455
Anal. Chem. 1981, 53,455-458 (3) Gilbert, P. T. Anal. Chem. 1982, 34, 1025-1026. (4) Lebedev, V. I,Zh. Anal. Khlm. 1969, 24, 236-239. (5) Willis, J. B. Anal. Chem. 1975, 47. 1752-1758. (6) O’Reilly, J. E.; Hale, M. A. Anal. Lett. 1977, 70, 1095-1104. (7) O’Reilly, J. E.; Hicks, D. G. Anal. Chem. 1979,57, 1905-1915. (8) Fry, R. C.; Denton, M. 6.Anal. Chem. 1977, 49, 1413-1416. (9) Fry, R. C.; Denton, M. 6. Appl. Spectrosc. 1979, 33, 393-399. (IO) Babington, R. S. U.S. Patents, 3 421 692; 3 421 699; 3 425 058; 3 425 059;3 504 859. (11) Popular Science, May 1973,p 102. (12) Suddendorf, R. F.; Boyer, K. W. Anal. Chem. 1978, 50, 1769-1771. (13) Wolcott, J. F.; Sobel, C. B. Appl. Spectrosc. 1978, 32, 591-593. (14) Garbarino, J. R.; Taylor. H. E. 20th Annual Rocky Mountain Conference on Analytical Chemistry, Aug 7-9, Denver, CO; Abstract No. 36. (15) Ward, A. F.; Wohlers, C. C. 1979 Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Cleveland O H Abstract No.
559. (16) Brinkrnann Instruments Inc., Westbury, NY, Technical Literature on Polytron Homogenizer.
(17) Brinkmann Instruments Inc., Westbury, NY, Applications Literature (BR 350).
(18) Newrnann, Steven, personal communication, Brinkmann Instruments Inc.,Westbury, NY.
(19) Fry, R. C.; Northway, S. J.; Denton, M. B. Anal. Chem. 1978, 50, 1719-1 722. (20) Langmyhr, F. J. Ana&st(London) 1979, 704,993-1016. (21) Kapar, S. T.; Tanner, J. T.; Friedman, M. F.; Boyer, K. W. Envlron. Sci. Techno/. 1978, 12, 785. Footnote e, Table 1 on p 788.
RECEIVED for review February 26,1980. Accepted September 17, 1980. This work was presented in part a t the 1979 Pittsburgh Conference, in part a t the 1979 Rocky Mountain Conference, and in part a t the 1979 FACSS meeting. This work was supported in part by the Kansas Agricultural Experiment Station and in part by FDA research Contract No. 223-80-2327.
Determination of Chlorine in Water by Luminol Chemiluminescence D. F. Marino’ and J. D. Ingle, Jr.’ Department of Chemistry, Oregon State University, Cowallis, Oregon 9733 1
The analytical utility of the iuminoi OCI- chemiluminescence (CL) reaction for the determination of OCI- in water is investigated. Optimization data for the reaction conditions in the absence of H202and interference data for over 60 species are provided. The limit of detection of OCI- by this method is determined to be 0.2 pg/L. The accuracy of the new CL procedure is demonstrated by comparison to the standard N,Ndiethyl-p-phenylenediamine (DPD) colorimetric technique. Response of the DPD and luminol systems to chioroamines is essentially identical. The preservation of dilute OCIsolutions is quantitatively investigated. Applications to the routine and trace determination of OCI- in natural waters are discussed.
The principal techniques now in use for the determination of “free chlorine” (HOC1, OC1-, and Clz) in water are listed in Table I (I). None possess a detection limit (DL) lower than 10 pg/L and several are subject to cation and/or chloroamine (combined chlorine) interferences. With increasing concern over water quality, a fast, precise, selective, and highly sensitive method for determination of free chlorine in natural waters would have wide applicability. The chemiluminescence (CL) reaction between OC1- and luminol in the presence of HzOzhas been investigated (2,3) and a DL for OC1- of 16 pg/L was reported, but no interference data were presented. In the absence of HzOz,OC1- also reacts with luminol to produce CL and studies of this reaction have been primarily mechanistic in nature (4, 5 ) . Although no detailed analytical studies have been made, approximate DLs of 0.5 pg/L ( 4 ) and 0.05 pg/L (6) have been reported. This investigation is concerned with optimization and evaluation of the OC1--1uminol CL system for the determi-
Table I. Standard Methods of OC1- Determination detection limit, method &/L interferences residual chlorine electrode I, titration amperometric titration o-tolidine colorimetric o-tolidine-arsenite colorimetric ferrous DPD titration DPD colorimetrica
50
NH,Cl
40 25
Fe(III), Mn(II), NO,Cu(II), Ag(I), NHCl,, ClO,, NCl,, NH,Cl Br,, ClO,, I,, MnO,-, NH,Cl Br,, ClO,, I,, Mn0,-
10 10 50
50
MnO;, Cu(II), NH,Cl MnO,-, Cu(II), NH,Cl
Apparent method of choice. nation of OC1- in water. This luminol-OC1- chemical system was chosen compared to the luminol-HzOz-OC1- system for several reasons. The luminol-0Cl- system had not been investigated for its analytical utility, but previous data indicated it provided superior detection limits. Because many cations in the presence of HzOzand base will activate the luminol CL reaction (61, it is desirable to eliminate HzOZto enhance specificity. Also even “reagent grade” HzOz contains ca. 0.5 mg/L of Fe(III), 0.1 mg/L of Cu(I1) and Ni(II), and 0.1 mg/L of other heavy metals (7)which may be at least a partial cause of the blank reaction when luminol and HzOzare mixed (2, 3,6). Since the preservation of dilute OC1- solutions presented somewhat of a problem due to OC1- disproportionation with time (3,8),concurrent studies were undertaken to minimize this problem through p H control of the sample solution. EXPERIMENTAL SECTION
‘Present address: E. I. du Pont de Nemoura and Co., Wilmington, DE.
All CL measurements were obtained with a discrete sampling CL photometer system reported earlier (9) and with the modi-
0003-2700/81/0353-0455$01.OW0 @ 1981 American Chemical Society
456
ANALYTICAL CHEMISTRY, VOL. 53, NO. 3, MARCH 1981
fications and approximate instrumental conditions previously described (IO). All solutions were prepared from Millipore water (MW) which is deionized water from a Millipore Milli-Q system fed by house distilled water. Luminol (5-amino-2,3-dihydro1,4-phthalazinedione,MCB) was used without further purification. In the optimization portion of this work, MeOH (reagent, Baker) was employed as a luminol solvent. For final work, luminol was dissolved in a 0.05 M pH 10.0 borate buffer. Standard OCl-solutions were prepared from 5% NaOCl (Baker) after standardization with KI and Na2S203( I ) . Chloramine T (l-CH3CsH4S02NCll\la.3H~O,reagent, Fisher) was used as a source of monochloramine. All monochloramine solution concentrations are expressed as micrograms per liter of NH2C1,calculated on a mole ratio basis. A 3% C102solution (technical,Crown Zellerbach) was used as a source of C102after standardization with KI and Naa203. Solutions of OC1-, NH2C1,and C102were prepared just prior to use via fast, successive dilutions in pH 4.0 (HN03)MW. During analysis, these solutions were stored in a brown glass, stoppered bottle. All glassware was carefully cleaned with 50% (v/v) HN03to prevent trace contamination that might accelerate the decomposition of OC1-. All other solutions (e.g.,interferent cations, anions) were prepared as previously described (IO). The general analysis procedure was modified during the course of this investigation. During the optimization studies, 0.5 mL of borate buffer and 1.0 mL of OC1- sample were added to the CL cell with Eppendorf pipets, followed by the injection of 0.5 mL of luminol solution (in MeOH) with an automatic dispensing syringe. Final hypochlorite determinations were made by adding 1.0 mL of OC1- sample to the CL cell, followed by the injection of 1.5 mL of luminol dissolved in pH 10.0 borate buffer. This change did not affect the optimization data. Between runs the cell was rinsed three times with MW. In all cases, it was found that fresh luminol solutions yielded a CL signal ca. 5 X smaller than the same solutions after 24 h of dark storage at room temperature. The reason for this situation is unclear. In any case, solutions of luminol prepared in pH 10.0 borate buffer were stable after the f i t day at room temperature for a minimum of 3 weeks. Colorimetric DPD (N,N-diethyl-p-phenylenediamine) determination of OC1- was carried out with a packaged reagent kit (Hatch Chemical No. 14070) by use of a Turner Model 330 single-beam spectrometer in conjunction with a four-digit digital voltmeter to improve readout resolution. The standard DPD ( I ) and luminol CL procedures were carried out for two tap water samples obtained on two consecutive days. In each case, OC1determinations were made within 3 min of each other. The same procedure was followed in the analysis of an artificial monochloramineOC1- “sample” of known concentration. DPD analyses were run on samples without dilution, while for luminol CL analyses, samples were run after a 1 : l O dilution in pH 4.0 MW, for reasons to be discussed. The definitions for the detection limit (cc) for the anale and other species and the interference limit (cc*) are the same as described previously (11). Basically the detection limit for any species is defined as the concentration yielding an analytical signal equal to twice the standard deviation of the blank. The interference level of a species is defined as the concentration of the species required to produce a change in the OC1- CL signal at the OC1- detection limit equal to twice the standard deviation in the blank. This is evaluated by testing solutions containing the potential interferent plus OC1- at ten times the DL of OC1- (2 pg/L in this case). RESULTS AND DISCUSSION Optimization Studies. The pH of the luminol-0C1- reaction mixture was adjusted with borate buffers of several different values in order to maximize both the CL signal and the signal to background ratio (S/B). To ensure that sufficient buffer capacity was being used, we took all pHs before and after completion of the CL reaction, and the final cell pHs were used for the optimization plots. CL signals in all plots are expressed in terms of the peak signal photoanodic current. Figure 1illustrates the results of the pH optimization. Note that a blank reaction between M W , buffer, and luminol begins to occur at pH 10.5, and steadily increases with increasing pH.
PH
optimization of the lumlnol-OCI- reaction: [OCl-J = 5 [luminol] = 1 X M.
Flgure 1. pH pg/L, 100
80
5
0 e 60 4
0
-I 2
40
L1 w
8 20
0
Flgwe 2. Stability of a 40 pg/L OCT solution. [luminol] = 1.8 X M.
CL conditions:
pH 10.0,
Thus, although a maximum in CL signal exists at p H 11.0, the S/B is unfavorable. Also, the CL signal drops off very rapidly from the maximum a t pH 9.0 as the pH is reduced. Small changes in pH cannot be tolerated at pH 9.0. As a consequence, pH 10.0 was chosen as an optimum value, since no blank reaction occurs and variations of up to ca. h0.5 pH unit can be tolerated with little change in the CL signal. I t is of interest to examine the general shape of the optimization curve. The right-hand portion of Figure 1 closely resembles Seliger’s (5) plot of the fluorescence quantum yield of aminophthalic acid, the suspected luminescing species in this reaction, as a function of pH. The left-hand portion of Figure 1 closely resembles a plot of the [OCl-]/ [HOC11 ratio as a function of pH ([OCl-]/[HOC11 = (3.4 X 10-8)/[H+])(8), leading one to conjecture that only the OC1- ion is effective in producing the luminescing species. With the pH of the reaction mixture fixed a t 10.0, the luminol concentration was varied from 3 X lo4 to 1 X lo-’ M. The CL signal increases with luminol concentration about 3 orden of magnitude over this range but is relatively constant above 1 X M luminol. For solubility reasons, a luminol cell concentration of 1.8 X M was chosen as optimum (precell concentration of 3 x M). As mentioned, the above optimizations were carried out with separate luminol and buffer solutions. The reason for the aforementioned combination a t the two solutions into a single solution at luminol dissolved in buffer will become obvious upon examination of Figure 2, a plot of percent Oc1remaining in a 40-pg/L sample after a 10-min “incubation” in MW of various pH values. Contrary to Chapin’s (12) conclusion that OC1- is most stable a t pH 13.0, it was determined that dilute solutions of OC1- are most stable a t pH 4.0. It is presumed that this is due to the tendency of the hypochlorite ion to disproportionate in basic solution: 30C1- * 2C1- + C103-, for which K is 1 X lon (8). With K, for HOC1
ANALYTICAL CHEMISTRY, VOL. 53. NO. 3, MARCH 1981
chemical effect. The mole ratio of OC1- to luminol at the point a t which curvature begins is about 2 to 1,indicating that the system is limited by the amount of luminol present. The 2:l stoichiometry observed agrees with Seitz’s proposed OC1mechanism for luminol CL (4), which has a first-order rate limiting step. Hence, to extend the dynamic range of this method, one could use a more concentrated luminol solution, although the solubility limit of luminol precludes any dramatic improvements. There is no blank reaction under the conditions used. The standard deviation in blank runs is due to dark current noise from the photomultiplier tube detector. Bubbling the buffer luminol solution with N2 to remove O2had no effect on CL signals. Interference Studies. Table I1 lists slopes, detection limits, and interference levels for various species. Of the species listed, those that could possibly cause interference in a typical water sample may be narrowed to Fe(III), NH2C1, CIOz,Fe(II), Mg(II), and Ca(I1) and MnO,. Fe(II) would only be present in reducing waters and destroy the OC1-. Fe(I1) has been reported to be one of the few metals which activates luminol CL in the absence of H202and analytical use has been made of this characteristic (13). Note for some transition metals and Ca(I1) and Mg(I1) that ct > > ct* and that the interference effect is depressive (negative M). Since no CL reaction with luminol was observed in the absence of OC1-, these metal ions probably decompose some of the OC1-. In real samples, this would occur naturally before the analysis so the OC1- procedure would still measure the desired remaining free chlorine. The CL activation by Fe(II1) may be due to Fe(I1) contamination since the Fe(I1) DL is so much lower. Thus, overall there are few interferences to the CL technique. This is especially true for drinking water analysis for two reasons. First, a 1 to 10 dilution of a tap water sample is required to bring the OC1- concentration (typically 0.5-1.0 mg/L) onto the linearportion of the calibration curve although alternately a 0.1-mL instead of 1.0-mL sample size could be used. This will reduce the typical concentrations of Fe(III), Ca(II), and Mg(I1) below their interference levels. Second, the interference levels indicate the effect on a OC1- determination at the DL of OC1-, while an actual determination in tap water is made at OC1- levels about 1000 times greater than the DL. C102is not often used in water chlorination due to the problems of on-site generation incurred with this explosive gas. C102 also appears to be a universal interferent for all approved OCl- methods (1). Monochloramine, however, is present in most natural waters and does interfere in the absence of OC1-, although its DL is 30 times higher than that of OC1-. In the presence of OC1-, however, it is not a serious interferent, as will be shown in the DPD-luminol comparison study that follows. Ethylenediaminetetraacetic acid (EDTA) was initially added to the buffer solution to reduce the effect of potential metal interferences. However, because EDTA is an amine, it reacted with OC1- and hence its use was discontinued. Analysis of Tap Water. Two tap water samples, taken on two separate days, were analyzed for OCl- via both the DPD colorimetric procedure (1) and the luminol CL procedure. In addition, a synthetic “unknown” containing 660 pg/L NH2Cl
Table 11. Interferents in the OCl--Luminol System“
interferent Fe( 111) Fe( 11) Cu(11) Zn(I1) Mn(I1) Mg(I1) Ca(I1) Co(I1) Cr(II1) Cr(V1) Ni(I1) Os(IV) Pt(1V) Mo(V1) Cl-
Io4-
Io3BrMn04C0,ZNH,C1 c10,
cl, mg/L
0.17 0.00006 0.11 >lo0 >lo0 >lo0 >loo 0.32
100 mg/L, anions with c , > 1000 mg/L: Na(I), K(I), AI(III), Sn(IV), Hg(II), Pb(II), Ag(I), V(IV), Ti(IV), Sc(III), Ru(III), Rh(IV), Ir(IV), In(III), Cd(II), As(III), As(V), W(VI), Pd(II), Ba(II), Rb(I), Sr(II), La(III), Ga(III), Ge(IV), Nb(V), Ta(V), Re(VII), Au(III), F - , ClO,-, S*-,SO4*-, I-, NH,’, humic acid. m= slope of calibration curve of CL signal vs. [interferent]. M = slope of enhancement or depressive curve, ACL signal of 2 pg/L OC1- vs. A [interferent]. equal to 3.8 X virtually all OC1- exists as HOC1 a t pH 4.0 (7).At pH 8.0 the [OCl-]/[HOCl] ratio is ca. 3/1, and the OC1- in dilute OC1- solutions virtually vanishes soon after preparation. In light of these data, addition of a pH 10.0 buffer to the OC1- sample prior to luminol injection could (and did) result in poor reproducibility of the CL signal due to the finite and somewhat variable time between buffer addition and luminol injection. Injection of luminol and buffer together served to reduce the relative standard deviation (RSD) in the CL signal from 11to 1% a t the 2 wg/L level. At the higher OC1- concentrations normally found in tap water, the rate of OC1decomposition is much lower. Providing that the sample is stored in a brown glass stoppered bottle, ca. 3% of the OC1decomposes in 10 min. With simultaneous luminol-buffer injection and optimized reagent concentrations, fast CL peaks with a half-width of about 0.4 s and a base width of about 0.8 s resulted. Fortunately, a check with a storage oscilloscope verified that our fast chart recorder was not attenuating these signals. It was necessary to use a 1-Hz cutoff frequency on the Spectrum 1021 noise filter to prevent attenuation of the peaks. The CL peak height is proportional to the OC1- concentration from a DL of 0.2 pg/L up to 500 wg/L, at which point the calibration curve starts to level off. This is primarily a Table 111. Results of CL Tap Water Analyses sample tap water 1 tap water 2 synthetic
[OCl-] DPD, [OCl-] CL, Mg/L pg/L 885 578 647
882 585 663
457
IXSD, %
DPD
CL
re1 % difference
6.2 5.1 4.8
1.3 0.9 2.0
0.3 1.2 2.5
-
abs error, % DPD CL
1.4
1.0
458
Anal. Chem. 1981, 53, 458-461
and 656 pg/L OC1- was analyzed for OC1- via the two procedures. Results appear in Table 111. It is apparent from these data that the luminol CL technique is suitable for the routine determination of OC1- in drinking water and that it gives results comparable to the standard DPD colorimetric method in terms of accuracy and freedom from major interferences but is more precise and has a detection limit over 2 orders of magnitude lower. Equal NHzCl and OC1- levels lead to only a ca. 1 % error in both methods. The CL technique requires only one reagent solution and is rapid in that about 90 samples/h can be run once solutions are prepared. The detection limit, lower than achieved with any-other standard method for OCl-, allows for studies of the fate of residual chlorine after reaction with water constituents. LITERATURE CITED (1) ”Standard Methods for the Examination of Water and Wastewater”, 14th 4.;American Public Health Association: Washington, DC, 1975; pp 304-349.
Isaccson, U.;Wettermark, G. Anal. Chlm. Acta 1976, 83, 227-239. Isaccson, U.; Wettermark, G. Anal. Lett. 1978, 7 7 (I), 13-25. Seitz, W. R. J . Phys. Chem. 1975, 79, 101-115. Seliger, H. H. “Llght and Life”; McElron, W. D., Gloss, B., Eds.; John Hopkins Press: Baltimore, MD, 1961; pp,,200-205. Seitz, W. R.; Hercules, D. M. Chemiluminescence and Bioluminescence”; Cormier, M. J., Hercules, D. M., Lee, J., Eds.; Pienum: New York, 1973; pp 427-449. Flsher Chemical Index 71C; Fisher Scientific Co.: Pittsburgh, PA. Cotton, F. A.; Wiikinson, G. “Advanced Inorganic Chemlstry”, 3rd ed.; Intersclence: New York, 1972; pp 477-479, 878-879. Hoyt, S.; Ingle, J. D. Jr. Anal. Chlm. Acta 1976, 87, 163. Montano, L. A; Ingle, J. D.,Jr. Anal. Chem. 1979, 57, 926-930. Marino, D. F.; Ingle, J. D., Jr. Anal. Chern. 1981, 53, 294. Chapin, R. M. J. Am. Chem. SOC.1934, 58, 2211. Seitz, W. R.; Hercules, D. M. Anal. Chem. 1972, 44, 2143.
RECEIWDfor review July 7,1980. Accepted December 9,1980. Acknowledgment is made to the National Science Foundation (Grant No: CHE 7616711 and CHE 7921292) for partial support of this research. Presented in part a t the 62nd Canadian Chemical Conference, 1979, Vancouver, British Columbia.
Rubber Disk Passive Monitor for Benzene Dosimeter Michael V. Sefton,” Ennlo L. Mastracci,‘ and John L. Mann’ Depaltment of Chemical Engineering and Applied Chemistty, University of Toronto, Toronto, Ontario, M5S lA4, Canada
A disk (3.75 mm X 12 mm dlameter) cut from a sheet of gum rubber has been designed and callbrated to act as a passive dosimeter for benzene in the workplace. The disk absorbs the amblent benzene and Is both ilmltlng resistance to mass transfer and collection element. There was a dlrect relationship between the benzene concentration in the CSz extract and the product of ambient concentration and the square root of exposure tlme over the range of 3-25 ppm (8 h exposure). While the 95% confidence interval Is less than f2 ppm at exposures corresponding to the TLV (10 ppm, 8 h) or hlgher, the error increases at lower concentratlons (f2.5 ppm at 5 ppm, 8 h) due to the apparent trace presence of benzene In blank disks and the llmlted sensitivity of the flame ionizatlon detector at these concentrations. Although field testing Is still required, the rubber disk may be a useful alternative to charcoal badge dosimeters for organic-vapor monitoring.
Passive monitors are becoming an increasingly desirable means of conducting industrial hygiene surveys in the workplace. For organic vapor dosimetry, these monitors typically consist of an activated charcol collecting element separated from the workplace air by a membrane or “draft shield” (1-5). The membrane acts as a limiting resistance to control the rate of mass transfer to the collection element and free the monitor from the effects of worker movement and air currents in the workplace. After exposure, the organic vapor is desorbed thermally or with carbon disulfide or other appropriate solvent and the amount collected quantified by gas chromatography. Unlike conventional active dosimeters which employ a pump to draw a sample of air through a bed of activated ‘Current address: Syncrude Canada Ltd., Ft. McMurray, Alberta, Canada. Current address: Imperial Oil Ltd., Toronto, Ontario, Canada. 0003-2700/81/0353-0458$01.00/0
charcoal, passive dosimeters are less cumbersome and therefore more readily tolerated by the worker. Since there is no pump to be calibrated nor battery pack to be recharged, they are more reliable in the field and easier to use than active dosimeters. However, they have yet to be approved by OSHA for routine compliance monitoring. A simpler, less expensive monitor, which does not suffer from the disadvantages associated with activated charcoal, has been designed and subjected to laboratory evaluation. For benzene dosimetry, this monitor consists of a small natural rubber disk, 3.75 mm thick and 12 mm in diameter, which acts as both collection element and rate controlling membrane. The rubber disk absorbs vapor at a rate which is proportional to the ambient concentration, governed by the diffusivity and solubility of the vapor in the rubber. The diffusion process within the rubber provides the limiting resistance to mass transfer, conventionally associated with the separate membrane. The dissolution of the vapor in the rubber (Le., absorption), rather than adsorption at specific sites as on activated charcoal, is the means of vapor collection or trapping. DOSIMETER DESIGN The size and shape of the dosimeter must be adjusted according to the nature of the vapor of interest and the material chosen for the disk. Natural rubber was a good choice for benzene because of the high diffusivity and solubility in the rubber and the ready availability of the material. HOWever, it would have been desirable to choose a material which contained no contaminating material extractable in carbon disulfide. By use of the properties of the material, the intended extraction and analytical procedures, and the desired sensitivty (as related to the recognized maximum exposure levels), the size and shape of the dosimeter and the constraints governing its use can be defined. The absorption of an organic vapor in a rubber disk can be described in the initial stages (i.e., with less than 60% of 0 1981 American Chemical Society