Anal. Chem. 1994,66, 1841-1846

Optical Fiber Sensor for Biological Oxygen Demand Claudia Preinlnger, Ingo Kllmant, and Otto S. Wolfbels' Institute of Organic Chemistry, Analytical Division, Karl-Franzens University, Heinrich Street 28, A-80 10 Graz, Austria

We describe the first fiber-optic microbial sensor for deter- mination of biochemical oxygen demand (BOD). The sensing membrane at the tip of the fiber consists of layers of (a) an oxygen-sensitive fluorescent material, (b) Trichosporon cu- faneum immobilized in poly(viny1 alcohol), and (c) a substrate- permeable polycarbonate membrane to retain the yeast cells. The layers are placed, in this order, on an optically transparent gas-impermeable polyester support. Tris(4,7-diphenyl-l,lO- phenanthroline)ruthenium(II) perchlorate is used as the oxygen indicator. Typical response times are 5-10 min, and the dynamic range is from 0 to 110 mg/L BOD when a glucose/ glutamate BOD standard is used. The fluorescent signal is affected by various parameters, including the thickness of the layers, the cell density of the yeast, and the rate at which the substrate is passed through the flow-through cell. BOD values estimated by this new biosensor correlate well with those determined by the conventional BOD5 method. The main advantages of this optical sensor are (a) a more rapid estimation of BOD (in comparison to the BOD5 method which requires 5 days), (b) the fact that opticaloxygen sensorsdo not consume oxygen, (c) the possibility of performing in situ monitoring using fiber optics, and (d) the option of designing inexpensive disposable sensor cells.

BOD5 is defined as the biochemical oxygen demand of waste water measured over 5 days under specified standard conditions. The parameter is based on the metabolic activity of aerobic microorganisms and gives an estimation for the amount of oxygen in waste-loaded water required for bio- chemical degradation of organic matter. Although BOD5 is a good indicator of the concentration of organic pollutants in the water, biochemical oxidation is a slow process, and the test, in its present form, takes 5 days until results areobtained. Thus, the conventional test is not suitable for process control and monitoring, where a rapid feedback is desirable. It is therefore of considerable interest to develop alternative methods that may replace this time-consuming test. This was achieved by immobilizing microbes at the tip of an amperometric electrode. Several kinds of microbial sensors for BOD5 have been reported, some based on measurement of a steady-state equilibrium,I4 others measuring in the kinetic m ~ d e . ~ - ~ They consist of microorganisms immobilized on a

(1) Hikuma, M.; Suzuki, H.; Yasuda, T.; Karube, I.; Suzuki, S . Eur. J . Appl.

(2) Karube, I. Biorechnol. Bioeng. 1977, 19, 1535-1547. (3) Kulys, J.; Kadziauskiene, K. Biofechnol. Bioeng. 1980, 22, 221-226. (4) Tan, T. C.; Li, F.; Neoh, K. G. S e w . Acruarors 1992, B8, 167-172. (5) Riedel, K.; Renneberg, R.; KOhn, M.; Scheller, F. Appl. Microbiol. Biorechnol.

Microbiol. Biorechnol. 1979, 8, 289-297.

1988, 28, 316-318. (6) Tan, T. C. Sens. Acruutors 1993, BIO, 137-142. (7) Riedel, K.; Lange, K.; Stein, H.; Kiihn, M.; Ott, P.; Scheller, F. Water Res.

1 9 9 0 , ~ 883-887.

0003-2700/94/ 0366- 184 1 $04.50/ 0 0 1994 American Chemical Society

porous membrane and an oxygen electrode. Various kinds of microorganisms have been used. These include Trichosporon cutaneum, 1.597,8 Bacillus s ~ b t i l i s , ~ J Hansenula a n ~ m a l a , ~ and a mixed culture of B. subtilis and Bacillus l icheniformi~.~~6 T . cutaneum is identical to Trichosporon beigeliiused in other work.

Conceivably, the amperometric measurement of oxygen may be replaced by optical (fluorescent) measurement of oxygen using an or706e (Greek; "theoptical way"). The major advantage of optodes over electrodes in the context of BOD is the fact that, unlike electrodes, they do not consume oxygen during measurement, so that no depletion of oxygen can occur, as occurs during electrochemical measurement. We therefore perceived that the use of some of the oxygen sensors developed by us in the past years would result in a sensor with improved performance. Two sensing schemes were envisaged: (a) placing an oxygen-sensitive membrane on the bottom of the sample vessel and monitoring oxygen over 5 days (in an instrument similar to a bacterial detection system using a carbon dioxide optodeg) or (b) performing the test using a biosensor arrangement using immobilized cells. We consid- ered the latter to be advantageous over the former mainly for the reason of being much faster and therefore providing a rapid feedback signal.

In this work we show that BOD indeed can be measured optically by using a microbial BOD biosensor membrane along with a measuring scheme resembling flow injection. We also show that this approach presents some attractive new features and advantages over electrochemical detection. Although the BOD measured with the biosensor (referred to as the BODS) is not identical to the conventional BOD5, it is shown to be a parameter that correlates acceptably well with the con- ventional test and, hence, is a useful parameter for rapid estimation of water quality.

EXPERIMENTAL SECTION Microorganisms and Cell Growth. The yeast T. cufaneum

(now known to be identical with T . beigelii; DSM, Brunswick, Germany) was grown under standard aerobic conditions in a rotating shaker at 30 OC for 36 h in a medium containing 0.25% malt extract, 0.25% peptone, 0.25% yeast extract, and 1% glucose. The culture broth was centrifuged at room temperature at 5000 rpm for 10 min, and the cell mass was washed twice with a 0.1 M phosphate buffer of pH 6.8.

Immobilization. The washed cell mass was mixed with a 10% aqueous solution of poly(viny1 alcohol) (pva) (MW

(8) Riedel, K.; Alexiev, U.; Neumann, B.; Kahn, M.; Renneberg, R.; Scheller, F. Biosensors: Applications in Medicine, Environmenral Protection and Process Conrrol; GBF Monographs; VCH: Weinheim, Germany, 1989; Vol. 13, pp 71-74.

(9) Swenson, F. J. Sew. Acruarors 1993, B l l , 315-321.

Analytjcal Chemisfty, Vol. 66, No. 17, June 1, 1994 1841

175vm l L - s

Figure 1. Cross-section of a sensing membrane for determination of BODS. 1, polycarbonate cover; 2, layer of yeast immobilized in PVA; 3, ca. 1-pm layer of charcoal acting as an optical isolator; 4, oxygen- sensltive fluorescent layer; 5, inert and gas-impermeable polyester support. Excitation light (from the bottom) passes the polyester support and excites fluorescence In the oxygen-sensltlve layer. Part of the emitted light is collected by the fiber bundle (not shown) underneath the polyester layer and guided to the photodetector.

100 000) in a ratio of 1:l (by weight) and spread in various thicknesses onto an optical oxygen-sensing membrane. The microbial membranes were dried at 4 O C for 24 h and stored at 4 O C until used. Both the oxygen sensor and the immobilized yeast were found still to work and to be useful for BOD determination after a 1-year storage at 4 "C.

Oxygen Sensor Membrane. In 5 mL of tetrahydrofuran (THF) were dissolved 13.5 mg of tris(4,7-diphenyl-l,10- phenanthroline)ruthenium(II) perchlorate [ (Ru(dpp)], pre- pared by a modification of a published method,IO 0.5 g of poly(viny1 chloride) (pvc) (Fluka, Buchs, Switzerland), and 0.5 g of 2-nitrophenyl octyl ether (NPOE) (Fluka). This solution was spread onto a 175-pm polyester film (Mylar, DuPont) acting as an optically transparent solid support. After solvent evaporation, the resulting clear oxygen-sensitive layer on the polyester film had a calculated thickness of around 10 pm. The concentration of the dye in the plasticized pvc film was approximately 12 mM. The red fluorescence of the ruthenium complex, which was reversibly quenched by oxygen, was the analytical information of this ~ y s t e m . ~ I - ' ~

Assembling the Sensor. A cross-section of the microbial optode is shown in Figure 1. A polyester support (Mylar, type GA-10, DuPont, Vienna), being impermeable to oxygen, served as a mechanical support onto which a 10-pm oxygen- sensitive fluorescent layer was spread. The polyester support enables a much easier handling of the sensing layers. The fluorescent layer was covered with a layer of commercial charcoal, which served as an optical isolator. The charcoal was spread evenly onto the pvc layer while still slightly wet, using a thin sieve. The optical isolation prevents ambient light from entering the optical system and blue excitation light from exciting fluorescence in the sample and makes the sensor insensitive to changes in the refractive index of the sample. The black layer was covered by a layer of immobilized yeast by spreading the suspension of yeast in a 10% pva solution onto the membrane, using a home-made spreading device. The preferred thickness of the yeast layer

(IO) Watts, R. J.; Crosby, G. A. J. Am. Chem. SOC. 1971, 93, 3184-3188. (1 1) Wolfbeis, 0. S.; Leiner, M. J. P.; Posch, H. E. Mikrochim. Acta (Vienna)

(12) Bacon, J. R.; Demas, J. N. Anal. Chem. 1987, 59, 2780-2784. (13 ) Carraway, E. R.; Demas, J. N.; DeGraff, B. A.; Bacon, J. R. Anal. Chem.

(14) Moreno-Bondi, M . C.; Wolfbeis, 0. S.; Leiner, M. J. P.; Schaffar, B. P. H.

1986, 3, 359-366.

1991, 63, 337-341.

Anal. Chem. 1990, 62, 2377-2380.

after drying on ambient air was 10 pm. Layer thicknesses were calculated from the volume spread and the amount of water that evaporated during drying. On this layer was placed a porous polycarbonate membrane of pore diameter 0.4 pm (Bio-Rad, Vienna), which was permeable to dissolved organic matter but retained the microorganisms.

Apparatus. The continuous flow system has been described in some detail previ0usly.1~ It consists of an optical sensor membrane, a peristaltic pump (Gilson Minipuls 3, Villiers- le-Bel, France), an automatic sampler (ND 12, Besta, Germany), a fiber-optic photometer (Oriel 3090, Chelsea Instruments, London, U.K.), a 150-W pulsed xenon lamp as a light source, and an R 928 photomultiplier (PMT) (Hamamatsu, Munich) as detector. A 480-nm interference filter was placed in front of the Xe lamp to isolate the appropriate excitation light, and a 560-nm long-pass filter was placed in front of the PMT to block scattered 480-nm light but to allow the red fluorescence, which has a maximum at 610 nm, to pass. Fluorescence intensity data were transferred to a data acquisition unit (Keithley 575) controlled by an

Standard Solution and Determination of the 5-Day BOD. A solution containing 150 mg/L glucose and 150 mg/L glutamate (resulting in a BOD of 220 mg/L) was employed as a standard solution (referred to as GGA solution) for calibration of the BOD sensor according to Riedel et al.' The BOD5 of waste water was determined by the standardized dilution method (DIN 38 409).

Sensor Conditioning. Microbial membranes stored at 4 "C were taken out of the refrigerator 2-3 days before measurement to allow the immobilized yeast to adapt to room temperatureconditions. A 3-mm-diameter spot was punched out of the large sensing membrane and covered with the polycarbonate membrane, and the sensing membrane was placed in the flow cell. Standard solutions of various concentrations and pH were injected into the system and passed over the sensing membrane until a constant signal was attained.

AT-PC.

RESULTS The sensing scheme is based on the measurement of oxygen

consumed by yeast, using an oxygen-sensitive fluorescent membrane similar to previous oxygen-sensitive materials based on luminescent ruthenium c o m p l e x e ~ ~ ~ - ~ ~ but using plasticized pvc as a matrix, which has advantages in terms of strong optical signal, efficient quenching by oxygen (the signal is quenched by 50% in going from nitrogen to air), rapid response, and ease of manufacturing. We therefore first characterized the oxygen sensor. Table 1 gives figures of merit. They show the sensor to have a dynamic range that covers the range of interest, a response time fast enough to monitor the bacterial metabolism, and an excellent long-term stability (except when in contact with samples containing detergent).

The change in the optical signal of this particular sensing material is shown in Figure 2 for the 0-1 50 Torr range, which is the one of interest in this context, along with the respective Stern-Volmer plot. The Stern-Volmer equation (eq 1) relates fluorescence intensity in the absence ( l o ) and presence (I) of oxygen, respectively, to the concentration of oxygen ([02]):

(1 5) Weigl, B. In Chemical, Biochemical, and Environmenral Sensors III; Liekerman, R. A., Ed.; Proc. SPIE-Int. SOC. Opt. Eng. 1991,1587,288-295.

1042 Analytical Chemistty, Vol. 66, No. 11, June 1, 1994

Table 1. Figures of MerH for the Oxygen Sensor Made from Plasilclzed Poly(viny1 chiorlde)

Stern-Volmer constanta quenching on going

from nitrogen to air oxygen sensitivity fluorescence intensity response time photostability

storage stability operational lifetime leaching of the dye

leaching of the plasticizer

reproducibility of K ,

0.0073 Torr'

0-200 Torr very high <1 s to gases, <1 min to liquids no bleaching detectable with a

6+ months 1 week minimum not detectable when stored in

buffer a t room temperature negligible in water, strong in

water containing detergent h2% (n = 6)

-52 %

blue LED after 2 weeks

nd interferences by

interferences by SOz (20 ppm)

0 Initial slope.

COZ (IO00 ppm), HzS (100 ppm), NH3 (50 ppm)

loll re1 intensity I I arb. units 2.5 1100

I I

2 -

1 1,5

1 0 50 100 150

pO2 / torr

+ Stern-Volmer plot + intensity plot

Figure 2. Stern-Volmer plot and intensity plot of the quenching of the fluorescence of Ru(dpp) by oxygen.

I 1

0 10 20 30 40 50 60 70 80 time (miid

Figure 3. Response time, relative signal change, and reversibillty of the microbial sensor when exposed to a standard BOD solution (GOA) having a BOD of 11 0 mg/L.

K , is the so-called Stern-Volmer constant. Most importantly, the resulting plot (Figure 2) is virtually linear, a fact that greatly facilitates calibration and, in principle, makes possible a 1-point calibration.

Figure 4 shows typical response curves of the optical biosensor using immobilized T. cutaneum. Initially, the Ru- (dpp) fluorescence is partially quenched by the oxygen dissolved in the buffer. This signal represents the fluorescence in absence of substrate and to some extent also reflects the

endogenous respiration of the immobilized microorganisms. After approximately 10 min, a synthetic sample solution containing glucose and glutamate was injected into the system. The organic compounds permeate through the porous mem- brane and are metabolized by the microbial cells, which thereby consume oxygen. As a result, the oxygen partial pressure inside the oxygen-sensitive pvc layer decreases until a steady- state oxygen concentration is reached. The change in signal can be related to BOD. When pure buffer is passed over the sensor system again (after ca. 25 min in Figure 3), fluorescence returns to the initial level. Response times are on the order of 5-10 min.

Calibration. A calibration curve was established for the microbial biosensor membrane using diluted GGA stan- dard solutions. As can be seen from Figure 4, a linear relationship was observed at below 110 mg/L BOD of the standard. The minimum detectable BOD is around 2-3 mg/ L. The signal was reproducible to within f4% of the mean value in a series of 10 samples of 5 5 mg/L BOD standard GGA solutions.

Effect of pH. The physiological state of microorganisms, in particular their respiratory activity, strongly depends on pH. Consequently, its effect on the optical biosensor was examined. The 55 mg/L glucose/glutamate standard was employed for these experiments. Figure 5 demonstrates that the maximum signal change is obtained at pH 6.8.

Effect of Thickness and Cell Concentration of the Microbial Layer. Clearly, one would wish to have a microbial layer as thick as possible in order to achieve a large signal change even at low BOD. However, increasing the thickness of the microbial pva layers to above 100 pm causes the response time to increase to up to 30 min, the relative signal change to decrease by up to 25%, and the conditioning of the sensing membrane to take much more time, typically 5-1 2 h. Figure 6 shows the effect of varying cell density of T. cutaneum in the pva layer on the response. It is obvious that sensing membranes containing high cell concentrations are more sensitive but have a smaller dynamic range. Thus, by varying the cell density, membranes with dynamic ranges up to 110 mg/L BOD may be produced. This does not come unexpected because (a) substrate diffusion is slower in highly loaded layers and (b) more oxygen is consumed by the cells through endogenous respiration.

Sensor Lifetime. Sensors were reactivated after various times of storage. Typically, there was a 30% drop out after 4 weeks of storage in a lab refrigerator at 4 "C. Of those sensors that could be reactivated, all had an operational lifetime of 1-2 weeks, sometimes even 30 days. Similar stabilities have been reported by A major problem is reconditioning. Even of those made in the same batch, some sensors could not be reactivated after storage, while others had no problem. As a result, all sensors had to be recalibrated after reconditioning.

Variation of Substrates. The response of the biosensor to various organic substrates was tested, and the resulting response curves are shown in Figure 7. The shape of all work functions is similar, but there are distinct differences between glucose and maltose on the one side and pyruvate on the other. The slope, and hence the sensitivity of the optode, is bigger for the former two than for L-glutamate, ethanol, and pyruvate.

Analytlcal Chemistry, Vol. 66, No. 11, June 1, 1994 1043

0

Figure 4. Optical signal changes of the BOD sensor when contacted with standard (GGA) solutions of increasing BOD, and the respective work function.

0 15 30 45 60 75 time (mid

Flgure 5. Response of the microbial sensor to GGA standard solutions (55 mg/L BOD) of various pH.

Figure 8. Effect of cell densityon the measured change in fluorescence. Membranes with 16 (curve 2) and 32 mg (curve 1) of dry yeast per milliliter of pva suspension were employed.

A distinct change in slope occurs at substrate concentrations of around 0.1 mM, but at this point we can only speculate as to the reason. Possibly, the effect is related to the endogeneous respiration of the cells, Le., consumption of oxygen in the absence of digestable substrate as a result of the normal (endogeneous) respiration of cells.

Application. The microbial optode sensor was applied for the estimation of BOD of untreated waste waters from sewage plant effluents and municipal sewage. The resulting oxygen demand is referred to as BODS in order to differentiate it from the standard BODS. The latter was determined by the conventional dilution method. As shown in Figure 8, a fairly good correlation between BODS and BOD5 was obtained ( r = 0.9704; n = 12). The ratios between optodes BODS and BOD5 ranged from 0.72 to 1.66. There is good reason to assume that this variation is caused by the different components

- 1 1 2 50 3

a 4 5 40 5 v, b 30

&20

$10

8! ' 0 2 ' o k 0 6 ' 0'8 ' i o concentration (mM)

Figure 7. Effect of varying substrates on the respiratory activity and response of pva-immobilized T. cufaneum in the BOD biosensor. 1, ffilucose; 2, maltose; 3, o,L-lactate; 4, ethanol; 5, L-glutamate; 6, pyruvate.

80 ,

Figure 8. Correlation of BODS data as determined by the microbial optode sensor (measurements in triplicate) with BOD5 values (obtained by the conventional 5day method). (- - -) Calculated line of slope 1; (-) regression line ( r = 0.9704).

contained in waste water, all of which are degraded at different rates. It is known, for example, that large molecules such as starch and cellulose are slowly degraded by the metabolic enzymes of immobilized yeast. Very large molecules do not even pass the polycarbonate cover membrane of the sensor presented here. Consequently, they escape BODS detection. The BODS data, therefore, are generally lower than the BOD5 data in case of high polysaccharide levels. A high BODS, in contrast, may be caused by the higher respiratory rate of T. cutaneum for certain substrates contained in sewage.

1844 Analytical Chemistry, Vol. 66, No. 11, June 1, 1994

Table 2. Flgureo of MerH for Varlow BODS Teeto upper limit of detection

ref microorganism sensing scheme response tim (mg/L BOD) 1 T. cutaneum amperometric, steady state 18 min 60 2 C. butyricum amperometric, steady state 15 min 300 3 H. anomala amperometric, steady state 15-20 min 5 B. subtilis amperometric, kinetic 15-30 20 5 T. cutaneum amperometric, kinetic 15-30 8 100

this work T. cutaneum fluorescence quenching 3-10 min 110 6 B. subtilis plus B. lichenijormis 7B amperometric, kinetic 15-30 s 80

DISCUSSION The results show that optical sensing of BOD is an attractive

alternative method to the conventional BOD5 test. At pH 6.8 and 30 O C , the BODS compares favorably with the BOD5 in terms of relative signal change and response times. The optode sensor is comparable in performance with electrochemical steady-state BOD sensors (Table 2). The time required for an assay is comparable with the steady-state amperometric methods (entries 1-3 in Table 2). BODS sensors using kinetic signal evaluation (i.e., by relating the first derivation of the current/time curve to the increase in yeast metabolism) are inherently faster because there is no need for establishing an equilibrium. Although not done so far, we have all reasons to assume that the optode sensor can be submitted to kinetic data evaluation as well. In each instance, BOD sensors give results in much shorter times than the conventional 5-day assay. Table 2 compares figures of merit of the various types of amperometric BOD sensors.

Steady-state electrochemical sensors are comparable in response time and upper limits of detection, while kinetic electrochemical methods are clearly faster and, in one instance, cover the low BOD range. Optical biosensing of BOD may have advantages over amperometric methods because no oxygen is being consumed by the oxygen transducer. Elec- trochemical reduction of oxygen can prevent BOD assay in cases of samples with low oxygen loading because this limits the amount of oxygen supposedly consumed by the microbes.

The BOD optode sensor spots described here can be manufactured at very low costs from standard plastic materials and minute amounts of a fluorophore. In our case, they are fabricated in standard-size sheets (A4 or B4), and 3-mm diameter spots are punched out of the sheet. Although not intended when starting the work, we now think that-in view of the costs-one application of the microbial biosensors is in low-cost disposable sensors.

The yeast T . cutaneum, although pathogenic,I6 was shown to be an efficient biodegrading agent for organics in aqueous solution, with good kinetics, sensitivity, stability, and repro- ducibility. Notwithstanding this, other microbes (some given in Table 2) may be used as well. Both bacterial and yeast species have been used previously, including Clostridiums (B.) and Trichosporon- and Hansenula-type yeast. More recently, a bacterium/ yeast combination (Rhodococcus eryth- ropolis/Zssatchenkia orientalis) has been implemented in a commercially available amperometric BOD meter. Two- bacilli systems (such as a B. subtilisllicheniformis combina- tion) have been applied as well. A highly practical sensor results from the useof activated sludge containing the kind of

(16) Leblond, V.; Bellefiqh, S. Cancer 1986, 58, 2399-2405.

microbes found in sewage and therefore being identical to the species that create a BOD. Soil bacteria have been used as well.

The relation between BODS and the measured fluorescence signal (I) can be deriyed from the Stern-Volmer equation. The situation resembles that of enzyme-based biosensors with oxygen optode transducers.14J7 The fluorescence intensity obtained with the sample in the absence of any oxygen consumption (la) is a measure for the initial oxygen concen- tration which can be calculated from la and IO via the Stern- Volmer equation (eq 1). Let us set the signal obtained before microbial action, i.e., (l0/Za - l ) , equal to a:

a = ~S,[O,l (2) Following microbial action, oxygen will be partially or totally consumed. An additional term [-A(O2)] has to be added to eq 2 to account for this. Under steady-state conditions, a new Z value will then be measurable, which we may call Ib. The respective signal is (ZO/Ib - l), and we set it equal to P:

P = ~S,[O,l - KS,[A(O2)1 (3) The difference in oxygen concentration [A(O2)] caused by microbial oxidation of organic matter can be calculated by subtracting the two equations:

a - P = - ~ s v [ A ( 0 2 ) l (4)

The relation between [A(O,)] and BODS can be described by

[A(O,)If= BODS ( 5 )

where f represents a conversion factor that accounts for the unknown rate of consumption of oxygen by bacteria. Factor f is governed by type and activity of the microbes, pH, temperature, and other parameters. If measurements are to be performed in a kinetic way, f also depends on the time period between the first (a) and the second measurement (P ) .

By combining (4) with ( 5 ) , we end up with a fairly simple relation between BODS and the two optical signals:

(6) This equation allows calculation of BODS from two fluo- rescence intensity data. It is independent of the initial oxygen concentration and, hence, allows determination of BODS even in cases of samples which are not air-equilibrated but already have suffered from oxygen depletion, a situation frequently encountered in practice. It is applicable, though, only over the linear part of the Stern-Volmer equation and under the

BODS = (P - a)/ fK, ,

(17) Wolfbeis, 0. S . Ed. Fiber Opfic Chemical Sensors and Biosensors; CRC Press: Boca Raton, FL, 1991; Vol. I, pp 93 ff.

Analytical Chemistry, Voi. 66, No. 17, June 1, 1994 1845

provision that enough oxygen is present so as not to end up with complete oxygen consumption.

Figure 8 demonstrates that the BODS is similar but not identical to the BODS. This has been known for quite some time and is due to the fact that immobilized microbes are slow in responding to large molecules such as starch and cellulose derivatives, which are species frequently encountered in domestic sewage and industrial effluents.

In conclusion, we think we have shown that optical biosensing of BOD is possible and has certain advantages over amperometric methods in that no oxygen is being consumed by the oxygen transducer, a fact that can prevent BOD assay in cases of samples with low oxygen loading because the amperometric sensor consumes the oxygen supposedly consumed by the microbes. Another interesting feature of BOD optodes is their cost. In contrast to conventional amperometric oxygen electrodes, the oxygen sensor spots

described here can be manufactured at very low costs from standard plastic materials and minute amounts of a fluoro- phore. Hence, low-cost disposable sensors for single use may be envisioned.

ACKNOWLEDGMENT Financial support by the FWF, project S 5702, is gratefully

acknowledged. The authors would also like to thank Mr. J. Rosmann (Styrian Government Laboratories) and Dr. Stuhl- bacher (Institute of Waste Technology, Graz University of Technology) for supplying sewage waste water samples and performing standard BOD measurements.

Received for review January 3, 1994. Accepted March 8, 1994.'

*Abstract published in Advance ACS Abstracts. April 15, 1994.

1846 Analytical Chemistry. Vol. 66, No. 11, June 1, 1994