(8) P. Grafstein and R. Goldberg, “Advances in Automated Analysis, Tech- nicon International Congress 1972”, Vol. 9, Mediad, Inc.. Tarrytown, N.Y., 1973, p 53.

(9) L. F. Cullen, J. G. Rutgers, P. A. Lucchesi, and 0. J. Papariello, J. Pharm. Scb, 57, 1857 (1968).

(10) C. E. Stevenson, L. D. Bechtel, and L. J. Coursen, “Advances in Auto- mated Analysis, Technicon International Congress 1969”, Vol. 2, Me- diad, inc., White Plains, N.Y., 1970, p 251.

(1 1) L. F. Cullen, D. L. Packman, and G. J. Papariello. J. Pharm. Sci., 59, 697

(12) C. W. Gehrke, L. L. Wall, and J. S. Absheer. J. Assoc. Off. Anal. Chem..

(13) T. Urbanyi, W. T. Brunsklll, and M. Lin, J. Assoc. Off. Anal. Chem.. 56,

RECEIVED’for review May 15,1975. Accepted July 8,1975.

(1970).

56, 1096 (1973).

1069 (1973).

Automated Determinations of Dissolved Organic Carbon in Lake Water

P. D. Goulden and Peter Brooksbank

Canada Centre for lnland Waters, Burlington, Ontario, Canada L7R 4A6

Automated methods are descrlbed for the determlnatlon of dlssolved organlc carbon In water from the Great Lakes. The lnorganlc carbonate Is removed in a heated packed column, the organlc carbon Is oxldlred and the resultlng carbon dloxlde measured by an Infrared analyzer. Two al- ternatlve methods of oxldatlon are used, ultraviolet Irradla- tlon and silver-catalyzed peroxydlsulfate at 95 O C . When ul- travlolet lrradlatlon Is used, the relatlve standard devlatlons obtalned at carbon levels of 50 pg/l., 1 mg/l. and 5 mg/l. are 4.0%, 1.2%, and 0.8% respectively, wlth sllver-cata- lyzed peroxydlsulfate the relatlve standard devlatlons are 3.2%, l.O%, and 0.7%, respectively. The silver-catalyzed peroxydlsulfate is the more convenlent and precise method but does not completely oxldlre all materlals In water; In a large number of natural water samples analyzed, the oxlda- tlon completeness averaged 97 %. The llmlt of detectlon for carbon Is 10 pg/l., the analysis rate Is 20 samples per hour.

The measurement of dissolved organic carbon (DOC) is an important part of many water quality studies. The dis- solved organic material may represent the degradation products of plant or animal life that lives or has lived in the water, or, alternatively, it may represent pollution by sew- age or industrial effluent.

The methods commonly used for DOC determination in- volve oxidation of the organic material with subsequent measurement of CO2. Oxidation can be carried out in the gas phase by passing the sample over a catalyst a t high temperature (1) or by wet oxidation. In its commonly-used form, where the sample is injected in a gas flow with a sy- ringe, the gas phase oxidation technique is convenient and rapid but the necessarily small sample size limits the sensi- tivity obtained. Wet oxidation, which uses larger sample sizes, gives higher sensitivity. In the method of Menzell and Vaccaro (21, oxidation is carried out by treating the sample with peroxydisulfate in a sealed tube in an autoclave; how- ever, this operation can be quite time consuming and Bal- dwin and McAtee (3) have used oxidation a t room temper- ature with silver-catalyzed peroxydisulfate. Oxidation can be carried out using irradiation with ultraviolet light and this has the advantage that it lends itself to an automated process such as that described by Erhardt (4).

In all these techniques, a distinction must be made be- tween the inorganic and the organic carbon present in the sample. The distinction can be made by measuring both the carbon dioxide liberated by acidification and also mak- ing a “total carbon” measurement. The organic carbon

then is represented by the difference between the two mea- surements. Alternatively, the acidified sample can be stripped of the “inorganic” carbon dioxide before the oxi- dation step. The validity of the organic carbon measure- ment then depends on the completeness of the stripping operation and is also subject to the possibility of volatile organic materials being lost.

In the present work, the different techniques have been examined in the development of automated methods for the analysis. An automated system is used in which the in- organic carbonate is removed by stripping the carbon diox- ide from an acidified sample in a heated packed column. This enables the stripping to be carried out without any problems of sample contamination and also gives opportu- nity to examine the stripped gas for volatile organics.

Automated methods using both gas phase oxidation and wet oxidation have been developed. To obtain high sensi- tivity, wet oxidation is used. In the systems described here, with the large sample flow (-6 ml/min) to the oxidation system, a limit of detection of 10 fig/l. C is obtained. The oxidation process of choice is ultraviolet irradiation, in the work of Erhardt ( 4 ) , Soier and Semenov (5), and, in the present work, this has been shown to oxidize all the organic materials of interest. However, an alternative system that uses silver-catalyzed peroxydisulfate a t 9P°C for the oxida- tion has some advantages. Although this treatment did not completely oxidize all organic materials studied, e.g., EDTA, it is a convenient and precise method and, in a study of waters from the Great Lakes, it oxidized practical- ly all (97%) of the naturally occurring organic material.

A non-dispersive Infrared Analyzer is used to measure the C02 resulting from the oxidation. This was chosen be- cause of its relative freedom from interferences and be- cause the laboratories for which this method was developed already have this equipment as part of a commercially available Total Carbon Analyzer. By making small piping changes, it is possible to utilize this existing equipment ei- ther in the original mode or coupled to the automated equipment described here.

EXPERIMENTAL Apparatus. The manifold used is shown in Figure 1. The sampler is an Industrial Sampler (Technicon Corp.) which

uses 25-mm X 100-mm borosilicate glass tubes to contain the sam- ples. The sampler is modified by the addition on the sampler arm control shaft of a cam controlling a three-way solenoid valve. In the wash cycle, the wash water after being de-bubbled, passes into a T at the top of the sampler arm. As the sampler arm starts to move to the sample tube, the three-way valve is switched so that the wash water flows directly to the wash receptacle. The cam is

ANALYTICAL CHEMISTRY, VOL. 47, NO. 12, OCTOBER 1975 1943

Lg - - - - _ - - - - u v - 1 ~ 1 i I I R

ANALYZER

GAS TO IR ANALYZER -7

-0-31 - >-o_zs_.-%lver _nitrat_e_ _-__ 5 8 0 ~ 0 ? i o * Wash water

580 1 0 tie* Wash water

4 41 ~ 0 030. 02

PROPORTIONING PUMP

TO WASTE

C TO

WASTE

FIRST STRIPPER SECOND STRIPPER

Flgure 2. Strippers

materials are made using either irradiated water or silver-peroxy- disulfate treated water. All the reagents are “Reagent-Grade”. The oxygen is “very dry grade” gas.

Procedure. The pump is started and the system set in opera- tion with wash water. The range for the levels expected is set by changing the “gain” on the analyzer. Standard solutions of potassi- um biphthalate are run through the system to obtain a calibration curve and the samples are then measured. Every few days, stan- dard solutions of biphthalate, carbonate, and EDTA of approxi- mately twenty times the levels of interest are sampled through the “calibration” line to confirm the completeness of the oxidation. (These samples are made up in non-acidified water.) This also cali- brates the biphthalate standard used to obtain the calibration curves vs. a carbonate standard.

The samples are preserved and pre-treated by the addition of 0.2% v/v HzS04 to the sample bottles. Before use, the sample bot- tles and sampler tubes are heated in an oven at 550 “ C for an hour to destroy any organic carbon.

RESULTS AND DISCUSSION Several hundred water samples from many different

locations in the Great Lakes have been analyzed by the two methods described here. Some of these samples were also analyzed by carrying out the oxidation in the gas phase, Le., pumping a small portion of the stripped sample into a furnace containing cobalt oxide catalyst a t 950 “C. Typical results are shown in Table I.

Batches of 40 samples were analyzed the same day by each of the three methods above; the average results are also shown in Table I together with the relative standard deviations obtained with natural water samples. Trea.ting the results on individual samples within each batch as paired observations, the lower average result found using the silver-peroxydisulfate oxidation compared with the gas phase oxidation (97% vs. 100%) is statistically significant a t the 95% confidence level. There is no statistically signifi- cant difference between the results obtained by the UV ox- idation and the gas phase oxidation (100.5% vs. 100%). It is calculated that a difference of 1.2% between these two would be statistically significant.

Waters containing a variety of organic materials (potas- sium biphthalate, glucose, sucrose, EDTA, humic acid) a t various levels have been analyzed; results are shown in Table 11. Of these materials, all gave essentially complete recovery except for EDTA when oxidized by silver-cata- lyzed peroxydisulfate.

The relative standard deviations obtained on natural wa- ters containing carbon at 50 ggh., 1 mgll., and 5 mgh., re-

1944 * ANALYTICAL CHEMISTRY, VOL. 47, NO. 12, OCTOBER 1975

Table I. Typical Results on Natura l Water Samples Method of analysis

Gas phase

W irrad., Ag-Per. , oxidation,

Sample No. mg/l . C 95 OC, mq,'l. C m g i l . C

1 2.05 1.96 2 1.81 1.76 3 1.63 1.66 4 3.08 2.85 5 1.50 1.52 6 2.45 2.30 7 2.30 2.18 8 1.80 1.75 9 2.95 2.90

10 2.86 2.88 Av. 40 samples. 1 1.99 1.91 Av. 40samples, 2 2.31 2.22

at 2 mg/L C 1.2% 0.9% at 5 mg/l. C 0.8% 0.7%

Av. 40 samples, 3 2.14 2.09 Re1 std dev"

Q Relative standard deviations based on 11 replicates.

1.90 1.83 1.70 2.90 1.49 2.30 2.13 1.91 2.84 2.94 1.97 2.33 2.09

3.4% 2.8%

spectively, were 4.096, 1.296, and 0.8% with UV oxidation; 3.2%, 1.0%, and 0.7% with silver-catalyzed peroxydisulfate oxidation. The better precision obtained with the silver- catalyzed peroxydisulfate is because of the shorter resi- dence time in the equipment, which gives less opportunity for surges to develop in the liquid flow.

The detection limit, the concentration which gives a sig- nal twice that of the background noise, is 10 pgh. C. These wet-oxidation techniques provide very sensitive methods that have enabled measurements to be made on all the lake water of interest and, in fact, will provide a measure of the carbon in laboratory-distilled water. The silver-catalyzed peroxydisulfate oxidation does not appear to oxidize all the naturally occurring organics. For monitoring purposes, however, the 97% recovery obtained on water from the Great Lakes is probably acceptable and the method re- quires less equipment and, as seen above, is somewhat more precise than the UV oxidation method.

Inorganic Carbon a n d Volatile Organics. In the strip- ping column, the acidified sample is heated and sparged with oxygen. To confirm the complete removal of inorganic carbon, synthetic samples containing 50 pgh. organic car- bon were spiked with sodium carbonate. There was no sig- nificant difference between the results of the samples con- taining 0,1, and 5 mg/l. added carbonate, respectively.

To determine if the sparging operation removed signifi- cant quantities of volatile organic material, the "waste" stream of oxygen and steam from the first stripper was passed to the infrared analyzer directly and also after pass-

ing through a combustion tube packed with cobalt oxide catalyst a t 950 "C. The difference between the two readings would represent the organic materials volatilized. Calibrat- ing the technique with ethanol, it was found that the limit of detection was 50 pgh. C. None of the natural samples an- alyzed showed levels of volatile organics greater than this.

Photochemical Oxidation. The photochemical oxida- tion of a number of organic materials has been studied by Soier and Semenov (5). They report that all of the 24 mate- rials studied were completely oxidized. In the present work, the organic materials added were oxidized completely and, in the natural water samples, the oxidation completeness obtained was the same as that obtained by gas phase com- bustion. From this, it is believed that the photochemical oxidation does convert all of the organic material in these lake-water samples to Con. In the process shown here, which is essentially that described by Erhardt ( 4 ) , the or- ganic materials used were completely oxidized in the 8- minute irradiation time. With natural samples, reducing the irradiation time to 4 minutes did not affect the oxida- tion completeness, nor did removal of the peroxydisulfate or the use of air rather than oxygen as the segmenting me- dium. With EDTA, however, with a 4-minute irradiation time, removal of the peroxydisulfate or the use of air rather than oxygen lowered the oxidation completeness by about 10%. To increase the certainty of the complete oxidation of natural samples, an irradiation time of 8 minutes and the use of peroxydisulfate and oxygen are maintained in the method.

Silver Catalyzed Peroxydisulfate Oxidation. Baldwin and McAtee ( 3 ) have used silver-catalyzed peroxydisulfate a t room temperature to oxidize naturally-occurring organ- ics and found that with real samples the recovery of CO? was as good as, and sometimes better than, that obtained with conventional oxidation methods. In the present work, this oxidation system was investigated and it was found that with water samples from the Great Lakes, practically all (97%) of all the organics were oxidized by a 30-minute treatment with silver-catalyzed peroxydisulfate. However, the same oxidation was obtained by exposure to the oxi- dant for one minute at 95 "C. Since in an automated sys- tem it is desirable to heat the solution to separate low lev- els of the resulting CO:! and to minimize precipitation problems, this treatment at 95 "C is used. Oxidation at room temperature is used to produce the "zero carbon" wash water. Of the organic materials studied, all were com- pletely oxidized with the exception of the EDTA which was approximately 85% oxidized. Repeated treatment of the EDTA solution with silver-peroxydisulfate did not result in any further oxidation, neither did treatment for much ex- tended periods either at room temperature or at 95 O C .

As described above, the precision of the silver-catalyzed peroxydisulfate method was somewhat better than the method using UV oxidation. That this was due to the

Table 11. Typical Recoveries of Organic Materials ,\mount recovercd,

>!aterial zdded Level, mgil . C

Potassium 0.05 biphthalate 1

5 EDTA 0.05

1 5

Glucose 1 Sucrose 1 Humic acid d1

mg, ' l . C ( a i . 11 replicates) Re1 std d e \ , :. ( 1 1 replicates)

UV irrad.

0.054 0.99 5.00 0.048 1.04 4.94 1.02 0.98 1.03

-19-Per., 95 oc

0.051 1.00 4.98 0.040 0.86 4.24 1.01 1.03 0.94

IJV irrad.

3.8 1.0 0.8 4.2 1.1 0.8 1.3 1.0 0.9

Ag-Per., 95 'C

4.0 0.9 0.6 3.8 0.6 0.7 0;9 0.8 1.1

ANALYTICAL CHEMISTRY, VOL. 47, NO. 12, OCTOBER 1975 1945

shorter residence time was confirmed by adding delay coils to the system. As the residue time increased, the variability increased, and corresponding flow surges were observed.

“Zero Carbon”. In determining low levels of organic carbon, the greatest problems are those of contamination and of obtaining water that contains “zero carbon” for base-line reference. In the present automated methods, the contamination problems are reduced by the completely en- closed automated system and the fact that the sample tubes, the sample bottles, etc., can be heated in an oven at 550 OC to destroy all organic materials. Once the sample is filtered and acidified in the sample bottle, contamination is negligible. I t should be noted that the relative standard de- viations quoted were determined by replicate sampling from the bottle into different sample tubes a t random times over a day, and, hence, include the variability intro- duced by the contamination in the handling procedure.

In the UV oxidation method, the water used for the sam- pler “wash” is irradiated for a somewhat longer time than that used in the analysis. This irradiated water is collected periodically and used to make the standard solutions and reagents. Irradiation of this water for several hours does not lower the reading seen on the infrared analyzer com- pared to that seen after a single pass through the irradia- tor, and it is believed that with this system, essentially “zero carbon” water is obtained. The first approach used to obtain “zero carbon” water was multi-distillations from a variety of oxidizing materials such as perchlorate, perman- ganate, and peroxydisulfate. In all cases, these waters gave positive peaks (corresponding to from 10-50 pg/l. C) when analyzed vs. the water irradiated and passed directly into the sampler arm. It appears that it is in the handling of the water after distillation that pick up of organic material oc- curs, since water treated, e.g., with silver catalyzed peroxy- disulfate and pumped directly into the sampler arm does not give peaks relative to the irradiated wash water.

Small amounts of carbon-containing materials in the re- agent or in the gas are of no concern in the method since these only elevate the base line of the determination.

Silicone rubber pump tubes are used to pump the sam- ples and reagents; these are used because it was found that there was an increase in base-line noise when regular Tygon pump tubes were used, particularly when newly in- stalled, This is presumably from the plasticizers, etc., leaching from the tubing material. However, this increase in the noise is a problem only if levels of organic carbon less than about 400 pg/l. C are of interest. For water samples from the Great Lakes which are normally above this level, regular pump tubes are satisfactory.

Sample Preservation. It was necessary to treat natural samples after filtering to prevent changes in organic carbon content during storage. The treatment that was found sat- isfactory for storage periods of up to two weeks was acidifi- cation and storage a t 4 OC. If a total carbon measurement is to be made on the sample, the acid cannot be added when the sample is taken. In this case, storage a t 4 OC and prompt analysis is the preservation technique used. The acid is then added immediately before the DOC measure- ment.

Interfering Substances. The effect on the method of the ions commonly found in the water of the Great Lakes was determined. Although gross amounts (-100 mg/l.) of certain metal ions, notably silver and mercury, were found to affect the photooxidation rate of certain organic materi- als, at levels ten times those normally occurring there were no interferences in the analysis of lake water samples by the UV oxidation method.

“Silver-precipitating” anions such as chloride do not in- terfere in the silver-catalyzed peroxydisulfate method since, although there may be some precipitation, the solids pass through the equipment without any problems and with natural waters there is sufficient excess of silver to provide the necessary catalysis.

One class of compounds that is not able to be analyzed by these automated methods is non-ionic surfactants. An attempt was made to analyze synthetic samples of both nonyl phenol-ethylene oxide condensates and linear alco- hol-ethylene oxide condensates to provide a tool for a study of their biodegradability. These materials adsorb so strongly on the walls of the tubing, pump tubes, etc., that, for example, a t the 100 pgh. C level, only about 40% of the material reaches the oxidation system. Studies using a batch oxidation system confirm that the problem is adsorp- tion rather than incomplete oxidation.

LITERATURE CITED (1) C. E. Van Hall, J. Safranko, and V. A. Stengen, Anal. Chem., 35, 315

(2) D. W. Menzell and R. F. Vaccaro, Limnol. Oceanogr., 9, 138 (1964). (3) J. M. Baldwin and R. E. McAtee, Michrochem. J., 19, 179 (1974). (4) M. Erhardt, Deep Sea Res., 16, 393 (1969). (5) V. G. Soier and A. D. Semenov, Gidrokhim. Mater., 56, 11 1 (1971).

(1963).

RECEIVED for review March 31, 1975. Accepted June 27, 1975. This paper was presented at the Pittsburgh Confer- ence on Analytical Chemistry and Applied Spectroscopy, Cleveland, Ohio, by P. D. Goulden, March 5 , 1975, (No. 288).

1948 ANALYTICAL CHEMISTRY, VOL. 47, NO. 12, OCTOBER 1975