but performed better on the rubber and cast iron, although still leaving significant residual
amounts.
4.3.6 Screening Test Summary
Table 4 summarizes the results from the screening tests regarding whether the
contaminant was observed to stick to the substrate and whether it was removed after
flushing with water for a particular duration of time. Plots of the measurements along
with a brief description of the data are provided in Figures C27 - C31.
Table 4 Results from Screening Tests for Exposure and Flushing
Contaminant
Diesel fuel
Gasoline
Strychnine
Cyanide
Phorate
Substrate
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Copper
Iron
PVC
Rubber
Did it stick?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Did it flush with water?
Partially in 24 h
Partially in 24 h
Partially in 24 h
Partially in 24 h
Not in 20 h
Not in 20 h
Not in 20 h
Partially in 24 h
Not in 20 h
Not in 20 h
Not in 20 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
Not in 24 h
As mentioned earlier, the Raman measurements do not provide absolute quantitative
information about the magnitude of contaminant accumulation. To provide some
estimates of contaminant layer thicknesses, static mass measurements were made using
diesel fuel to associate the measured Raman intensities with film thicknesses. The static
50
-------
measurements were made by comparing the mass of substrate samples before and after
soaking them in liquid contaminant for 24 h.
The initial film thickness that was associated to zero flushing time was calculated from
the measured surface area and the mass of diesel adsorbed to the pipe surfaces. For
diesel fuel, the copper and PVC tube samples had an adsorbed film thickness of 56 |j,m ฑ
10 |j,m and 47 |j,m ฑ10 |j,m, respectively. Using these values, the thicknesses of the
diesel layer for copper can be estimated:
[(mass of diesel/density of diesel)]/surface area of coupon
At the beginning of the screening tests this turned out to be 56 |j,m, which corresponds to
a measured Raman intensity of 100. As flushing proceeds, the relative change in Raman
intensity can be used to estimate the thickness of the remaining diesel layer.
These results are plotted Figure C32 along with other measurement results obtained using
an in situ fluorescence technique presented in Subsection 3.2. In those measurements, a
diesel/water mixture was circulated until diesel accumulation reached a steady value,
followed by flushing with clean tap water. It is interesting to note the similarity in initial
measured film thicknesses for the two exposure techniques, although the static tests are
somewhat higher. The removal rate, as expressed in jim/hr of accumulated diesel fuel, is
faster for the screening tests, likely due to flushing at a higher Reynolds number (30,000
versus 5,000).
It should be noted that there is a discrepancy between the results of the screening tests for
diesel and the flushing results from the small scale dynamic tests. As summarized earlier,
screening tests showed that the flushed surface still had about 4% of adsorbed diesel,
while the small scale dynamic tests showed that after flushing, diesel was removed from
copper, iron and PVC pipe material. This could be due to the different levels of
sensitivity for the two different methods: Raman spectroscopy for the screening tests and
fluorescence for the small scale dynamic tests. Alternatively, this could be because the
coupons in the screening tests were contacted with pure diesel, while for the small scale
dynamic tests, the pipe material was contacted with relatively dilute (2,000 and 3,000
ppm diesel) mixtures of diesel in water. The different exposure conditions would likely
lead to different flushing results. It should be noted that the screening tests were for the
purpose of aiding the decision on which contaminants and pipe materials should be used
in the full scale tests. They were not mainly for the purpose of evaluating the validity of
the small scale dynamic tests.
4.4 Hot Water Heater Tests
The measurements consistently showed that most of the contaminants did stick to the
sediments of the hot water heaters after the initial exposure. Some contaminants, such as
diesel fuel and strychnine, showed a partial reduction in the sediment after flushing with
clean tap water. Bacillus spores required the addition of high levels of chlorine for
effective removal. Samples of the bulk water were not typically taken for analysis of
51
-------
chemical contaminants since the sediment and tank surfaces were seen as the tank
features that would be the most difficult to clean. It is not clear if water flushing alone
can effectively remove strychnine, phorate or cyanide contamination from tank sediment,
but even if it was effective, it is conjectured that the required water volumes would be so
large as to be impractical. A better approach might be to flush with hot water, water with
high chlorine levels, or water with detergent. Methodology for this approach is discussed
in Section 6. The test results for hot water heaters are summarized in Table 5. The details
of the hot water heater tank measurements follow.
Table 5. Results from Hot Water Heater Tests
Contaminant
Diesel fuel
Strychnine
Sodium Cyanide
Bacillus
thuringiensis
spores
Results
Diesel adhered to sediment.
Flushing with water (either hot or cold) was ineffective in
removing diesel from sediment.
Laboratory detergent appeared to be effective in removing
diesel from sediment.
Strychnine was still present in sediment after flushing.
Flushing results were inconclusive due to cyanide's tendency
to form complexes with metal substrates which hindered
analysis.
Flushing removed spores associated with bulk water.
Flushing ineffective in removing spores from sediment.
4.4.1 Diesel Fuel
Approximately 8 L of commercial diesel fuel were placed into a previously used 189 L
(50 gallon) hot water heater, which was then filled with cold water, resulting in a
nominally 4 percent solution. This mixture was circulated from the drain to the cold
water inlet for 8 h to mix the solution, and to promote wetting of the tank interior and the
sediment in the bottom of the tank. The tank was drained and a sediment sample
extracted from the bottom of the tank. The tank was re-filled with cold water and
allowed to stand for 4 weeks, at which time it was drained again, and another sediment
sample was taken. The tank was re-filled again with water and flushed with clean cold
water at 0.13 L/s (2 gpm) for a total of 24 h, corresponding to a water volume of
approximately 10,900 L (2,880 gallons) being passed through the tank. The tank was
then drained, re-filled and flushed with hot water at 0.25 L/s (4 gpm) for 75 min,
followed by another sediment sample extraction. Finally, in an attempt to clean the
sediment more thoroughly, the tank was drained, six cups of powdered laboratory
detergent were put in, and the tank was filled with hot water and allowed to soak (no
flushing) for 4 h and then drained before another sediment sample was taken. Aside from
the static mixing that occurred when the detergent and water refilled the tank, no other
mixing (e.g., mechanical mixing) occurred.
52
-------
Figure C33 shows the measured results for the testing sequence. The higher point at the
left of the graph corresponds to the initial Raman spectrum for the sediment following the
exposure. The lower point at the left of the graph was taken after the tank was drained
and refilled, which showed little change in the Raman intensity. The data point at the
right of the graph corresponds to a sediment sample taken after completion of all of the
flushing, i.e., after the cold water flush for 24 hr and the hot water flush for 75 min. It is
seen that the Raman intensity was somewhat lower than it originally had been but that it
was only reduced to about 30 percent of the initial value. Following the cleaning
operation with the laboratory detergent, the sediment sample showed no characteristic
Raman peak for diesel fuel.
These results suggest that simple draining and refilling of the hot water tank is not
particularly effective at removing the diesel fuel, in large part because of the lack of
solubility of diesel and its tendency to float on top of the water. A better approach might
be to extract it from the top of the tank if possible, and avoid possible contact between the
diesel fuel and the sediments at the bottom of the tank. While cold water flushing was
not effective, cleaning with hot water and detergent was successful at removing all
measurable traces of diesel fuel from the tank. In practice, this method would require
injecting the cleaning solution upstream of the tank, or through the drain valve.
A test was also conducted with the tank heating elements energized so that the water was
heated to approximately 60 ฐC. A heated hot water tank was used to try to be more
representative of actual operation, although the process of flushing introduces cold water
to the tank in any case. The tank was partially drained to allow the addition of diesel fuel
at a concentration of 4 % by volume, which was then mixed using a circulating pump,
and allowed to stand for 24 h. This was followed by flushing at 0.25 L/s (4 gpm) with
clean tap water which occurred via normal operation of the heated hot water tank, and
periodic water samples were taken from the tank outlet and drain for analysis. Because
the tank used in this particular test was sealed, it was not possible to collect sediment
samples. Figure C34 and C35 indicate that flushing was only partially effective at
removing the diesel contamination from the tank as a significant amount of diesel
remained present in water samples following 24 h of flushing. Another challenge
associated with flushing hot water tanks is their large volume relative to flush water flow
rates, which causes low flushing velocities in the tank and correspondingly low removal
rates.
4.4.2 Strychnine
Raman spectra were obtained for the sediment samples at 0, 1,3, and 24 h of flushing at
0.13 L/s (2 gpm). As shown in Figure C36, strychnine was still present in sediment
samples after 24 h of flushing, at about 10 % of the initial levels. The initial level of
strychnine in the tank was 600 ppm.
4.4.3 Sodium Cyanide
The initial level of sodium cyanide in the tank was 200 ppm. Sodium cyanide was
detected in the initial sediment sample following exposure, but was not detected in
53
-------
subsequent samples. However, other cyanide measurements indicate that this compound
reacts with substrates (e.g., metal pipe) to form complexes that confounded
measurements. Thus, a definitive conclusion could not be reached.
4.4.4 BT Spores
As mentioned in Chapter 2, a concentrated solution of Thuricide BT spores was diluted
out into 3.8 L (1 gallon) of water. This mixture was then added to the 189 L (50 gallon)
hot water heater tank that had been filled to 50 % capacity. After the BT solution was
added, the tank was filled up to the 189 L (50 gallon) mark.
The hot water tank containing the spores was allowed to sit for 24 h at ambient
temperature. The tank was then mixed with a long rod and then flushing with cold water
started at 7.6 L/min (2 gal/min). The water was sampled at various times by collecting
the outflow of the tank using a valve. The tank was flushed for a total of 24 h (a total of
48 h elapsed after the spores were added). Water samples and sediment samples were
then collected. Sediment was collected from the bottom of the tank using a long tube
inserted into the tank, and the sediment suctioned up to a receiver flask by connecting a
gentle vacuum. The sediment samples were approximately 10 ml in volume and 30 mL
of PBS containing 0.01% Triton X100 was added and the tube vortexed for several
minutes. The sacrificial anode (previously removed) was suspended into the tank using a
piece of wire and was immersed during the experiment in the tank. The sacrificial anode
was removed at the end of the experiments and the surface scraped with a cell scraper and
rinsed three times with 20 mL of PBS containing 0.01% Triton X-100.
The results showed the BT spore concentration remained the same in the hot water heater
after it was allowed to sit for 24 h (Figure C37). In both trials (#1 and #2,) the spore
concentration at the tank outflow after 10 min of flushing was similar to the spore
concentration in the tank. This indicates the spores were being flushed out during this
time and the concentration had not yet significantly decreased. A 60 min sample was
taken in Trial #1, and it showed that there was an approximate a 2 log decrease in the
number of spores (Figure C37). The sample taken from Trial #1 after 2.5 h of flushing
revealed no spores, indicating that the spores in the bulk water had been effectively
removed. In trial #2, a sample was taken after 10 minutes of flushing (Figure C37) and
the next sample was taken after 100 minutes of flushing. The data for trial #2 showed
that the 10 min flush was similar to the initial concentration (Figure C37) and the sample
taken at 100 minutes had no detectable spores, indicating that the spores which were not
associated with the sediment were essentially flushed out of the tank at this time.
Even though a majority of the spores were flushed out of the hot water heater, a fraction
of the spores remained in the sediment. (The initial volume of sediment was not known
in these trials, so only qualitative statements about the presence of spores in the sediment
can be made.) The data in Figure C38 show the concentration of BT spores in the water
initially, after 24 hr and in sediment samples taken after 24 h (only for trial #1), after 48
h, and on the surface of the sacrificial anode. The small number of samples and the
uncertainties associated with the sampling methods only allow the conclusion that a
measurable amount of spores were still associated with the sediments and sacrificial
54
-------
anode surfaces after 48 h of flushing. This suggests that the sediment at the bottom of a
hot water tank sediment could be a long term source of contamination even after
significant flushing.
4.5 Full-Scale Tests
The measurement results for the full-scale plumbing loop are described in this section.
4.5.1 Diesel Fuel
Diesel fuel was mixed with water at a nominal concentration of 8.3 % (83,000 ppm), then
circulated through the pipe loop and allowed to stand for eight days. The water lines
where the pipe samples were located were then flushed at a flow rate of 0.25 L/s (4 gpm).
Pipe samples were collected from Floor 1 (3/4 inch copper pipe) and Floor 3 (1/2 inch
CPVC) and Raman spectra were obtained for all of the samples taken. The
measurements showed that flushing with cold tap water was not very effective at
removing the accumulated diesel fuel from the pipe samples for either the copper (Figure
C39) or CPVC (Figure C40) pipe at this flow rate for flushing times of 13 h, although
there was some reduction for the CPVC pipe. The difference in diesel removal
effectiveness between this test and the previously described, smaller scale tests could be
due to the different flushing duration times and/or exposure conditions. The small scale
dynamic tests mentioned in Section 4.2 had a much longer flushing time (150 hours) and
a much lower diesel concentration (2,000 - 3,000 ppm) during the exposure phase of the
tests.
4.5.2 Strychnine
Samples were measured for Floor 1 ( % inch copper pipe) and Floor 2 (1A> inch copper
pipe). For Floor 2, no Raman spectrum was obtained for the sample taken before
flushing began and reasonable spectra were only obtained for the samples taken after
approximately 1 h and 2 h of flushing: The measurement results are shown in Figure C41
for % inch copper pipe and Figure C42 for l/2 inch copper pipe, which indicate that cold
water flushing was not effective at removing the accumulated strychnine from the pipe
samples.
4.5.3 BT Spores
A solution containing BT spores was prepared in the following manner:
Add 2 Liters RO water to 3.78 L (1 gallon) jug
Add 4 mL 10 % Triton X-100, mix
Mix bottle of Bionide (#4) and add 100 ml to jug and mix well
Add 1.67 L RO water to jug and mix well again
Titerjug
This solution was circulated through the pipe test loop and left overnight to soak. This
was followed by flushing and sampling at 1 h, 3 h, and 13.3 h.
Pipe samples were analyzed in the following manner:
55
-------
Received samples and sampled the next day
Remove pipe from container and dip pipe in beaker with lOOmL RO water
Repeat 2 more times using a new beaker with lOOmL RO water
Scrape pipes using cell scraper into 5 mL PBS/Triton in 50 ml conical tubes
Wash pipe 4 times with buffer in tube
Vortex for 30 sec
liter
The results of these measurements are shown in Figures C43, C44,and C45 for the three
floors, which had different pipe sizes and materials. The 3/4 inch copper pipe Figure C43
showed less than a one log reduction after 3 h of flushing, and a three log reduction after
13.3 h of flushing. While this is a significant reduction, it is possible that the remaining
spores would detach over time and be a long term source of contamination, thus posing a
hazard to human health.
Figure C44 shows the results for /^ inch copper pipe, which showed a 2.5 log reduction
after a 3 h flush. The 13.3 h flush sample could not be analyzed, and thus the curve fit
for the data likely underestimates the time (-12.2 hours) for 5 log removal of the spores.
Figure C45 shows the results for the /^ inch CPVC pipe, which showed a 2 log reduction
after 3 h of flushing, and a three log reduction after 13.3 h. Substantial reductions in the
number of spores were observed, which is surprising given that chlorine was not added to
the flushing water. The bench scale disinfection/flushing, reported in Section 4.1.2.2 and
in more detail in EPA and NIST, 2011, achieved similar (3 to 4) log reduction at high
levels (10 - 100 mg/L) of chlorine. Flushing without chlorine, as reported in this section,
was not expected to yield comparable log removal. One significant difference between
the two tests was the amount of biofilm on the pipe surfaces. For the bench scale testing,
a purposed effort was made to ensure a well developed biofilm prior to spore addition,
whereas for the full scale tests a similar effort to develop a biofilm was not done. If there
was less biofilm in the full scale system, then the spores would likely not have adhered as
strongly to the wetted pipe surfaces. This would lead to a greater percentage of spores
being removed in the full scale system compared to the bench scale studies. In any event
for complete eradication of spores from actual plumbing systems would likely require
additional decontamination procedures, such as shock chlorination and/or germination.
Table 6 summarizes the results for the full-scale plumbing system tests for diesel fuel,
strychnine, and BT spores. The results showed diesel fuel, strychnine and BT spores
adhered to all of the surfaces of this study. Diesel fuel was only minimally removed and
BT spores were partially removed by flushing after approximately 12 h. However,
strychnine was not removed in 14 h of flushing.
Table 6 Results from Full Scale Plumbing System Tests
Contaminant
Diesel fuel
Strychnine
Substrate
copper
CPVC
copper
Did it stick?
Yes
Yes
Yes
Did it flush with water?
Minimal removal in 12 h
Partially in 12 h
Not in 14 h
56
-------
BT Spores
PVC
copper
PVC
Yes
Yes
Yes
Not in 14 h
Partially in 13 h
Partially in 13 h
Chapter 5 MODEL DEVELOPMENT
5.7 Semi-empirical
Two models were developed based on the theory and results of the dynamic fluid/surface
interface studies described in sections 3.2 and 4.2. These models were developed, in
part, because of the logistical difficulty in quantifying the surface concentration of a
contaminant on a particular length of plumbing. This difficulty necessitated an
alternative approach to aid in responding to an event that impacted a building's plumbing
system. The approach assumed a worst case scenario in which the maximum possible
amount of a contaminant adhered to the pipe. Two models were used in this approach
and are referred to herein as the flushing model and the contamination model. For a
given contaminant mass fraction in water, the contamination model predicts the
maximum thickness of the contaminant layer on a pipe surface. The flushing model
predicts the time to flush a contaminant from a plumbing surface. In this approach, the
contamination model was used to predict the maximum layer thickness, and then the
flushing model was used to predict the time to flush this maximum estimated thickness
from the surface. The result of the former model served as input to the latter model.
In the discussion below the flushing model is discussed first since it is the more
foundational of the two. The contamination model is actually a special case of the
flushing model.
The derivations of the models to predict the thickness of the contaminant excess layer on
plumbing surfaces for the contamination and the flushing conditions are presented in
Kedzierski, 2008 and reproduced here. Each model assumes that transfer of mass to and
from the surface occurs solely in a direction that is perpendicular to the surface.
5.1.1 Flushing Model5
The model for flushing with contaminant-free water is based on the conservation of
contaminant mass within the excess layer (7e). Because the excess layer is thin, it is
approximated as stagnant in the axial direction such that the net motion of diesel fuel is
solely perpendicular from the surface, i.e., in the y-direction. The mass flow rate of
diesel (md ) per unit area (A) is governed by diffusion and turbulent convection of
contaminant from the surface:
A dt 2 - BT
5 More detail on the flushing model development can be found in Kedzierski (2008).
57
-------
Here the effect of diffusion is modeled by the ratio of the diffusion coefficient (Aiw) to
the transition depth (Bi) over which the concentration difference occurs. The turbulent
convection is modeled by the contaminant's kinematic viscosity (Vd), the friction velocity
(it*), and an entrainment constant (Ki) that relates the average entrainment velocity
(vmax / 2) to the local axial-velocity (u) in the viscous sublayer as:
K3 = Vmax/W (5.1.1.2)
The law of the wall is used to approximate the average entrainment velocity as 2vd
(The law of the wall states that, for turbulent flow, the average velocity at a specific point
in the fluid is proportional to the logarithm of the distance from that point to the wall or
other boundary.)
Separating and integrating eq. (5.1.1.1) with respect to time (f) results in an equation to
predict the contaminant excess layer (/e) as a function of flushing time (f) and initial
excess layer thickness (7eo) as:
e
eO
2V;
1 + J_) -^r-' 1
V KD;
K
D
(5.1.1.3)
where the dimensionless constant ^D is a ratio of the convective to the diffusive
influences:
D ^ (5.1.1.4)
For XD = 1, convection and diffusion fluxes of contaminant from the surface are equal at
the beginning of the flushing. Values of KV larger than 1 indicate that convection is more
important than diffusion of contaminant from the surface.
The friction velocity, it*, is calculated from equation 5.1.1.5, which was given by Kays
and Crawford, 1980, using the average (bulk) axial velocity, V, and where the Reynolds
number was calculated using the properties of the flushing liquid:
(5.1.1.5)
58
-------
Flushing measurements for contamination levels larger than those of the scope of the
present project were taken in order to establish changes in the measured excess layers that
were large enough to fit to eq. (5.1.1.3). These experiments flushed diesel fuel from a
PVC surface with tap water flowing at a Reynolds number of approximately 5000. The
initial diesel excess layer was approximately 31.4 |j,m. Figure C46 shows the flushing
measurements that were used to obtain the Kj and Ddw/^i constants from a least squares
regression of the data to eq. (5.1.1.3). The regression constants are:
K} = 0.66 x!0~8ฑ 0.05x10"
DA
m
s
ฑ0.1x10
-10
m
s
(5.1.1.6)
(5.1.1.7)
Figure C46 plots eq. (5.1.1.3) using the regressed eq. (5.1.1.6) and eq. (5.1.1.7) constants.
The model predictions are within ฑ 4 (im of all measured /e. This translates into
predictions being within 12 % of the measurements at the beginning of the flushing tests
and more than 100 % larger than measurements near the end of flushing. This could
seem to be an unacceptably large prediction error for the film thickness; however, the
primary goal of the model is to predict flushing times to clean. To put it in perspective,
the more than 100 % error in the prediction of the film thickness results in an over
prediction of the flushing time by approximately 9 h, which is an acceptable error
considering that a safety factor of at least two would be applied adding more than 50 h to
the required total flushing time.
The entrainment constant, Kj, is expected to be less of a function of the properties of the
contaminant/flushing pair than is the diffusion constant. The ratio of the axial bulk
velocity to its peak fluctuating component could be nearly constant because the bulk
velocity is the potential for the fluctuating velocity. Considering this and that Kays and
Crawford, 1980 show that the ratio of the peak fluctuation turbulent velocity components
is constant, the K} can be relatively constant for a particular flushing fluid and for various
Reynolds number. It is expected that the Kj presented here would be valid for water and
liquids with similar kinematic viscosity.6 Conversely, the D^IBi depends on the
properties of both the contaminant and the flushing fluid.
The required time (f) to flush the surface clean is derived in Kedzierski, 2008 and is
presented here:
t= ~' ฐ In (5.1.1.8)
However, Kj may be altered by the adhesion forces that must be overcome to remove
the contaminant from the wall.
59
-------
Table 7 compares predictions using eq. (5.1.1.8) with observed values for diesel fuel. For
copper and for one of the trials with iron, the model underestimated flushing times. In all
the other trials, predictions were conservatively overestimated. The data in this table was
meant to provide a measure of accuracy for the flushing model only, and not a measure of
how effective the two models worked together in predicting both the maximum layer
thickness and the subsequent required flushing time.
For the majority of the modeling results in Table 7, there was not good agreement
between predicted and observed flushing times, however, the fact that some of the
flushing times were in good agreement suggests that the models could be improved with
additional development. In addition, the poor agreement may have had more to do with
the observed measurements than the predictions. The observed flushing times required
the measurement of a contaminant on a wetted surface, which is a challenging matrix to
analyze. Improvements made in these types of measurements would also help
development of improved predictive models.
60
-------
Table 7 Comparison of Flushing Model Predictions to Measurements
Surface
PVC
Iron
Copper
Reynolds number
(during exposure
testing)1
0
3200
7000
0
3200
7000
4600
4600
/eO
(Urn)
1.5
1.5
2.5
1.0
1.0
0.5
6.6
1.9
^D
0.29
0.29
0.49
0.20
0.20
0.10
1.40
0.40
Predicted Flushing
time (h) for clean
eq. (5.1.1.7)
6.4
6.4
9.7
4.3
4.3
2.3
20
7.5
Observed flushing
time (h) for clean
~02
~02
~8
^O2
^O2
~5
<60
-15
Flushing was conducted at Re = 5,000. The Re values presented in the table refer to the exposure
tests where the pipe materials were exposed to contaminant solutions at the above Re values.
2
A value of zero is given as an approximation based on the measurements described in Section 4.2.2
and the associated uncertainty of these measurements. See Section 4.2.2 for a more detailed
discussion.
5.1.2 Contamination Model
To model contamination, an equilibrium was assumed where a balance between
deposition of contaminant on the surface and removal of the contaminant from the
surface is achieved. For this equilibrium case it is necessary that a distinction be made
between the velocity of the contaminant toward the surface (v;) and that away from it (v0).
Very near the wall there is an additional resistance to flow away from the surface as
caused by the attraction of contaminant molecules to the pipe surface. Likewise, for the
same region near the pipe, the attractive forces induce a reduction in the resistance of
flow toward the surface. Considering that it is flow that is being modeled, a simple way
to approximate this behavior is via an effective viscosity For example, the entrainment
velocity (evaluated using the properties of the contaminated flow) given in Kedzierski,
2008 for flow approaching the pipe surface becomes:
l
2 VA tanh -*
(5.1.2.1)
Here, the effective kinematic viscosity is Vdtanh(le/5) and is less than the kinematic
viscosity (vd), of the bulk phase for distances from the surface less than the penetration
depth, 8. This is because the adhesive forces assist the flow of contaminant to the surface
within this region. The term, tanh(4/5) can be seen as a correction factor for kinematic
viscosity to account for how the adhesive forces decrease the viscosity of the fluid
moving toward the wall. The magnitude of the penetration depth is determined by the
61
-------
affinity of the contaminant for the pipe surface and coincides with a region near the wall
where the viscosity differs from the viscosity of the bulk liquid. Consequently, each
contaminant/pipe combination will have its own value of 8. The hyperbolic tangent was
chosen for its simplicity and because it closely matched what was believed to be the
required relationship with respect to /e.
Likewise, surface adhesion acts to deter the flow of contaminant from the wall according
to an assumed hyperbolic cotangent relationship with respect to 7e. Here the leaving
velocity is approximated as a local increase in kinematic viscosity, above the kinematic
viscosity in the bulk phase for distances from the wall that are less than 8.
[-1
V =
0 J v=L
(5.1.2.2)
For this case the effective kinematic viscosity is VdCoth(4/5) and is greater than the
kinematic viscosity (VA). The term coth(4/5) can be seen as a correction factor for
kinematic viscosity to account for how the adhesive forces increase the effective viscosity
of the fluid moving away from the wall.
Figure C47 provides an example of the ratio (referred to herein as &) of the effective
kinematic viscosity to the kinematic viscosity in the bulk phase. In this example, the
contaminant is diesel, and the pipe material is iron. As the figure shows and as the text
above supports, this ratio is a function of distance from the iron surface. The kinematic
viscosity in the bulk phase is considered constant for a given temperature, and the
effective kinematic viscosity is partially dependent on the adhesion forces at the
liquid/wall interface. For this example, the excess layer is 0.5 |j,m and the penetration
depth is 14 |j,m. The viscosity ratio is plotted as a function of_y to illustrate the waning
and waxing influences of the adhesive forces between iron and diesel fuel as a model
concept. However, as far as the contamination model is concerned, & is evaluated only at
4 = 0.5 urn.
Considering that the exposure time of a pipe surface to a contaminant might not always
be known, a conservative decontamination response would be to flush for the maximum
contaminant film thickness (excess layer) for the given concentration of contaminant in
the flow. The maximum contaminant excess layer, which occurs at steady state where
the contaminant deposition balances the removal at the wall, can be determined by setting
the partial derivative of the excess layer with respect to time to zero (see Kedzierski,
2008) and solving the resulting equation for 4, yielding:
= *ln
max o
(5.1.2.3)
62
-------
Equation (5.1.2.3) can be used to determine the maximum possible contamination level
and used as input to eq. (5.1.1.7) in order to calculate the required flushing time to obtain
a clean surface. Mathematically, eq. (5.1.2.3) is valid for values of x\, between zero and
1; however, it has been validated experimentally for dilute solutions only.
Average values of the penetration depth, 8, were found by back-substituting the measured
diesel bulk mass fraction and the measured [/e]max into eq. (5.1.2.3) and solving for 8.
The ฃwas found to be surface dependent: 150 |j,m, 42 |j,m, and 16 |j,m for copper, PVC,
and iron, respectively. Figure C48 shows the maximum contamination layer for the three
pipe surfaces as a function of x\> as predicted by eq. (5.1.2.3). Because of the presumed
stronger affinity between copper and diesel, the copper surface has greater contamination
levels for a given x\, as compared to the PVC and iron surfaces. The greater affinity is
modeled via the larger penetration depth.
Figure C49 is a preliminary validation of eq. (5.1.2.3) and its theory in that it compares
well against independent measurements and exhibits an exponential decay with respect to
flushing time. Figure C49 demonstrates the agreement between eq. (5.1.2.3) predictions,
as shown by the solid lines, and two different measurement techniques. Raman intensity
measurements from the technique in Subsection 3.3 are represented by the filled circle,
triangle, and square. The second technique was the traversing fluorescence technique
provided in Subsection 3.2 and depicted by open circles. The Raman screening tests
were for copper, PVC and cast iron at a Reynolds number of 30,000. The fluorescence
tests were for PVC at a Reynolds number of 5000. Removal at a Reynolds number of
5000 was slower than at the higher Reynolds number, and the time required to remove
most of the diesel fuel was on the order of 50 h. Good agreement between the measured
flushing times and the model was achieved. Each prediction coincides with
measurements from both techniques for at least one value above zero. In addition, each
prediction falls on zero between measured points that indicate a clean surface. It is not
likely to measure the surface at the precise time that it has become clean. Consequently,
it is not likely to have a measured point coincide with the model at zero film thickness.
5.1.3 Future Work on Semi-empirical Model
The goal of this project was to form a foundation for the development of tools necessary
for adequately responding to a contamination event. It was envisioned that these tools
would consist of measured data, predictive models, and a computer program that
embodies the data and models. The software would allow the user to input information
that is specific to a contamination event and receive estimates of contamination levels
and required flushing times. Such an all-encompassing goal requires a significantly large
effort to ensure that the software makes reliable predictions for all possible contaminants.
To achieve this task in a reasonably short time, the model must be physically based, yet
simple, so that it can be easily adjusted and extrapolated for different contaminants.
In this light, the preliminary models developed in this project require additional
enhancement to ensure that the best possible predictions can be provided. Possibly, the
63
-------
major effort toward this end will be using measured data to calculate the constants Kj,
D&JBi, and ฃfor potential contaminants like toluene and pipe surfaces; (see Kedzierski,
2008.) In the course of this effort, it will be determined whether or not K] is merely a
function of the properties of the flushing fluid. In addition, it can be determined if <5is
constant for a particular contaminant/surface pair. In the short term, more flushing
measurements at different initial 4 are required to validate the flushing constants
considering that only one data set was used to obtain them. It could be appropriate to
lump the diffusion that occurs during contamination into the adhesion effect. On the
other hand, inclusion of a diffusion component in the contamination model could be
necessary. These questions can be answered with further experimentation and model
enhancement. In addition, refinement of surface measurement methods, e.g., Raman
spectroscopy and dynamic fluorescence measurements, would be needed to overcome the
observed limitations of detection.
5.1.4 Software to Predict Contamination Levels and Required Flushing Times
Preliminary versions of computer software were developed to predict contaminant levels
and required flushing times. The flushing and contamination models described in
Subsections 5.1.1 and 5.1.2, respectively, served as the basis of the computer program.
The general concept is based on modeling the interaction between the contaminant and
the pipe walls under different flow conditions to predict the contaminant accumulation
and removal. Certain parameters are used as inputs to the software program, such as
mass fraction of contaminant and exposure time, flushing Reynolds number, and the
software tool provides, as output, a recommended flushing time. Currently, the software
tool has been validated using detailed measurements of diesel fuel, but it is under
development to expand its capabilities to cover a wider range of contaminants. The
complete computer software may be downloaded from: ftp://ftp.nist.gov/pub/bfrl/squid/.
Figure C50 shows an example of graphical user's interface (GUI) that is used to input
data on the pipe material, flushing flow rate (Reynolds number), pipe diameter,
contaminant, and other information relative to the contaminant incident. The GUI also
shows the output in terms of the contaminant thickness on the pipe wall as a function of
flushing time. Overall flushing time and estimated maximum contaminant thickness are
also given.
In Figure C51 the flushing model given in Subsection 5.1.1 illustrates the effect of the
building floor area on the required flushing times. It was assumed that only a single
entrance and a single exit were used to flush the building plumbing. As a result, piping
downstream of the entrance is flushed with water that has entrained a small amount of
contaminant in the flow. The contaminated flushing fluid might redeposit on the surface
or reduce the effectiveness of flushing due to a loss in concentration gradients for
diffusion. For such an assumption, the required flushing time should increase with the
building area because of effects like re-contamination of the surface and reduction in
diffusion concentration gradients.
The total length of plumbing (Zj) in a building with floor area, A, was estimated using the
following equation.
64
-------
LT = 8VJ
(5.1.4.1)
This equation was based on the assumption that that for a square building, Lt is
equivalent to wrapping hot and cold water lines once around the building perimeter.
Only the effect of diminished concentration gradient on contaminant removal was
modeled because it is likely to be the most significant effect. The mass fraction of the
flushing fluid leaving the building was calculated from a Lagrangian control volume
analysis7 on a fluid element as it progressed from the pipe entrance to its exit:
*ป =
L ,
T
The terrnKp is the diffusion term corrected to account for the fact that the contaminant is
in a mixture and not in its pure form. The average contaminant mass fraction was used to
correct the diffusion term, KD as shown in the following equation.
The average contaminant mass fraction itself was calculated through the following
equation:
- = *out-*in (5.1.4.4)
In the above equation, x;n = 0, and xout = Xb, and the average mass fraction simplifies to
0.5xb.
Figure C51 shows the effect of the building area for three flushing Reynolds numbers. In
general, as the Reynolds number increases, the effect of building area is diminished. In
fact, for turbulent Reynolds number, the building area has a relatively negligible effect on
flushing time. For the three Reynolds numbers evaluated, only the results at a Reynolds
number of 3 showed a significant change in flushing time for increased building area.
For this flow condition, the contaminant diffusion was the same order of magnitude as
convection in terms of contaminant removal. At the higher flow rates turbulent flushing
resulted in convective entrainment of the contaminant dominating, and thus building area
had little effect on predicted flushing time.
7 Mass of diesel in control volume at tube exit is
0
65
LT IV .
\ ~
-------
5.2 Flow in idealized pipe geometries
In Section 5.1 two models were described based on the theory and results of the dynamic
fluid/surface interface studies described earlier. However neither model took into account
the effect that obstructions and various pipe bends (e.g., elbows, U shaped turns) can
have on contaminant transport and deposition. This section (5.2) describes modeling that
addresses the effect of such flow disruptions. Unlike the previous models, the model in
this section gives a close up view of the flow paths occurring in specific plumbing
obstructions and bends. It does not predict flushing times but instead gives a detailed
picture of how phenomena such as vortices formed in a pipe elbow can affect
contaminant removal.
5.2.1 Introduction
The traditional approach for decontaminating plumbing systems is to flush pipes with
water at high volumetric flow rates, primarily because this methodology is most easily
implemented, and there are few simple alternatives. Although previous sections of this
report do show examples of removal of some contaminants via flushing, this technique
has not been fully demonstrated. In fact, for several reasons, it might not always be the
best way to remove contaminants. For example, computer generated data in this section
show that because high velocity water flow in piping systems are in the turbulent regime,
vortices are generated that tend to inhibit the transport and removal of contaminants. So
while turbulence, in general, improves mass transfer, vortices cause such phenomenon as
localized reverse flow which can impede flushing efforts
Further, it is not understood to what degree the location of the contaminant in the pipe
system will affect its removal rate. For example, upon initial analysis of the water, post
flushing, it could appear that the contaminant had been completely removed, but later
sampling and analysis could reveal that the contaminant had reappeared in the pipe
system at a later time because it was located outside the main flow path. That is, the
contaminant could have entered the main flow path at a later time via diffusive transport.
For the work described in this section, various possibilities were investigated by
comparing transport in different flow geometries at low (laminar) and high Reynolds
numbers using numerical modeling. An effort was made to separate the relationship
between contaminant transport and the bulk flow rate, from the relationship between
contaminant transport and the actual flow path the contaminant takes, which will change
with Reynolds number. The former relationship will provide a more global picture of
contaminant flushing, while the latter provides a more local picture of contaminant
flushing in specific pipe structures. In the numerical model simulated patches of
contaminant were placed in different locations of the pipe system to help understand the
effect of the local flow field on contaminant removal
It should be pointed out that the actual efficiency of contaminant removal for any
particular pipe component cannot currently be predicted. This would entail detailed flow
modeling for that specific pipe geometry with a correct understanding of the contaminant
interaction with the pipe wall. In addition, the approach used in this study needs further
66
-------
experimental validation. Hence, the results presented in this work should be thought of
as qualitative in nature with the hope of identifying potential problem areas in
contaminant removal such as a contaminant being temporarily trapped in a vortex formed
in an elbow or tee fitting.
5.2.2 Flow Around Bends, Past Cavities, and Over Obstructions
To investigate the effect of pipe geometry on the movement of contaminants, four
idealized flow geometries that could represent typical pipe system components are
considered. For comparison purposes, the base system studied is a straight square pipe
with a square cross-section. Figures C52 and C53 show the second system studied, a
pipe fashioned as a U shaped bend. This geometry could, for example, be representative
of a glove valve in the pipe system. Figure C54 shows the third pipe system, which has a
cavity or pocket (analogous to a shunt or trap) out of the line of the main flow. Figure
C56 shows the fourth pipe geometry, a simple rectilinear obstruction in the flow path.
This geometry could be similar to a gate valve or an obstruction in the main flow path.
The computational fluid dynamics approach used in this study was based on a finite
difference approximation of the Navier Stokes equations and is fully described in Martys,
2001. This simulation approach has been validated for a variety of flow scenarios ranging
from simple Poiseuille flow to obtaining correct solutions for boundary layer flow. For
this study, the Navier-Stokes equation was used to generate flow fields in the different
pipe geometries. The flow fields are then used to determine the contaminant transport via
numerical solution of the advection diffusion equation (Martys, 1994). For this report,
two Reynolds number values, 30 and 3000, which are representative of typical slow and
fast flow rates in pipe systems respectively were considered. The Reynolds number is
given by Re=F7/v where Fis the average velocity at the inlet, / is the channel width, and
v is the kinematic viscosity. In these simulations, Reynolds number was adjusted by
varying the kinematic viscosity while fixing the flow velocity at the inlet. Although for
narrow temperature ranges, kinematic viscosity is essentially constant, by fixing the inlet
velocity and varying the viscosity, the effects on contaminant removal due to simulated
changes in flow trajectory as Reynolds number is increased can be separated from
changes in removal due to a simple increase in flow velocity.
Figure C52 and 53 illustrate the main flow features obtained from the simulations for the
U shaped pipe at Reynolds numbers of 30 and 3,000, respectively. Figure C54 shows a
simulation for flow near a cavity. In all three figures the fluid enters from the left side and
exits out the right side. As expected, the flow fields at low and high Reynolds number
were quite different. At the lower Reynolds number (Figure C52), the flows appeared
laminar with little rotational flow. In the case of the U shaped pipe at higher Reynolds
number (Figure C53), significant rotational flow developed, especially near the pipe's
corners. In the case of flow past a cavity, at low Reynolds number (not shown in figure),
a single rotational flow field developed near the opening of the cavity but at the higher
Reynolds number (Figure C54) two counter rotational flows developed. In general, as
Reynolds number increased, the local velocity reached a higher value nearer the pipe
surface.
67
-------
5.2.3 Contamination Model
The contaminant is treated as a scalar field whose time evolution is controlled by
advection rather than diffusion, which was considered to have negligible effect on
contaminant transport. The transport of the contaminant was considered passive. That is,
the contaminant does not affect the flow field. Contaminants are introduced into the pipe
system either as a rectangular patch along the pipe surface or at the inlet as a constant
source. The egress time for different flow scenarios is then compared.
In the case of the contaminant patch, a simple linear coupling of contaminant with the
pipe wall where the rate of adsorption is proportional to the concentration difference in
the fluid and on the wall is assumed.
Here,
dc
= -k(c-cs),
dt
(5.2.2.1)
where k is a is a mass transfer coefficient and cs is the equilibrium concentration on
surface. The simulation model was not limited to this surface interaction but it was used
as a first approximation.
One goal of the modeling of contaminant migration and removal was to separate the
effect of a change in overall flow rate from the effect of a change in the actual flow path
of the contaminant. Both these parameters, flow rate and flow trajectory, will change
with changing Reynolds number, but the change in contaminant transport is different for
the two cases. For fixed flow rate it was found that the Reynolds number can have a
dramatic effect on the trajectory of the contaminant. At higher Reynolds number the flow
near the boundary of straight paths is much greater allowing for more contaminant to be
carried away into the main flow streams. For other geometries, higher vorticity, i.e.,
greater eddy formation and intensity can be near some corners so that the contaminant
has a tendency to stay in that region. Contaminants starting from a corner but closer to
the outlet took longer to egress in the main flow path compared to contaminants on a
straight section of pipe.
Figure C55 compares the amount of contaminant as a function of time for the different
flow geometries and at different Reynolds number. The variable, C, is the total
contaminant in the pipe system and is normalized by Qnit, the value for the initial
contaminant concentration. The black curves correspond to the straight pipe flow. The
blue (Re = 3000) and green (Re = 30) curves correspond to the case of flow in the U
shaped region. The dashed lines correspond to the placement of contaminant in the region
near the bottom right hand corner of the U shaped pipe, whereas the solid line correspond
to a contaminant being placed at the midpoint of the pipe along the bottom. In general,
placement of the contaminant away from the main flow path can make the egress time
longer. Clearly, at both Reynolds number, the contaminant left the pipe system much
more efficiently when placed in the bottom straight section. At low Reynolds number,
when the contaminant is placed in the corner, a significant increase in time for its
removal is simply due to the lower flow rate i.e., the low flow rate yields a high hydraulic
68
-------
retention time and thus a low rate of contaminant flow through the pipe section. In the
case of the U shaped pipe and the high Reynolds number, the delay of removal is greater
because of the slow circulatory flow in that corner. Essentially, eddies of fluid form in
the corners and these volume of fluid are temporarily trapped as they swirl around in the
corners. Any contaminant associated with the water in the eddies will likewise be
delayed in leaving the pipe section.
The violet lines correspond to contaminant removal from the cavity. In this case, the
contaminant was initially placed along the bottom of the cavity. As can be seen, the
contaminant remains for a significant time in the cavity. Here the slow rotational flow
inside the cavity has a tendency to suppress the contaminant's egress. At either Reynolds
number the cavity was the most difficult to clean and presents the biggest challenge as
the time scale for contaminant removal was significantly higher.
Figure C56 for low Reynolds number, and Figure C57, for high Reynolds number, show
flow past a rectilinear obstruction with a contaminant patch immediately behind the
obstruction. At the lower Reynolds number the fluid simply moves up, over, and down
the back of the obstruction, and no appreciable vortex formed. However at the higher
Reynolds number, a vortex formed downstream from the obstruction and is shown in
Figure C57. Figure C58 shows the time dependence of the contaminant concentration for
the flow conditions shown in Figure C56 and Figure C57. Clearly the concentration
decreases more rapidly for the higher Reynolds number case. This is a consequence of
the existence of relatively higher velocity near the pipe wall causing the contaminant to
be carried away more quickly. Thus for this latter case, even though the direction of flow
adjacent to the contaminant patch is in the opposite direction of the flow through the pipe
section, the velocity at this point, due to the vortex, is much higher, resulting in a faster
removal time.
5.2.4 Ingress of Contaminant
Although the placement of the contaminant plays an important role in its removal, it rates
a second consideration in the history of how the contaminant actually enters and passes
through the pipe. In the Figure C59 the time evolution of a contaminant entering a U
shaped pipe system is shown for a Reynolds number of 3000. The visualization shows
that regions which take longest to remove contaminant are also the same regions that take
longest to reach from the inlet (see Figure C59). In other words, locations that take longer
to decontaminate are also less accessible to contamination. Thus concern over difficult to
clean areas such as cavities can be offset due to the likelihood that these areas would be
less contaminated to begin with.
It is important to note that the only cases considered were those in which hydrodynamics
controlled the motion of the contaminant. However, if the contaminant enters a slow
flow region, or else the water is stagnant for a period of time, then diffusion would play a
larger role regarding which areas become contaminated, and removal time could be
significantly longer compared to situations where water was more continuously flowing
through the plumbing system.
69
-------
5.2.5 Effect of Flow Rates
While it was found that the flow fields associated with higher Reynolds number can
strongly affect the trajectory of the contaminant it was also found that this phenomenon
was confined to regions of higher vorticity. Note, the actual flow rates needed to achieve
the higher Reynolds number will, in general, more than compensate for the reduced
efficiency of removal. For example, consider the U shaped pipe study. The contaminant
placed near the corner (CP), at a Reynolds number of 3000, took nearly 40 % longer in
time to reach the same concentration than when the Reynolds numbers was 30. However,
these Reynolds number numbers were achieved by adjusting the viscosity so that the
average flow rate was the same for those two cases. But, in real applications Figure C60
shows the viscosity is fixed, and the flow rate would have to increase by a factor of 100
(i.e., 3000/30 = 100) to reach the same Reynolds number value. As a consequence, the
time to reach the same level of decontamination was about a factor of 70 times faster at
the higher Reynolds number. However another consideration is that the higher Reynolds
number flow will require more water to reach the same level of decontamination.
It is clear that flow rates dominate the removal of contaminant over time. Delays in
contaminant removal, due to eddies or vortex motion in flow, do not have as big effect as
actually increasing the flow speed in the cases studied.
To further help clarify the effect of flow rate vs. geometry and contaminant placement,
Table 8 gives a relative comparison of contaminant removal times.
Table 8 Comparison of Contaminant Removal Corrected for Flow Rates
Re Number
30
3000
Straight Pipe
1
0.009
U-Shape (MP)
1.5
0.012
U-Shape (CP)
2.5
0.03
Cavity
10
0.18
CP, corner patch of a U-shaped pipe; Re, Reynolds number; MP, middle patch (straight part) of a U-shaped
pipe
Results are normalized to the case of the straight pipe at a Reynolds number of 30. Table
8 shows, for example, the removal of the contaminant patch located at MP in the U-
Shaped pipe at a Reynolds number of 30. This took 1.5 times as long to reach the same
concentration as a similar patch in the straight pipe for a Reynolds number of 30. But at a
Reynolds number of 3000, removal of the same contaminant patch in the U-Shape pipe
took 0.012 times as long as the straight pipe case with a Reynolds number of 30.
Note that in the case of high for flow in the straight portions of pipe or near the MP of the
U shaped pipe, the removal of contaminant is enhanced (111 and 125 times faster,
respectively) compared to the same geometry at the lower Reynolds number. In contrast,
the removal of contaminants is not as enhanced when the contaminant is near corners
(CP) or in a cavity, i.e., regions away from the main flow path, (83 and 56 times faster,
respectively) at the higher Reynolds number.
70
-------
The model development for flow in idealized pipe geometries has shown that pipe
geometry, the location of contaminants, and flushing rates can have a significant effect on
the decontamination of pipe systems. The higher the flush rate, the faster the removal of
contaminant; however, in some cases, the effect of the vortices could reduce the
efficiency of removal. That is, it will take a greater amount of water to reach the same
contamination level when using high flush rates because higher flows cold cause produce
vortices which could temporarily trap a contaminant and prevent it from moving forward
at the same rate of the bulk fluid. However, in general, if only hydrodynamics is
considered, parts of a pipe system that are hard to decontaminate are generally less likely
to be reached by contaminants in the first place. However, this analysis does not take into
account diffusion and mechanisms of wall interactions that are clearly important and
could significantly affect removal times. Further study, such as, validation of the
simulation model by close comparison with experiment is still needed to make this
approach a predictive tool for contaminant removal.
71
-------
Chapter 6 DISCUSSION OF RESULTS AND RECOMMENDATIONS FOR
CONTAMINANT REMOVAL
6.1 Inactivation and Flushing of Spores and Bacteria
General recommendations for the decontamination of biological threats in water systems
are difficult to formulate because of the different conditions encountered in different
systems and lack of data in real life systems. For example, corroded plumbing surfaces,
which are highly variable in terms of physical and chemical composition, have been
found in some cases, to increase the persistence and susceptibility to disinfectants of
Bacillus spores (Szabo et al., 2007) and bacteria (Szabo et al., 2006). Also, the age of the
plumbing system and local water conditions will affect the corrosion present on the pipes
and, therefore, the type and amount of biofilms likely to occur.
Data on chlorine inactivation of biological threats in water have been published in Rice et
al., 2005; Rose et al., 2005; and Hosni et al., 2009. This data, generated under carefully
controlled laboratory settings, is valuable in the design of disinfection systems, but
models must take into account the different conditions that could be present in real world
systems. Temperature, pH, solution ionic composition, and the presence of
microorganisms in the water will all affect the disinfection process. A large safety factor
for disinfection will have to be built into the design of any disinfection process. Aside
from the targeted microorganisms, many other substances present in the system to be
disinfected will also react with and thus deplete the amount of chlorine available for
inactivation of the targeted microorganisms. Thus if a Ct value of 100 is recommended
for a certain log kill, a reasonable safety factor would result in a Ct of 200 or 300,
although it is beyond the scope of this report to recommend specific safety factors as
these would be highly site specific.
Due to the complexities and the uncertainties associated with the conditions and nature of
the biological threats to be encountered in the real world, methods to monitor the actual
disinfection process in real time are especially valuable. In situ monitors to measure
chlorine demand and free chlorine in a disinfection process are valuable to provide
feedback. Free chlorine levels can be adjusted as needed, assuring that a free chlorine
level is maintained while avoiding excessive amounts of chlorine added to the
environment. Monitoring, via grab sampling and subsequent and analysis, for the
decrease in native fluorescence due to inactivation of spores (Alimova et al., 2005) and
protein toxins (Cole et al., 2008) also provides rapid feedback about the state of the
biological target. Fluorescence is a real time measurement that can be done in a
noninvasive manner and does not require additional reagents or complex sample
manipulation. However these fluorescence processes have only been studied in laboratory
conditions and not with real water systems. It will be important to prove the utility of
monitors with real world systems.
Chlorine dioxide was found to be a more effective disinfectant than chlorine for Bacillus
globigii spores in solution (Hosni et al., 2009) and should be investigated for the
disinfection of spores associated with biofilms in water systems. Two promising
72
-------
approaches, high flow rate and the use of germinants, should also be tested in large-scale
systems using real world conditions. The use of high flow increased the effectiveness of
disinfection in the pipe section bioreactors. Presumably the turbulent flow resulting from
the higher flow rates increased both the permeation of the biofilm by chlorine and the
mechanical scouring of the pipe walls. Achieving turbulent flow in small diameter pipes
and all sections of a building water system can be difficult in some configurations. The
high flow and germinant techniques for removal of Bacillus spores are discussed at
length in the Phase 1 report for this project (EPA and NIST, 2011).
The use of germinants is also a promising approach that does not involve the use of
environmentally damaging chemicals, or at least allows lower levels of such chemicals to
be used. The germinants used in this research were highly purified reagent grade amino
acids, alanine and inosine. It should be investigated to determine if low cost bulk
chemicals, such as food supplements could work as well as the more expensive reagent
grade chemical in the full-scale plumbing system test facility described earlier. The use of
such biological stimulants should be tested for disinfection efficiency in the test facility
to determine the efficiency of disinfection in a near to real life system. As in the
contacting experiments the sampling and analysis of these experiments will require
significant planning and coordination of effort because of the large number and nature of
the samples. Because of the complexity of the test facility, the disinfection stage of these
experiments will need to be carried out for extended periods of time until the system
reaches a level judged to be reasonable. The efficiency of the disinfection process in the
test facility will be very valuable because of the relevance to the situation in real-life
systems.
6.2 Dynamic Flushing of Diesel
As described earlier, detailed dynamic tests were done to measure the adherence and
removal of diesel from plumbing material. Only tests with diesel fuel were performed
due to the significant effort involved in calibrating a particular fluid. In general, the
measurements suggest that diesel fuel is readily flushed from copper, iron, and PVC pipe
substrates. Because the measurement technique is a direct measure of the contaminant
thickness, the uncertainty of the measurement implies an uncertainty as to whether or not
the diesel fuel has been completely removed from the surface by flushing with water.
Subsection 3.2.1 shows the resolution or uncertainty to be on the order of tens of
nanometers. It is likely that water flushing does not completely remove diesel fuel from a
pipe surface and that a layer less than the sensitivity of the technique (tens of nanometers)
exists. By the same token, it is likely that recontamination of the water would be minimal
since the amount of residual contaminant is much less than the initial contaminant layer,
and thus there is much less contaminant available for recontamination. Further study is
required to quantify the residue of diesel fuel left behind after flushing and the potential
for the residue to re-contaminate a cleaned plumbing system.
Despite the aforementioned shortcomings, the computer software presented in Section
5.1.4 can be used to estimate the required flushing time for diesel fuel, but a safety factor
73
-------
of two or more should be incorporated when making calculations for flushing time for
straight sections. In addition, as was shown in Subsection 5.2, longer flushing times
would likely be required for obstructed locations such as valves and cavities. Based on
simulation results described in Subsection 5.2, flushing time multipliers of 10 to 20 might
be required to ensure contaminant removal from locations that are difficult to flush.
Verification of the efficacy of the flushing should be confirmed at desired intervals. In
addition, hot water and/or cleaning solutions could be desirable in flushing efforts.
6.3 General Considerations for Real World Scenarios
In this section general recommendations for decontamination are given for a number of
scenarios. These recommendations are based in part, on the studies and modeling
described in this report. In addition knowledge of plumbing systems and sound
engineering principles also played an integral part in developing recommendations for
decontamination.
For the contaminants studied essentially all of them adhered to plumbing surfaces. The
differences were in the strength of that adherence. Compounds such as diesel did not
bond as strongly to surfaces compared to cyanide and mercuric ions, both of which
bonded to copper pipe in an irreversible and complex manner. In the following
discussion it is assumed that a contaminant will stick to most plumbing surfaces.
In general, a commons sense approach to the restoration of a building plumbing system
following a contamination scenario leads to two basic steps:
(1) safely purge the system of the contaminated water and (2) flush or treat the system to
eliminate of any accumulated contaminant. The most effective methods for removing the
accumulated contaminants will be a function of the contaminant, plumbing system
materials and design. The preferred method, in terms of simplicity and cost is
conventional flushing using water directly from the water distribution system. However
other methods might be faster or more effective at cleaning. Three choices that need to
be made are as follows:
Which flushing fluid should be used?
How should the flushing fluid be introduced?
How should the system be flushed?
The fluid could be:
water as is available from the distribution system
water that has been amended with additional chlorine or other disinfectant
water that has been amended with surfactants or other chemicals to neutralize
otherwise interact with the contaminant
hot water or steam
germinant solution to promote germination of bacterial spores
74
-------
The fluid source could be:
the distribution system
a reservoir supplied by the distribution system
a tanker truck
The flushing method could be:
conventional flushing, effuse down drain
conventional flushing, effuse collected for disposal
flood system and let stand, then drain effuse
flood system and let stand, then collect effuse for disposal
high velocity pumping
steam injection
pulsating flow
An effective method for removing chemical contaminants is by continuous flushing.
This is concluded based on the general premise that accumulated contaminants tend to
become entrained in the water due to turbulence, advection, and diffusion. The work
done with dynamic fluid/surface interface measurements support this premise. In light of
this conclusion, removal is primarily a function of the amount of clean water passing
through the system. This approach works best for pipe sections, but, as was seen in the
hot water heater studies discussed earlier, it is less efficacious for water tanks or
reservoirs, especially those that have a top outlet. Flush water velocities are very low in
tanks with large diameters relative to inlet and outlet pipe diameters. Thus, the
interaction of flush water with any contaminant that has accumulated on surfaces exposed
to the water is slight. Given enough time/water, water soluble contaminants will tend to
be removed from the plumbing system. The amount of time/water required to reduce
contaminant residuals to an acceptable level depends upon many factors. These include
the type and severity of the contamination, the design of the plumbing system, and the
residual levels considered to be acceptable. The complete evaluation of these factors is
beyond the scope of this report. Previous contamination events and measurements with
immiscible organic substances suggest that flushing times on the order of days could be
required in some cases (Kedzierski, 2006; Morbidity and Mortality Weekly Reports,
1981).
A software tool has been developed to assist in the determination of recommended flush
times for contamination events. The details of the software tool are described in
Subsection 5.1.4. The general concept is based on modeling the interaction between the
contaminant and the pipe walls under different flow conditions to predict the contaminant
accumulation and removal. Certain parameters are used as inputs to the software tool.
These include mass fraction of contaminant and exposure time, flushing Reynolds
number. The tool provides a recommended flushing time as the output. Currently, the
software tool has been validated using detailed measurements of diesel fuel/water
75
-------
mixtures, but is under development to expand its capabilities to cover a wider range of
contaminants.
For less-miscible contaminants, flushing with hot water will dissolve the contaminant
compared to cold water and can thus shorten required flushing times. Immiscible
contaminants that float are best removed from water tanks by flushing out through the
top, rather than draining from the bottom. The latter procedure allows a high
concentration of the contaminant to come into direct contact with the sediments that tend
to accumulate at the bottom of the tank, which makes removal more difficult. In contrast,
immiscible contaminants that sink should be drained from the bottom of the tank, and the
sediments flushed out if possible.
Bacterial contaminants are best attacked by flooding the plumbing system with
disinfectants such as chlorine, and letting it stand to kill or otherwise disable the bacteria.
A short flushing to remove the chlorinated water and inactivated bacteria could then be
done to make the plumbing system operational. (Reipa et al., 2006).
Based on the work described at length in EPA and NIST, 2011, bacterial spores could be
treated by flooding the affected plumbing system with germinant solution, such as an
inosine or L-alanine solution, which would encourage the spores to germinate making
them susceptible to inactivation by disinfection in the same manner as vegetative
bacteria. It was found that the use of hot water enhances this type of decontamination
process. Also, since the disinfection and growth media solutions only need to flood the
system, relatively small amount of solution is required. This is in contrast to continuous
flushing, which requires a much larger volume of water.
Flushing with water from the water distribution system is the simplest flushing method.
However, it is difficult to control water velocities since they depend on water pressure,
which varies with location and function of demand. Sequential flushing of individual
water lines will provide the highest water velocities, and, therefore, the greatest scour.
Water saving devices, such as faucet aerators and showerheads, should be removed and
cleaned or disposed of before flushing to allow for greater flow rates. It is important to
ensure that the number of water lines being flushed simultaneously does not overwhelm
the capacity of the building drainage system, since the size of drainage lines are specified
based on the assumption that only a percentage of water fixtures and appliances discharge
to the drain system at a given time. Additional chlorine or other disinfectant or cleaner
could be introduced into the flushing water, either directly from the distribution system or
from an auxiliary source closer to the contaminated water system. Care should be taken
to ensure that the concentrations of disinfectant or cleaner will not damage the plumbing
system materials.
Another concern is that a contaminant might become an inhalation hazard due to
volatilization into the air after exiting a faucet or other fitting above a sink or tub. Where
that is a consideration, avoiding this problem might require special flushing precautions,
such as a direct connection between water supply outlets and drain lines. Following
76
-------
flushing, all sinks, tubs and other surfaces that have been exposed to the contaminated
water should be thoroughly cleaned.
Water-using appliances present another set of challenges regarding decontamination.
The greatest concerns lie with those that involve water that might be consumed or come
into contact with building occupants. Chief among these are water tanks, such as hot
water heaters, that tend to accumulate sediments and deposits. In general sediments and
deposits in tank bottoms were areas which accumulated contaminants. These were
difficult to flush due to the large tank volume and corresponding low velocities.
Cleaning water tanks can require direct draining and filling with special cleaning
solutions using one of the techniques described below. In some cases, it might not be
possible to eliminate all of the accumulated contaminant, and water lines, fittings,
fixtures or appliances might need to be replaced. Appliances that have no drain
provision, such as residential ice makers, will not be able to be flushed so will need to be
removed, cleaned offline, or replaced. Appliances such as dishwashers and clothes
washers cannot be flushed, but they can be cleaned through operation or disconnected to
allow their supply lines to be flushed.
Figure C61 shows a schematic depiction of a possible configuration for back flushing a
building water supply system. This is done without using water directly from the
distribution system but instead a separate source of flushing water is connected via an
external water spigot or other similar connection point. It could be necessary to bypass or
disable any back-flow prevention device to allow water to flow in the reverse direction.
The flushing water can be pre-treated as is appropriate with disinfectant or cleaner before
injection. If the valve at the water meter is closed, the flushing water will be forced
under pressure through the water heater and both the cold and hot water supply lines to
any fitting or fixture (e.g. faucet, shower, etc.) that is open. In this configuration, the
flush water would be directed into the sanitary drains. It could also be collected and
transported for disposal. The water supply line leading from the service connection
would need to be flushed separately in the normal flow direction.
Figure C62 shows a similar configuration for decontaminating a water heater tank or for
using the water heater as an injection point for flushing the building water supply system.
This operation would be similar to that described previously. Although in this
configuration, cleaning fluid could be both injected and extracted directly from the water
heater via the drain valve. This could be repeated as many times as necessary.
6.4 Recommendations for Contaminant Removal from Plumbing Surfaces
This report presents an overview of measurements and analysis of contaminant
accumulation. It includes removal in building plumbing systems and methods for
decontaminating building plumbing systems, and restoring their operation based on both
specific and generic contaminant characteristics. Some of the measurements were used to
develop fundamental models to predict maximum contamination levels and required
77
-------
flushing times. These models were, in turn, used to develop computer software that can
be used as part of a response to a contamination event.
Measurements typically showed that the contaminants did stick to the plumbing material
substrates after the initial exposure. (There were some exceptions to this. As described in
Appendix A, Raman spectroscopy showed that neither strychnine nor mercuric chloride
adhered appreciably to rubber pipe material, and mercuric chloride did not adhere
appreciably to either brass or PVC. However Raman spectroscopy is not as sensitive
compared to some of the other methods used which did show evidence of compound
adherence.) Lastly, diesel fuel and toluene showed a substantial reduction from flushing
with clean tap water. Some other contaminants required the addition of high levels of
chlorine to effect removal (phorate, gasoline, biologicals).
For the various measurements that were conducted, both exposure and flushing
conditions varied considerably. The results were surprisingly consistent regarding the
tendency of the particular contaminants to stick and the difficulty of removing them by
flushing with tap water. It is not clear if water flushing alone can effectively remove
strychnine, phorate or cyanide contamination. Even if it did, it is likely that the required
water volumes would be so large as to be impractical. A better approach might be to
flush with hot water with high chlorine levels or water with detergent.
Specific recommendations that link decontamination procedures to particular
contaminants or groups of contaminants with similar characteristics are given in Table 9.
It is envisioned that these recommendation will prove useful as a starting point for a set
of comprehensive guidelines that support general response plans for effective recovery
from water supply system contamination events. Although some of the procedures
recommended in Table 9 were not effective or only partially effective in some of the
actual testing, the general approach for specific scenarios is judged to be a reasonable
approach from a practical point of view. For example, continuous flushing is seen as a
sound approach for water soluble contaminants. Further testing using additives such as
detergents or non-toxic solvents may yield more effective procedures.
78
-------
Table 9 General Decontamination Procedures Based on Contaminant Type
Contaminant
Category
Soluble chemicals
Immiscible
chemicals with
specific gravity less
than one
Immiscible
chemicals with
specific gravity
greater than one
Sediments or
particles
Bacteria
Bacterial spores
Toxins
Example
from
Category
Strychnine,
Cyanide,
Salts
Diesel fuel,
Gasoline
Phorate
Foreign
particles
E. coli
0157:H7
Bacillus
anthracis
Ricin
Key Methods
For Pipes
Continuous flushing with
water, water buffered with
chlorine, or water mixed
with cleaner
Continuous flushing with
water, water buffered with
chlorine, or water mixed
with cleaner
Continuous flushing with
water, water buffered with
chlorine, or water mixed
with cleaner
Continuous flushing with
water, drain from
cleanouts where available
Flood system with water
and disinfectant and let
stand, followed by short
flush. Repeat as needed
Flood system with
germinant solution and let
stand to allow spores to
germinate, followed by
bleach disinfection
Continuous flushing with
water, water buffered with
chlorine, or water mixed
with cleaner
For Tanks
Continuous flushing
with water, water
buffered with
chlorine, or water
mixed with cleaner
Flush through fitting
at top of tank
Drain through drain
valve at bottom of
tank, and fill with
cleaning solution.
Repeat as needed
Drain and flush from
bottom
Flood system with
water and
disinfectant and let
stand, followed by
short flush. Repeat as
needed
Flood system with
germinant solution
and let stand to allow
spores to germinate,
followed by bleach
disinfection
Continuous flushing
with water, water
buffered with
chlorine, or water
mixed with cleaner
79
-------
The following is a summary of the recommended list of steps for dealing with a potential
contamination scenario involving a building plumbing system following a complaint or
other indication of a problem:
Collect and analyze water samples to determine if the complaint is associated with
the presence of a contaminant. Identify the contaminant and measure the
contaminant concentration.
Determine the extent of the contamination.
Isolate the contaminated water piping to prevent propagation to uncontaminated
piping.
Locate the source or point of introduction of the contaminant.
Determine if the contaminated water can be flushed into the waste water system.
Assess volatilization potential of contaminant if exposed to atmospheric pressure
within a building.
Determine maximum drainage water flow rate per building to prevent overloading
the drainage system.
Run predictive computer software to estimate contaminant accumulation within
the plumbing system.
Run predictive computer software to determine required flush rate and flush
duration.
If deemed safe, flush with water while simultaneously exploring alternatives.
Select appropriate decontamination procedure.
If it is water, continue flushing.
If it is cleaning agent or shock chlorine, select injection point, flush with solution.
If waste water cannot be discharged into the drainage system, collect waste water.
Circulate cleaning solution.
Verify effectiveness of decontamination effort.
Analyze water samples.
Analyze pipe samples.
Determine if remedial measures are needed to restore plumbing system
components.
Clean/replace faucets, valves, aerators, tanks, and hoses.
Evaluate possible pipe surface restoration using mechanical or chemical
procedures.
Replace and dispose of any components that could not be decontaminated.
80
-------
7.0 REFERENCES
Alimova, A., Katz, A., Siddique, M., Minko, G., Savage, H. E., Shah, M. K., Rosen, R.
B., and Alfano, R. R., 2005, "Native Fluorescence Changes Induced by Bactericidal
Agents," IEEE Sensors Journal 5(4): 704-711.
Alliot, L., Bryant, G., and Guth, P.S, 1982, "Measurement of Strychnine by. High-
PerformanceLiquidChromatography," J. Chromatogr. 232(2): 440-442.
Almeida, J., Wang, L., Morrow, J.B., and Cole, K.D, 2006, "Requirements for the
Development of Bacillus anthracis Spore Reference Materials Used to Test Detection
Systems," J. Res. Natl. Inst. Stan. 111(3): 205-217.
Almeida, J.L., Harper, B., and Cole, K.D., 2008, "Bacillus anthracis Spore Suspensions:
Determination of Stability and Comparison of Enumeration Techniques," J. Appl.
Microbiol. 104(5): 1442-1448.
Amadeo, J. P., Rosen C., and Pasby, T. L., 1971, Fluorescence Spectroscopy: An
Introduction for Biology and Medicine, Marcel Dekker, Inc., New York, p. 153.
ASTM, 1999, Method Dl 125-95, "Standard Test Methods for Electrical Conductivity
and Resistivity of Water."
ASTM, 1995, Method D5790-95, "Standard Test Method for Measurement of Purgeable
Organic Compounds in Water by Capillary Column Gas Chromatography/Mass
Spectrometry."
ASTM, 1996, Method D1253-86, "Standard Test Method for Residual Chlorine in
Water."
AWWA (American Water Works Association), 199la, Method 2320, "Alkalinity."
AWWA, 1991b, Method 5310 C "Total Organic Carbon by Persulfate-UV or Heated
Persulfate Method."
AWWA, 1990, Method 2310, "Acidity."
Budavari, S., O'Neil, M.J., Smith, A., Heckelman, P.E., andKinneary, J.F., Eds., 1996,
The Merck Index, 12th Edition, Merck and Co., Inc., Whitehouse Station, NJ.
Camper, A.K., Brastrup, K., Sandvig, A., Clement, J., Spencer, C., and Capuzzi, A. J.,
2003, "Effect of Distribution System Materials on Bacterial Regrowth," J. Am. Water
Resour. As. 95(7): 107-121.
Cole, K.D., Gaigalas, A.K., and Almeida, J.L., 2008, "Process Monitoring The
Inactivation of Ricin and Model Proteins by Disinfectants Using Fluorescence and
Biological Activity," Biotechnol. Prog. 24(3): 784-791.
81
-------
EPA (U.S. Environmental Protection Agency) and National Institutes of Standards and
Technology (NIST), 2011.
"Development and Testing of Methods to Decontaminate a Building's Plumbing System
Impacted by a Water Contamination Event: Decontamination of Bacillus Spores," EPA
600/R-09/089.
EPA, 2005, National Primary Drinking Water Regulations,
http://www.epa.gov/safewater/mcl.html#mcls (accessed December 12, 2011)
EPA, 1996, Method 9213, "Potentiometric Determination of Cyanide in Aqueous
Samples and Distillates with Ion-Selective Electrode."
EPA, 1995a, Method 551.1, "Determination of Chlorination Disinfection Byproducts,
Chlorinated Solvents, and Halogenated Pesticides/Herbicides in Drinking Water by
Liquid-Liquid Extraction and Gas Chromatography with Electron-Capture Detection."
EPA 1995b, Method 524.2, "Measurement of Purgeable Organic Compounds in Water by
Capillary Column Gas Chromatography/Mass Spectrometry."
EPA 1989a, Method 502.1, "Volatile Halogenated Organic Compounds in Water by
Purge and Trap Gas Chromatography."
EPA 1989b, Method 503.1, "Volatile Aromatic and Unsaturated Organic Compounds in
Water by Purge and Trap Gas Chromatography."
EPA, 1978, Method 180.1, "Turbidity (Nephelometric)."
Gaigalas, A.K., Cole, K.D., Bykadi, S., Wang, K., and DeRose, P., 2007, "Photophysical
Properties of Ricin." Photochem. Photobiol. 83(5): 1-8.
Hamid, Z. A. and Aal, A. A., 2009, "New Environmentally Friendly Noncyanide Alkaline
Electrolyte for Copper Electroplating." Surf. Coat. Tech. 203(10-11): 1360-1365.
Hawkins, C.L., Pattison, D.I., Davies, andMJ., 2003, "Hypochlorite-Induced Oxidation
of Amino Acids, Peptides and Proteins," Amino Acids 25(3-4): 259-274.
Hong, F. andPehkonen, S., 1998, "Hydrolysis of Phorate Using Simulated
Environmental Conditions: Rates, Mechanisms, and Product Analysis," J. Agric. Food
Chem. 46(97): 1192-1199.
Hosni, A.A., Shane, W.T., Szabo, J.G., and Bishop, P.L., 2009, "The Disinfection
Efficacy of Chlorine and Chlorine Dioxide as Disinfectants of Bacillus globigii, a
Surrogate for Bacillus anthracis, in Water Networks," Can. J. Civ. Eng. 36(4): 732-737.
82
-------
Ismail, I, Abdel-Monem, N., Fateen, S.E., and Abdelazeem, W., 2009, "Treatment of a
Synthetic Solution of Galvanization Effluent Via the Conversion of Sodium Cyanide into
an Insoluble Safe Complex," J. Hazard. Mater. 166(2-3): 978-983.
Kays, W.M., and Crawford, M.E., 1980, Convective Heat and Mass Transfer, McGraw-
Hill, New York.
Kedzierski, M. A, 2008, "Diesel Adsorption to PVC and Iron During Contaminated Water
Flow and Flushing Tests," NISTIR 7520, U.S. Department of Commerce, Washington,
D.C.
Kedzierski, M. A, 2006, "Development of a Fluorescence Based Measurement Technique
to Quantify Water Contaminants at Pipe Surfaces During Flow," NISTIR 7355, U.S.
Department of Commerce, Washington, D.C.
Kedzierski, M.A., 2002, "Use of Fluorescence to Measure the Lubricant Excess Surface
Density During Pool Boiling," Int. J. Refrigeration, 25(8): 1110-1122.
Ku, Y. and Lin, H., 2002, "Decomposition of phorate in aqueous solution by photolytic
ozonation ," Water Res. 36(16): 4155-4159.
Martys, N.S., 2001, "A Classical Kinetic Theory Approach to Lattice Boltzmann
Simulation," Int. J. Mod. Phys. C 12(8): 1169-1178.
Martys, N. S., 1994, "Fractal Growth in Hydrodynamic Dispersion Through Random
Porous Media," Phys. Rev. E 50(1): 335-342.
Matz, L.L., Beaman, T.C., and Gerhardt, P., 2001, "Chemical Composition of
Exosporium From Spores of Bacillus cereus" J. Bactedol. 101(1): 196-201.
Mays, L., 2000, Water Distribution Systems Handbook, McGraw-Hill, New York.
Miller, J.N., 1981, Standards in Fluorescence Spectrometry, Volume 2, pp. 44-67.
Chapman and Hall, London.
Morbidity and Mortality Weekly Reports, 1981, "Chlordane Contamination of a Public
Water Supply- Pittsburgh, Pennsylvania," MMWR 30(46):571-572, 577-578.
Morrow, J. B., Almeida, J. L., Fitzgerald, L.A. and Cole, K.D., 2008, "Association and
Decontamination of Bacillus Spores in a Simulated Drinking Water System," Water Res.
42(20): 5011-5021.
Morrow, J.B., and Cole, K.D., 2009, "Enhanced Decontamination of Bacillus Spores in a
Simulated Drinking Water System by Germinant Contact," Environ. Eng. Sci. 26(5): 993-
1000.
83
-------
Nightingale, Z.D., Lancha, A.H., Handelman, S.K., Dolnikowski, G.G., Busse, S.C.,
Dratz, E.A., Blumberg, J.B., and Handelman, G.J., 2000, "Relative Reactivity of Lysine
and Other Peptide-Bound Amino Acids to Oxidation by Hypochlorite," Free Radical Bio.
Med. 29(5): 425-433.
Radnedge, L., Agron, P.O., Hill, K.K., Jackson, P.J., Ticknor, L.O., Keim, P., and
Andersen, G.L., 2003, "Genome differences that distinguish Bacillus anthracis from
Bacillus cereus and Bacillus thuringiensis" Appl. Environ. Microbiol. 69(5): 2755-2764.
Reipa, V., Almeida, J. and Cole, K., 2006, "Long-Term Monitoring of Biofilm Growth
and Disinfection Using a Quartz Crystal Microbalance and Reflectance Measurements,"
Journal of Microbiol. Meth. 66(3): 449-459.
Rice, E.W., Adcock, N.J., Sivaganesan, M., and Rose, L.J., 2005, "Inactivation of Spores
of Bacillus anthracis Sterne, Bacillus cereus., and Bacillus thuringiensis Subsp.
israelensis by Chlorination," Appl. Environ. Microbiol. 71(9): 5587-5589.
Rose, L.J., Rice, E.W., Jensen, B., Murga, R., Peterson, A., Donlan, R.M. and Arduino,
M.J., 2005, "Chlorine Inactivation of Bacterial Bioterrorism Agents," Appl. Environ.
Microbiol. 71(1): 566-568.
Schwartz, T., Hoffmann, S., and Obst, U., 2003, "Formation of Natural Biofilms During
Chlorine Dioxide and U. V. Disinfection in a Public Drinking Water Distribution
System," Journal of Appl. Microbiol. 95(3): 591-601.
Szabo, J.G., Rice, E.W. and Bishop, P.L., 2007, "Persistence and Decontamination of
Bacillus atrophaeus on Corroded Iron in a Model Drinking Water System," Appl.
Environ. Microbiol. 73(8): 2451-2457.
Szabo, J., Rice, E. and Bishop, P.L., 2006, "Persistence of Klebsiella Pneumoniae on
Simulated Biofilm in a Model Drinking Water System," Environ. Sci. Techno. 40(16):
4996-5002.
Treado, S.J., 2007, "The Decontamination of Building Plumbing Systems- Analysis and
Procedures," NISTIR 7448, National Institute of Standards and Technology,
Gaithersburg, MD 20899.
Wingender J. and Flemming, H.C., 2004, "Contamination Potential of Drinking Water
Distribution Network Biofilms," Water Sci. Technol. 49(11-12): 277-286.
84
-------
Appendix A Chemlmage: NIST Pipe Contamination Study
Return to text Section 4.1.1.4
Researcher: John Maier
Chemlmage
73Q1 Penn Avenue, Pittsburgh, PA 1520B
Tel 412.241 7335 Fax 412.241.7311.
www.chemimage.com
rtiliie your vision.
REPORT
Title: NIST pipe contamination study
Type of Report: Contract Research
Principal Investigator:
John Maier
maier@chemimage.com
Date of Publication: June 17, 2008
Contracting Officer's Representative:
ATTN: Stephanie Watson
Address:
Building & Fire Research Laboratory 100 Bureau Drive, Stop 8615 Building 226, Room
B344 Gaithersburg, MD 20899-8615 Tel: 301-975-6448 E-mail:
Stephanie.Watson@nist.gov
A-l
-------
Contractor Name: Chemlmage Corporation Contractor Address: 7301 Penn Avenue, Pittsburgh, PA 15208
Executive Summary
This report focuses on the methods used to evaluate a series of pipe samples
exposed to two different contaminants. The work was performed for Stephanie
Watson on NIST in May of 2008. In this project fluorescence spectral imaging
with UV illumination in conjunction with PCA image analysis was used to study
the contamination of different types of pipe material with different contaminants.
The limitations of this approach include no sensitivity to contamination that does
not change the luminescence of the sample under UV light illumination and the
lack of specific chemical information about the chemical nature of the residues
left on the pipes.
With those limitations in mind, this analysis shows that the rubber pipe material
does not suffer significant residual contamination for either contaminant. Brass
and PVC similarly do not show any evidence of residual contamination with
mercuric chloride. Brass shows evidence of complete residual contamination
with strychnine while the PVC has 78 % surface area contaminated with
strychnine. The iron samples show evidence of complete residual contamination
with both the mercuric chloride and strychnine contaminants used in this study.
The copper samples show residual contamination for both contaminants, but not
complete contamination in either case. For the case of strychnine 66% of the
surface appears to have residual contamination while for the case of mercuric
chloride only 9% of the surface appears to have residual contamination.
Background
NIST is involved in a research project to study the interaction of two fouling
contaminants on 5 different types of pipe. The basic question is: after an
extended exposure of the contaminants and a washing step with water, how
much of the surface remained contaminated. After discussion with Chemlmage,
an experimental approach was designed including a preliminary set of
experiments to assess what spectral imaging methods might help answer this
question and a second set of more focused experiments chosen based on
performance in the preliminary work.
A-2
-------
Materials and Methods
Samples (provided by NIST):
D 5 unexposed pipe samples
D o Copper
D o Brass
D o Iron
D o Rubber
D o PVC
D 5 pipe samples exposed to mercuric chloride (HgCb)
D 5 pipe samples exposed to Strychnine
D Samples of chemicals exposed to pipe and precipitates from pipes
D o Strychnine
L o HgCl2
D o Precipitate formed on iron exposed to mercuric chloride
D o Precipitate formed on brass exposed to mercuric chloride
D o Precipitate formed on copper exposed to mercuric
chloride
Preliminary work
(No unexposed samples were available for this phase of work)
Raman spectroscopy and Imaging
Raman spectroscopy was performed using Chemlmage's Falcon II Raman
microimaging system. The spectra were acquired using 532 nm excitation. Data
collection parameters varied between samples. Raman spectra were collected
for the pure components and the precipitate samples. The pure component
materials, mercuric chloride and strychnine, produced very strong Raman
scattering with a high signal to noise (SNR) ratio; however the precipitate
samples were highly fluorescent and the Raman bands were obscured by the
abundant fluorescence. Figure A1 shows a Raman spectral library of the pure
component and precipitate samples provided by NIST. The precipitates were
analyzed in situ, that is, they were not scraped off prior to analysis, but the
entire sample, the precipitate and the underlying pipe material, were placed in
the instrument's sample platform where the precipitate's Raman spectrum was
then evaluated.
A-3
-------
BROWN: Precipitate formed on copper exposed to HgCI2
LIGHT BLUE: Precipitate formed on brass exposed to HgCI2
GREEN: Precipitate formed on iron pipe exposed to HgCI2
BLUE: HgCI2, pure compound
RED: Strychnine, pure compound
Raman Shift cm1
Figure A 1. Raman spectra of pure components and precipitate samples.
The spectra shown in Figure A1 have not been baseline corrected, but have
been normalized using vector normalization. They are displayed on the same y-
axis with an offset to facilitate visual comparison. The Raman peaks are sharp
and distinct while the luminescence is a broad background spectral response.
Fluorescence and IR macro imaging
In the preliminary phase of this study, the exposed pipe samples were analyzed
using Macroscopic Near-Infrared Chemical Imaging, Visible Absorbance
Chemical Imaging and Fluorescence Chemical Imaging. The five pipe samples
exposed to mercuric chloride and the five pipe samples exposed to strychnine (a
total of 10 samples), were placed in a single field of view and a dataset was
collected for each mode. In addition to the sample data collection, a background
image was collected (for each mode) using the same respective parameters.
The parameters used for each data collection are as follows:
Near Infrared Imaging Spectral Range: 850 - 1800 nm Step Size: 5 nm
Exposure Time: 16.36 msec Averaging (per frame): 10 Binning: None
Processing: Divide by background image, Convert to Absorbance, Extract
wavelengths 950 - 1650 nm, Normalize
A-4
-------
Visible Reflectance Imaging Spectral Range: 420 - 720 nm Step Size: 5 nm
Exposure Time: 0.5 sec Averaging (per frame): 1 Binning: None Processing:
Divide by background image, Normalize
Fluorescence Imaging Spectral Range: 420 - 720 nm Step Size: 10 nm
Exposure Time: 60 sec Averaging (per frame): 1 Binning: None Processing:
Divide by background image, Normalize
Only the fluorescence chemical image of the 10 exposed pipe samples
yielded a promising result. In order to better understand the information
contained in the fluorescence chemical image, it was necessary to analyze
unexposed pipe samples along with the exposed samples.
Phase 2 work
Phase 2 work was started after discussing the preliminary results with Stephanie
Watson of NIST. For phase 2 of the project, the control (un-exposed) pipe
samples were placed on the same platform as the exposed pipe samples (15
samples total in the field of view). A fluorescence chemical image dataset was
collected for the samples according to the following parameters:
Spectral Range: 420 - 720 nm Step Size: 10 nm Exposure Time: 60 sec
Averaging (per frame): 1 Binning: 2x2
Analysis
1. Goals
Discussion with Stephanie Watson regarding the preliminary results led to a
second phase of work focused on using spectral fluorescence imaging to get
data which would be evaluated to estimate the percent area of the surface that
underwent change based on spectral differences. Principal Component Analysis
(PCA) was chosen as an established tool for analysis of the spectral images. As
described below, percent area changed was calculated based on first measuring
the area of each sample in the image and second using PCA images of the
samples of a given type of pipe to estimate the area of each sample which is
spectroscopically different from normal.
One limitation of this approach is the absence of the specific chemical
information about what compounds were causing the spectral differences in the
contaminated samples. A second limitation is that a change would not be
detected using this approach unless the interaction of the contaminant and the
sample manifested itself as a change in the sample luminescence.
A-5
-------
2. Estimation of total area of samples
The area of each sample was estimated using a frame from the visible
reflectance image and a two step masking process. A frame was selected in
which all of the samples were visible with positive or negative contrast compared
to the background. Figure A2 shows a single frame from the visible reflectance
image of the samples. A Region Of Interest (ROI) selection tool in Chemlmage
Xpert was used to select a region containing samples of similar intensity and
contrast relative to the background. A first binary mask was created in the shape
of the selected region of interest and multiplied by the image. This results in an
image of the same size as the original image, but only containing information
within the boundaries of the selected region. A histogram of this image shows
two distinct non-zero populations, one for the background and one for the
sample. A second mask is created using a histogram based threshold tool in
Chemlmage Xpert. In this second mask a pixel is set to one if the pixel is in the
population which corresponds to the samples and zero if it is not.
Strychnine
Control
HgCL
Copper Iron Brass Rubber PVC
This process is carried out so that ultimately there is a mask for each sample.
Figure A 2. Visible reflectance image of samples.
The mask is used subsequently for two parts of the process. First, the sum of
the intensity in a region of the mask containing a sample is a value proportional
to the area of the sample. Second, the mask is also used to multiply the spectral
fluorescence image yielding a fluorescence image which only contains data at
the site of the specific sample. In this fashion the fluorescence spectral image of
only the three copper samples (for example) can be analyzed together, yet
separate from the other types of pipe samples. Figure A3 shows the same
A-6
-------
image as in Figure A2 with a red overlay of the mask created using the above
approach. This image demonstrates how accurately the method described
above captures the area of the samples.
Strychnine
Control
HgCL
Copper Iron Brass Rubber PVC
Figure A 3. Overlay of area masks with samples.
3. Estimation of affected area of samples
In order to estimate the contaminated area of a set of samples the fluorescence
spectral image is masked to contain only information from the set of samples. A
mask for the two contaminated samples and the uncontaminated sample of one
type of pipe is selected from the mask image of all the samples. This mask image
is multiplied by the fluorescence spectral image of all the samples yielding an
image which contains information only where the three samples of a given
material are present. For the pixels which are not on one of the samples the data
values in the image are zero for all spectral frames.
ChemImage Xpert was used to perform PCA on such images of each set of
pipe material. The result of the PCA is a set of principal component loadings
and a set of score images. The loadings are the orthogonal basis vectors,
which result from the principal component analysis. The score images are
A-7
-------
maps of the weight of each of the principal components, or loadings, at each
pixel. PCA is an objective analysis of the mathematical variability in a set of
spectra. In the case of this application of PCA, the set of spectra are the
spectra from each pixel in the image of a set of samples of one pipe type.
Without knowing the chemical origin of changes which cause the spectra to
be different, PCA image analysis allows objective assessment of whether
spectra from pixels in either exposed sample are similar to, or different from,
the unexposed sample. This manifests itself in the score image.
For instance, in a set of spectrally identical samples, the score image would be
expected to be have the samples be bright and dark in the same frames. There is
a frame of the score image for each principal component of the analysis.
Alternatively, two samples of one type and a third sample of another type would
lead to score images which would consistently show the like samples having like
contrast.
The principal component spectra, or loadings, are mathematically determined,
and are not necessarily determined in such a way that specific identifiable
spectral signatures are present in the loadings. In fact, if a distinctive spectral
signature is present in the sample set, it may mathematically fall as a linear
superposition of one or more of the loadings determined through PCA. Various
approaches exist to rotate the loadings in a fashion that maintains the
orthogonality of the basis set, while aligning the basis vectors with more
interpretable spectra. This analysis was not performed in this project because
there was not a specific goal of establishing the spectra of different components.
The score images can be evaluated using the same histogram based threshold
approach used to determine the masks for estimating the area of each sample. A
difference in intensity in a given score image indicates a difference in the
material, especially when it is correlated by spatially similar observations in other
score images. Evaluation of the histogram of a score image shows graphically
when there is obviously more than one component present. Setting bounds on
the intensity values for inclusion in a mask will yield a binary mask with ones
where pixels with a particular spectral type appear.
4. Example of application of contaminated area estimation approach:
The subject of this study is how much of the surface is contaminated, not just
whether there is a difference overall. Spatial variation of the spectral change
which indicates interaction with the contaminant can be analyzed by considering
the individual spectra of each pixel, and the degree to which they are different
from the spectrum of the normal sample.
An example of the analysis described above and associated data is presented
A-8
-------
below.
In the first step the mask which selects for the copper samples is multiplied
by the spectral fluorescence image yielding the image shown in Figure A4.
A-9
-------
Figure A4 is the frame at 560 nm of the masked fluorescence spectral image.
For reference, the unexposed sample is in the middle, the strychnine exposed
sample is on the top, and the HgCb exposed sample is on the bottom (see Figure
A2). The gray scale is chosen to span the range of non-zero data in the frame at
560 nm. All of the pixels in this frame which are black have a value of zero, and
will therefore not contribute to the subsequent analysis of spectral variability.
i:-
c.s:
:.ฃ:-
D.TU-
:.ฃ:-
D.M-
D.4B-
D.13-
D.23-
D.10-
c.c:-
5ic fit =v: ^o: 53D EEO 630 722
Figure A 5. Frame from spectral image data Figure A 4. Spectra extracted from image of
for copper samples. samples.
Figure A5 shows the mean spectra of the ROIs shown in Figure A4. Note that
ROI3 (the strychnine sample) has a different spectrum from the HgCb and
unexposed sample indicating the difference of this sample from normal. We
assume this difference is due to the exposure. It is clear that the mean spectrum
of the strychnine exposed sample is different from normal.
A principal component analysis is performed using Chemlmage Xpert chemical
imaging software. The first frame of the resulting score image is shown in Figure
A6 (left). Figure A6 (right) shows the histogram of intensity values for this frame.
This histogram plots the number of pixels (y-axis) as a function of the intensity
value (x-axis) where the intensity value is the weight that results from the
principal component analysis for the first principal component. This weight is the
relative weight of the spectral loadings which also result from PCA. The loadings
are the orthogonal basis set which are chosen to account for the variability in the
set of spectra (in this case the spectra from the pixels where there is sample
material).
A-10
-------
The weights in the first principal component for the spectra in this case are
negative numbers. This is because the PCA performed here is not run with
constraints that the loadings, or principal components, appear like physical
spectra (have only positive values). Because the spectra can have what
appear as "negative going peaks", the weights are sometimes negative
numbers. While this is not relevant for the present analysis, the reason for the
negative weights in the histogram can be a point of confusion.
1000
900
800
0) 70ฐ
ra
LL 600
i 500
2 400
o
:* 300
200
100
0
-3360
-3320 -3300
Intensity
Figure A 6. Histogram of score image for PCI (copper).
The image shows that while the normal sample is homogeneously gray (having
similar weights), the strychnine exposed sample is not homogeneous. Similarly,
the HgCb exposed sample has some dark areas and is also distinct from the
normal sample.
The histogram is consistent with two populations of for the weights, one with a
peak at about -3295 and one with a peak at about -3310. The histogram can be
used to select a population of pixels in the image, by choosing a range of values
for construction of a binary mask. In this case, two masks, each encompassing
predominantly one distribution, can be generated.
Figures A7 and A8 show masks (left) generated from the upper and lower
distributions. The white areas in the mask image are the pixels which have
weights in the histogram highlighted by gray. In Figure A7 the lower values (less
than -3305) are highlighted in the histogram. The resulting mask has white pixels
in the strychnine sample in the spatial distribution which appears like the dark
areas seen in the score image shown in Figure A6. The areas are dark in Figure
A6 because they have lower weights from the PCA evaluation. The histogram
analysis allows identification of those regions based on the observation that there
A-ll
-------
are two distributions in the histogram. By setting the threshold based on
observed distributions in the histogram, an image can be generated with less
bias than if the image itself is employed by the user.
1000
90O
800
700
600
500
400
300
200
100
0
-3360
-3340 -3320 -3300
htensity
-3280
Figure A 7. Mask 1 based on PCI score image from copper.
1000
900
800
700
600
500
400
300
200
100
0
-3360
-3340 -3320 -3300 -3280
Intensity
Figure A 8. Mask 2 based on PCI score image of copper.
The resulting mask images can be evaluated in a fashion similar to the original
mask images which were used to estimate the sample area for each sample. In
this case the number of white pixels in the mask of a specific characteristic
A-12
-------
(same as control or different from control for example) are added up to
determine, in pixels, the interacted area. For instance the upper sample in this
example has a significant number of pixels which are not like the normal. These
pixels appear white in Figure A7 (note in this mask the normal sample is black).
The total number of white pixels in the top sample is 2966. From the analysis of
sample area described above the total area of the sample is 4504 pixels. Thus,
the contaminated area (the area which is not like normal) is 66%.
In all cases all of the score images were reviewed. In some cases there was
evidence of change in one type of pipe sample from normal in one score image,
and another exposure in a second score image. This is to be expected as there
could be different spectral responses to the exposures for the two different
exposing chemicals for a given type of pipe sample.
In this fashion the combination of spectral imaging of samples and controls along
with masks generated from histogram analysis of PCA score images can give an
estimate of the percent interaction of the pipe samples with the chemicals they
were exposed to.
A-13
-------
Results / Conclusion
As described above, each set of samples of a particular type were evaluated.
The table shown in Figure A9 shows the estimated total area and interacted area
of the samples.
Sample 1 Total Area 1 affected area | % affected
Strychnine
Copper
Iron
Brass
Rubber
PVC
4504
1924
3559
8154
5199
2966
66%
1924 100%
3567 100%
0 0%
4057 78%
Untreated
Copper
Iron
Brass
Rubber
PVC
4064
3568
3553
4674
5732
Mercuric Chloride
Copper
Iron
Brass
Rubber
PVC
5311
2435
2241
7541
5928
460
2434
4
0
0
9%
100%
0%
0%
0%
Figure A 9. Results table.
In conclusion, fluorescence spectral imaging with UV illumination in conjunction
with PCA image analysis was used to study the contamination of different types
of pipe material with different contaminants. The limitations of this approach
include no sensitivity to contamination that does not change the luminescence of
the sample under UV light illumination and the lack of specific chemical
information about the chemical nature of the residues left on the pipes.
With those limitations in mind, this analysis shows that the rubber pipe material
does not suffer significant residual contamination for either contaminant. Brass
and PVC similarly do not show any evidence of residual contamination with
mercuric chloride. Brass shows evidence of complete residual contamination
with strychnine while the PVC has 78% surface area contaminated with
strychnine. The iron samples show evidence of complete residual contamination
A-14
-------
with both the mercuric chloride and strychnine contaminants used in this study.
The copper samples show residual contamination for both contaminants, but not
complete contamination in either case. For the case of strychnine 66% of the
surface appears to have residual contamination while for the case of mercuric
chloride only 9% of the surface appears to have residual contamination.
A-15
-------
Appendix B Determination of Soluble Copper Cyanide Complexes in
Tap Water
Return to text Section 4.1.1.5
Tuesday, July 8, 2008
Ivy Grimm
William R. LaCourse, Ph.D
University of Maryland Baltimore County
Department of Chemistry and Biochemistry
1000 Hilltop Circle
Baltimore MD, 21250
Purpose:
To identify soluble metal cyanide complexes that form in tap water treated with copper
pipe and a cyanide salt, and to determine the concentration of copper cyanide complexes
using ion chromatography (1C) with ultra-violet detection (UV).
Experimental:
Method: CuCN2.met
Run Time: 35 min.
Gradient Elution:
Eluent A: 20 mM NaOH/150 mM NaCN
Eluent B: 20 mM NaOH/300 mM NaClO4
Eluent C: 20mMNaOH
Pump Program:
Time (min)
Initial
0.0
18.0
22.0
25.0
35.0
Flow Rate:
Detection:
%A %B
10 10
10 10
10 45
10 45
10 10
10 10
1 mL/min
UV absorbance
%C
80
80
45
45
80
80
at 215 nm
Sample treatment: Filtered through 0.45 um LC Acrodiscฎ syringe filter (Waters)
Samples from the adsorption experiments were analyzed immediately after the respective
experiment. Blank: Tap water from NIST lab
Method Reference: Application Update 147, Dionex Corporation, available at
www.dionex.com
Results:
Results obtained from the IC-UV method provided quantitative determination of
concentration levels of the copper-cyanide complex, [Cu(CN)3]2", present in samples.
The following data and results were determined using a [Cu(CN)s] " standard calibration
B-l
-------
as a reference. These standards were prepared following the procedure in Application
Update 147 (Dionex Corporation). Interpretation of the results is not included.
Analytical Figures of Merit:
Peak Area
Calibration Curve for [Cu(CN)3]2" y = 2661 08x + 1 9739
1 Annnnnn
finnnnnn
j^
^^
^^
^^
^^
^^
^^
**^
0.00 10.00 20.00 30.00 40.00 50.00 60.00
Cone, ppm
2-
Figure B 1. Shows the calibration curve for the [Cu(CN)s] " standard.
Sensitivity = 266108, R2 = 0.9999, LOD = 0.108 ppt
B-2
-------
26.33 ppm standard (Rt = 7.75 min)
OAC
OAC\
Ooc
^ U.OU
8 0.25
TO U.ZU
o n -i c;
0.15
< n -in
One
.00
One
II ;
0 5 10 15 20 25 30 35
Time (min)
Figure B 2. Shows a chromatogram of 26.33 ppm [Cu(CN)3]2- standard.
Sample #4c (Rt = 7.92 min)
0.03
0.00
10
15 20
Time (min)
25
30
35
Figure B 3. Shows a chromatogram of sample #4-50 ppm KCN + Cu pipe (old) mean
concentration = 0.963 ppm, (zoomed in to enhance peak).
B-3
-------
1.00 -i
0.80
S 0.60
u
| ฐ'4ฐ
o
in
3 0.20
0.00
-0.20
C
Sample #11 a (Rt = 7.67 min)
1 v- I V ~^_
1
) 5 10 15 20 25 30 35
Time (min)
Figure B 4. Shows a chromatogram of sample #11-50 ppm KCN + Cu pipe (fresh)
mean concentration = 59.49 ppm.
Appendix B Table 1. Concentration of Soluble Copper-Cyanide Complexes formed
in Aqueous Solutions of Sodium Cyanide in Presence of Copper Pipe
Sample
#1 - 3 ppm KCN + Cu (a)
#l-3ppmKCN + Cu(b)
#1 - 3 ppm KCN + Cu (c)
#2-10ppmKCN + Cu(a)
#2- 10ppmKCN + Cu(b)
#2-10ppmKCN + Cu(c)
#3 - 20 ppm KCN + Cu (a)
#3 - 20 ppm KCN + Cu (b)
#3 - 20 ppm KCN + Cu (c)
#4 - 50 ppm KCN + Cu (a)
#4 - 50 ppm KCN + Cu (b)
Cone. [Cu(CN)3]2"
(ppm)
ND
ND
ND
ND
ND
ND
12.53
12.57
12.55
0.88
1.12
Mean Cone.
(ppm)
12.55
0.963
% RSD
(N = 3)
0.17
14.01
B-4
-------
#4 - 50 ppm KCN + Cu (c)
#5 - 3 ppm NaCN + Cu (a)
#5 - 3 ppm NaCN + Cu (b)
#5 - 3 ppm NaCN + Cu (c)
#6-10ppmNaCN + Cu(a)
#6- 10ppmNaCN + Cu(b)
#6-10ppmNaCN + Cu(c)
#7 - 20 ppm NaCN + Cu (a)
#7 - 20 ppm NaCN + Cu (b)
#7 - 20 ppm NaCN + Cu (c)
#8 - 50 ppm NaCN + Cu (a)
#8 - 50 ppm NaCN + Cu (b)
#8 - 50 ppm NaCN + Cu (c)
#9-20gKCN+ IgCu T
#10-10 ppm KCN + Cu fresh (a)
#10-10 ppm KCN + Cu fresh (b)
#10-10 ppm KCN + Cu fresh (c)
#11-50 ppm KCN + Cu fresh (a)
#11-50 ppm KCN + Cu fresh (b)
#11-50 ppm KCN + Cu fresh (c)
3 ppm KCN + tap water (in duplicate)
10 ppm KCN + tap water (in duplicate)
50 ppm KCN + tap water (in duplicate)
0.89
ND
ND
ND
1.04
1.01
0.82
ND
ND
ND
5.54
6.97
5.36
ND
18.76
18.99
18.52
59.52
59.57
59.37
ND
ND
ND
0.956
5.95
18.76
59.49
12.47
14.82
1.26
0.18
'This sample had a broad interfering peak that overloaded the detector. Diluting the
sample to overcome this interference would cause the analyte peak to become not
detectable.
B-5
-------
Appendix C Tables Cited in the Text
Analysis of potential
contamination
scenarios
Selection of initial
contaminants and substrates
Bench scale tests
Chemical
Biological
Screening tests of different
contaminant/substrate
combinations
Controlled dynamic
tests using
fluorescence
method
r\
Development of new measurement methods
Bench scale
Fluorescence
Raman Spectroscopy
Full-scale dynamic tests
Plumbing loops
Water heaters
Development of
decontamination procedures
and recommendations
Detailed modeling of
contaminant dispersal
Development of software
tool for flushing
recommendations
Figure C 1. Flowchart depicting process of research.
Return to text Section 1.1
C-6
-------
Direction of water flow
Contaminant Injection Locations
1-Remote water main
2-Water main near building
3- Water mai n di rectly before bu ilding
4-Building water system
ฃ
Building
flush
J
i
d
Contaminant could
be introduced at
various locations.
Collecting water
samples at
different outlets
will help identify
the effected water
lines to be flushed.
Collection points
K Contaminated
Clean
Figure C 2. Potential contamination scenarios for building plumbing systems.
Return to text Section 1.2
C-7
-------
(a)
Y/////7//7/////7///7/^^^ Y//////7///////////////^^^
t
Contaminant
Contaminant
Pipe wail
(0
Y//////////////^^^
(d)
Pipe wall
Water-Contaminant -
Mixture
Pipe wall
Pipe wall
Contaminant/substrate exposure can vary with solubility and density
Figure C 3. Contaminant/substrate interactions and the exposure conditions.
Return to text Section 1.2
C-8
-------
reservoir
turbine acoustlc
flow meter flow meter
diesel filter
for (flushing)
decontamination
tests
check
valve
pressure transducer
PG pressure gauge
thermopile
differential pressure
transducer
acidity/alkalinity
measurement
n temperature
measurement
ฉ
conductivity meas.
oxygen level meas.
Figure C 4. Schematic of dynamic flow test loop.
Return to text Section 3.2.1.2
C-9
-------
corrected excitation
module
84 optical fibers
for excitaion
quartz tube
stainless sheath
optical fibers
(168 send and receive)
} cross section of
bifurcated optical
bundle
84 optical fibers
for emission
Digital displays
COD
O O DPZO
voltage outputs
spectrofluorometer
linear positioning
device
test chamber
96 mm x 1.6 mm
flow cross section
test surface
Figure C 5. Schematic of spectrofluorometer, test section, and linear positioning device.
Return to text Section 3.2.1.2
C-10
-------
Flexible tube
Water flow
Coupon
Figure C 6. Schematic of screening test setup.
Return to Text Section 3.3
C-ll
-------
water supply
pump
drain
Figure C 7. Schematic view of the full-scale plumbing system test facility.
Return to text Section 3.4
C-12
-------
sink
drain piping
reservoir
j
lap water
supply
pump
Figure C 8. Schematic of full-scale test loop.
Return to text Section 3.4.1
C-13
-------
water exit
water inlet
observation window
and access port
overflow pipe>
upper element
lower element
sediment
T< flowmeter
evacuation port
^sample chamber
exposed metal tank
and insulation
cutaway for illustration
insulation
drop tube to
sample chamber
drain valve
Figure C 9. Schematic of hot water heater testing apparatus.
Return to text Section 3.4.2
C-14
-------
A
Cu Bltflln PlfiA (nsrtSBJ Er|KiKAi1 rn 1
B
Cu 3iofi m Pipe (58369) Expasscto 12.1 nfl Pho^ale
After 124 mgyL Phorate treatment (A posed to T2A niฃ/L Hioraite
ftFfer17J mgJ P ho rate treat me nt (4 days)
After 1120 decontamination
After 10 ml bleach decani ami nation
Pure Prorate
ISO]
1600 1400 1200
'ifli .^vpn i rrherc (":m"^J
1000
Figure C 10. IR micro spectroscopy for Cu biofilm pipes exposed to 12.4 mg/1 phorate.
Return to text Section 4.1.4
Optical images of the Cu pipe are shown above the IR spectra where the y-axis shows IR
absorbance and the x-axis shows wave number. Figure A represents IR spectra taken on
pipe sample 68068 decontaminated in 2.6 mL bleach in water. Figure B represents IR
spectra taken on pipe sample 68069 decontaminated in 5.3 mL bleach in water. Figure C
represents IR spectra taken on pipe sample 680611 decontaminated in 10 mL bleach in
water. Data in blue represent Cu pipe exposed to 12.4 mg/L phorate. Data in pink
represent Cu pipe decontaminated in water. Data in red represent Cu pipe
decontaminated in the various bleach concentrations. Data in light blue present an IR
spectrum of pure phorate. The black arrows show disappearance of the phorate peak.
-------
B
PVC
PVC + 30 g/L toluene in water
2000
1500
1000
Raman Shift (cm-1)
ฃ00
Figure C 11. Thirty g/1 toluene in water exposed to PVC pipe.
Return to text Section 4.1.1.2
(A) Raman spectra for PVC pipe (blue) and PVC + 30 g/L toluene in water (red). Arrows
represent the toluene Raman peaks. (B) and (C) optical images at 50x magnification for PVC
and PVC + 30 g/L toluene in water, respectively.
C-16
-------
RT 000.1JT,
Gasoline
ฃ
o a 4 ..*..*ซ
Figure C 12. Gas chromatograms for gasoline and diesel.
Return to text Section 4.1.1.3
The x axis is Time (min) and the y axis is Relative Abundance.
C-17
-------
o
ง
tl
I
c a
m a
Ct
CuFlat
3000
SOD
2QOD 1500
Raman Shift [rrr1)
10QD
SDQ
Figure C 13. Raman spectra of Cu flat pipe exposed to 2000 mg/L diesel in water.
Return to text Section 4.1.1.3
Black spectra represent pure diesel and gray spectra represent diesel on Cu flat pipe. Circled
peaks are those peaks from diesel found in both spectra.
C-18
-------
Strychnine
Control
pm n
EULJ LJ
^^^^^^^^^^^^^^"^^^^^^^^^^H^^H^H ^^H
Figure C 14. Optical image of control pipes and those exposed to strychnine.
Return to text Section 4.1.1.4
(a) Fluorescent images of exposed pipe materials
(b) From left to right, Cu in-service, brass, iron, rubber, and PVC.
C-19
-------
Figure C 15. SEM image at 2600x magnification of mercury precipitate on Cu
pipe.
Return to text Section 4.1.1.6
Precipitate formed when mercuric chloride in tap water contacted Cu pipe material.
Crystals are copper oxide and droplets are mercury metal.
C-20
-------
30
25 I
'23
.ฃ
C
o
I
15
3
10
500 mg/L HgCI2 + Pipe Material
n
n
3 4
Tirne(Days)
Figure C 16. Mercury concentration profile of 500 mg/1 HgC12 solution with PVC and Cu
in-service pipe materials.
Return to text Section 4.1.1.6
PVC (squares) and Cu in-service (circles) pipe materials.
C-21
-------
t.3CTU3 -
4nc j.nn -
.ub+uy
3f.Cj.no
. Dt+uy
_ฃ 3.0E + 09 '
3
LL.
^ z.ab+uy
H
1
o) "? ncj-nci -
_ /.ut + us
E
v
O -1 KCj-RQ
,_ i .Dt+uy
CD
1 riF+nci -
5ncj.no -
.Ub + UH
O.OE + 00^
1
QCopper
PVC
1
Pipe section 1 mUmin CDC baffle 120 rpm Pipe section 2.5 Urrin
Figure C 17. Biofilm organism levels in the reactors with different conditions for growth of
the biofilm.
Return to text Section 4.1.2.1
C-22
-------
0
fl
1.2E+09
1.0E+09
8.0E+Q8
6.0E+Q8
o
CD
4.0E+Q8
2.0E+08
O.OE+00
Q Copper
DPVC
Pipe section 1
CDC no baffle CDC baffle 60 rpm Pipe section 2.5
L/min
Figure CIS. Bacillus thuringiensis (BT) spore levels adhered to biofilm condition pipe
surfaces in the reactors with different spore contacting conditions.
Return to text Section 4.1.2.2
C-23
-------
1 U
9
8
7
6
5
t
_ | 1 1 1 1 1 1 1 1 1 1 1
| Diesel layer on PVC disk from 0.15
- air-gap test SV1
svm. after Re - 0
<-
if contam. c
: * flush A
~- A
''-- *
* $
1 ijib^sP^?^
ฃ *
0 j
L ^
? v *
-1 V
n l~ 1 1 L 1 1 L 1 1 L I 1 L J 1 L J 1 L 1 I L J 1 L J
i i ii
ii i r _
% bulk freestreain |
n. Re
0
3200
7000
gpm
0
3
6
1,/s
0
.19
.38
flushing
sym.
A
after
0
Re
3200
7000
J
i
-
;
-^
J
"
_
1 L J 1 L J 1 1
;
j i i j i i
100
200
300
time (h)
400
Figure C 19. Effect of exposure time and flow rate on thickness of the diesel excess layer
for a 0.15 % bulk free stream mass fraction on a PVC disk.
Return to text Section 4.2.1
C-24
-------
1 U
9
8
7
6
4
3
2
::!
-
- 1
_
5-
^-
-
5-
1:
~ 1 1
i i rn
Diesel layer on i
air-gap test
sym. after 6 gpm
* contam.
* flush
air-gap test
sym. after 3 gpm
^ flush
***
|fc *
i.i
1 ' ' ' ' :
ron disk from 0. 15 % bulk frecstrcam |
flushing svm Rc epm L/S
.
svm. atter Re ~ A A
o 0 U 0
0 n 3200 3 .19 J
3200 A 7000 6 .38 j
A 7000 :
-^
u*. , I
ฐ J
i i i i i i 1 i i i i i i i i i 1 i i i i i i i i ~
100
200
300
time (h)
400
Figure C 20. Effect of exposure time and flow rate on thickness of the diesel excess layer
for a 0.15 % bulk free-stream mass fraction on an iron disk.
Return to text Section 4.2.1
C-25
-------
1 U
9
8
7
6
5
4
3
_r i i r i i r i i r i i i i i i i i i i r i i
' "1" ' "1"" "T rl" '
: | Diesel layer on PVC 0. 1 5% bulk frccstrcam
_
sym Re gpm
: 0 0 0
n 3200 3
f A 7000 6
;
-
2 i i
1 1 *
0
-1
-2
:
-
;-
~i 1 i i 1 i i 1 i 1 1 i i 1 i i 1 i i i i [ i
0 1000 2000
L/s
0
.19
.38
.'JL /WM
'3PH^
/
-* x jflB^A
IB1 '
r
i i | i r i i iTC
__
-
peak/e _=
1 ] L 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 I L 1 L
1 1 1 1 1 1 1 1 1 1 1 r
3000 4000 5000 6000 7000 8000
Re
Figure C 21. Diesel excess layer thickness as a function of Re (Reynolds number) for PVC
surface and water/diesel (99.95/0.15).
Return to text Section 4.2.1
C-26
-------
10
9
8
7
6
5
4
3
2
ill i i i i i i i ill ill ill
Diesel laer on iron 0.15% bulk frccstrcam
sym.
O
n
A
Re
0
3200
7000
gpm
0
3
6
L/s
0
.19
.38
X
PVC data -
-1
ZLJ i i MI i i i i MM iii iii IM ii ii ii MI MI MI i r
0 1000 2000 3000 4000 5000 6000 7000 8000
Re
Figure C 22. Diesel excess layer thickness as a function of Re (Reynolds number) for iron
surface and water/diesel (99.95/0.15).
Return to text Section 4.2.1
C-27
-------
30
25
20
15
-5
-10
i 1 I I I
filled symbols represent data
fora drained test chamber
50
100
time (h)
Diesel layer on oxidized Cu disk from 0.2 % bulk frecstream
sym.
*
flushing
after Re
4600
sym.
0
0
D
V
A
Re
0
1900
3200
4600
7000
gpm
0
1.5
3
4.5
6
Us
0
.09
.19
.28
.38
150
200
Figure C 23. Effect of exposure time and flow rate on thickness of the diesel excess layer
for a 0.2 % bulk free-stream mass fraction.
Return to text Section 4.2.1
C-28
-------
30
25
20
10
-5 -
-10
- ' 1 ' ' ' ' ' '
I [Diesel layer on o\kli/ed
j9
-
%
ฃ 'rA
"W rffl) V
$ ffl sym. flushing
after Re
JK~
ff ft 4600
+ 7000
: AAA
1 |f ft
1 ' 1 ' ' ' ' ' ' ' ' ' 1 '
C'u disk from 0.3% hulk freestream ]
-
Q
IjjftP D
ซ ^
w
sym.
0
0
D
V
A
Re
0
1900
3200
4600
7000
gpm
0
1.5
3
4.5
6
L/s
0
.09
.19
.28
.38
-
"
j
-
-
0 A
rj 4H^
A^3 ฃ
filled symbols represent data
'. for a drained test chamber ~
i i i i i i i i i i i i i i i i
i i I i i i i i i i i i I i i i i i i i i i
50
100
150
200
time (h)
Figure C 24. Effect of exposure time and flow rate on thickness of the diesel excess layer
for a 0.3 % bulk free-stream mass fraction.
Return to text Section 4.2.1
C-29
-------
ov
20
15
10
0
-5
10
t
I I I
- | Diesel layer on oxidized Ciuiisk from 0.2 % bulk freestream
i 1 r~
sym. Re gf
0 0 (
I ,*\ peak /for V ซ 190ฐ l
0.3% data \ 0 320ฐ :
\ v 4600 4
1 , ' A 7000 <
: ' ฃ9p
t ^^ 1 \
f 4B^J
- peak/e
i i i i i i i i i I i i i i i i i i i I i i i i i i i i i I i i i i
) 2000 4000 6000
Re
m l./s
) 0
5 .09
J .19
5 .28
i .38
ihiw
1 =
,
8000
Figure C 25. Diesel excess layer thickness as a function of Re (Reynolds number) for
water/diesel (99.8/0.2).
Return to text Section 4.2.1
C-30
-------
30
25
20
15
10
5
ฐ
0
-5
-10
Diesel layer on Cu
0.3% bulk freestream
2000
4000
Re
6000
8000
Figure C 26. Diesel excess layer thickness as a function of Rre (Reynolds number) for
water/diesel (99.7/0.3).
Return to text Section 4.2.1
C-31
-------
1000
100
CD
CO
E
CD
DC.
10
Diesel
Used PVC
Used cast iron
Used copper
New rubber
0 4 8 12 16 20 24
Flush time, hr
Figure C 27. Raman intensity of coupons soaked in 100 % diesel fuel.
Return to text Section 4.3
Coupons of copper, PVC, iron, and rubber were soaked in 100 % diesel fuel for approximately
140 h, except for the rubber, which was soaked for approximately 283 h. The data points
furthest to the left represent the Raman intensity at a selected wave number for the specimens
immediately after soaking. The data points just to the right of those points correspond to the
specimens having been flushed with cold water for 1 h. The data points to the far right were
taken after approximately 24 h. of flushing.
C-32
-------
1000
c
V
"c
c
CD
E
CD
a:
100
10
Gasoline
A Used PVC
Used cast iron
Used copper
New rubber
0 4 8 12 16 20 24
Flush time, hr
Figure C 28. Raman intensity of coupons soaked in 100 % gasoline.
Return to text Section 4.3
Coupons of the materials shown to the right of the graph were soaked in 100 % gasoline for
approximately 24 h, except for the rubber, which was soaked for approximately 480 h. The data points
furthest to the left represent the Raman intensity at a selected wave number for the specimens
immediately after soaking. The data points just to the right of those points correspond to the specimens
having been flushed with cold water for 1 h. The data points to the far right were taken after
approximately 20 h. of flushing, except for the rubber specimen, which was flushed for approximately
24 h.
C-33
-------
I
c
CO
CD
01
1000
100
10
Strychnine
+
Used PVC
Used cast iron
Used copper
New rubber
0 4 8 12 16 20 24
Flush time, hr
Figure C 29. Raman intensity of coupons soaked in strychnine (0.5% mass
fraction).
Return to text Section 4.3
Coupons of the materials shown to the right of the graph were soaked in a 0.5 % mass
fraction solution (5000 ppm) of strychnine in water for approximately 140 h., except for
the rubber, which was soaked for approximately 480 h. The data points furthest to the
left represent the Raman intensity at a selected wave number for the specimens
immediately after soaking. The data points just to the right of those points correspond to
the specimens having been flushed with cold water for approximately 1 h. The data
points to the far right were taken after approximately 24 h. of flushing.
C-34
-------
1000
100
V
"c
CD
E
CD
-A
A"
10
Cyanide
Used PVC
Used cast iron
Used copper
New rubber
0 4 8 12 16 20 24
Flush time, hr
Figure C 30. Raman intensity of coupons soaked in sodium cyanide (1 % mass
fraction)
Return to text Section 4.3
Coupons of the materials shown to the right of the graph were soaked in a 1% mass fraction
solution (10000 ppm) of sodium cyanide for approximately 260 hr., except for the rubber, which
was soaked for approximately 480 hr. The data points furthest to the left represent the Raman
intensity at a selected wave number for the specimens immediately after soaking. The data
points just to the right of those points correspond to the specimens having been flushed with cold
water for approximately 1 hr. The data points to the far right were taken after approximately
24 hr. of flushing.
C-35
-------
1000
100
V
"c
CD
E
CD
10
Phorate
Used PVC
Used cast iron
Used copper
New rubber
0 4 8 12 16 20 24
Flush time, hr
Figure C 31. Raman intensity of coupons soaked in 100 % phorate.
Return to text Section 4.3
Coupons of the materials shown to the right of the graph were soaked in a solution of 100%
phorate (a systemic insecticide, CyHnC^PSs) for approximately 144 h, except for the rubber,
which was soaked for approximately 283 h. The data points furthest to the left represent the
Raman intensity at a selected wave number for the specimens immediately after soaking. The
data points just to the right of those points correspond to the specimens having been flushed
with cold water for approximately 1 h. The data points to the far right were taken after 24 h. of
flushing.
C-36
-------
Fluorescence
Measurements
(Re =5000)
O I'VC
Raman Measurements
(Flushing Re = 30000)
- A Used PVC
- Used copper
- Used cast iron
I
I
I
I
I
time (h)
250
200
150
100
50
10 20 30 40 50 60 70 80 90 100
CD
E
^ *
f*
Figure C32. Comparison of flushing test results using the fluorescent measurement
technique and the Raman measurement technique.
Return to text Section 4.3.6
C-37
-------
1000
(U
-I
- 100
c
(0
CO
10
\ i i i r
Diesel test
Hot water tank No. 1
Sediment sample
i i i i i
8 12 16 20
Flush time, hr
24 28
Figure C 32. Measurement results for a hot water tank exposed to diesel fuel.
Return to text Section 4.4.1
C-38
-------
zuuuu
10000
-i <
V 1000
03
ro 100
H n
I i i i I i i i I i i i I i i i I i i i I i i i I
_V A
_ v
= V
:Ai *
Diesel test
= Hot water tank
I Top output
i i i i i i i i i i i i i i i i i i i i i i i i i
V Max
Mean
A Min
0 4 8 12 16 20
Flush time, hr
24
Figure C 33. Heated hot water heater test with diesel fuel, sampling from outlet.
Return to text Section 4.4
C-39
-------
zuuuu
10000
0) 1000
.c
c
03
ro 100
H r\
I i i i I i i i I i i i I i i i I i i i I i i i I
i A
8
:
'_ A
Diesel test
-- Hot water tank
i Bottom drain
i i i i i i i i i i i i i i i i i i i i i i i i i
V Max
Mean
A Min
0
8 12 16 20 24
Flush time, hr
Figure C 34. Heated hot water heater test with diesel fuel, sampling from drain.
Return to text Section 4.4
C-40
-------
zuuu
1000
'(75
c
"c
g 100
E
CD
H n
I i i i I i i i I i i i I i i i I i i i I i i i I
- * Strychnine test
:ป Hot water tank
_ V
V
A
- A
:A
V Max
Mean
A Min
0 4 8 12 16 20 24
Flush time, hr
Figure C 35. Hot water heater test with strychnine.
Return to text Section 4.4.2
C-41
-------
OQ
Hot Water Heater
Initial
Dilution Hot
Water
Heater
Heater after
sitting one
day
Flush-10
min
Flush- 60
min
Figure C 36. Hot water heater measurements with BT spores - water samples.
Return to text Section 4.4.4
C-42
-------
OQ
Hot Water Heater
Initial Heater Sediment-
Dilution Hot after sitting 24 Hours
Water one day
Heater
Sediment-
48 Hours
Anode- 48
Hours
Figure C 37. Hot water heater measurements with BT spores sediment
and anode samples.
Return to text Section 4.4.4
C-43
-------
10000
1000
CD
CD
DC.
100
Cumulative flow volume, gal.
0 500 1000 1500 2000 2500 3000
2nd diesel test
3/4 in. Cu
0 2 4 6 8 10 12 14
Flush time, hr
V Max
Mean
A Min
Figure C 38. Pipe loop measurements for copper pipe and diesel fuel.
Return to text Section 4.5.1
C-44
-------
Cumulative flow volume, gal.
0 500 1000 1500 2000 2500 3000
IUUUU
&
c
0)
-1
- 1000
c
CD
E
CD
01
\ nn
l l l l l l l
: 2nd diesel test
'. 1/2 in. CPVC
V
V
V
; 0 V
: * v
A
A
A
l i l i l i l i l i l i l i
V Max
Mean
A Min
0 2 4 6 8 10 12 14
Flush time, hr
Figure C 39. Pipe loop measurements for copper pipe and diesel fuel.
Return to text Section 4.5.1
C-45
-------
Cumulative flow volume, gal.
0 1000 2000 3000
100000
v
v
V
A
CD
-i<
- 10000
CD
E
CD
DC.
1000
0 2 4 6 8 10 12 14 16
Flush time, hr
Figure C 40. Pipe loop measurements for % inch copper pipe and strychnine.
Return to text Section 4.5.2
Strychnine test
3/4 in. Cu
V Max
Mean
A Min
C-46
-------
100000
c
0)
-I
- 10000
c
CD
E
CD
1000
Cumulative flow volume, gal.
0 1000 2000 3000
V
v A
0
Strychnine test
1/2 in. Cu
V Max
Mean
A Min
4 6 8 10 12 14 16
Flush time, hr
Figure C 41. Pipe loop measurements for 1A inch copper pipe and strychnine.
Return to text Section 4.5.2
C-47
-------
O
o
en
o
3/4 in. Cu pipe
5 10
Flush time per floor, hr
15
Figure C 42. Pipe loop measurements for % inch copper pipe and BT spores.
C-48
-------
,o
O
5 10
Flush time per floor, hr
Figure C 43. Pipe loop measurements for 1A inch copper pipe and WY spores.
Return to text Section 4.5.3
C-49
-------
-------
Flushing diesel layer from PVC disk
with tap water at Re = 5000
10 20 30 40 50 60 70 80 90 100
time (h)
Figure C 45. Flushing measurements used to fit coefficients of model.
Return to text Section 5.1.1
C-51
-------
1,8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
59 Till IT'lPTl 6 ITT II ITTTT1 FT j Til rTTTI 1 [11 T 1 T T I
Modified kinematic viscosity to account for
adhesive forces between diesel and iron
contaminant flow away from surface
contaminant flow toward surface
4
8
= 0.5 Jim
= 14 |im
0 10 20 30 40 50 60 70 80 90 100
Figure C 46. Ratio of the effective kinematic viscosity to the kinematic viscosity in the bulk
phase.
Return to text Section 5.1.2
Shows the influence of adhesive forces between iron and diesel derived for two flow conditions (Re = 3200
and Re = 7000)
C-52
-------
500
400
300
200
100
Maximum contamination diese
layer on various pipe surfaces
.
= In
copper
(8= 150 urn)
iron
(8= 16jLim)
0
0.25
0,5
0,75
Figure C 47. Maximum contamination diesel layer on various pipe surfaces as predicted
by eq. (3.12).
Return to text Section 5.1.2
C-53
-------
150
130
110
90
70
50
30
10
Fluorescence
Measurements
(Re =5000)
o pvc
Raman Measurements
( Fl ushin^ej=ฃ000{yL
* Used PVC
* Used copper
* Used cast iron
\ ,
I ,
250
200
150
100
50
10 20 30 40 50 60 70 80 90 100
time (h)
Figure C48. Preliminary validation of semi-empirical flushing model.
Return to text Section 5.1.2
C-54
-------
EPA/NIST Building Plumbing Contamination Model
Input Data
Pipe Surface Copper
Contaminant Diesel
Surface condition New
Contaminant
Mass Fraction
Event Date
Event Date
Dec. 7,1941
Event Name Fed.Bldg.
Floor Area of Residence or
Building Unit in square meters
1200
-Predictions
200
150
Contaminant
Thickness 100
(microns)
50
0
Copper
200 400 600
Flushing time (hrs)
Maximum
Contaminant
Thickness 173.4672
(microns)
Required
HUSH 436.9862
Time (hrs)
Figure C 49. Sample graphical users' interface for EPA/NIST building plumbing
contamination model.
Return to text Section 5.1.4
C-55
-------
600
500
400
300
200
100
imp
\ I I I I I I
irrT
Effect of Building Area on Diesel
Hushing time for 6mm piping
building area (m2)
Figure C 50. Effect of building area on diesel flushing time for 6 mm piping.
Return to text Section 5.1.4
C-56
-------
300
250 -
200 -
150 -
100 -
50 -
400
Figure C 52. Velocity fields, as indicated by the arrows, for a Reynolds number of
about 30 in the U-shaped pipe system.
Return to text Section 5.2.2
Contaminants were placed near the lower midway part of the flow path (MP) and at the
bottom right hand corner (CP) of the pipe. X and Y represent the coordinates and length is in
units of lattice spacing. Note that the simulation resolution is 6 times higher than that indicated
by the velocity vectors.
C-57
-------
300
250 -
200 -
150 -
100 -
400
Figure C 53. Velocity fields for a Reynolds number of about 3000 for the U-shaped
pipe system.
Return to text Section 5.2.2
(Note the regions of rotational flow near the corners and bends. Patches are located near MP
and CP.)
C-58
-------
30.
20.
50
Figure C 54. Flow near a cavity.
Return to text Section 5.2.2
(Note the rotational flow near the opening of the cavity. A secondary rotation pattern developed near
the bottom of the cavity. For this geometry, the contaminant was placed along the bottom of
the cavity.)
C-59
-------
1.1
i.o
0.9-
0.8 -
0.7-
I 0.6 H
0.5-
0.4 -
0.3 -
0.8-
0.1 -
0.0
RE = 30 dAVlTY
RE=3ooo CAVITY
- *.%
--. "fiE^SOOO U PIPE OP
3000 U PIPE MP
3 PIPE
RE-3000 PIPE
100 200 300 400 500 600 700 800 900 1000
time
Figure C 55. Normalized pipe contamination, C/Cinit.
Return to text Section 5.2.3
Time is given in relative units where 100 corresponds to the time it takes for the fluid to pass
through the pipe system. The black lines (Re=30 dashed, Re=3000 solid) correspond to straight
pipe flow (S), green and blue are for the U shaped pipe and the violet and purple for the cavity.
The dashed (green or blue) lines correspond to case where the contaminant was placed in a corner
(CP) and the solid lines all correspond to the contaminant placed midway (MP) in the U shaped
pipe.
C-60
-------
100 -
50 -
400 500
600
Figure C 56. Low Re (Reynolds) number flow past a rectilinear obstruction.
Return to text Section 5.2.3
(The red area indicates the location of the contaminant patch, and designated by
the letter "P" on the x-axis.)
C-61
-------
100 -
50 -
0
100
200
400
500
600
Figure C 57. High re flow past a rectilinear obstruction.
Return to text Section 5.2.3
(Note the well-defined vortex downstream. The red area indicates the location of the contaminant
patch and designated by the letter "P" on the x-axis.)
C-62
-------
1.00
0,75 -
0.50
0.25 -
0.00
100 200 300 400 500 600 700 800 900 1000
time
Figure C 58. Normalized total concentrations for the rectilinear cases.
Return to text Section 5.2.3
C-63
-------
Figure C 59. Time sequence of ingress of contaminant.
Return to text Section 5.2.4
Note contaminant reaches the middle patch (MP) before it reaches the corner patch (CP).
C-64
-------
1,1
1.0-
O.fl -
o.e -
^ 0.7-
,
,2 o.e -
"0
0.3 -
0.2 -
0.1 -
0.0
-1
RE-3000 dAVITY
'3'9 U PIPE CP
1 ;
RE-3Qป0\U PIPE HP RE=go ฃ ^jpfe
\ V ^ป i
PIPE HP
Iog10(time)
3
Figure C 60. Re-scaled total concentration adjusted for actual flow rates.
Return to text Section 5.2.5
(Note that higher Re flow removes the contaminant sooner for the cases studied.)
C-65
-------
Inject
flush
water
O
Exterior water spigot
Faucet
Hot water
Cold water
Drain valve
Water heater
O
o
Shut off valve
Wbter meter
^unnlv
Figure C 61. Schematic of injection of flush water through an exterior water spigot.
Return to text Section 6.3
This set up allows flushing of both hot and cold water lines and water heaters.
C-66
-------
Shut off valve
Water meter
Exterior water spigot
IAJ,*~. .-...,!.,
Figure C 62. Schematic showing injection of flush water through drain valve of a hot water
heater.
Return to text Section 6.3
Allows direct flushing of tank and water lines.
C-67
-------
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
------- |