United States
Environmental Protection
Agency
Municipal Environmental
Research Laboratory
Cincinnati OH 45268
Research and Development
EPA-600/S2-84-136  Sept. 1984
Project Summary
Pilot-Scale Evaluation  of
Photolytic  Ozonation for
Trihalomethane  Precursor
Removal

William H. Glaze, Gary R. Peyton, Brigitte Sohm, and Dean A. Meldrum
  Ozone used  in combination  with
ultraviolet (UV) radiation was studied
on  the pilot  scale  for  removing
trihalomethane (THM) precursors from
potable water. The effects of varying
parameters such as ozone dose rate and
UV intensity were first studied using a
synthetic feedwater reconstituted from
a humic concentrate. This concentrate
was  obtained by sodium hydroxide
elution of  granular activated carbon
(GAC) columns used for the adsorption
of humic material from  water in a
previous project. The pilot plant was
then  operated for  8 months at the
Sabine River Water Treatment Plant in
Longview, Texas.
  A laboratory investigation was also
conducted  on the mechanism of the
ozone/UV  process.  Kinetic  analysis
was performed on data collected during
ozone photolysis experiments in the
presence  and  absence  of  known
hydroxyl radical scavengers. This anal-
ysis indicates that ozone photolysis in
aqueous solution leads directly to the
formation of hydrogen peroxide, which
then   produces hydroxyl  radical by
secondary  reaction of peroxy  anion
(HO2~)  and subsequent species with
ozone.
  The mechanistic  results predict a
maximum yield of % hydroxyl radical
for each ozone molecule  used in the
proposed mechanism. The mechanistic
results also predict a plateau in treat-
ment efficiency as a function of the
fraction of ozone photolyzed and thus
as a  function of UV power input.  A
simple model was developed for ozone
mass  transfer  with  chemical  and
photochemical reaction. Correlation of
this model with mechanistic and pilot
plant data is ongoing.
  Operating  data from the pilot plant
and analysis of the residual trihalometh-
ane formation potential (THMFP) after
treatment  provided the  basis  for
projecting costs for removing THMFP
by photolytic ozonation. At a 1-mgd
plant capacity and an initial THMFP of
300 fjg/L, projected treatment costs
ware $0.55, $0.70, and  $0.91  per
thousand  gallons for 60%, 70%, and
80%  THMFP removal,  respectively,
assuming an electrical cost of $0.10 per
kWh.  At  this   capacity, projected
treatment costs were still decreasing
rapidly with increased plant capacity.

  This Project Summary was developed
by EPA's Municipal Environmental
Research  Laboratory. Cincinnati, OH,
to announce key findings of the research
project  that  is fully documented in a
separate report of the same title (see
Project Report ordering information at
back).

Introduction
  The chlorination of natural water for
potable  use is now  known to produce
trihalomenthanes  (THM's), which  have
been  shown  to  be carcinogenic in
laboratory animals. A  preliminary
laboratory study reported by this research
group  in  1980  (EPA-600/2-80-110)
showed  that  the potential of  natural
waters  for  producing   THM's  upon
chlorination could essentially be elimi-

-------
•*''"f
   •A
    nated by simultaneous treatment with
    ozone  and  ultraviolet (UV)  radiation.
    However, the ozone doses required in the
    laboratory experiments were rather high.
    The present report describes an attempt
    to  understand and refine the ozone/UV
    system;  also described is  a pilot-scale
    evaluation using photolytic ozonation to
    destroy THM precursors in  water.
      The project  was  conducted in four
    parts: (1) a  laboratory and theoretical
    investigation of the mechanism of photo-
    lytic ozonation, (2) construction and initial
    testing of the pilot plant using a reconsti-
    tuted lake water concentrate for f eedwa-
    ter consistency, (3) continuous operation
    of the pilot plant using feedwater from the
    Sabine  River  Water Treatment  Plant
    located  in Longview, Texas,  and  (4) a
    laboratory study of byproducts formed
    upon photolytic ozonation  of  natural
    waters. In addition to these four phases, a
    simple model  for ozone mass transfer
    with photochemical  reaction  has  been
    developed to use with the mechanistic
    results for interpreting the experimental
    results.  Data obtained during pilot-scale
    operation were used in  treatment cost
    projections for a full-scale  system.

    Procedures

    Laboratory Study
      Ozone photolysis was studied in the lab-
    oratory by monitoring the accumulation
    of  ozone, hydrogen peroxide, and total
    oxidants  in a  laboratory scale, stirred-
    tank reactor, with and without irradiation
    by UV light and in  the presence and
    absence  of  known  hydroxyl  radical
    scavengers. The experiment was set up
    so that the ozone concentration in the gas
    stream into and out of the  reactor  could
    be measured. The analyses  indicatedthat
    in the absence of organic compounds, the
    total oxidant concentration was equal to
    the  sum  of the ozone  and  hydrogen
    peroxide  concentrations.  Ozone  was
    measured  using   indigo  disulfonate
    bleaching;  hydrogen peroxide  was
    measured by the colorimetric detection of
    the peroxide-titanium complex, and total
    oxidants by molybdate-catalyzed iodimetry.

    Pilot-Scale Studies
      Pilot studies were conducted in a pilot
    plant housed in a 42-ft-long drop-frame
    trailer. The  plant had been previously
    used in  a pilot  study of ozonation  com-
    bined with  GAC  for THM  precursor
    removal. This facility was  moved to the
    SumX  Corporation  in Austin, Texas,
    where it was modified by the addition of
    UV lamps to the existing ozone contactor
and the addition of tubular photochemical
reactors (bubble columns with UV lamps
installed  in  quartz  lamp  wells)  for
comparison studies.
   Operating parameters were varied in
the initial  phase, which used synthetic
lake   water  reconstituted  from   a
concentrate.  The  concentrate  was
obtained by sodium hydroxide elution of
several GAC columns that had been used
for several weeks to treat water with a
high   natural  organic  content.  The
consistency of feedwater provided by this
concentrate  during  a  several-month
study  to determine optimum treatment
parameters  outweighed  arguments
about  the  concentrate's representative
nature. A stirred-tank  reactor equipped
with a turbine  contactor was compared
with the tubular photochemical reactors
(TPR's) while the ozone dose rates and UV
intensities  were varied in both.
   Following the study to determine opti-
mum  treatment  parameters,  the  pilot
plant  was  moved to the Sabine River
Water Treatment  Plant in  Longview,
Texas, and operated for 8 months. Feed-
water  was of  two types:  alum-settled
water  before and after lime clarification.
The stirred-tank reactor and the  TPR's
were compared while ozone dosages and
UV intensities  were varied.  Treatment
effectiveness of various configurations
was evaluated  by withdrawing  water
samples after  treatment,  chlorinating
them for 7 days in the laboratory at  an
applied chlorine dose  of 30 mg/L, and
measuring them  for THM formation  by
liquid-liquid extraction and gas chroma-
tography/electron   capture   detection
(GC/ECD) analysis.


Byproduct Studies
   Treated and untreated water samples
from the pilot plant were passed through
columns of  the  macroreticular resins
XAD-4 and XAD-8 to trap organic bypro-
ducts of photolytic ozonation. In the first
of two  experiments,  the  water was
acidified to  pH  1.8,  a  water-to-resin
•volume ratio of 20 was used, and the
resin  was  eluted  with methanol.  To
derivatize  carboxylic acids,  the sample
was evaporated  to  dryness by gentle
warming in a stream of inert gas after
adding tetrabutylammonium hydroxide
(TBAH) to  pH 9. The residue was taken
back up in  a small quantity of acetonitrile
containing a  slight excess  (based  on
TBAH) of ethyl bromide and allowed to
stand  at room temperature  for 1 hour or
longer. This procedure produces the ethyl
esters of carboxylic acids.
  In the second experiment, the water
was passed through the resins after it
was adjusted to pH 10, the resin was
eluted with ethyl ether, the water was
adjusted to pH 2 and put back through the
column, and the column was again eluted
with  ether.   Extracts  from  both
experiments were  analyzed by capillary
gas  chromatography/flame  ionization
detection  (GC/FID)  and  by  gas
chromatography/mass spedrometry
(GC/MS) at overall concentration factors
of 1300 and 1200, respectively, for the
first and second experiments.
Results and Discussion

Laboratory Study of
Active Species
  The main thrust of this portion of the
project was to identify the principal active
specie(s) in photolytic ozonation  and, if
possible,  to  determine the  reaction
mechanism.  The  primary  barrier  to
elucidating  the  mechanism   was
identifying the products of the first step,
ozone photolysis. The literature indicated
that this reaction would produce either
two  hydroxyl  radicals  or  hydrogen
peroxide, followed in  either case  by a
complex network of secondary reactions.
  Ozone  photolysis  studies  were
designed to differentiate between the two
possibilities  for the  primary step  in
photolytic  ozonation. The studies led  to
conflicting results when two  different
known hydroxyl radical scavengers were
used. When 0.015M acetic acid/acetate
was  present,  hydrogen  peroxide
accumulated in  solution  upon continued
ozonation and irradiation at  pH values in
the range of 3  to 8.  However, peroxide
levels were too low to be quantified in the
presence of 0.015M sodium bicarbonate
at pH 7. Experiments run in distilled water
with no scavenger  added gave results
intermediate between the two scavenger
experiments, indicating the presence  of
some additional complicating effect  in at
least  one set  of experiments. Kinetic
analysis of the complex reaction system,
using rate constants from the literature
and assuming first one initiation step and
then the other,  indicated that the initial
step of the reaction  mechanism is the
photolysis of aqueous ozone to produce
hydrogen peroxide. The ensuing second-
ary reaction system is primarily that of the
ozone/hydrogen peroxide system.  This
system  includes the  dissociation  of
hydrogen  peroxide to yield the  anion,
which  in turn reacts with ozone to yield
hydroxyl  radical and  superoxide (O2~).

-------
Superoxide  reacts  very  quickly  with
additional  ozone  to  yield  ultimately
another hydroxyl radical. Thus in the net
reaction, three ozone molecules and one
photon yield two hydroxyl radicals. The
implication   of this  net  reaction
stoichiometry is that, in a system where
the  hydroxyl  radical  is  effectively
scavenged by  the  substance  to  be
destroyed  (organic   substrates),   the
optimum   hydroxyl  radical   yield  is
achieved  when  one-third  of  the ozone
transferred into  solution is photolyzed.
On the other  hand,  if  ozone  photolysis
resulted directly  in the production of two
hydroxyl  radicals, the maximum  yield
would be realized when all  the trans-
ferred ozone was photolyzed.


Pilot Studies
  The  studies  to determine  optimum
operating parameters indicated that the
stirred-tank reactor  performed  better
than the tubular photochemical reactors,
even  on the  basis of ozone consumed.
Thus  the greater effectiveness  in  the
former case was not  simply the result of
better mass transfer. These studies also
indicated  that an optimum level of UV
power existed, above which performance
increased very little. These data support
the proposed mechanism.
  The  studies of optimum  operating
parameters also  indicated that  in  the
stirred tank with 0.1 to 0.3  watt/L of
applied UV radiation, an ozone dose of 30
mg/L was required to reduce the  total
trihalomethane  formation  potential
(TTHMFP) of the reconstituted lake water
concentrate to 1/e (37%) of its original
value.  (Original  THMFP values ranged
from 106 to 192 /jg/L in this phase of the
study).  In the  tubular photochemical
reactors, 40 mg/L ozone was required to
achieve  the  same  extent  of removal,
using applied UV radiation of 0.5 to 0.7
watt/L of reactor volume.lt was found in
most   cases  that  upon ozone/UV
treatment, the THMFP first  increased
from its initial value, then dropped expo-
nentially with continued treatment. This
is attributed to the formation of precursor
sites from nonprecursor organic material.
During pilot operation in Longview, the
stirred tank also proved superior to the
tubular  photochemical  reactor.  In an
attempt to test whether gas-phase ozone
photolysis was detrimental to process
efficiency,  delayed  irradiation  (Dl)
configurations were also tested. In these
configurations,  ozone  contacting  took
place in the stirred tank reactor, followed
by ozone photolysis in a separate vessel.
The Dl configurations were less effective
than  simultaneous  ozone  contacting
and  photolysis in  the  stirred tank, but
more effective than the tubular reactors
alone, based on the  amount of ozone
removed from the gas stream. The ozone
doses required for TTHMFP removal to
1/e  of the  original value are listed in
Table 1, along with those required for the
reconstituted  lake  water used  in the
parameter  studies. Correspondence of
results  obtained on the two  different
feedwaters is  good when expressed in
terms of the ozone dose required to reach
a particular extent  of removal. Note that
the desired TTHMFP  removal could be
reached using ozone alone, but that the
required ozone dose is prohibitive.
  The pilot-scale operation at Longview
also  snowed that  the  lowest dose rate
that  could achieve the desired removal
proved  to  be  the most efficient.  In
addition, the efficiency with which ozone
was  used in the ozone/UV reactions was
relatively  insensitive  to  pH and other
matrix components in  clarified, filtered
Sabine River water.

Cost Analysis
  Data for removing 60%, 70%, and 80%
of the TTHMFP in the optimum configura-
tion was used  to project an approximate
treatment cost, using cost data supplied
by the  U.S.  Environmental Protection
Agency (EPA) for small systems (0.1 and
0.5 mgd). These data were extrapolated to
other system sizes using the equation T —
LQ n, where T is the cost, Q is the plant or
unit capacity, and L and n are parameters
evaluated  by fitting the equation to the
EPA-supplied  data  for   capital  and
operating costs for  ozone generation and
feed,  UV light, stirring, etc., at 0.1- and
0.5-mgd capacities. Capital costs were
subjected to a multiplicative factor of 1.4
to cover engineering, contractors, over-
head and profit, site preparation, interest
during construction, and legal and admin-
istrative fees. An interest rate of 8% was
assumed over 20 years.
  Cost  projections for various  levels of
removal strongly  depended on plant
capacity in the 0.1 - to 0.5-mgd range, and
a  method  was sought  to   estimate
treatment costs over a larger  range of
plant size.  To  assess the validity of
extrapolation to larger systems, suppliers
of ozonation equipment were contacted
and  asked about ozone generation  and
feed equipment costs for systems in the
size  range of 20 to  1000 Ib/day.  The
hardware costs thus obtained were fit to
the  same  exponential equation given
above and then extrapolated upwards (in
plant capacity) for comparison  with the
installed costs for actual large  systems
installed  in  recent  years.  A  ratio of
hardware costs to installed costs for large
systems was obtained from this compari-
son. When this ratio was applied to the
manufacturer-supplied hardware costs
applicable to 0.1 - and 0.5-mgd plants, the
resulting  value  predicted the  EPA-
supplied cost figures (including the factor
of 1.4) within 10% to 15%, thus indicating
the validity  of the above extrapolation.
Projected treatment costs  are  given in
Table 2 for  0.1- to 5.0-mgd capacities.
Treatment costs are still strongly  capacity-
dependent  in  the  range  of  5.0-mgd
capacity.

Byproduct Study
  The methanol extract derived  from the
first experiment was analyzed by GC/FID
andGC/MS. Several peaks were reduced
to a few percent of their original size by
ozone/UV treatment (17.7 mg/L of ozone
used),  and  only two new  peaks were
found. The latter corresponded with 9 and
Table 1.    Ozone Dose Required to Reduce TTHMFP of Sabine River Water and Reconstituted
           Lake Water to 1/e of Its Original Value*

                                                      Ozone Dose
Reactor
Stirred Tank
Stirred Tank
Tubular
Stirred Tank
Mode
Oi/UV-CPj
O3/UV- Dl\
O3/UV-CP]
Ozone only
Sabine River Water,
Longview, Texas
27 mg/L
34 mg/L
41 mg/L
48 mg/L
Reconstituted
Lake Water
30 mg/L
§
40 mg/L
§
* 1/e = 37%; initial THMFP values ranged from 106to 192 g/L in reconstituted lake water and 120
  to 437 g/L in Sabine River water.
f CP = simultaneous ozone contacting and photolysis.
t Dl = ozone contacting followed by photolysis in a separate vessel.
§ Not run.

-------
Table 2.   Projected Treatment Costs for THM Precursor Destruction Using Photolytic Ozonation*

                                         Dollars/1000 gallons
Plant Capacity
0. 1 MOD
0.5 MOD
1.0 MOD
5.0 MOD
60% Removal
$2.16
0.83
0.55
0.21
70% Removal
$2.43
1.02
0.70
0.29
80% Removal
$2.75
1.27
0.91
0.42
 * An electrical cost of $0.10 per k Wh was used, but costs are only 5% to 10% lower at an electrical
 cost of $0,05 per kWh.
2 //g/L, assuming compound response
factors equal to that of the internal stand-
ard d,0-anthracene. In the second experi-
ment,  GC/FID analysis of the  ether
extract revealed several compounds in
the treated water that were not present in
the untreated sample. The concentration
of these compounds  ranged from  a few
/ug/L to several tens of /ug/L, assuming
response  factors equal to that of  the
internal standard. No GC/MS identifica-
tion of these compounds has yet to be
made.
  Also in  that sample,  a small group of
peaks  present  in  the  untreated  water
disappeared upon treatment, reconfirm-
ing the ability of the ozone/UV process to
remove trace  quantities of potentially
harmful  organic  pollutants  during
treatment  for  THM precursors.
Comparison with the internal standard
indicated that the compounds were origi-
nally present in the water at less than 1
ppb. The mass spectra indicated that the
compounds were probably a homologous
series of alkyl benzenes. Further analysis
of  the  GC/MS data reported  here  is
ongoing.
Conclusions
  The  laboratory study shows that the
primary  mechanism  of  photolytic
ozonation with UV irradiation at 254 nm
is the photolysis of ozone  to  produce
hydrogen peroxide, which  dissociates to
some degree to produce  peroxy anion.
The reaction between peroxy anion (and
in a later step,  superoxide) and ozone
produces hydroxyl radical, which is the
primary active species of the ozone/UV
process. The above conclusions predict
that (1) the maximum yield of hydroxyl
radical per ozone molecule is 2/3, and (2)
when more than V3 of the available ozone
is photolyzed, little  or no  increase in
hydroxyl  radical  production  should  be
obtained.
  In the pilot-scale study,  an ozone dose
of 27  mg/L (transferred  from  the gas
phase  to the  liquid)  was sufficient to
reduce the THMFP of clarified and filtered
Sabine River water to 1/e (37%) of its
original value (300 /ug/L) using a stirred
tank reactor  with  0.27 watts of UV/L
applied in the reactor. If UV application
was delayed  until after  initial ozone
contacting, or if the stirring or  UV was
omitted, required dosages (still based on
that amount of ozone  transferred out of
the gas)  were 34, 41, and  48 mg/L,
respectively.   The  experimental  data
supported the  concept of  minimal in-
crease in treatment efficiency  above  a
certain UV intensity. The lowest rate of
ozone addition that effected the desired
THMFP removal was the most efficient in
terms of ozone used.
  The projected treatment  costs of the
process at a  1-mgd plant for  60%, 70%
and 80% THMFP removals from clarified
Sabine River water were $0.55, $0.70,
and $0.91 per thousand gallons, respec-
tively, at an  assumed electrical cost of
$0.10  per kWh. Treatment  costs still
strongly depended on plant capacity at
this size.  The  increase  in  process
efficiency by  the addition of stirring and
UV greatly outweighed the increased cost
over  ozonation  alone  in these  cost
projections. The dominant term  (65%) in
the  projected  treatment   cost  is the
amortized capital cost of the ozone gen-
eration and feed equipment.
 Recommendations
   Further laboratory-scale investigations
 should  be made of  several points that
 were either  beyond the  scope of this
 study or emerged too late  in the study to
 be systematically investigated. Ozonation
 of natural organics  in water should be
 studied to determine how or whether it is
 possible to suppress the formation of new
 THM  precursor  material during  the
 destruction of that material initially pres-
 ent. If that could  be accomplished, the
 possibility would exist for considerably
 lowering the  ozone dose requirement for
 satisfactory THMFP removal. The model
 developed for mass transfer with simul-
 taneous  chemical  and  photochemical
 reaction should be verified in the labora-
tory  under  conditions  where careful
oxidant measurements can be made. The
use of a simple model substrate rather than
natural organic material would increase
the probability of success, as would the
use of a photolyte that is less susceptible
to side reactions than is ozone.  Finally, a
survey of ozone dose requirements for
water samples from different geographi-
cal locations would provide  information
concerning the transferability of projec-
ted  treatment  costs  from  one  water
source to another.
  Two further   studies  are necessary
before  photolytic   ozonation  can  be
recommended   as  a  potable   water
treatment process. Thorough toxicity and
mutagenicity testing should  be done on
concentrates of treated  and untreated
water  from  several  different   water
sources to help ensure that public health
problems do not arise later. The process
itself should be tested at the field scale to
verify ozone and UVdose requirements in
larger systems and  to  address  the
problems associated with scale-up of the
complicated,  photochemically  initiated
reaction.
  The full  report  was  submitted  in
fulfillment  of  Cooperative  Agreement
CR808825 by the University of Texas at
Dallas under the sponsorship of the U.S.
Environmental Protection Agency.
*USGPO:  1984-759-102-10686

-------
     William H. Glaze, GaryR. Peyton, andBrigitte Sohm are with University of Texas at
      Dallas, Richardson,  TX 75083; and Dean Meldrum is with SumX Corporation,
      Austin.  TX 78761
     J. Keith Carswell is the EPA Project Officer (see below).
     The complete report, entitled "Pilot Scale Evaluation ofPhotolytic Ozonation for
      Tnhalomethane Precursor Removal." (Order No. PB 84-234 517; Cost: $20.50.
      subject to change) will be available only from:
            National Technical Information Service
            5285 Port Royal Road
            Springfield. VA 22161
            Telephone. 703-487-4650
     The EPA Project Officer can be contacted at:
            Municipal Environmental Research Laboratory
            U S  Environmental Protection Agency
            Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
                                                                              AGtMCY
                                                          1L
                                                                     oTwetT

-------