Formation of Condensation Products in Advanced Oxidation

Formation of Condensation Products in Advanced Oxidation Technologies: The Photocatalytic ... Environmental Science & Technology 2002 36 (16), 3618-36...
0 downloads 0 Views 989KB Size
Environ. Sci. Techno/. 1995, 29, 2226-2234

Formation of Condensation Products in Advanced Olridation Technologies: The Photocatsllytir: Degradation of Dichlmphenols on Ti02 CLAUD10 MINERO,+ E Z I O P E L I Z Z E T T I , * # +P I E R R E P I C H A T , $ MICHELA SEGA,+ AND MARC0 VINCENTI’ Dipartimento d i Chimica Analitica, Universitd degli Studi d i Torino, via P. Giuria 5, 10125 Torino, Italy, and Ecole Centrale de Lyon, Ecully, France

Photocatalytic treatment of dichlorophenols (DCPs) leads to the formation of condensation products. These products were identified as polyhydroxyPCBs. Depending on the initial DCP isomer, different extents and yields of polyhydroxyPCB formation (up to 1%) and different types of polyhydroxyPCBs were observed. The formation of dehalogenated hydroxyphenols and hydroxytrichlorobiphenyls in the presence of oxygen suggests that, in addition to oxidative pathways, reductive pathways are also operating. In contrast to direct photolysis by UV irradiation, no evidence of hydroxypolychlorodibenzo-p-dioxins or dibenzofurans was found. PolyhydroxyPCBs are destroyed in the same time window as the initial DCP. The similarities between this photocatalytic process and other advanced oxidation processes (AOP) based on *OH radical chemistry are discussed.

Introduction Chemical oxidation of contaminated surface water and groundwater represents an important method for the destruction of biocidal or non-biodegradable pollutants ( 1 , 2). Recent developments in the domain of oxidative degradation include catalytic and photoassisted methods (3, 4).

The use of oxidants for water remediation and disinfection can lead to the formation of byproducts in the treated water through the reaction of the oxidant with organic and inorganic matter (5-7). Some of these byproducts have been found to be of human health concern, and their presence in drinking water has been ascertained. These undesired byproducts will very probably have to be subjected to regulatory laws (8). Examples of undesired processes are the formation of trihalomethanes and chlorinated aromatics during disinfection with chlorine (913);of oxidized products by the interaction of chlorine with +

Universita degli Studi di Torino. Ecole Centrale de Lyon.

2226

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 9. 1995

organics on activated carbon (14);of aldehydes and ketones from ozone treatment (15-1 7); of dibenzofurans, dibenzodioxins,and related compounds from oxidation of phenol in supercritical water (18); of dichlorobiphenyls from chlorobenzenes treated with Fenton’s reagent (19);and of polychlorodibenzo-p-dioxinsand polychlorodibenzofurans by W photolysis of pentachlorophenol (20). In the last decade, several studies have examined Ti02mediated photocatalysis as a possible treatment process for water decontamination (21-23). In this process, light of suitable energy generates electron/hole pairs. Both species can contribute to the organic substrate degradation since the potentials of the hole and the conduction band electron is sufficient to oxidize or reduce many organic molecules (24). In aerated solution, dioxygen is an effective scavenger of the conduction band electrons. Either the holes or the ‘OH radicals formed from the reaction with water or hydroxide ions (24) are the major active species for oxidative degradation. Interestingly, complete mineralization into carbon dioxide and eventually into inorganic ions has been assessed for an extensive range of organic molecules (22). Limited attention has so far been devoted to the formation of dangerous major intermediates during the TiOz-mediated process (25)or to the occurrence of noxious byproducts. Previous works on photocatalytic transformation of dichlorobenzenes (26) and chlorophenols (27) on zinc oxide have shown the formation of traces of chlorohydroxy derivatives of biphenyls. Recently, a communication reported the formation of similar products during the photocatalytic transformation of 1,2,4-trichlorobenzeneon TiOz (28). Respiration rate measurements in activated sludge on wastewater contaminated with 2,4-DCP showed that this compound is toxic and that the activity of the sludge is reduced even more by degradation products formed in the photocatalytic treatment (29). In this paper, we report a detailed examination of the condensation products formed during the photocatalytic degradation on titanium dioxide of the six dichlorophenol isomers, whose complete mineralization by means of photocatalysis has already been reported (30).

Experimental Section Reagents and Materials. The six dichlorophenol isomers (2,3-,2,4-,2,5-,2,6-,3,4-,and 3,5-dichlorophenols)(DCPs) were purchased from Aldrich and analyzed by gas chromatography/mass spectrometry to determine their impurities. DCP purities ranged from 97% to 99%. The main impurities found were DCP isomers, whereas no PCBs nor their hydroxylated derivatives were found in the starting materials. The 2,2’,5,5‘-,2,2’,6,6’-, and 3,3’,5,5’-tetrachlorobiphenyls (TCBs) were obtained from Ultra Scientific (North Kingstown, RI). The 1,4,7,8-tetrachlorodibenzodioxin (1,4,7,8-TCDD)was from Cambridge Isotope Laboratories (Woburn, MA). The 4,4’-dihydroxy-3,3’,5,5’tetrachlorobiphenyl was from Eastman Kodak. Dichloromethane for residue analysis from Merck was utilized for extraction procedures. Ti02 Degussa P25 was used as a photocatalyst; in order to avoid interferences from adsorbed organic and inorganic species, the powder was suspended

0013-936W95/0929-2226$09.00/0

G 1995 American Chemical Society

in water (about 10 g L-9, preliminarily irradiated for 24 h using the light source describedbelow, washed several times with bidistilled water, and dried at 80 "C. Irradiation and Sample Treatment. Each DCP was separately dissolved in CH2C12 dispersions of TiOz under sonication. After desiccating the slurry in a rotary evaporator at 20 "C, aqueous suspensions were prepared by dispersing the TiOzpowder loaded with DCP in water. The final concentration of DCPs and the completeness of CH2Clz removal was checked by hexane extraction and GC analysis (see below). Final slurries contained 20 mg L-' (20 ppm) of the substrate and 250 mg L-I TiOz (250 ppm). A different procedure for preparing the samples was also tested. A 40 mg L-' aqueous solution of DCP was mixed with an equal volume of an aqueous suspension of TiOz (500 mg L-l). The two procedures were equivalent, provided that both suspensions were left to equilibrate for 30 min in the dark before being irradiated. The suspension pH at the start of the irradiation experiment was 5.4. Samples of 5 mL of these dispersions or DCP water solutions (in the experimentswithout TiOd were introduced into cylindrical Pyrex glass cells, with flat parallel optical windows (9 = 40 mm, h = 30 mm) and were irradiated for variable time periods (for most substrates 0, 1,3, 6, 10,20, 30, and 60 min) using a 1500-WXenon lamp (CO.FO.MEGRA, Milan) equipped with a 340-nm cutoff filter and simulating AM1 solar light (31). Subsequently, the content of nine cells was collected together (45 mL) and acidified (pH 5 2, HC1). The suspension was then extracted for 40 min using 30 mL of CH2Clzand a continuous liquid extractor. The procedure proved to recover '98% of DCPs, phenol, 1,4,7,8-TCDD, and 4,4'-dihydroxy-3,3',5,5'-tetrachlorobiphenyl(DTCB) from the suspension. A toluene solution (300pL)containing 500 ppb of 2,2',5,5'-TCB, 2,2',6,6'-TCB, and 3,3',5,5'-TCB as internal standards was added to the organic extract for quantitative measurements by mass spectrometry. The organic extract was concentrated to small volume under vacuum and then almost to dryness (about 300 pL) in a conical vial under a gentle stream of argon. Analytical Determinations. Organic solutions were analyzed by gas chromatography/mass spectrometry (GCI MS) in electron impact (EI) and, in part, also in negative ion-chemical ionization (NI-CI), using methane as the moderating gas. GUMS experiments were performed on a Finnigan-MAT 95 Q triple-focusing instrument with magnetic, electrostatic, and quadrupole analyzers mounted consecutively; ions were collected at the first detector located after the electrostatic sector. Samples were injected splitlessly into a Varian 3400 gas chromatograph equipped with a S.G.E. BP-5 30 m x 0.25 mm capillary column. The oven was programmed as follows: isothermal at 90 "C for 3 min, from 90 to 300 "C at 12 "C min-', and isothermal at 300 "C for 10min. In the E1 experiments,the electron energy was set at 70 eV, and the electron current was set at 1 mA; in the CI experiments, they were set at 200 eV and 0.2 mA, respectively. The ion source temperature was maintained at 240 "C in E1 and at 200 "C in CI. Quantitative determinations for condensation products were carried out in E1 by integrating, along the chromatographic peak width, the reconstructed ion current for the masses of the two most abundant isotopic ions within their molecular cluster. This integrated signal was matched against the average corresponding signal of the three isomeric TCBs (2,2',5,5'-TCB, 2,2',6,6'-TCB, and 3,3',5,5'-

*

- A

'

*

*

r

.2110'

I

'

Id00

ld30

1dOO

ld30

20;OO

retention time, min

FIGURE 1. Mass chromatograms of some Condensation products obtained by 6-min irradiation of 20 mg L-' solution of 2,4-DCP in the presence of 250 mg L-I TiOz. PolyhydroxyPCBs(see text and Figure 2) are marked by an asterisk.

TCB) used as internal standards. The response factors for the three standards and 4,4'-dihydroxy-3,3',5,5'-tetrachlorobiphenyl (the only available polyhydroxyPCB standard) were accurately determined. Their relative values varied from 78%to 100%. A response factor equal to that averaged fromthe internal standards was assumed for each individual condensation product. From the instrumental response of the internal standards, a detection limit of 4 ng L-' (4 ppt) was estimated. The molecular weights recorded in this work were calculated taking into account only the contribution of the 35Clisotope.

Results and Discussion The formation of condensation products from irradiated DCPs was confirmed by the present experiments. Condensation products were found in high abundance, particularly in those samples irradiated for short time periods. In the GC run, condensation byproducts from all DCPs can be easily recognized as they are eluted at long retention times, and they are well-separated from the initial DCPs and from the other degradation intermediates. Under the experimental conditions adopted, the retention times of intermediates are comprised between 15 and 22 min. Condensation products are present for all DCP isomers. An example is provided in Figure 1, which shows the reconstructed total ion current (RIC)and the reconstructed current for the molecular ions of some specific condensation products formed during the irradiation of 2,4-DCP in the presence of TiOz. In the retention time window shown, eight byproducts (marked with an asterisk) were detected. They corresponded to three molecular ion masses only. The other peaks in the total ion current chromatogram corresponded to impurities that were unrelated to the original DCP. From the analysis of GC retention times and mass spectra, it follows that the sets of condensation VOL. 29, NO. 9, 1995 /ENVIRONMENTAL SCIENCE &TECHNOLOGY m 2227

zt :::

3,5-DCP

C

V G 0 0 7

vw

0

v

0

100

0

A MW 306 3

rn

2.6-DCP 2,s-DCP 2,4-DCP 2,3-DCP

3nvJ v

A

0 a35

1

A

, , ,

AAA

,;-, 17

m

a0 0

0

0

v

0

19

20

21

a0

I -

0

; d _ T v v , A ^ , o , 18

*

c

a

0vom

nMuWjT?OOv ocO0

,

0

,,

I

cn

C

0

.-5

60

9)

> ‘g - 40

?!

20

retention time, min FIGURE 2. Condensation products observed in the photocatalytic degradation of six DCP isomers as a function of GC retention time. For experimental conditions see text.

0 100

150

200

250

300

350

252

products originated from the individual DCPs are different from one another. This is shown in Figure 2, which gives an overview of the molecular weights and the retention times of the condensation products. Identification of CondensationByproducts. The condensation products were identified from interpretation of their E1 and NI-CI mass spectra and on the basis of chromatographic and chemical considerations. As an example, Figure 3 gives the E1 mass spectra of three condensation products with MW 322 Da, originating from 2,4-DCP and 2,6-DCP. In general, E1 mass spectra of condensation products originating from the same DCP can be quite different, although their retention times are similar (Figure 3a,b); in addition, intermediates formed from different DCPs can have almost the same mass spectrum, although the difference in their retention times rules out the compounds being identical (in Figure 3b,c, a difference of 23 s, far above the experimental uncertainty). Typical features of these spectra are (i)the presence of an abundant molecular ion; (ii)the existence of several peaks within the molecular ion cluster and fragment ion clusters, mainly due to the two natural isotopes of chlorine; and (iii) a fragmentation corresponding to consecutive losses of HCl and/or Clz. From the relative intensity of the peaks within the molecular ion cluster and from its fragments, it is possible to infer how many chlorine atoms are present in the structure. For example, the spectra in Figure 3 indicate that four chlorine atoms are present in the molecular structure. From the knowledge of the molecular weight and the number of chlorine atoms present, it is possible to hypothesize a molecular structure for the different condensation products. However, without comparison with a standard, the E1 mass spectrum does not provide information either on the position of the functional groups in the two aromatic rings or on the linkage between the two rings.

The linkage between the two rings (C-C or C-0-C) determines which class of condensation products originates from the photocatalytic treatment, either polyhydroxyPCBs or polyhydroxypolychlorodiphenyl ethers (polyhydroxy PCDEs) and, as a consequence, the byproduct toxicity and hazard. PCDEs can be converted to dioxins by’OH radical attack in ortho position to the C-0-C bridge followed by a condensation leading to ring closure. Although with a slightly different mechanism, direct photolysis of pentachlorophenol by Wirradiation (20)leads to the formation of hydroxypolychlorodibenzo-p-dioxinsand dibenzofurans. 2228

ENVIRONMENTAL SCIENCE &TECHNOLOGY

/

VOL. 29, NO. 9. 1995

I

c 0

I

1

322

l

I

IIlllr

Y 2o

0 100

150

200

’ O Oi

250

300

350

300

350

2

54

0

,

-‘.

100

150

200

250

mass FIGURE 3. El mass spectra of condensation products of mass 322 Da obtained in the photocatalytic degradation of 2,4-DCP (a) tR = 18.30 min and (b) Ik = 18.40 min and of 2,6-DCP (c) tR = 18.08 s.

About 100 congeners of DCPEs have been synthesized, and their E1 mass spectra are often very similar to one another and to those of PCBs (32). Standards for polyhydroxyDCPEs are not available nor to our knowledge have been synthesized. In conclusion, the E1 mass spectra of Figure 3 do not provide unambiguous information on the molecular structure and can either be assigned to a dihydroxytetrachlorobiphenyl or to a hydroxytetrachlorodiphenyl ether. From a set of experimental data, it is possible to conclude that the formation of polyhydroxypolychlorodiphenylethers is unlikely. We have identified 4,4’-dihydroxy-3,3’,5,5’tetrachlorobiphenyl, for which a standard was available, as one of the condensation products from 2,6-DCP. Other

TABLE 1

Type of Condensation Products and Number of Isomers Found in Photocatalytic Degradation of Dichlorophenols on Ti02 CI

CI

CI

CI

OH

OH

compd

MW288

2,3-DCP 2,4-DCP 2,5-DCP 2,6-DCP 3,4-DCP 3,5-DCP

2 2

MW 304

$

CI

OH

2 ‘ 1CI

MW 306

1

research groups, using different AOPs, have found PCBs not polychlorodiphenyl ethers as condensation products (18,19,26,27). Unfortunately, some analytical techniques used by the research groups just referred to, and in particular NMR, are not adequate for the purposes of the present study in view of the very low concentration of condensates formed. It was also demonstrated (33)that the phenoxy1 radical, when produced by pulse radiolysis at concentrations > lo-* M from basic solutions of phenol, combines to give dihydroxybiphenylsas predominant products. Electronic disproportionation or coupling to form diphenyl peroxide is relatively unimportant. Phenoxyphenol was also measured at lower yields. In contrast, in neutral or acidic solutions the yield of dihydroxybiphenyls and probably that of phenoxyphenol is reduced. y-Irradiation of creosol for long times (3 h) gave the diary1 ether only in the presence of high alkalinity and 10%oxygen purging (34). In neutral or slightly acidic conditions, such as the ones applied in this study, no evidence of the C-0-C condensates was reported. However, the strongest evidence against formation of polyhydroxypolychlorodiphenyl ethers as condensation products proceeds from experiments run in NI-CI-MS, where PCBs were found, as expected, to produce an extremely high response, owing to their large n system of conjugated aromatic rings, while polyhydroxypolychlorodiphenyl ethers and polychlorodibenzodioxins, in which ether bridges interrupt the conjugation, are known to have a much lower cross section for electron capture. We used 1,4,7,8-TCDD as a test compound. Under the present experimental conditions, 1,4,7,8-TCDDgave very poor response in NI-CI-MS. In contrast, all GC peaks classified as polyhydroxyPCBs from GC-EI-MS were found also in GC-NI-CI-MS experiments where they exhibited high intensity and comparable relative abundances. This supports the former identification. Thus, all condensation products listed in Figure 2 are likely to belong to the class of polyhydroxyPCBs. Under the present experimental conditions, polychlorodibenzodioxins and polychlorodibenzofurans could have been determined if they had been present at concentrations > 4 ng L-l (the estimated detection limit). In fact, their molecular weights are 2 mass units below those of the corresponding polyhydroxyPCBs. No experimental evidence was found for their formation. Table 1shows the structures, exclusive of isomery, of all

CI

OH

CI

HO MW OH 338 CI

MW 322

5

7 2 2

r:: x::

OH

6

7 2 5

5

the different polyhydroxyPCBs formed from the six DCPs and gives the complete outline of the qualitative analysis of condensation products originated from the substrates studied, including the number of the isomers found. Examination of Table 1 and Figure 2 reveals that polyhydroxyPCBs carry 1-4 hydroxyls and 3-4 chlorine atoms. Several possible isomers of the structures listed in Table 1 exist, even if some reasonablerestrictions associated with their origin are introduced. As Figures 2 and 3 suggest, different substrates should produce sets of polyhydroxyPCB isomers with little if any overlapping with one another (except for mass 338). Attempts to link isomer structures with retention times is also extremely difficult in the absence of standards. Since this is beyond the scope of the present study, we avoided identifymg each isomer uniquely or inferring its structure from a possible mechanism. QuantiflcationofCondensation Products. Quantitative determination of condensation products has been performed in EI, using the described procedure, based on the addition of internal standards. Several circumstances limit the precision of quantitative determinations and lead to a possible underestimation of some polyhydroxyPCBs. For instance, among these byproducts the most polar ones could possibly not be fully extracted from the reaction medium. In addition, these polar products are likely to have a high number of hydroxyl groups,which decreasetheir volatility. Therefore, very polar condensation products could possibly be discriminated in the GC injector and column. The determinationswere also affected by the uncertainty in the response factors of the condensation products for which authentic standards were not available. As the structures of condensation products are relatively homogeneous and correspond to mono-, di-, and trihydroxy derivatives of tri- and tetrachlorobiphenyls, and since response factors may vary over a wide range in NI-CI but over a much more limited range in EI, their response factors were assumed to be equal and as having the value of unity. Although this is clearly an oversimplified assumption, the lack of authentic standards as well as the absence of data in the literature preclude the possibility of using correction factors for the different classes of condensates. Despite all the limitations mentioned, the quantitative determination of polyhydroxyPCBs provides an estimation of how extensively the condensation process occurs. VOL. 29, NO. 9 , 1 9 9 5 /ENVIRONMENTAL SCIENCE &TECHNOLOGY

2229

0:

0: f

0

10

20

30

2.3-DCP

m

f\

.: 2,5-DCP

322

338

40

50

0

60

10

20

30

40

50

irradiation time, rnin

irradiation time, min

b 0:

m

2,4-DCP

b/20L

a

15

0:

E -0

5

m 15

288 304 322 *: 338 0. 8: 0:

C

10

2,S-DCP

za e -

0

0

c 10

0

10

20

30

40

50

0

c 0

5 F

s

". N I

x

0 0

60

-e

c

I

0

-E-

10

20

30

40

50

60

irradiation time, min

irradiaton time, min

FIGURE 4. Photocatalytic degradation of 2,3-DCP (a) and 2,CDCP (b): disappearance of DCP and time evolution of all condensate products found. The products correspond to the molecular weight indicated. Concentrations of different isomers having the same mass were added for the sake of simplicity.

FIGURE 5. Photocatalytic degradation of 2.5-DCP (a) and 2.6-DCP (b): disappearance of DCP and time evolution of all condensate products found. The products correspond to molecular weight indicated. Concentrations of different isomers having the same mass were added for the sake of simplicity.

The concentrations of polyhydroxyPCBswere estimated in the absence and in the presence of Ti02 at various irradiation times. In the absence of TiOz (direct light photolysis), the amount of condensates detected was very low (in the ng L-I range) and slightly increasing with the irradiation time. For example, a 40-min irradiation of a 20 mg L-I water solution of 2,6-DCP led to the formation of 20 ng L-' trichlorodihydroxyPCB (MW 288, one isomer detected), of 44 ng L-I trichlorotrihydroxyPCB (MW 304, four isomers detected), and of 37 ng L-' tetrachlorodihydroxyPCB (MW 322, two isomers detected), respectively. In the presence of TiOz, the concentrations of polyhydroxyPCBs were estimated at various irradiation times, starting from t = 0 min (the blank) and ending at t = 30, 40, or 60 min, depending on when the condensates ceased to be detectable. In this way, it has been possible to follow the kinetics of formation and subsequent decomposition for all the condensation products. The corresponding curves are depicted in Figures 4-6. The estimated concentrations of all the isomeric compounds have been added together for the sake of simplicity, although they were determined separately. For each set of DCP isomers, the result of quantitative determinations of polyhydroxyPCBs is expressed as micrograms per liter (parts per billion, ppb) of the slurry. The amount of condensation products (from 5 x lo-'' to 2.5 x M) is small compared to that of the initial DCP (1.2 x M), but significantly higher than that measured in the absence of Ti02 (tens of ng L-l). If the sensitivity of the procedure in E1 is almost the same also for hydroxypolychlorodibenzo-p-dioxins and dibenzofurans, it follows from the foregoing discussion that the

presence of these noxious compounds is ruled out, at least at the level of parts per trillion or higher. The disappearance of the various DCP isomers and the evolution of condensation products is completed in the same time window. The completeness of the mineralization process, as determined from TOC measurements (301, seems not to be dependent on the amount of polyhydroxyPCBs formed. For all DCP isomers, mineralization was completed after 120-min irradiation, and condensates were no longer detectable after 60-min irradiation. The maximum concentration of condensation products was reached after 6- 10-min irradiation, well before the total disappearance of the DCP, and does not depend on the specific DCP studied (Figures 4-6) except for mass 338 in the case of 2,3-DCP. The formation of hydroxyPCBs is closelyrelated to the first steps of the degradative attack on the initial DCP. As the oxidation process proceeds, also the type and concentration of condensation products may evolve, possibly giving products with more hydroxyl groups. Condensation products of mass 338, which usually appear after other hydroxyPCBs,can then be originated in a subsequent step by hydroxylation of monohydroxy- and dihydroxytetrachlorobiphenyl (MW 306 and 322, respectively). Results obtained from different DCP isomers are markedly uneven in many respects. The most striking evidence is that the concentration of condensation products varies by 2 orders of magnitude. While irradiation of 2,3-DCP and 2,6-DCP produced relatively high amounts of polyhydroxy-PCBs, for 2,4-DCP, 3,4-DCP, and 3,5-DCP, only a few parts per billion of condensates were observed. The total maximum concentration of condensation products,

2230 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29. NO. 9,

1995

7

1.2

J

m

a 1.0

k

).cI In

0

0:

3.4-DCP

0:

288

4:

322

m: 304 A: 306

g 0.8

c-0

+: 338

.-0 0 " k 0.6

E

" C E O

2O E

0.4

"g 0.2 c

0 0

0.0 0

10

20

30

40

50

SO

irradiation time, min

7

2.0

'

20

4

c

I

m

4

a

m . 15 E

1.5

-

.c-

og

0

c-0

c

.-

0 0

P)

; k 1.0

10

ZE

2 80

-c

eo w 'E ua

0

2 0.5

- 5

UP)

.L!

-0

I

-0

c 0

.o

0.0 0

10

20

30

40

50

x

60

irradiation time, min

FIGURE 6. Photocatalytic degradation of 3,4-DCP (a) and 3.5-DCP (b): disappearance of DCP and time evolution of all condensate products found. The products correspond to molecular weight indicated. Concentrationsof different isomers having the same mass were added for the sake of simplicity.

obtained by summing up the concentrations of condensation products at the irradiation times when they exhibited their maximum concentrations, can be compared with the initial DCP concentration. Taking into account that two DCP molecules are necessary to produce one polyhydroxyPCB molecule, it turned out that under the experimental conditions adopted at least 1%of 2,3-DCP and 2,6-DCPmolecules was involved in condensation processes. The percentage of molecules undergoing dimerization declined to at least 0.2% in the case of 2,5-DCP and to only 0.02-0.07% in the case of 2,4-DCP,3,4-DCP,and 3,5-DCP. It may be noted that both 2,3-DCPand 2,6-DCPcarry three functional groups contiguous to one another and that their main condensation products still contain four chlorine atoms (molecular weights 322 and 338). Therefore, the formation of polyhydroxyPCBs is probably related to the presence of unsubstituted aromatic carbons, and the steric hindrance might prevent these processes from occurring. The yield of PCB formation in photocatalytic treatment using Ti02 was not far from that obtained by the generation of the 'OH radical by means of the Fenton reagent (19). Experiments carried out on chlorobenzene and chlorophenols using the latter reagent in the presence of oxygen showed that about 0.4% of the initial compound was transformed to dichlorobiphenyl isomers (19). 2,4-DCP produced a low amount of condensates. A relatively high abundance of condensates with MW 288 and 304 as compared to M W 322 was observed (Figure 4). Since on the former condensates there are only three chlorine atoms, a dechlorination step is probably competi-

tive with processes leading to products of M W 322. This can be correlated with the presence of two chlorines in ortho and para positions in the initial DCP, hence suggesting that electronic factors also play an important role in the observed condensation yield. The dihydroxybiphenyls produced by pulse radiolysis of phenol under a variety of experimental conditions showed that the radical combination at the ortho and para positions of the ring occurs statistically, the pura site being favored. No reaction at the metu site was observed (33). In general, condensation products still carrying four chlorine atoms (MW 322 and 338) were preferentially formed from all DCP isomers, with the exception of those substituted in position 4, namely, 2,4-DCP and 3,4-DCP, for which dechlorination (MW 288 and 304) or dehydroxylation (MW306)were observed,respectively. This behavior closely matches the type of main degradation intermediates found, namely, monochlorobenzenediolsand chloroquinones for 2,4-DCP and 3,4-DCP and dichlorohydroquinones and dichloroquinones for the other DCP isomers (30). Proposal of a Reaction Pathway. Scheme 1 rationalizes the foregoing observations and depicts some possible pathways for the initial transformations of DCPs, leading either to condensation products (this work) or to chlorobenzenediolslchloroquinonesand other more oxidized species (30). Unlike other AOPs based only on 'OH radical chemistry, the photocatalytic process simultaneously produces 'OH radical-like oxidant species and valence band electrons (24). These reactive species can concurrently oxidize or reduce the aromatic ring, forming semiquinone radicals that dimerize or condense following standard mechanisms (35). H-abstractionby'OH radical (freeat the surface or bound to it) (36) produces radical I. This is well-documented by radiation chemistry experiments (33, 37). Hydroxylated hydroxycyclohexadienyl radicals are relatively stable in neutral or acidic solutions (38). Radical I can dimerize (33, 3 9 to form the stable tautomer in equilibrium with dihydroxytetrachlorobiphenyl(MW322). Correspondingly, *OH addition to radical I leads to the formation of dichlorobenzenediols, from which radical I1 is formed by further H-abstraction. Condensation of radical I with radical I1 leads to trihydroxytetrachlorobiphenyl (MW338). Since the same path is followed, higher abundances of masses 322 and 338 are correlated with the formation of dichlorobenzenediols, as previously noted. This path implies the addition of surface-generated'0H radical to an organic radical to form dichlorobenzenediols. Since DCPs are likely to react at the catalyst surface with photogenerated active species (39),giving radicals that in their lifetime are not allowed to leave the surface, the probability of the addition of a surface-generated 'OH radical to an organic radical is also very high (radical-radical hydroxylation). This high probability is in contrast with accepted mechanisms in solution, where organic radicals preferentially react with molecular oxygen to give radical organic peroxides (40). In the photocatalytic process, oxygen is probably scavenged more extensively by conduction band electrons rather than by organic radicals. The maximum transient concentration reached by dichlorobenzenediols during photodegradation was about 10% of the initial DCP (30). If the probability of Habstraction were the same as for DCP, radical I1 would be formed to an extent of approximately 10% of radical I. This leads to the following: VOL. 29. NO. 9,1995 / ENVIRONMENTAL SCIENCE & TECHNOLOGY m 2231

SCHEME 1

Pathways for Formation of HydroqPCBsa

-

...........

..

a Their molecular weight is shown in a box. Compounds marked with an asterisk are present in different possible tautomer structures. Radical intermediates, marked with roman numerals, are given in open dotted boxes. A number of other possible reaction pathways are not indicated but are discussed in the text.

(i) a small amount of condensate with mass 338, in contrast to the observed concentrations. As previously suggested, mass 338 can be formed also from mass 322 by hydroxylation (path shown in Scheme 1with a dotted line). Thus, high concentrations of mass 338 can be observed whenever mass 322 is present and probably evolves from the latter. In addition, inspection of Figure 2 shows that for condensates of MW338 onlythere is a close coincidence of their retention times when they originate from different DCPs. For mass 338 isomers, there is the superposition of retention times ( t ~ with ) , a difference of less than 2 s, at t~ =19.72 i 0.02 for 2,3-DCP, 2,5-DCP, 3,4-DCP, and 3,5DCP; at t~ = 20.12 i 0.02 for 2,3-DCP, 2,5-DCP, 2,6-DCP, and 3,4-DCP; at t~ = 20.80 i 0.03 for 2,3-DCP, 2,5-DCP, 3,4-DCP, and 3,5-DCP; and at t~ = 21.16 f 0.03 for 2,3DCP, 2,5-DCP,2,6-DCP,3,4-DCP, and 3,5-DCP,respectively. This fact suggests that mass 338 can also be formed by double hydroxylation of mass 306 (see Figure 6 and below), for which the internal rearrangement leading to its formation can favour chlorine translocation to give more stable isomers. The oxidative rearrangements of hydroxyquinones is also well-documented (41). (ii) A very small amount, if compared with that of the intermediate of MW 322, of tetrahydroxytetrachlorobiphenyl (TTCB)was produced by dimerization of radical 11. These condensates were not found in the GC runs. Last, radical I is likely to be formed when ortho and para positions are free (33). When this is not the case (2,4-DCP), this path is no longer allowed, and the concentration of oxidized condensation products (upper part of Scheme 1) is markedly decreased (see also Figure 4). The initial reduction of DCP to form radicals 111 and IV (40)opens the way to the formation of condensates of M W

-

2232 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL.

29, NO. 9, 1995

306, by condensation of radical 111 with radical I, and of condensates with MW 288 and 304, by condensation of radical IVwith radicals I and 11, respectively. 'OH addition to radical IVforms chlorobenzenediols, which can undergo further degradation. Since the same initial reductive path is followed, a high abundance of condensates with molecular weights 288 and 304 is associatedwith the formation of chlorobenzenediols, as previously noted (30). In the present experimental conditions, the experiments were carried out in the presence of 02, which is an effective scavenger of electrons. In this case, reductive paths would play a secondary role. In fact, the concentration of the dechlorinated condensates (with MW 288 and 304) was invariably about 10% of the hydroxylated ones (with MW 322 and 338). Experiments carried out in the absence of 0 2 (by exhaustive purging of the test solution with He) on 2,6-DCP showed that, after 10-min irradiation in the presence of 250 mg L-l TiOz,the condensates were formed in lower amount with respect to oxygenated solutions (see Figure 5) and the importance of reductive paths was increased. Concentrations of 0.28, 0.92, and 1.38 pg L-I were measured for Mw 288 (two isomers), for M W 304 (four isomers),and for MW322 (two isomers), respectively. Mass 338 was not detected. Relativelyhigh amounts of condensates of mass 288 and 304 were present only in the case of 2,4-DCP, for which electronic factors that prevent the oxidative path are likely to favor electron addition in the free metu position. Condensates of MW 306 were predominant in the degradation of 3,4-DCP (Figure 6). The same condensates were present at about the same concentration only for 2,5-DCP (Figure5). These two DCPs have in common a considerable steric hindrance, which is partly reduced in the transition

Ol

1

m-

200

m m

1

150

2 c

0

100

VI

0

c (II

50 v 0

O

O

0

10

20

30

40

irradiation time, rnin

FIGURE 7. Time evolution of condensates with MW 338 in the photocatalytic degradation of 2,3-DCP (different initial concentrations) in the presence of 300 mg L-' TiOz.

condensate to form a monohydroxyPCB by water elimination. The foregoing discussion suggests that fewer than the maximum number of isomers may be observed in some cases (2,4-DCP and 3,4-DCP). For 2,5-DCP and 3,5-DCP, more isomers are observed than those predictable from direct condensation of two molecules, but this is accounted for by the hydroxylation of isomers of mass 306 on the ring still carrying one hydroxyl. HydroxyPCBs formed by the dimerization of radicals I11 and IV were not detected, probably owing to their very small formation yield. Accordingly, also a further reduction giving chlorophenol is not probable. Experimentally, chlorophenol was not detected. Scheme 1 is not exhaustive of all the possible reaction pathways. A n alternative reaction pathway may involve condensation between the organic radical -and the DCP molecule owing to the high concentration of the starting material. Such a mechanism may lead to the same condensation products upon the attack of primary radicals ('OH or e-/H+) and final loss of a water molecule. The dependence of the observed condensation products (Mw338)from the initial 2,3-DCPconcentration is reported in Figure 7. The experiment shows that both the amounts formed and the time at which maximum concentrations are reached increase with increasing initial substrate concentration. The condensates of masses 288, 304, and 322 follow the same trend. The corresponding DCP disappearance is also delayed as the initial concentration is increased. The increase of condensates formed can be ascribed either to the increase of radical-radical reaction rate, due to the increase of radical concentration, or the radical-DCP reaction rate, due to the increase of the initial DCP concentration. Akinetic analysisfor the two alternative reaction pathways under the hypothesis of constant concentration of primary active species (holes, hydroxyl radicals, and valence band electrons) at the surface of TiOe could not clarify this point. Almost the same dependence of the condensate concentration on the square of the initial concentration of the substrate is obtained for the two alternative reaction pathways. In favor of the reaction pathway proposed in Scheme 1 is the location of the reaction in the limited domain of the catalyst surface (39). Local concentrations of organic radicals at the catalyst surface, where reaction with primary active species takes place (39),are likely to be higher than in solution. This hypothesis is able to justify the yields of

condensates that are comparable to those observed with the Fenton reagent (19). Many other reasons against the radical-molecule reaction pathway exist, although only speculative. Semiquinoneradicals are reported to dimerize or condense, not to react with the starting materials, even when their concentrations are higher (33-3537,401 than those used in the present experiments. Secondly, the high reactivity of the species involved may largely favor the rate for a radical-radical reaction rather than that of processes forming new radicals followed by further molecular rearrangement. Last, a radical-molecule reaction would restore the aromatic character of the rings only after the final loss of a water molecule. Under such a mechanism, the formation of the intermediate with MW306would imply the final loss of two water molecules. The water loss is likely to be a kinetically slow step. In conclusion, the proposed reaction pathway is able to justify all the experimental observations. Because of the similarities of photocatalysis with other AOPs based on 'OH radical chemistry, for all AOPs attention must be paid to the identification and the quantitation of side products.

Acknowledgments C.M., E.P., and M.V. are grateful to CNR, CIEMAT, MURST, and EEC for financial support.

Literature Cited (1) DorC, M. Chimie des Oxidants et Traitement des Eaux (Oxidant Chemistry and Water Treatment); Lavoisier: Paris, 1989; in French. (2) Ramalho, R. S. Introduction to Wastewater Treatment Processes; Academic Press: New York, 1983. (3) Legrini, 0.;Oliveros, E.; Braun, A. M. Chem. Rev. 1993, 93, 671. (4) Photochemical Conversion and Storage ofsolar Energy; Pelizzetti, E., Schiavello, M., Eds.; Kluwer: Dordrecht, 1991. (5) Harrington, G. W. Developinga Disinfection By-product Control Strategy; AWWA Annual Conference, Philadelphia; American Water Works Assoc.: Denver, 1991. (6) Oxenford, J. L.; Brink, D. R. Prepr. Pap., Natl. Meet-Am. Chem. Soc., Diu. Environ. Chem. 1993, 33 (2), 139. ( 7 ) Miller, S. Environ. Sci. Technol. 1993, 27, 2292. (8) Occurrence, Assessment for Disinfectants and Disinfection ByProducts (Phase VI a) in Public Drinking Water; Final Report; U S . Environmental Protection Agency: Washington, DC, Aug 3, 1992. (9) Larson, R. A.; Rockwell, A. L. Environ. Sci. Technol. 1979,13,325 (10) Norwood, D. L.; Johnson, J. D.; Christman, R. F.; Hass, J. R.; Bobenrieth, M. J. Environ. Sci. Technol. 1980, 14, 187. (11) Boyce, S. D.; Homing, J. F. Environ. Sci. Technol. 1983, 17, 202. (12) Oyler,A. R.; Liukkonen, R. J.; Lukasewycz, M. T.; Heikkila, K. E.; Cox, D. A,; Carlson, R. M. Environ. Sci. Technol. 1983, 17, 334. (13) Singer, P. C.; Chang, S. D. J. Am. Water WorksAss. 1991, 81 (E), 61. (14) McCreary, J. J.; Snoeyink,V.L.: Larson, R. A. Environ. Sci. Technol. 1982, 16, 339. (15) HoignC, J.; Bader, H. Water Res. 1988, 22, 313. (16) Miltner, R. J.; Shukairy, H. M.; Summers, R. S. J.Am.Water Works Assoc. 1992, 84 ( l l ) , 53. (17) Weinberg, H. S.; Glaze, W. H. Prepr. Pap., Natl. Meet-Am. Chem. Soc., Diu. Environ. Chem. 1993, 33 (21, 199. (18) Thornton. T. D.; LaDue, D. E., 111; Savage, P. E. Environ. Sci. Technol. 1991, 25, 1507. (19) Sedlak, D. L.;Andren, A. W. Environ. Sci. Technol. 1991,25,777. (20) Vollmuth, S.; Zajc, A.; Niessner, R. Environ. Sci. Technol. 1994, 28, 1145. (21) Ollis. D. F.; Pelizzetti. E.; Serpone, N. Environ. Sci. Technol. 1991, 25, 1522. (22) Bahnemann, D.; Cunningham, J.; Fox, M.A.; Pelizzetti, E.; Pichat,

P.; Serpone, N. In Surface and Aquatic Photochemistry; Helz, G. R., Zepp, R. G., Crosby, D. G., Eds.; LewisPublishers: BocaRaton, FL, 1993. (23) Photocatalytic Purification and Treatment of Water and Air; OKs, D. F., Al-Ekabi, H., Eds.; Elsevier: New York, 1993. (24) Photocatalysis. Fundamentals and Applications; Serpone, N., Pelizzetti, E., Eds.; Wiley: New York, 1989.

VOL. 29, NO. 9, 1995 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

2233

Glaze, W. H.; Kenneke. I. F.; Ferry, J. L. Environ. Sci. Tecltnol. 1993, 27, 177. Sehili,T.; Boule, P.; Lemaire, 1.1.Photochern. Phorobiol. A: Cliern. 1989, 50, 103. Sheili. T.; Boule, P.; Lemaire, 7. Chemosphere 1991, 22, 1053. Pelizzetti, E.; Minero, C.; Sega, M.; Vincenti, M. In Photocatalytic PuriJcatiori and Trearrnent of Water and Air; Ollis, D. F., AlEkabi, H., Eds.; Elsevier: New York, 1993; p 291. Manilal, V. B.; Haridas, A,; Alexander R.; Surender G. D. Water Res. 1992, 26, 1035. Minero, C.; Pelizzetti, E.; Pichat, P. J. Pliotockem. D'Oliveira, J.C; Pliotobiol. A: Chem. 1993, 72, 261. Minero, C.; Pelizzetti, E.; Malato, S.; Blanco, J. Chemosphere 1993, 26. 2103. Nevalainen, T.; Koistinen, J.: Nurmela, P. Enitiron. Sci. Technol. 1994, 28, 1341. Ye, M.; Schuler, R. H. 1. Phys. CIiern. 1989, 93, 1898. Cierer, J.; Yang, E.; Reitberger, T. Holzforschiing 1992, 46, 495. Thornson, R. H. In The Chernisny ofthe Quinonoid Compounds; Patai, S., Ed.; Wiley: New York, 1974; Part 1, Chapter 3.

2234

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 9, 1995

Pelizzetti, E.; Minero, C. Comments Inorg. Chern. 1994, 1.5, 297. Fendler, J. H.; Fendler, E. J. In The Chemistry oftlie Quinonoid Compounds; Patai, S . Ed.; Wiley: NewYork, 1974:Part 1,Chapter 10. Sun, Q.; Schuler, R. H. 1. Plzys. Chem. 1987, 91, 4591. Minero, C.; Catozzo, F.; Pelizzetti, E. Langmctir 1992, 8, 481. Sommeling, P. M.; Mulder, P.; Louw, R.; Avila, D. V.; Lusztyk, 1.; Ingold, K. U. J. Pliys. Cliern. 1993, 97, 8361 and references cited therein. Moore, H. W.; Wikholm, R. 7. In The Chemisrryofrhe Quinonoid Compounds; Patai, S . Ed.; Wiley: NewYork, 1974; Part 1, Chapter 8.

Received for review October 25, 1994. Revised manuscript received May 9, 1995. Accepted May 23, 1995." ES940662M Abstract published in Advance ACS Abstracts, July 1, 1995.