Integration of Photocatalytic
Oxidation with Air Stripping of
Contaminated Aquifers
by
Rajnish Changrani and Gregory B. Raupp
Arizona State University
Tempe, AZ 8528T-6006
and
Craig Turchi
Zentox Corporation
Ocala, FL 34470
CR 821100-01-0
Project Officer
Norma M. Lewis
Sustainable Technology Division
National Risk Management Research Laboratory
Cincinnati, OH 45268
National Risk Management Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH 45268
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and
Development funded and managed the research described here under contract C R
821100-01-0 to Arizona State University. It has been subjected to the Agency's peer
and administrative review and has been approved for publication as an EPA document.
ii
-------
Foreword
The U.S. Environmental Protection Agency is charged by Congress with protecting the
Nation's land, air, and water resources. Under a mandate of national environmental
laws, the Agency strives to formulate and implement actions leading to a compatible
balance between human activities and the ability of natural systems to support and
nurture life. To meet this mandate, EPA's research program is providing data and
technical support for solving environmental problems today and building a science
knowledge base necessary to manage our ecological resources wisely, understand how
pollutants affect our health, and prevent or reduce environmental risks in the future.
The National Risk Management Research Laboratory is the Agency's center for
investigation of technological and management approaches for reducing risks from
threats to human health and the environment. The focus of the Laboratory's research
program is on methods for the prevention and control of indoor air pollution. The goal
of this research effort is to catalyze development and implementation of innovative,
cost-effective environmental technologies; develop scientific and engineering
information needed by EPA to support regulatory and policy decision; and provide
technical support and information transfer to ensure effective implementation of
environmental regulations and strategies.
This publication has been produced as part of the Laboratory's strategic long-term
research plan. It is published and made available by EPA's Office of Research and
Development to assist the user community and to link researchers with their clients.
E. Timothy Oppelt, Director
National Risk Management Research Laboratory
iii
-------
Abstract
Bench-scale laboratory studies and pilot-scale studies in a simulated field-test situation
were performed to evaluate the integration of gas-solid ultraviolet (UV) photocatalytic
oxidation (PCO) downstream of an air stripper unit as a technology for cost-effectively
treating water pumped from an aquifer contaminated with chlorinated volatile organic
compounds (VOCs). Chlorinated ethylenes such as trichloroethylene (TCE) can be
destroyed in a wide process window, although chlorinated ethanes such as
trichlorethane (TCA) are non-reactive. Water vapor significantly inhibits the chlorinated
ethylene destruction rate. For this reason, PCO units should be placed downstream of
a dehumidification unit located between air strippers and the PCO unit, with targeted
reduction of the relative humidity in the contaminated air stream to less than 50%.
Principal carbon-containing products of PCO identified experimentally at the bench-
scale include carbon dioxide, carbon monoxide, and phosgene (COCI2). Failure to
close carbon mass balances under some process conditions suggests that not all
byproducts were identified. Further studies are needed in this area. A panel bed was
identified as the preferred photoreactor configuration. This unit is characterized by
simplicity of construction, ease of maintenance, and high UV photon utilization
efficiency.
This report was submitted in fulfillment of contract number C R 821100-01-0 by Arizona
State University under the sponsorship of the United States Environmental Protection
Agency. This report covers a period from October 1, 1993 to June 30, 1997, and work
was completed as of June 30, 1997.
iv
-------
Contents
Foreword jjj
Abstract iv
Figures vi
Tables vii
Abbreviations and Symbols viii
Chapter 1 Introduction 1
Chapter 2 Conclusions 4
Chapter 3 Recommendations 6
Chapter 4 Experimental Methods 7
Chapters Results and Discussion 13
Chapter 6 References 36
Appendix
Bibliography 38
V
-------
Figures
1 -1. Integrated Air Stripping with Photocatalytic Oxidation Process 1
1-2. Photocatalytic Oxidation Reactor Configurations 2
4-1. Bench-scale photocatalytic oxidation apparatus 8
5-1. Interaction effect between Relative Humidity and
UV Intensity on VOC conversion 18
5-2. Interaction effect between Relative Humidity and
Temperature on VOC conversion 18
5-3. Response surface for VOC conversion versus
residence time and inlet VOC concentration 19
5-4. Response surface for VOC conversion versus residence time and RH 19
5-5. Influence of VOC concentration and temperature on selectivity 21
5-6. Influence of UV intensity and temperature on selectivity 21
5-7. Selectivity versus TCE conversion for Set 1 experiments 23
5-8. Influence of pseudocomponent ratios R, and R2 on Selectivity 28
5-9. TCE conversion, selectivity, and CI atom balance
versus time on stream 31
5-10. Integrated Photon Efficiency versus Inlet TCE Concentration
for the two pilot reactor configurations 35
vi
-------
Tables
4-1. Minimum Detection Limits for Gas Analysis Methods 9
4-2. QA Objectives for Gas Analysis Methods 12
5-1. Process Parameters and Levels 13
5-2. Methods to Control and Measure Independent Process Variables 14
5-3. Process Responses and Measurement Techniques 15
5-4. Results for the Factorial Block of the fee Design 16
5-5. Carbon Mass Balance for the Set 1 Experiments 24
5-6. Mixture Experimental Design 26
5-7. Mixture Experimental Results 27
5-8. Carbon Mass Balance for Mixture Experiments 29
5-9. TCE Conversion Results for Reactor Configuration 1 32
5-10. TCE Conversion Results for Reactor Configuration 2 34
vii
-------
Acronyms and Abbreviations
C, volumetric concentration of species / in air
Cl2 chemical formula for molecular chlorine
CO chemical formula for carbon monoxide
C02 chemical formula for carbon dioxide
COCI2 chemical formula for phosgene
DOE Design of Experiments
FID flame ionization detector
GC gas chromatograph
HCI chemical formula for hydrogen chloride
MDL minimum detection limit
mfc mass flow controller
PCO photocatalytic oxidation
ppmv parts per million (volumetric basis)
QA Quality Assurance
QAPP Quality Assurance Project Plan
R1 pseudocomponent ratio (trichloroethylene mole fraction to sum of
chlorinated ethylenes mole fractions)
R2 pseudocomponent ratio (trichloroethylene mole fraction to
dichlorobenzene mole fraction)
RH relative humidity
RSM response surface methodology
SCFM standard cubic feet per minute
seem standard centimeters cubed per minute
SC02 carbon dioxide selectivity
T temperature
TCA trichloroethane
TCE trichloroethylene
Ti02 chemical formula for titania (also known as titanium dioxide)
x gas residence time
UHP ultra high purity
UV ultraviolet
VOC volatile organic compound
X, conversion of species /' or conversion of pseudocomponent /'
x, mole fraction of species /
viii
-------
Chapter 1
Introduction
Background
The global objective of this work was to evaluate the integration of gas-solid ultraviolet
(UV) photocatalytic oxidation (PCO) downstream of an air stripper unit as a technology
for cost-effectively treating water pumped from an aquifer contaminated with chlorinated
volatile organic compounds (VOCs). Experimental work was performed at the bench-
scale in the laboratory during the first phase of the project, and at the pilot-scale in a
simulated field-test situation during the second phase.
The photocatalytic oxidation process [1-10] is shown schematically in Figure 1-1. In the
integrated air stripper/photocatalytic oxidation configuration, the air stripper off-gases
are fed directly to the PCO reactor without pretreatment. In the continuous flow PCO
reactor, the contaminated air stream contacts the surface of a near-ultraviolet irradiated
titania (Ti02) catalyst, causing photochemical destruction of the contaminants at or near
room temperature. Exhaust gas from the PCO reactor is fed to a dry scrubber for
removal of HCI and Cl2 generated during the oxidation of chlorinated solvents.
A commercial scale reactor should be designed to achieve reasonably high integrated
photon utilization efficiencies, while providing low pressure drop, high throughput
operation in a reasonably compact physical configuration [11,12]. The preferred reactor
configuration is a key technical issue addressed in this research.
Skid-mounted PCO Unit
VOC-laden
humid air
Air
Stripping
Towers
VOC-eontamlnated
groundwater
l- -
Surge
Tank
Dry scrubber
Stripped
water
Utility
air
Figure 1-1. Integrated Air Stripping with Photocatalytic Oxidation Process
l
-------
Figure 1-2 compares the two principal reactor configurations tested in this research
project. Configuration 1 is the packed bed design, in which a cylindrical photoreactor
containing a number of equally-spaced parallel UV lamps is packed with a catalyst-
coated support material. The air flows parallel to the lamps that irradiate the supported
catalyst. Configuration 2 is panel bed design, in which a rectangular duct-shaped
photoreactor contains a number of alternating UV lamp banks and a removable panel-
type catalyst support material. In this design, the air flows perpendicular to the lamps
that irradiate the panels.
Catalyst-coated packing
UV lamps
Catalyst panels
Figure 1-2. Photocatalytic Oxidation Reactor Configurations. Configuration 1
(top) is a packed bed design, and Configuration 2 (bottom) is a panel
bed design.
2
-------
Project Objectives
The primary objectives of this research project were as follows:
• Define the gas residence time, catalyst characteristics, UV light intensity, and
photoreactor configuration that achieves greater than 95% destruction of the primary
VOCs in air stripper off-gases.
• Characterize the overall performance of the integrated air stripper/PCO process by
quantifying contaminant destruction during short-term transients generated by air
stripper startup, shutdown, and upsets, and by quantifying destruction performance
over an extended test period.
The secondary objectives of this project included:
• Quantify catalyst lifetime independent of factors related to integration of the PCO
unit with an air stripper,
• Identify reaction byproducts and conditions that inhibit or enhance their formation
rate.
Most of these objectives were successfully achieved through bench-scale and pilot-
scale controlled testing described in the body of this report. The first objective was met
through the following strategy. Statistically designed experiments enabled efficient
process performance characterization as a function of key process variables. The
derived empirical response surface models allowed prediction of conditions for which
high VOC destruction efficiency could be achieved. These predictions were
subsequently confirmed in a long-term bench-scale test and in several short-term pilot-
scale tests.
Because we were unable to obtain actual field-test data, the second primary objective
listed above could not be achieved.
3
-------
Chapter 2
Conclusions
This research led to key findings in two areas: (i) preferred photocatalytic reactor
configuration, and (ii) reactor performance under conditions expected at a groundwater
remediation site employing air strippers. This section summarizes these key findings.
Further details can be found in Chapter 5.
Preferred Photocatalytic Reactor Configuration
The original configuration designed and tested was an annular photoreactor based on
the "flow-by" configuration conceived and prototyped by IT Corporation and described
in our QAPP. In the "flow-by" design, the commercial Degussa P25 titania catalyst is
coated on the inside surface of a glass tube and the gas to be treated flows in the
annular space created by the tube and a concentric fluorescent UV lamp inserted in the
tube. This "flow-by" design provides an efficient distribution and utilization of the UV
light, and yields low pressure drop operation. However, both laboratory-scale and pilot-
scale performance screening tests revealed that this configuration was prone to
diffusional mass transport limitations, and for this reason this design was abandoned
early in the project and no performance results for this design are presented. To
minimize mass transport limitations, while distributing the light efficiently throughout the
reactor volume, a multi-lamp, packed bed, "flow-through" configuration was designed
and tested. This packed bed design is referred to in this report as Configuration 1, and
is shown in the top of Figure 1-2. All bench-scale testing was accomplished with a
single lamp, annular photoreactor, with internals that were consistent with, although not
identical to, the packed bed pilot unit.
Following completion of all bench-scale testing, and subsequent to pilot-scale testing of
Configuration 1, professionals at Zentox designed, constructed, and subsequently
tested a second "flow-through" photocatalytic reactor. This second reactor type
employs a removable panel-type catalyst support material for the active commercial
Degussa P25 titania, and the gas-flow is perpendicular to the lamps that irradiate the
panels. This panel bed design is referred to in this report as Configuration 2, and is
shown in the bottom of Figure 1-2.
Controlled testing of the two pilot-scale reactor systems revealed that they exhibit
qualitatively similar performance characteristics. Increasing water vapor content in the
feed decreases trichloroethylene (TCE) destruction efficiency. Lower feed rates yield
enhanced TCE destruction for otherwise fixed process conditions. Integrated photon
utilization efficiencies depend on inlet TCE concentration, with higher efficiencies
observed for higher inlet concentration. The abilities of the two reactor types to
distribute and employ photons efficiently throughout their respective reactor volumes
appear to be similar. The primary advantages of the second configuration lie in ease of
4
-------
fabrication and maintenance. In particular, the second or panel configuration utilizes a
catalyst support that is simple to produce, and that allows simple, rapid installation or
exchange of the active catalyst loaded panels.
PCO Process Performance
Based on the results of our bench-scale studies, we can make the following
conclusions on PCO process performance for the target application:
• Although a large process conditions window exists for which TCE can be destroyed,
trichlorethane (TCA) is non-reactive. Thus PCO as presently employed is not a
viable technology for treating TCA-contaminated air streams.
• Increasing incident UV light intensity and mean gas residence time increase TCE
conversion. The quantitative operating window (combination of process conditions)
required to yield 95% TCE destruction can be predicted through an empirical
response surface model. In general, UV intensities greater than 3.5 mW/cm2 and
residence times greater than 2.5 s will yield greater than 95% TCE destruction.
• Water vapor significantly inhibits the VOC destruction rate for chlorinated ethylenes
such as TCE. For high VOC conversion operation in the PCO unit, PCO units should
be placed downstream of a dehumidification unit located between air strippers and
the PCO unit; these dehumidifiers should reduce the RH to less than 50%.
• Principal carbon-containing products of PCO identified experimentally at the bench-
scale include carbon dioxide, carbon monoxide, and phosgene (COCI2). Failure to
close carbon mass balances under some process conditions suggests that not all
byproducts were identified. Further studies are needed in this area.
• The presence of secondary chlorinated ethylenes and chlorobenzene did not
significantly affect the conversion of the primary VOC TCE. However, selectivity to
the desired complete combustion product carbon dioxide decreased with increasing
co-contaminant concentration, suggesting that the presence of the co-contaminants
enhanced the production of undesirable partial oxidation products.
• TCE destruction activity could be maintained for up to thirty days on stream.
Results suggested that chlorine atoms build up on a fresh catalyst surface during an
initial period of operation until the surface is saturated. During this time frame only
small amounts of chlorine-containing products are detected in the reactor effluent.
After this initial period of operation, Cl-containing compounds, including HCI, Cl2 and
COCI2 are evolved from the surface. Failure to completely close the chlorine mass
balances under any process conditions suggests that not all Cl-containing
byproducts were identified.
5
-------
Chapter 3
Recommendations
Based on the research described in this report, the following recommendations are
made for further investigation:
• The hypothesis that the lack of carbon mass balance closure under some
conditions was due to undetected/unidentified volatile partial oxidation
intermediates should be tested. Hypothesis testing will require byproduct
detection studies with gas chromatography / mass spectrometry and subsequent
PCO testing under select conditions.
• The preferred PCO reactor Configuration 2 should be tested at an acceptable field
site so that the objectives related to practical integration of PCO with air stripping,
and perhaps vacuum soil vapor extraction can be achieved.
• First principles modeling of the preferred PCO reactor configuration should be
undertaken to allow reactor optimization to be performed in a rapid, efficient and
scientifically-sound manner. The modeling effort should include submodels for the
UV radiation field and intrinsic VOC destruction rate expressions, as well as the
reactor-scale advection-diffusion-reaction model.
The following additional recommendations are made for practical implementation of
PCO technology:
• Focus testing efforts on sites that are contaminated with chlorinated ethylenes.
Avoid sites contaminated with such recalcitrant compounds as aromatics and
chlorinated ethanes.
• Continue to focus equipment development efforts on advanced reactor designs.
Based on previous experience, it is likely that a large margin for process
performance improvement exists through this avenue.
6
-------
Chapter 4
Experimental Methods
Bench-scale Apparatus
The bench-scale PCO system [12] is shown schematically in Figure 4-1. Contaminated
air is prepared by mixing gases from pressurized gas cylinders using Tylan Corp. mass
flow controllers (mfc). Because N2 and 02 are used to produce a synthetic air mixture
of 0.79:0.21 N2:02(vol:vol), the feed gas stream contains no C02. Water vapor is
added to the feed flow by passing a separate nitrogen flow through a saturator; the
degree of saturation is determined by the gas-liquid contact time and the temperature in
the saturator. After mixing, the gases are fed to the photoreactor. All reactant lines are
heated to a temperature above the source cylinders' and saturator's temperatures to
ensure that no condensation takes place within the gas transport system. Individual
mass flow controllers can be calibrated or the total gas flow rate can be measured with
the bubble flow meter at the system outlet.
The annular photoreactor incorporates a 1.5 inch OD 20 Watt UV black lamp in a 2.25
inch ID glass tube. The annular space is filled with titania-coated packing; bed length is
90 cm. Prior to loading the reactor, the packing was coated with an adherent, uniform
titania film using a sequential washcoat method employing commercial Degussa P25
titania in a slurry of isopropyl alcohol. Following each washcoat, the packing was dried
in air at 150 °C. for two hours to evaporate the solvent. No other catalyst pretreatment
was employed. Ultraviolet light fluxes at the lamp and at the reactor outer wall in the
range of 310 - 400 nm wavelength were measured with a Minolta integrating
photometer.
The system incorporates three in-line sensors (Sensl, Sens2 and Sens3) and three
discrete sampling locations (Sam1, Sam2 and Sam3). Sam1 and Sam2 are automatic
multiple port GC sampling valves (Valco) which send a well-defined volume of gas to
one of two columns in the GC for VOC and carbon dioxide analysis, respectively. The
third sampling location (Sam3) allows gas samples to be withdrawn for off-line analysis
by gas detection tubes. Sensl is an in-line relative humidity sensor (Vaisala). Sens2
and Sens3 are in-line electrochemical sensors for HCI and CO measurement,
respectively. The feed gases may be analyzed with the sensors and sampling location
on the outlet line by bypassing the reactor. Flow, temperature and pressure measuring
devices are also included as indicated in the figure. All measurements are recorded
using automatic data acquisition driven by a Hewlett-Packard 486 processor based PC.
Because all bench-scale measurements are made on line in real time, sample holding
time is not an issue.
7
-------
GC
Carrier
Gas #1
GC
Carrier
Gas #2
£4
mfc
mfc
Pure
02
Pure
Pure
N2
VOC
in Air
£
mfc
mfc
mfc
mfc
n
£
Saturator
KD
Feed Preparation Section
"To GC
Bubble
Flow
Meter
Sens2 Sens3
PCO Feed
m i»
Sensl
Sam1 Sam2 Sam3
PCO Reactor Effluent
UV Lamp
Annular PCO Reactor
KEY
Sam1 - 3
Sampling locations
Sensl - 3
Sensing locations
mfc
mass flow controller
P
Pressure gauge
T
Temperature sensor
Figure 4-1. Bench-scale photocatalytic oxidation apparatus.
On-line analysis of VOCs at concentrations as low as 0.5 ppmv is performed with a
Varian model 3700 gas chromatograph fitted with a 6' x 1/8" stainless steel column
packed with 0.1% SP-1000 on 80/100 Carbopack C and equipped with a flame
ionization detector (FID). The carrier gas is ultra high purity (99.998%) helium, which is
further purified upstream of the GC with an in-line oxygen trap and an Alltech in-line gas
purifier containing mole sieve pellets and indicating drierite. Carrier gas flow rate is 20
seem. The GC column oven and the detector are maintained at 130 °C. at 248 °C.,
respectively. Three-point calibration using certified gas standards from Scott Specialty
gases covering VOC concentrations from 1 to 30 ppmv was performed prior to analysis.
Further system and operating details can be found in Appendix I: Standard ASU
Operating Procedures of the Quality Assurance Project Plan for this study.
On-line carbon dioxide analysis is performed with an isothermal mole sieve column and
an external methanizer/ FID unit (Valco). As the name implies, the unit converts C02 in
the air stream to methane over a heated nickel catalyst bed, and the methane formed is
then detected by the FID. The carrier gas is 18 seem ultra high purity (99.998%)
8
-------
nitrogen, which is further purified upstream of the GC with an in-line oxygen trap and an
Alltech in-line gas purifier containing mole sieve pellets and indicating drierite. The
methanizer catalyst bed, GC column oven and the detector are maintained at 380 °C.,
120 °C., and 297 °C., respectively. Three-point calibration using a certified gas
standard from Matheson covering C02 concentrations from 9 to 50 ppmv was performed
prior to analysis. Note that because synthetic air is employed in these experiments, our
measurements are not masked by the relatively large concentrations of carbon dioxide
present in ambient air.
Table 4-1 summarizes the minimum detection limits (MDL) of the analytical tools
employed in the bench-scale studies.
Table 4-1. Minimum Detection Limits for Gas Analysis Methods.
Measurement
Technique
MDL
Reporting Unit
VOCs
GC
0.5
ppmv
C02
GC
0.8
ppmv
CO
sensor
0.5
ppmv
HCI
sensor
0.5
ppmv
COCI2
gas tube
0.1
ppmv
Cl2
gas tube
0.1
ppmv
RH
sensor
0.1
%RH
Bench-Scale Procedures
Prior to a set of experiments employing a given combination of VOCs, the VOC in N2
cylinder is prepared by first filling the cylinder with 1 atm UHP N2, and then injecting a
known, pre-determined amount of VOC or VOCs into the cylinder through a gas-tight
septum and valve. The cylinder is then pressurized to 500 psi using UHP N2, and
allowed to mix for several hours before being used. Actual concentrations are
determined by analyzing the cylinder contents by flowing the gas through the system
and using the on-line GC. Feed composition and total flow rate to the photoreactor are
independently selected by appropriately setting individual flow rates of reactants
(organic in UHP N2, UHP N2, and zero-grade air or UHP 02) from the individual
pressurized gas cylinders using mass flow controllers (Tylan). The flow rate through the
water saturator relative to other flow rates contributing to the total, coupled with
saturator conditions control the water vapor content in the feed stream. Actual RH in
the feed gas is measured with the in-line humidity sensor.
9
-------
Once flows are set and stabilized, the feed gas is analyzed with the GC until a steady
reading is achieved. At this point, the UV lamps are illuminated, and the reactor outlet
is regularly analyzed with the sensors and GCs described above. Once steady
operating conditions are achieved (typically several hours with fresh catalyst beds,
several minutes with conditioned beds), measurements are recorded for several more
hours. It is these average steady-state values, typically comprising six to eight
individual datapoints taken over two hours, that are reported.
Quality Assurance Metrics, Procedures and Objectives
Quality Assurance (QA) metrics employed in the bench-scale studies include: Precision,
Accuracy, Completeness, and Mass Balance (both carbon and chlorine). Only data
taken during the steady-state portion of a run is subject to these objectives.
For purposes of this project, Precision is a measure of the mutual agreement among
individual measurements of the same property, usually under prescribed similar
conditions. When there are sufficient individual data points (i.e., more than 3),
Precision is expressed in terms of relative standard deviation (RSD). Because a typical
steady-state condition is monitored for several hours, we were able to make six to eight
measurements for each process variable and subsequently determine RSDs for all
critical measurements (including VOC concentrations). The relative standard deviation
is defined as:
RSD = —.100%
Uave
where s is the standard deviation and uave is the mean of replicate analyses. Standard
deviation is defined as follows:
where u, is the measured value of the /'-th replicate, uave is the mean of replicate
analyses, and n is the number of replicates.
Accuracy is the degree of agreement of a measurement (or an average of replicate
measurements) with an accepted reference or true value. Accuracy for process
measurements is expressed as a percentage of the reference or true value as
determined by the measurement system. Reference or true values are external
calibration standards that are independent for the standards used to prepare the
instrument for normal operation. Determination of the accuracy for certain process
10
-------
measurements could not be made by this approach because reference or true sources
can not be applied to the measuring sensor. Measurements falling in this category
include the following: air flow rate, light intensity, temperature, and pressure. Assessing
the accuracy of these measurements was made solely on the measurement
instruments performance against calibration standards.
Accuracy is quantified as percent recovery (%R). For measurements in which matrix
spikes are used, percent recovery is calculated as follows:
%r = (S~U) .100%
C*,
where S and U are the measured concentrations in the spiked and unspiked aliquot,
respectively, and C$a is the actual concentration of the spike added.
When a calibration standard or standard reference material (SRM) is used, percent
recovery is given by:
%R = -100%
r
SRM
where C,A is the measured concentration and C: is the actual concentration of the
M orcm
SRM.
Completeness is a measure of the amount of valid data obtained from a measurement
system compared to the expected amount to be obtained under correct normal, steady
state operations. Non-valid data or outliers are indicative of instability in either the
treatment process or the measurement systems.
Completeness (%C) is defined quantitatively as the percentage of all measurements of
a given type that are judged to be valid:
%C = —100%
n
where V is the number of measurements judged to be valid and n is the total number of
measurements made.
Mass balances are performed during predetermined mass balance intervals for both
carbon and chlorine. These balances are calculated according to the following
expressions:
li
-------
mass C in PCO reactor outlet stream
• lOO/o
mass C in PCO reactor feed stream
Mass CI in PCO reactor outlet stream jqq0//
Mass CI in PCO reactor feed stream
Objectives for these metrics were that the balances should close to within 75 and 125
percent. These project specific QA objectives were not met, as described in detail in
Chapter 5. Failure to meet these mass balance goals was most likely due to our
inability to detect unknown byproduct species in the photoreactor outlet.
Table 4-2 summarizes the Quality Assurance (QA) objectives and aggregate results for
the analytical tools employed in the bench-scale studies. For Precision and Accuracy,
the range of RSDs achieved and percent recoveries achieved, respectively, are
tabulated. Completeness includes all data collected. All data presented in this report
met Precision, Accuracy and Completeness QA objectives.
Table 4-2. QA Objectives and QA Results for Gas Analysis Methods
Precision
Precision
Accuracy
Accuracy
Complete-
Complete.
Technique
Goal
Achieved
Goal
Achieved
ness Goal
Achieved
(RSD)
(RSD)
(%R)
(%R)
(%C)
(%C)
VOCs
GC
25%
4 - 9%
50- 150%
87-110%
90%
100%
co2
GC
25%
12-19%
50- 150%
72- 123%
90%
100%
CO
sensor
25%
6-15%
50- 150%
77-119%
90%
100%
HCI
sensor
25%
7 -14%
50-150%
84-115%
90%
100%
COCI2
gas tube
25%
15-22%
50- 150%
67-128%
90%
100%
Cl2
gas tube
25%
11 -21%
50- 150%
70-131%
90%
100%
RH
sensor
20%
2 - 4%
70-125%
93 -105%
90%
100%
C
mass balance
CI
mass balance
12
-------
Chapter 5
Results and Discussion
Bench-scale Experiments: Set 1
Experimental Design
Set 1 experiments are designed to characterize the performance of the bench-scale
unit under process conditions that are representative of those found in the field at the
site that was to be tested. Groundwater at the target industrial field site is pumped and
treated by a 660 gpm remediation plant employing air stripping and carbon bed
adsorption of the stripper off gases. In the planned field test, the PCO unit would have
been employed to treat a slipstream of the air stripper off gases. The expected total
VOC concentration ranged from 10 to 18 ppmy, with typically two-thirds of the total
attributed to TCE, An in-line dehumidifer reduces the RH in the off-gases to about 50%
upstream of the carbon beds.
PCO process characterization was efficiently achieved by employing statistical design
of experiments (DOE) surface response methodology (SRM). This approach
maximizes empirical information obtained for a minimum of experimentation.
Independent process variables are total VOC concentration at the reactor inlet, relative
humidity, incident UV intensity, reactor temperature, and mean gas residence time.
Table 5-1 summarizes the independent process variables and their low, center point,
and high level values; coded levels appear in the top-most row of the table. Because
the study was performed over a fairly narrow range of conditions, it was not necessary
to set five discrete levels for each independent variable. Instead, a 1/2 fraction of a 25
factorial design augmented with axial runs and center point replicates was chosen.
This "face centered cubic" design provides a complete response surface and a
reasonably stable variance over a large portion of the design region.
Table 5-1. Process Parameters and Levels
Factor
Total inlet VOC concentration (ppmv)
Relative Humidity (%)
UV lntensity|wai, (mW/cm2)
Temperature (°C)
Residence Time (s)
Code Low(-1) Center (0) High (+1)
A
10
20
30
B
20
50
80
C
0.5
2.0
3.5
D
40
60
80
E
0.5
1.0
1.5
13
-------
To maintain a high signal-to-noise ratio in our GC measurements of VOC concentration,
we chose an inlet VOC range on the high side of expected operation. Recall that the
pilot unit was to be sited between a de-humidifier and downstream carbon beds, and
that expected relative humidity (RH) at this location was on the order of 50%. The RH
range is therefore centered around a value of 50%. Temperatures were centered
around the expected operating temperatures of the PCO unit; these modestly elevated
temperatures result through thermal energy dissipation from the high intensity UV
lamps employed in the pilot-scale unit. Table 5-2 summarizes the methods used to
control and measure the independent process variables.
Table 5-2. Methods to Control and Measure Independent Process Variables
Process Variable
Process Control
Process Measurement
VOC concentration
Relative Humidity
UV intensity
Temperature
Residence Time
VOC in N2 flow rate
Wet N2 flow rate
Manual lamp dimmer
Omega PID temperature
controller
Total feed rate
Varian 3700 GC with FID
Vaisala Humidity Sensor
Minolta UV photometer
K-type thermocouple
Mass flow controller
The experimental runs were conducted in a randomized manner according to a
structure suggested by Design Expert (a statistical software package). The aliasing
relationship used in generating the factorial block is NABCDE, therefore it is important
to keep in mind that two-factor interactions are not aliased with each other but are
aliased with three-factor interactions.
The principal measured responses are conversion Xvoc of the VOC and selectivity SC02
of the reaction towards producing carbon dioxide. The conversion is a direct measure
of the destruction efficiency, while the selectivity is a measure of the percentage of the
VOCs destroyed that are converted to the desired complete combustion product carbon
dioxide. These responses are defined as follows:
\-out
•100%
—(- CO} |out —C-C02 (j>;
Sco2 = Nvqc -100%
X (C; \in ~C;\out )
i = l
14
-------
where
Cj = concentration of the i-th VOC (nmol/m3)
Cc02 = concentration of carbon dioxide (nmol/m3)
n, = number of carbon atoms in the i-th VOC
and |,„ and \out refer to values at the PCO reactor inlet and outlet. To determine these
derived responses, individual VOC and inorganic gas-phase product concentrations in
the outlet stream were measured. Table 5-3 summarizes the direct (as opposed to
derived) process responses and the measurement methods employed.
Table 5-3. Process Responses and Measurement Techniques
Process Response Measurement Method
Outlet VOC concentration Varian 3700 GC with FID
Outlet C02 concentration Varian 3700 GC with a methanizer/FID
Molecular chlorine in outlet stream Gas detector tubes (Sensidyne)
Phosgene in outlet stream Gas detector tubes (Sensidyne)
HCI concentration in outlet stream On stream sensor (Detcon)
CO concentration in outlet stream On stream sensor (Sierra Monitor Corp.)
Set 1 Results
Table 5-4 summarizes the results of the 16 runs that comprised the factorial block, the
four center point replicates, and the ten axial runs of the face centered cubic
experimental design. Note that under the conditions of these experiments, TCA
conversion is essentially negligible, with measured values ranging between about 2 and
6%. For purposes of brevity, subsequent DOE models for VOC conversion described
below and the accompanying discussion are for TCE conversion only.
The data of Table 5-4 were analyzed using Design Expert, a statistical software
package for response surface studies. Linear, quadratic and cubic models were
employed in an attempt to fit the data. The adjusted R2 increased from 0.75 for the
linear model to 0.97 for the more complex quadratic model. The latter value implies
that 97% of the measured response can be captured by the quadratic model. The fit
does not significantly improve if a cubic model is employed.
15
-------
Table 5-4. Results for the Factorial Block of the fee Design
RSM Run A B C D E TCE TCA C02
ID No. VOC (RH) (UV) (T) (x) Conv. Conv. Select.
(%) (%) (%)
1
17
10
20
0.5
40
1.5
64
3.8
71
2
5
30
20
0.5
40
0.5
48
3.2
57
3
8
10
80
0.5
40
0.5
37
2.7
70
4
1
30
80
0.5
40
1.5
43
3.5
42
5
19
10
20
3.5
40
0.5
75
4.1
84
6
10
30
20
3.5
40
1.5
83
5.4
49
7
14
10
80
3.5
40
1.5
50
4.3
100
8
15
30
80
3.5
40
0.5
40
2.6
51
9
12
10
20
0.5
80
0.5
50
4.5
93
10
13
30
20
0.5
80
1.5
58
4.8
47
11
4
10
80
0.5
80
1.5
62
4.7
85
12
3
30
80
0.5
80
0.5
43
2.6
56
13
9
10
20
3.5
80
1.5
85
5.5
100
14
11
30
20
3.5
80
0.5
67
3.9
67
15
16
10
80
3.5
80
0.5
63
4.2
100
16
20
30
80
3.5
80
1.5
70
4.6
77
0
2
20
50
2
60
1
66
3.8
69
0
6
20
50
2
60
1
62
3.7
64
0
7
20
50
2
60
1
63
3.5
64
0
18
20
50
2
60
1
61
3.7
66
17
21
10
50
2
60
1
65
4.0
86
18
28
30
50
2
60
1
60
3.2
54
19
24
20
20
2
60
1
70
3.9
86
20
22
20
80
2
60
1
60
3.5
97
21
27
20
50
0.5
60
1
51
2.5
72
22
26
20
50
3.5
60
1
69
4.0
81
23
29
20
50
2
40
1
58
2.7
47
24
23
20
50
2
80
1
63
3.0
100
25
25
20
50
2
40
0.5
51
3.9
78
26
30
20
50
2
40
1.5
65
4.2
78
16
-------
The quadratic model for conversion as a function of the significant effects is given by:
Xvoc = 111-02 - 0.22 ¦ VOC - 2.81 ¦ RH + 9.02¦ UV - 0.23-T + 42.0 • r
- 15.11-t2 - 0.07 ¦ RH ¦ UV + 0.008 ¦ RH ¦ T
where VOC = VOC concentration (ppmv), RH = Relative Humidity (%), UV = UV
intensity (mW/cm2), T = Temperature (K), and r - residence time (s).
The magnitudes of the parameter coefficients reveal that all five independent factors
explored are significant. Residence time, UV intensity and relative humidity are the
principal single factor effects. Increasing residence time and UV intensity increase
VOC conversion, whereas increasing relative humidity decreases VOC conversion.
Inhibition of the TCE photocatalytic oxidation rate by water vapor is a well-established
phenomenon first reported by our laboratory [1-3,13], and then reproduced
independently elsewhere [14], The effects of VOC concentration and temperature have
relatively modest negative effects on VOC conversion.
The model contains the following two significant two-factor interaction terms: (i) relative
humidity and UV intensity, and (ii) relative humidity and temperature. Figures 5-1 and
5-2 are the respective effect plots for these interactions. This empirical interaction can
be explained consistently by a mechanistically-based model in which the apparent
water vapor rate inhibition is largely caused by competitive adsorption between water
vapor and TCE [1,13]. Figure 5-1 shows that RH has a stronger inhibiting effect at high
UV intensity, reflecting the ability of ultraviolet irradiation to enhance desorption of water
[15] and significantly reduce the steady state surface concentration of adsorbed water
at low RH. Figure 5-2 shows that the effect of water vapor inhibition is more significant
at low temperature. As temperature is increased, water desorption rate (which is
strongly thermally activated) increases while the non-activated adsorption rate remains
unaffected. This shift in adsorption-desorption equilibrium will decrease the steady
state adsorbed water surface concentrations. We therefore expect a stronger rate
inhibition at lower temperatures where, for otherwise identical process conditions,
adsorbed water fractional surface coverages are higher.
A quadratic terms in residence time explains the observed curvature in the conversion
response. Three-dimensional response surfaces, which visually show the quadratic
behavior, appear in Figures 5-3 and 5-4. Figure 5-4 dramatizes the magnitude of the
apparent water vapor inhibition.
17
-------
90
g 80
| 70
<5
§ 60
o
O
g 50
40
0 20 40 60 80 100
Relative Humidity (%)
Figure 5-1. Interaction effect between Relative Humidity and UV Intensity on
VOC conversion.
High UV
Low UV
i
I
1
¦
80
g 70
I 60
<5
I 50
o
O
O 40
30
0 20 40 60 80 100
Relative Humidity (%)
HighT
Low T
Figure 5-2. Interaction effect between Relative Humidity and Temperature on
VOC conversion.
18
-------
25 >
VOC concentration (ppm
VOC conversion
Residence time (s)
Figure 5-3. Response surface for VOC conversion versus residence time and
inlet VOC concentration. 20% RH, 3.5 mW/cm2 UV intensity, 40 °C.
Relative Humidity (%)
60 VOC conversion
70 1 i -**1 Residence time (s)
05
Figure 5-4. Response surface for VOC conversion versus residence time and
RH. 10 ppmv VOC concentration, 3.5 mW/cm2 UV intensity, 40 °C.
19
-------
Selectivity
Selectivity data were analyzed using Design Expert. Probability test values and
analysis of variance for the various rival models (linear, quadratic, etc.) clearly indicated
that the following linear model adequately describes the observed selectivity behavior:
SC02 = -23.90 - 1.61-VOC -0.19- RH + 3.93 -UV + 0.43 -T - 12.51 ¦ r
+ 0.22 ¦ RH ¦ r
where variables are as defined previously. The model shows that the strongest effects
are VOC concentration, temperature and UV intensity. Increasing VOC concentration
decreases selectivity, while increasing light intensity (UV) and increasing temperature
(T) increase selectivity. The model suggests that relative humidity (RH) and residence
time (x) do not by themselves have strong effects on selectivity, but that the interaction
of these two independent variables is significant and in a direction opposite to the main
effects. This somewhat illogical outcome is likely due to the nature of the design. In
particular, the design employed is of Resolution V, implying that two-factor interactions
are aliased with three-factor interactions. Therefore the apparent relative humidity -
residence time two-factor interaction is in all likelihood a complex three-factor
interaction between the other 3 factors, namely VOC concentration, UV intensity and
temperature. In other words, it is valid to replace the two-factor interaction term in the
linear model with a three-factor "VOC x UV x T" term. We can therefore rewrite the
linear model for selectivity as follows:
SC02 = - 23.90 - 1.61 ¦ VOC - 0.19 ¦ RH + 3.93 • UV + 0.43 • T - 12.51 ¦ r
+ 0.0002 ¦ VOC - UV -T
The magnitude of the interaction term, while statistically significant, is small relative to
the main factor effects. Figure 5-5 shows the VOC and T response surface for
selectivity: Figure 5-6 is the UV and T response surface for selectivity. The relatively
planar response surfaces revealed in these figures confirm that the three-factor
interaction term is not a major contributor to the observed selectivity response.
The main selectivity responses can be explained within the framework of a surface-
mediated, free-radical sequential oxidation mechanism [5,16]. In this model, hydroxyl
radicals created through UV excitation of the titania photocatalyst initiate oxidative
attack on adsorbed VOCs to create VOC radicals. The radical sites are then attacked
by oxygen to create peroxy radicals. These reactive peroxy radicals readily decompose
to produce partial oxidation products on the catalyst surface. This series hydroxyl
radical attack / peroxy formation through oxygen addition sequence continues until
complete combustion products are produced. In the well-documented case of
20
-------
100-
90-
80-
Selectivity (%) 70 -
60.
50
353
343
40
7 333
Temperature (K)
< 323
313
VOC Concentration (ppm)
10
Figure 5-5. Influence of VOC concentration and temperature on selectivity.
Selectivity (%)
Temperature (K)
UV Intensity (mW/cm2)
Figure 5-6. Influence of UV intensity and temperature on selectivity.
21
-------
isopropyl alcohol photocatalytic oxidation, the alcohol is converted to acetone, which is
subsequently oxidized to acetaldehyde, formaldehyde and formic acid. Rapid formic
acid decomposition produces C02 [16]. Under certain conditions, acetone desorbs into
the gas phase, and selectivities below 100% are observed [12].
By analogy with isopropyl alcohol oxidation, we expect that TCE is first oxidized to
dichloroacetaldehyde, then to dichloroacetic acid, and ultimately to chlorinated formic
acid intermediates. Under dry conditions (< 10% RH), hydroxyl radicals may not be the
primary initiator of oxidation, and an alternative reaction path through phosgene
becomes dominant [17,18], In addition to phosgene, dichloroacetyl chloride may be
evolved into the gas phase. Hung and Marinas have detected the presence of carbon
tetrachloride and chloroform under dry conditions [14], With increasing relative
humidity, the production of these undesirable intermediates decreases. For sufficiently
high humidity (>40-50%) and residence time, essentially complete conversion to carbon
dioxide is observed [14,19].
Within the framework of Langmuir-Hinshelwood-Hougen-Watson (LHHW) surface
reaction kinetics previously documented for TCE PCO at the bench-scale [1], an
increase in TCE concentration increases the average surface coverage by TCE. For
otherwise fixed conditions, the production rate of reactive hydroxyl radicals will be
essentially the same. At steady state, there will therefore be relatively fewer hydroxyls
to attack adsorbed partial oxidation intermediates, leading to a reduced integrated
selectivity to C02. A similar argument can be posed to explain the UV dependence.
With increasing UV intensity, the production rate of hydroxyl radicals increases. For
fixed TCE surface concentration, a relatively greater amount of hydroxyls will be
available to attack partial oxidation intermediates, leading to an increased integrated
selectivity to C02. The principal affect of temperature is likely associated with
decreased surface coverage by TCE.
Intuitively, one might expect to observe a correlation between the two dependent
response variables selectivity and conversion. Figure 5-7 plots selectivity for
conversion for the Set 1 experiments. The figure shows that there exists only a mild
positive correlation (correlation factor« 0.09) between C02 selectivity and TCE
conversion. This finding is consistent with the mechanistic argument outlined above.
22
-------
o° j? o
a 40
y = 0.4641x +45.15
0.0907
20
40 60
Conversion (%)
80
100
Figure 5-7. Selectivity versus TCE conversion for Set 1 experiments.
Byproduct Detection, Mass Balances and Other Data Quality Indicators
Based on the literature, potential gas-phase byproducts formed during the
photodegradation of TCE are COCI2, Cl2, CO, HCI, and dichloroacetyi chloride
[14,17,18]. in an effort to close the mass balance for carbon and chlorine, the detection
of these chemical species (with the exception of dichloroacetyi chloride, which was not
detectable with our analysis tools) was carried out as described in the QAPP. No VOCs
other than TCE and TCA were observed in the GC chromatograms. Carbon monoxide
levels were between 0.5 to 1.5 ppmv for all experimental runs. Molecular chlorine (Cl2)
and phosgene (COCI2) levels were observed at levels < 0.2 ppmv; no significant
hydrogen chloride was detected.
Table 5-5 reports the calculated C mass balances for each of the experimental runs.
Carbon mass balance closures range from ca. 60 to 100%. In general, highest percent
closures were obtained for low inlet VOC concentration. Low inlet TCE concentrations
also favor higher selectivity. Lack of closure could not be attributed to inaccuracy in
C02 measurements, but is instead most likely due to non-detect of volatile partial
oxidation products including, but not necessarily limited to, dichloroacetyi chloride.
23
-------
Table 5-5. Carbon Mass Balance for the Set 1 Experiments
Design Inlet TCE Outlet TCE Outlet C02 % closure
ID concentration concentration concentration
(PPmv) (PPmv) (ppmv)
1
8
2.9
7.3
88.1
2
24
12.5
13.1
81.5
3
8
5.0
4.1
94.4
4
24
13.7
8.7
77.3
5
8
2.0
10.1
94.3
6
24
4.1
19.5
59.8
7
8
4.0
8
100
8
24
14.4
9.8
82.5
9
8
4.0
7.5
100
10
24
10.1
6.5
57.7
11
8
3.0
8.4
96.3
12
24
13.7
11.6
83.3
13
8
1.2
13.6
100
14
24
7.9
21.5
79.8
15
8
3.0
10.1
100
16
24
7.2
25.9
86.0
0
16
5.4
14.6
82.3
0
16
6.1
12.7
80.9
0
16
5.9
12.9
80.3
0
16
6.2
12.9
82.2
17
8
2.8
8.9
96.9
18
24
9.6
15.6
74.6
19
16
4.8
19.3
93.4
20
16
6.4
18.6
100
21
16
7.8
11.8
88.8
22
16
5.0
17.9
90.3
23
16
6.7
8.7
72.2
24
16
5.9
20.1
100
25
16
7.8
12.7
91.6
26
16
5.6
16.2
88.8
A CI atom balance is not reported here since, except for unconverted TCE, only very
small amounts of chlorine-containing compounds were detected in the reactor outlet for
all runs. Researchers at the National Renewable Energy Laboratory (NREL) have
reported that chlorine builds up on the surface of a fresh catalyst, and that breakthrough
24
-------
of chlorine-containing compounds typically does not occur until after many hours on
stream [20], It is therefore likely that chlorine buildup is responsible for the low levels of
chlorine-containing compounds detected in the exit stream from the reactor. This
outcome highlights the need for long-term activity testing.
The data for the four center point replicates show that reproducibility in these
experiments is high. Conversions, selectivities, and carbon mass balances vary by no
more than ± 5% from their mean values. This level of reproducibility falls well within the
acceptable reproducibility bounds for these experiments as outlined in the QAPP.
Bench-Scale Experiments: Set 2
Experimental Design
Set 2 is designed to efficiently characterize the performance of the bench-scale unit for
conversion of complex contaminant mixtures. In particular, we were most interested in
the effects of the presence of the co-contaminants on the conversion of the primary
contaminant (TCE) and on overall selectivity to the desired complete combustion
product (carbon dioxide).
At the target Superfund site, co-contaminant or secondary VOCs present along with the
primary VOCs are 1,1-dichloroethylene, cis 1,2-dichloroethylene, vinyl chloride,
tetrachloroethylene (also known as perchloroethylene), and 1,2-dichlorobenzene. Each
of these co-contaminants is present in varying amounts depending on air stripper
operation, but under no circumstance have levels higher than 1 ppmv been detected.
Because a mixture experimental design employing six individual co-contaminants would
yield an exceptionally large number of experimental runs, we chose to lump together
the compounds that should behave in a chemically similar fashion. Specifically, the
chlorinated ethylenes (1,1-dichloroethylene, cis 1,2-dichloroethylene, vinyl chloride, and
perchloroethylene) were lumped together to create a "secondary chlorinated ethylenes"
pseudo-component. This strategy yields the following three-component mixture:
trichloroethylene (mole fraction x,), secondary chlorinated ethylenes (mole fraction x2),
and 1,2-dichlorobenzene (mole fraction x3). The initial concentrations for each of the
chlorinated VOCs in the lumped secondary chlorinated ethylenes category were held in
the same ratio for all experiments, at levels designed to approximate the average ratios
found at the Superfund site.
A 32factorial mixture experiment was designed to capture the maximum pseudo-
component ratios R1 = x, / x2 and R2 = x, / x3 expected at the Superfund site. Table 5-6
summarizes these ratios and the corresponding pseudo mole fractions of the three
pseudocomponents for the nine experimental runs performed in this set.
25
-------
Table 5-6. Mixture Experimental Design
Design Run R1 R2 x, x2 x
3
ID
1
2
6
0.60
0.30
0.10
2
4
6
0.70
0.18
0.12
3
6
6
0.75
0.13
0.12
4
2
11
0.63
0.31
0.06
5
4
11
0.74
0.19
0.07
6
6
11
0.80
0.13
0.07
7
2
16
0.64
0.32
0.04
8
4
16
0.76
0.19
0.05
9
6
16
0.81
0.14
0.05
During these mixture experiments, other process conditions were fixed at
representative values as follows:
TCE inlet concentration 10 ppmv
Relative Humidity 20%
UV Intensity 3.5 mW/cm2
Temperature 40 °C.
Residence time 0.5 s
These process values were chosen to yield a TCE conversion that was reasonably high
(order of 70%) but not so high (i.e., >95%) that conversion sensitivity to the presence of
co-contaminants would be masked by poor measurement statistics.
Set 2 Results
Table 5-7 summarizes the experimental results for the nine mixture runs. Included in
the table are the respective fractional conversions X2, and X3for the three
pseudocomponents, as well as the selectivity to the complete combustion product
carbon dioxide. Also summarized for reference purposes are the CO and C02 outlet
concentrations.
26
-------
Table 5-7. Mixture Experimental Results
Design ID
Ri
R2
X,
X2
X3
Selectivity
(%)
CO
(ppmv)
C02
(PPmv)
1
2
6
0.68
0.66
0.28
24
0.88
5.5
2
4
6
0.66
0.60
0.20
52
0.33
8.2
3
6
6
0.68
1.00
0.26
75
0.02
13.3
4
2
11
0.66
0.60
0.28
41
0.27
7.3
5
4
11
0.68
0.60
0.30
59
0.15
10.5
6
6
11
0.71
1.00
0.22
89
0.02
13.2
7
2
16
0.66
0.40
0.20
50
0.88
8.5
8
4
16
0.67
1.00
0.60
59
0.09
8.2
9
6
16
0.71
1.00
0.40
79
0.02
14.2
The TCE conversion X, appears to be only mildly affected by the presence of the co-
contaminants. A statistical analysis of the experimental data revealed that nearly the
entire sum of squares is explained by the sum of squares for the mean, i.e., that the
effect of the co-contaminants on TCE conversion is not statistically significant.
For Selectivity, statistical analysis of the results revealed that the effect of the presence
of the co-contaminants is significant. The selectivity as a function of the
pseudocomponent ratios can be described adequately by a quadratic model given by:
SC02 = - 3919 + 10 69¦ Rl + 9-23 -r2 +0-7J- Rf - 0.26 ¦ R22 - 0.525 • Rr R2
Figure 5-8 shows this response visually. The response shows a complex dependence
on the pseudocomponent ratios. As the ratio (R^ of TCE to secondary chlorinated
ethylenes increases, Selectivity increases regardless of the chlorobenzene level. The
increase is most dramatic at the lowest ratio of TCE to chlorobenzene. The Selectivity
tends to go through a broad maximum as the ratio (R2) of TCE to chlorobenzene
increases, with the location of the maximum shifting to lower R2 ratios as the ratio of
TCE to secondary chlorinated ethylenes increases.
The apparent complexity of the Selectivity behavior is related to the conversion
behavior of the co-contaminants. Aggregate conversion data for the three
pseudocomponents is included in Table 5-7. In general, conversion of the chlorinated
ethylenes increases with increasing ratio (R,) of TCE to secondary chlorinated
ethylenes, while the conversion of chlorobenzene increases with increasing R2 up to a
value of R2 of about 11, at which point the Selectivity becomes insensitive to further
ratio increases.
27
-------
100
80-
70
Selectivity {%) 60
R1 (TCE: Ci-
ethylenes)
R2 (TCE: Aromatic)
Figure 5-8. Influence of pseudocomponent ratios Rt and R2 on Selectivity.
In general, there is a scarcity of literature on the effects of co-contaminants on PCO
performance; the overwhelming majority of experimental work has focused on single
contaminant studies. The results described above that show an increase in co-
contaminant conversion with increasing relative TCE concentration are consistent with
several literature studies that revealed the rather unique ability of TCE to enhance co-
contaminant conversion [21]. For example, Ollis and coworkers found that including
TCE in an air stream contaminated with toluene increased the toluene conversion
dramatically relative to a TCE-free stream [21]. This enhancement was attributed to the
creation of highly reactive chlorine radicals from TCE destruction. These radicals,
perhaps present in the gas phase as well as on the catalyst surface, initiate an
oxidative attack on the co-contaminant, or/and create other reactive radicals that attack
the co-contaminant.
Mass Balances and Other Data Quality Indicators
Table 5-8 reports the calculated carbon mass balances for each of the experimental
runs in the mixture design. Carbon mass balance closures range from ca, 60 to near
100% (basically the same range observed for the factorial experiments in Set 1). As for
28
-------
Set 1, a chlorine atom balance is not reported since essentially no chlorine-containing
compounds other than those found in unconverted VOCs were detected in the reactor
outlet for all runs.
Table 5-8. Carbon Mass Balance for Mixture Experiments.
Inlet C
Outlet C
Outlet C02
%C
Design
concentration
concentration
concentration
balance
ID
(PPmv)
(PPmv)
(ppm¥)
closure
1
41.4
24.6
5.5
60.0
2
36.2
21.0
8.2
83.4
3
36.1
18.3
13.3
90.3
4
31.3
13.1
7.3
68.4
5
32.6
14.9
10.5
81.0
6
29.8
15.1
13.2
98.3
7
29.4
12.4
8.5
74.5
8
27.4
11.0
8.2
73.7
9
37.8
14.6
14.2
78.8
Bench-scale Experiments: Set 3
Set 3 experiments were designed to test the long-term stability of the photocatalyst in a
reactive environment. The TCE inlet concentration, RH, UV intensity and temperature
listed below were chosen because the Set 1 Bench-scale experiments (specifically, Run
no. 9) showed that they should result in high TCE conversion (85%) and selectivity
(100%) during the initial stages of the run. In addition, 100% C atom mass balance
closure was obtained under these conditions.
TCE inlet concentration 10 ppmv
Relative Humidity 20%
UV Intensity (initial) 3.5 mW/cm2
Temperature 80 °C.
Residence time 2.5 s
29
-------
Note that to achieve greater than 95% TCE conversion initially, a 2.5 s residence time
was employed (1 s longer than the residence time employed in the factorial
experiments). To simplify gas analysis, TCE was the only VOC in the contaminated air
stream. A freshly-prepared catalyst bed was employed. Figure 5-9 shows the TCE
conversion, selectivity, and chlorine mass balance as a function of time on stream. The
outlet gases were analyzed daily during this long-term run. The C mass balance was
essentially closed in all measurements («100%).
Over the course of the first six days of operation, TCE conversion held steady at
greater than 95%, while Selectivity was constant at approximately 100%. The CI atom
mass balance was only about 3-4%. These results are consistent with those expected
based on the Set 1 experiments. After six days on stream, a dramatic change in
performance is observed. The CI atom balance closure jumps over the next few days
to the 70-80% closure range. This improved closure (which falls within allowable limits
specified in the QAPP) is associated with the appearance of substantial concentrations
of HCI and Cl2, as well as lesser amounts of COCI2. Selectivity falls below 100% at this
point as a portion of the inlet C appears in the COCI2 byproduct. Over the remainder of
the run, the TCE conversion gradually falls, although it remains above 90% over the
entire duration. The selectivity also gradually falls, although it remains above 95%.
The findings in this experiment are consistent with a model in which the fresh surface
becomes chlorinated as TCE is dechlorinated and oxidized in the PCO process. Initially
this deposited CI is non-volatile and an inventory of adsorbed CI builds up on the
surface as TCE-containing air is treated. At the point of CI atom saturation on the
surface, TCE conversion continues but chlorine produced subsequently is volatile,
appearing in the gas phase as HCI, Cl2, and COCI2. We have estimated the available
surface area of the titania in the reactor by measuring the amount of titania loaded onto
the support through a weight gain measurement, and then assuming a nominal surface
area of 80 m2/g (manufacturer's specification). Comparison of this estimate to the
integrated amount of chlorine converted through TCE destruction over the six day
induction period reveals that breakthrough occurs after about one monolayer equivalent
of CI is produced through dechlorination of converted TCE.
The origin of the gradual decline in selectivity and conversion after the point of CI
breakthrough is unknown. Besides the implied chemical changes of the active
photocatalyst surface nature, we are aware of an additional systematic change in the
reactor system that could conceivably be responsible for degradation in performance:
the UV lamp output gradually decreases with lamp illumination time. We have not
attempted to quantify this degradation in lamp output. However, the measured UV
intensity at the end of the 30 day run was approximately 5% lower than the value
measured at the start of the run. For purposes of cost estimation, commercial suppliers
typically assume a 6 month replacement interval for UV lamps.
30
-------
100
©
o
c
JS
re
CO
o
k.
O
w
x
80
70
60
50
40
30
20
10
0
Conversion
Selectivity
CI balance
-H-
¦ ¦ ¦
¦ ¦ |
¦ ¦ ¦
¦ ¦
10 20
Time On Stream (days)
—i
30
Figure 5-9. TCE conversion, selectivity, and CI atom balance versus time on
stream.
31
-------
Pilot-Scale Experiments
Set 1: Reactor Configuration 1
At the pilot-scale, we have the capability to control UV intensity (by controlling the
illumination of different numbers of installed lamps) and the gas flow rate. Because the
system employs ambient air, the inlet VOC concentration, the inlet relative humidity and
the average reactor temperature are dependent on the prevailing ambient conditions
and therefore cannot be tightly controlled by the experimenter. An in-line dehumidifier
was used to reduce the incoming humidity on some of the runs. However, the operator
does not have complete independent control over temperature and RH, since the
operating dehumidifier reduces the moisture content but also increases the air
temperature slightly. Thus the temperature and RH are coupled in these pilot-scale
tests. It is therefore not possible to design the type of well-defined experiments
employed at the bench-scale level. Nonetheless, the pilot-scale experiments provide
valuable information on the performance capabilities of different configurations. VOC
concentration measurements are made off line by collecting a gas sample and then
immediately analyzing the sample with a stand alone GC. Maximum sample holding
times were met in all cases.
Table 5-9 summarizes TCE conversion performance for a set of short term, controlled
pilot-scale runs at Zentox's production facility with the packed bed, parallel flow PCO
reactor configuration (Reactor Configuration 1). Note that the temperature reported
refers to the ambient air temperature, not the temperature of the reactor, which should
be substantially higher (20 to 40 °C.) due to thermal energy dissipation of the UV
lamps.
Table 5-9. TCE Conversion Results for Reactor Configuration 1
UV
Air Flow
Temp.
TCE
RH
TCE
Photon
i No.
Lamps
(scfm)
(°C.)
inlet
(%)
conversion
efficiency
(PPmv)
(%)
(%)
1
22
28
21
22
76
55
1.2
2
22
28
33
19
23
90
1.7
3
22
28
24
21
45
90
1.9
4
22
20
24
31
45
96
2.1
5
22
38
24
15
44
90
1.9
6
0
39
22
67
68
5
-
7
22
39
22
69
68
77
7.4
8
22
39
32
64
25
90
8.0
32
-------
Several conclusions can be made on the basis of these results. Assuming that
temperature has only a modest effect on conversion based on the bench-scale RSM
studies, one can compare runs with otherwise similar or identical UV intensities, air
flows, and TCE inlet concentrations to verify the water vapor inhibition effect previously
identified at the bench-scale. Indeed, comparison of Run 1 with Run 2, and
comparison of Run 7 with Run 8 shows that the inhibition of the VOC conversion rate
by the presence of water vapor is substantia! at the pilot-scale. The inhibition effect
appears to be more severe at the lower TCE concentrations (Runs 1 and 2). Second,
higher destruction efficiencies are achieved for lower air flow rates (compare Runs 3
and 4). This behavior is essentially a residence time effect expected for rate-based
processes.
Apparent photon efficiency is a performance metric adopted by PCO equipment
suppliers to allow comparison of the inherent energy efficiencies of different
photoreactor designs. Apparent photon efficiency is estimated by assuming that the UV
lamps are themselves 20 % efficient (a 40 W bulb emits 8 Wof useable UV energy).
The integrated or average efficiency is estimated as the global molar VOC destruction
rate divided by the assumed UV delivery rate, times 100%. The results in Table 9
shows that this reactor configuration uses approximately 2% of the incident photons at
low VOC concentrations, but that the photon utilization increases to 7-8% at high VOC
concentrations. This behavior is consistent with that observed in our independent
intrinsic kinetics studies, in which the TCE destruction rate is dependent on incident UV
intensity to the !4 power at low VOC concentrations, with a shift to a first order
dependence on UV intensity at high VOC concentrations.
Although this reactor configuration yielded acceptable performance characteristics with
respect to throughput and light utilization, the packed bed design used a support
material that was difficult to coat with catalyst, and the loosely packed bed was
cumbersome to exchange if a catalyst replacement was necessary or desired.
Moreover, because the support was organic polymer based, it tended to degrade under
UV irradiation. An improved reactor configuration was therefore designed, constructed
and tested based on experience gained in the laboratory and in these pilot-scale tests.
Set 2: Reactor Configuration 2
Table 5-10 summarizes TCE conversion performance for a set of short term, controlled
pilot-scale runs at Zentox's production facility with a layered, perpendicular flow PCO
reactor (Reactor Configuration 2). This design employed consecutive, alternating
banks of lamps and catalyst-coated panels. The coated panels in this reactor can be
easily exchanged by sliding them through an access door on the side of the reactor.
33
-------
In a general qualitative sense, this reactor configuration exhibits the same performance
behavior as the first reactor configuration. Water vapor inhibits the VOC destruction
rate, as evidenced by the somewhat lower TCE conversion in Run 17 versus Run 18
(40% vs. 20% RH, respectively), and by the dramatically lower TCE conversion in Run
19 versus Run 20 (70% vs. 21% RH, respectively). Note that the actual effect of RH is
partially masked by the fact that the temperatures were somewhat different in these
runs (higher in Run 18 vs. Run 17, and higher in Run 20 vs. Run 19); TCE
concentrations were slightly different in the respective pair of runs compared. The
bench-scale SRM studies showed that TCE conversion decreases modestly with
increasing temperature. On this basis one can conclude that the observed conversion
inhibition is largely attributed to differences in RH.
Lower flow rates (longer residence times) yield higher TCE destruction efficiency. The
integrated apparent photon utilizations are higher for runs in which the inlet VOC
concentration is higher.
Table 5-10. TCE Conversion Results for Reactor Configuration 2
Inlet
TCE
TCE
Photon
n No.
UV
Air Flow
Temp.
inlet
RH
conversion
efficiency
Lamps
(scfm)
(°C.)
(PPmv)
(%)
(%)
(%)
10
18
28
28
85
24
90
9.4
11
18
43
29
55
24
81
8.4
12
18
27
23
81
47
85
8.0
13
18
43
24
29
45
77
4.3
14
18
22
24
29
43
89
2.5
15
18
41
26
53
60
62
6.0
16
18
25
26
76
45
87
7.1
17
46
50
29
45
40
91
3.5
18
46
50
32
46
20
99
3.9
19
50
75
27
52
70
36
2.2
20
50
75
41
49
21
92
5.3
A direct performance comparison of the two reactor configurations is not possible since
the reactors were not tested under identical conditions of flow rate, VOC concentration,
relative humidity, and temperature. With this disclaimer in mind, Figure 5-10 is a visual
34
-------
comparison of the effective photon utilization in each configuration versus inlet VOC
concentration. Although there are significant run-to-run variations in process
parameters, the figure shows that in terms of effective photon utilization, the two
configurations behave roughly the same at equivalent inlet VOC concentrations. The
lower fabrication cost and simpler maintenance make the second configuration a
superior design.
Zentox has field-tested this design at a U.S. Navy site containing air contaminated with
nitroglycerin vapor. Currently, two pilot-scale units have been constructed that are
capable of treating air flows of 25 and 100 SCFM, respectively. Zentox will continue to
seek suitable sites for field trials and eventual full-scale application of PCO technology.
10
o Configuration 2
O Configuration 1
+
+
+
+
+
+
+
+
0 10 20 30 40 50 60 70 80 90
Inlet VOC Concentration (ppm)
Figure 5-10. Integrated Photon Efficiency versus Inlet TCE Concentration for the
two pilot reactor configurations.
35
-------
Chapter 6
References
1. Dibble, L. A., and Raupp, G. B., Catal. Lett. 4, 345, (1990).
2. Raupp, G. B. and Dibble, L. A., Gas-Solid Photocatalytic Oxidation of
Environmental Pollutants, U.S. Patent No. 5,045,288, September 3, 1991.
3. Dibble, L. A., and Raupp, G. B., Environ. Sci. Technol. 26, 492, (1992).
4. Dibble, L. A., Ph.D. Dissertation, Arizona State University, 1989.
5. Phillips, L. A., and Raupp. G. B„ J. Mol. Catal. 77, 297, (1992).
6. Raupp, G. B., and Junio, C. T., Appl. Surf. Sci. 72, 321, (1993).
7. Raupp, G. B., J. Vac. Sci Technol. B 13(4), 1883, (1995).
8. Peral, J., and Ollis, D. F„ J. Catal. 136, 554, (1992).
9. Sauer, M. L„ and Ollis, D. F„ J. Catal. 158, 570, (1996).
10.Berman, E., and Dong, J., in "The Third International Symposium on Chemical
Oxidation: Technology for the Nineties. Vanderbilt University. Nashville. Tennessee.
1993" (W. W. Eckenfelder, A. R. Bowers, and J. A. Roth, Eds.) pp. 183-189.
Technomic Publishing, 1993.
11.Miller, R., and Fox, R., in "Proceedings, First International Conference on Ti02
Photocatalytic Purification and Treatment of Water and Air, London, Ontario,
Canada. 8-13 Nov. 1992" (D. F. Ollis and H. Al-Ekabi, Eds.) Elsevier, Amsterdam,
1993.
12. Raupp, G. B., Nico, J. A., Annangi, S., Changrani, R., and Annapragada, R., AlChE
J. 43, 792, (1997).
13. Annapragada, R., Leet, R., Changrani, R., and Raupp, G.B., Environ. Sci. Technol.
31, 1898 (1997).
14.Hung, C.-H., and Marinas, B. J., Environ. Sci. Technol. 31, 1440 (1997).
15.Misra, D. N„ Nature 240, 14 (1972).
36
-------
16.Ameen, M. M., Burrows, V. A., and Raupp, G. B., submitted to J. Catal. (1998).
17.Nimlos, M. R., Jacoby, W. A,, Blake, D. M., and Milne, T. A,, Environ. Sci. Technol.
27, 732 (1993).
18. Jacoby, W. A., Nimlos, M. R., Blake, D. M., Noble, R. D., and Koval, C. A., Environ.
Sci. Technol. 28,1661 (1994).
19.Berman, E., and Dong, J., Proc. 1st Intl. EPRl/NSF Symp. Adv. Oxid., EPRI TR-
102927-V2, p. 2-19 (1993).
20. Jacoby, W. A., "Gas Phase Destruction of TCE and PERC: Factors Affecting
Products and their Time Dependence", Photocatalytic Oxidation Research Review
Meeting, NREL/CP-471-20577, Copper Mountain, CO, October 1995.
21.Sauer, M. L., Hale, M. A., and Ollis, D.F., J, Photochem. Photobiol. A 88, 169
(1995).
37
-------
Appendix
Bibliography
Review Articles
Hoffmann, M. R., Martin, S. T., Choi, W., and Bahnemann, D. W., Environmental
Applications of Semiconductor Photocatalysis, Chem. Re v. 95, 69-95 (1995).
Linsebigler, A. L., Lu, G., and Yates, J. T. Jr., Photocatalysis on Ti02 Surfaces:
Principles, Mechanisms, and Selected Results, Chem. Rev. 95, 735-758 (1995).
Peral, J., Domenech, X., and Ollis, D. F., Heterogeneous Photocatalysis for Purification,
Decontamination and Deodorization of Air, J. Chem. Technol. Biotechnol. 70, 117
(1997).
Bibliographies
Blake, Daniel M., Bibliography of Work on the Photocatalytic Removal of Hazardous
Compounds from Water and Air, NREL/TP-430-6084; DE94006906. National
Renewable Energy Laboratory, Golden, CO., May 1994.
Blake, Daniel M., Bibliography of Work on the Photocatalytic Removal of Hazardous
Compounds from Water and Air; Update No. 1, NREL/TP-473-20300; DE95013148.
National Renewable Energy Laboratory, Golden, CO., October 1995.
Blake, Daniel M., Bibliography of Work on the Photocatalytic Removal of Hazardous
Compounds from Water and Air; Update No. 2, NRELH"P-430-22197; DE97000084.
National Renewable Energy Laboratory, Golden, CO., January 1997.
38
-------
TECHNICAL REPORT DATA
xtdmtjt r-TM Kin Am (Pltme rt«i Immicittm a* iht revtm befort tomph
1. REPORT NO. a.
EPA 540/SR-98/504
a
1
4. TITLE ANO SUBTITLE , . , . . „ fti„
Integration of Photocatalytic Oxidatiqi wi-th A*r
Stripping of Contaminated Aquifers .
9. REPORT OATE
Tlonomhpr 1 QQR ,
t. PERFORMING ORGANIZATION CODE
l^aTn'i'sh Changrani, Gregory B. Raupp and
Craig Turchi
8. PERFORMING OROANIZATTION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Arizona State University
Tempe, AZ 85287-6006
10. PROGRAM ELEMENT NO.
TD1Y1A
11. CONTRACT/GRANT NO.
CR 821100-01-0
12. SPONSORING AGENCY NAME ANO AOORESS
National Risk Management Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency, Cincinnati
OH 45268
13. TYPE OF REPORT AND PERIOD COVERED
SITE Emerqina Report
14. SPONSORING AGENCY CODE
EPA?600/14
15. SUPPLEMENTARY NOTES
Project Officer: Norma Lewis - 513-569-7665
16. ABSTRACT
In a recently completed test program, bench scale laboratory studies at Arizona State University
in Tempe, Arizona, and pilot scale studies in a simulated field-test situation at Zentox Corporation
in Ocala, Florida were performed to evaluate the integration of gas-solid ultraviolet (UV)
photocatalytic oxidation (PCO) with air stripping of an aquifer contaminated with chlorinated
volatile organic compounds (VOCs). Chlorinated ethylenes such as trichloroethylene (TCE) can
be destroyed in a wide process window, although chlorinated ethanes such as trichlorethane
(TCA) are non-reactive. Water vapor significantly inhibits the chlorinated ethylene destruction
rate. For this reason, PCO units should be placed downstream of a dehumidification unit located
between air strippers and the PCO unit, with targeted reduction of the relative humidity in the
contaminated air stream to less than 50%. Principal carbon-containing products of PCO identified
experimentally at the bench scale include carbon dioxide, carbon monoxide, and phosgene
(COCl2). Failure to close carbon mass balances under some process conditions suggests that not
all byproducts were identified. Further studies are needed in this area. A panel bed was identified
as the preferred photoreactor configuration. This unit is characterized by simplicity of
construction, ease of maintenance, and high UV photon utilization efficiency.
t?. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b.lOENTIFIERS/OPEN ENOED TERMS
c. COS ATI FieM/Croup
Titanium Dioxide
Oxidation
Photochemical Oxidation
Gas Phase Oxidation
Photocatalytic Oxidation
1«. DISTRIBUTION STATEMENT
Release to Public
IS. SECURITY CJ-AJS m,isReport)
Unclassified
21. NO. OF PAGES
9
20. SECURITY CLASS (Thltpogri
Unclassified
22. PRICE
EPA For* 2220.1 (fUv. 4-77) previous edition is obsolete
------- |