was constantly operated a t 350 "C f 3. The column used for gas chromatography was 1-m x 3-mm stainless steel, packed with 3% OV-1 on 100-mesh Chromosorb W. The pressure of the helium carrier gas was 0.7 kg/cm2. The mass spectrometer was operated a t 10,700-G magnetic field and 70-volt ionizing current. The time interval of rapid recording between 0 and 400 mass units was 2 seconds. Chemicals. The methyl esters of gibberellins A1 and & were prepared by the method reported by Takahashi et al. (9). The methyl ester of gibberellin As was supplied by Dr. MacMillan's Laboratory. Ten micrograms of each of these compounds were mixed and dissolved in 0.5 gram of acetone solution. Regenerator Assembly. The instrumental assembly for regeneration of the molecular separator by ultra-shortwave plasma is shown in Figure 1. A used separator, from which the organic materials should be removed, is connected a t the center part of the assembly with rubber tube joints. The two ring-shaped copper electrodes are rolled around the separator glass tube and are connected to a 200-W microwave generator, Itho-Chotanpa Co. KSF type 27 MHz. After the oxygen gas flow is adjusted a t 200-600 ml/min a t 7-2 Torr, the ultra-shortwave discharge is ignited and operated continuously for 3 hours. The electrodes can be placed a t any part of the separator with the desired gap. After 3 hours, if the pale blue color of the glow due to Swan band (Gband) is still observed, discharge must be continued further. When organic compound residue on the porous glass tube is removed completely, the color of the glow turns into pink. When it is completed, a used dark brown porous glass turns into the snow-white original color. The temperature of the molecular separator during the discharge is lower than 200 " C .
RESULTS AND CONCLUSION The mass spectra of the methyl ester of gibberellin & purified from the mixture of the methyl esters of gibberellins AI, &, an.d & is shown in Figure 2. The spectrum in Figure 2A is obtained when the molecular separator in the Takahashi, H . Kitamura, A. Kawarada, Y . Seta, M . Takai, S. Tarnura, and Y . Sumiki, Bull. Agr. Chem. SOC.Japan, 19, 267 (1955);
(9) N .
23, 405 (1958).
apparatus had been used repeatedly for a long time. It is seen that the spectrum is completely different from that of the pure methyl ester of gibberellin & which has a parent peak at 344 mass number. Peaks at 346 for gibberellin & and 362 for gibberellin A1 are also recorded on the spectrum, indicating many kinds of materials were impregnated prior to use in the porous glass of the separator. On the other hand, the spectrum of the same gibberellin A5 shown in Figure 2B is taken after cleaning the same molecular separator by ultra-shortwave discharge with oxygen. This spectrum shows only the sharp peaks of the pure methyl ester of gibberellin &, without any residual impure spectra. From this example, it is clear that this cleaning method is more useful for the Watson-Biemann type molecular separator. From other examples similar to those presented in this paper, the method of regeneration of used molecular separators for the GC-MS spectrometer by microwave plasma is found effective for most of the organic materials. It is simple and perfect without causing any damage to the apparatus. Although at present we are using a cleaning assembly separate from the spectrometer, the principle can be applied to the on-line system, so that the molecular separator would be cleaned up without being detached from the apparatus. The method can also be applied for other types of separator rather than the Watson-Biemann type studied in the present paper, provided the separator is made of glass,
ACKNOWLEDGMENT The authors wish to acknowledge the valuable suggestion of Hitoshi Kamada and Jake MacMillan for donating the methyl ester of gibberellin A g . Received for review October 18, 1973. Accepted December 26,1973.
Device for the Automatic Compensation of Current Transducer and Integrator Errors for Chronocoulometry and Related Techniques W. S. Woodward, T. H. Ridgway, and C. N. Reilley The William R. Kenan, Jr., Laboratories of Chemistry, University of North Carolina, Chapel Hill, M.C. 27574
Chronocoulometry is a valuable technique for the study of chemical kinetics (1-7) and adsorption phenomena (71 1 ) whose utilization will increase as more laboratories implement the computerized data acquisition, control and analysis systems (5, 10, 12) required for highly precise re(1) J. H. Christie, J . Electroanal. Interfacial Electrochem.. 13, 79 (1967). (2) S. W . Feldberg, in "Electroanalytical Chemistry," Vol. 3, A. J. Bard, E d . , Marcel Dekker, inc., New York, N . Y . , 1060, pp 199-296. (3) H . N . Blount and H . E . Herman, J . Phys. Chem., 72, 3006 (1968) ( 4 ) R . P. Van Duyne, P h . D . Dissertation, University of North Carolina, Chapel Hill, N.C.. 1971. (5) T . H . Ridgway. P h . D . Dissertation, University of North Carolina, ChaDel Hill. N.C., 1971. (6) T H Ridgway, R P Van Duyne, and C N Reilley, J Electroanal Interfacial Electrochem 34. 267 11 , 9721 , (7) R. P. Van Duyne, T. H . Ridgway. and C . N . Reilley, J. Electroanal lnterfacial Electrochem., 34, 283 (1972) (8) J. H . Christie, R A Osteryoung, and F . C. Anson. J . Electroanal. Interfacial Electrochem., 13, 236 (1967) (9) F. C. Anson. J. H . Christie, and R. A. Osteryoung, J. Electroanal, Interfacial Electrochem., 1 3 , 343 (1967) (10) G. Lauer, R Abel. and F . C. Anson. Anal. Chem.. 34, 765 (1967). (11) F. C. Anson and D. J. Barclay, Anal. Chem.. 40, 1791 (1968). (12) S. P. Perone, Anal. Chem., 43, 1288 (1971) and references there-
in.
sults. A major practical problem encountered with either the single- or multiple-step form of this method arises from the drift of the charge integrator. This drift error can manifest itself both as a non-zero output a t the initiation of the experiment and as a time-varying component during the experiment (13, 24). The first error is readily eliminated by means of FET switches which short the integrator capacitor until initiation of the experiment ( 5 ) . The dynamic error is much more difficult to eliminate and may arise from sources other than the integrator itself. For short-time experiments-Le., ten seconds or lessusing modern FET, super-beta, chopper, or electrometer operational amplifiers, the drift is essentially linear with time. This linear ramp is either added to or subtracted from the signal arising from the cell process and obviously distorts the measured response. The monitored charge response for single potential step (SPSCC) and the forward step of double potential step (DPSCC) chronocoulometry is theoretically linear with t 1 t 2 for planar diffusion and in (13) L. L. Schick, IEEESpectrum, April, 36 (1971). (14) E. E. Bonnelycke, Anal. Chem.. 43, 610 (1972)
ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974
1151
r - - - - -:--i
r
I
----
- -c-i
r------i--i
Figure 2. Sources of error in current transducers and integrators
E;
Figure 1. Circuit configurationsfor current integration The compensation circuitry shown in dashed boxes consists of a track and hold device activated by S and with a gain of -G.
the absence of mechanisms which involve multiple charge transfer steps or kinetic paths linked to the initial solution species. Deviation from linear behavior is usually taken as indicative of some form of kinetic complication. An analysis of the data acquisition system in use a t this location indicates that the correlation coefficient for such a regression should exceed 0.99990, lower values implying the existence of contributions from sources other than planar diffusion. In practice, this criterion was met only at the expense of continuous “fine tuning” of the current transducer and integrator. In an attempt to obviate this experimental nuisance, a study of the prominent error sources was made and from that analysis an auto-compensation circuit was developed.
ANALYSIS OF ERROR SOURCES Given that the potentiostat itself is ideal, with infinite compliance, zero rise time, no overshoot or ringing and perfect iR compensation, then the only possible error sources are offset currents and voltages in the transducer section of the instrument. There are three common, basic ways, shown in Figure 1, to construct the transducer portion of the instrument: connect the working electrode (WE) to the summing point of an operational amplifier (OA), Figure IA;connect WE to ground through a small resistance and monitor the potential drop and, hence, the current flowing through Rm with a voltage follower OA, Figure 1B; place the resistor before the auxiliary electrode as in Figure 1C. This last circuit imposes a stringent com1152
ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, JULY 1974
mon mode rejection criterion on the transducer and will not be treated in detail below. The additional circuitry enclosed in the dashed boxes of Figure 1 represents the placement of the compensation scheme described below for the three basic circuits. In the first method, one may place either a resistor or a capacitor in the feedback loop of the transducer OA. While use of a capacitor permits monitoring the charge directly, this scheme eliminates the direct indication of the current flow, Et required for iR compensation and, hence, will also not be treated in detail. We will now consider several prominent factors which lead to errors at the integrator output with a view toward eliminating their contribution through a suitable compensation circuit. For the circuits in Figure 1, Et = icRm. However, as indicated in Figure 2A and B, the presence it) east where it of east and i+ or i- causes Et = R, (i, is the negative input bias current i- for the current follower form in Figure 2A, or the positive input bias current i, for the voltage follower in Figure 2B. If resistor R, is inserted between the + input and ground in Figure 2.4, Et = Rm(& + i-) - R,i, + cost. The effects of the individual bias currents i, and i- can be minimized by proper selection of R,. The output of the integrator, Figure 2C, preceded by current transducer A or B is given at time T by
+ +
E , = e,,,
+
( ~ / R , C , ) J - [ (+ L ?L J R , - z,R, - (eosl- e,,,dldt
(1)
where the initial voltage across C, was zero, The error associated with this measurement is, thus,
E,,,
= eosi
- ( 1 / R C , ) i - ( i , R-, i t R m+
cost
- e,,,)dt
(2)
While the current and voltage offset errors are a function of both time and temperature, once the OA’s warm up, the change in these terms over a few tens of seconds is usually negligible and, hence, they will be treated as static quantities. It is highly desirable that once one balances the instrumentation at the beginning of a series of experiments, no further “fine tuning” should be necessary for a reasonable period of time. Almost any OA can be trimmed so that its voltage offset will be less than 1 mV several hours after balancing, and we can ignore the eosi term outside the integral in Equations 1 and 2. Most FET, chopper, or electrometer OA’s will have bias currents in the range of 100 pA or less and even the super-beta OA’s achieve less than 1 nA with current balancing. Excluding super beta OA’s and assuming a worst case bias current of 100 pA and 100 K as a typical maximum value for Ri leads to a voltage equivalent, ill$, of 10 pV. For the transducer amplifier even using a super beta OA with 1 nA bias current and a maximum value of R, of 5 Kohms, a voltage equivalent i,R, of 5 pV results. R , is generally even smaller to allow rapid double layer charging. The only OA’s which can maintain at e,, value in the 1- to 10-pV range for around one half hour after offset trimming are almost uniformly characterized by high cost, lower power, and/or low slew rate and frequently cannot be used in the follower mode. Thus, the e,, contribution is the major factor contributing to E,,, in Equation 2. AUTO-COMPENSATION SCHEME The value of E,,, can be minimized by applying the principle of chopper stabilization to the transducer-integrator system itself as indicated in Figures l.4 and 1B. This procedure is quite effective and results in no more than 0.5 mV/second drift for periods of less than 10 seconds and with no critical adjustments required for periods of several hours. The drift of the system is eliminated by applying a drift compensation current into the current path of the system and holding the most recent correction during an actual experiment. This is most simply implemented by monitoring the integrator output with a trackand-hold amplifier with a gain of -G (Le., inverting) which goes into the hold mode under control of S during the experiment and tracks at all other times. The output of this device is fed through a resistor R, into the system current path. In order to perform a cancellation, the track and hold must be inverting for current injection with the configuration in Figure L4 and non-inverting for Figure 1B. In the track mode, the current injected will be given by i, = ( G ) ( E ) / R , which stabilizes E , at a value near zero volts. For a high degree of compensation, G should be large. Any inexpensive track and hold is based upon capacitive storage so the correction circuitry itself will drift during the hold period. This drift is voltage-independent and is primarily due to bias currents in the circuit. To minimize the ensuing error, it is advantageous for the output voltage of the track and hold to be large. Figure 3 shows the circuitry used to implement this function. Application of a TTL low or “track” voltage (0 to 1 V) a t S produces +0.3 volt a t the gates of Q2 and $3 while a high or “hold” voltage (2 to 4.5 volts) at S produces -15 volts a t the gates. In track mode, both FET’s are “on,” and the OA has a gain determined by R7 and T1. In track mode, the output E,,, charges C1 to a value equal to E,,, with a time constant of l/C1R6. One must be in track mode for at least 5 time constants. Note that the response of the circuit in track mode (EOut/Eln) is fast, essentially independent of the time constant l/C1R6 but also that the voltage across C1 is filtered by this time
S
Eout
03
45 Figure 3.
Compensator circuit. All resistors are 10% ’/4 watt
RI = 3.3 Kohms; R Z = 820 ohms: R3 = 470 ohms; Rq = 10 Kohms; Rs = 33 Kohms. R s = 1 Megohms: R7 = 100 Kohms; Q1 = 2N3702; Qz = MPF 102; QJ = MPF102; OA1 = Philbrick 1009 C Y = 1.0 I f polystyrene: T, = 1 Kohms 20-turn: S = 0 5 low 5 1.0 V (track mode); 2.0 5 high 5 4.5 V (hold mode)
Table I. Integrator Drift Rates for the AutoCompensated and Offset-Compensated Circuitso Average % error for 5 replicates
Min. after balance
0
10 20
30 40
Auto-compensation
Offset compensation
0.05 0.12 0.05 0.05 0.06
0 .oo 0.84
1.49 1.61 2.27
Rr = 500 ohms, Ri = 1000 ohms, R, = 10 Megohms and Ci = 0.1 pf. Rd = 1000 ohmn, c d = 1.0 pf, i= 2.5 seconds, step = 1.0 v.
constant. In the hold mode, this filtered voltage across C1 maintains an E,,, equal to the value it had at the end of track mode. This voltage will, of course, droop at a rate determined by the bias current of the OA and the leakage current of the FET’s (about 1 nA in this case). For a Philbrick 1009 OA, the FET leakage is the dominant contribution and the droop will be about 1 mV/second with the parts specified. The value of G (adjusted by 2’1) and R, should be chosen so that the stable track voltage E,,, is in the range 2-5 volts ensuring a compensation loss of less than 0.l%/second . PERFORMANCE EVALUATION The circuit was evaluated by conducting a number of computer-controlled experiments with the circuit configuration of Figure l.4 using a dummy cell consisting of a 1.0 pf capacitor in series with a 1000-ohm resistor. The control program locked the compensation circuit into hold mode 10.8 psec before the potentiostat executed a 1.0-volt step. The charge was monitored for 2.5 seconds and the potentiostat then drove the dummy cell to ground and the charge was monitored for an additional 2.5 seconds a t which time the compensator returned to track mode. The transducer and integrator OA’s used were PN 1009’s. In the first series of experiments, the OA’s were voltage and current balanced as well as possible manually. Five runs with the auto-compensator circuit were made. This protocol was repeated at 10-minute intervals for 40 minutes and the results are given in Table I. The experimental conditions corresponded to a very high integrator gain situation which requires high performance from the compensator. The worse case error using the auto-compensator was 0.12% which corresponds to slightly over 1 ADC count since the average value for the response was 1040 ADC counts. This is within the range of usual system noise and is considered quite adequate. ANALYTICAL CHEMISTRY, VOL. 46, NO. 8, J U L Y 1974
1153
Table 11. Cell Compensation of Background C u r r e n t s
Using Auto-Compensationa Average % error for 5 replicates Current, nA
Auto-compensation
Offset compensation
1.o 5 .O 10 .o 20 .o 50 . O
0.02 0.05 0.02 0.12 0.08
0.50 1.20 2.42 5.12 23.25
a lif = 500 ohms, Ri = 1000 ohms, R , = 10 Megohms, l i d = 1000 ohmil, c d = 1.0 pf, 7 = 2.5 seconds, step = 1.0
v.
and Ci = 0.1 pf.
Prior to the development of the auto-compensator, our standard scheme made use of an offset compensation principle. This scheme used two FET’s to implement a N.O. switch between the integrator input resistor and the summing point and a N.C. switch across the feedback capacitor. In this circuit, the two switches changed states 10.8 lsec prior to the step which ensured that the integrator was set to zero prior to the experiment. For comparison, an analogous series of experiments was performed utilizing this form of offset compensation. The results also summarized in Table I indicate the superiority of the auto-compensator. In electrochemistry, situations do exist where the value of i, in Figure 1 is non-zero at the potential imposed prior to the potential step. These currents can arise from a number of sources and frequently it would be desirable to null out their contribution automatically. For this reason, a second set of experiments was undertaken to test the performance of the auto-compensation circuit in eliminating finite i, values. Several values of i, have been simulated by injecting a known current into the transducer amplifier summing point using the circuit in Figure l.4.The
auto-compensation circuit does an excellent job of eliminating response from these currents. The offset compensation scheme is not designed to correct for this effect, and the results using this scheme are included in Table I1 solely for magnitude comparison. CONCLUSIONS The auto-compensation circuit described appears to do an excellent job of eliminating integration drifts for times of up to 5 seconds. The basic input specifications for the transducer, integrators, and compensation amplifiers can be met by almost any FET input operational amplifier. This removes many limitations from the choice of instrumentation for transient chronocoulometry. Thus, very high power and high slew rate operational amplifiers can now be used for the transducer amplifier since only the dynamic error terms contribute to the experimental error with the use of the auto-compensation circuit. This circuit also may be applied in other applications such as derivative pulse polarography. In this case, the gain set by R7 and T1 in Figure 3 should be centered a t unity and the filter time constant must be lowered to a value consistent with the time scale of the experiment. ACKNOWLEDGMENT The authors would like to thank C. Michael Elliott for extensive practical testing of this circuit and many helpful suggestions. Received for review November 12, 1973. Accepted January 7, 1974. Various aspects of this work were supported by the Air Force Office of Scientific Research (AF0SR)USAF, under Grant AF-AFOSR-69-1625 and by the Materials Research Center, UNC, under Grant Number GH33632 from the National Science Foundation.
New Sampling Device for the Recovery of Petroleum Hydrocarbons and Fatty Acids from Aqueous Surface Films Russell Miget, Howard Kator, and Carl Oppenheimer Marine Science Institute. University of Texas, Port Aransas, Texas 78373
John L. Laseter’ and Enoch J. Ledet Department of Biological Sciences. Louisiana State University in New Orleans. New Orieans, La. 70122
Replacement of coal by petroleum as the major world source of energy has resulted in increasing amounts of petroleum products being released into the environment. The need to quantify and to identify petroleum hydrocarbons in natural waters has presented methodological difficulties-especially with hydrocarbon films a t the air/ water interface. Since petroleum spills generally result in surface slicks of non-uniform thickness which can cover relatively large areas, a surface film sampler was required which would permit rapid, consistent, and efficient retrieval of surface hydrocarbons. We have developed an inexpensive sampler which fulfills these requirements, and which can be operated under moderately rough surface conditions. Field comparison of this sampler with the screen technique ( 2 ) and the sorTo whom correspondence should be addressed. ( 1 ) W . D. Garrett, Limnoi. Oceanogr.. 10,
1154
602 (1965).
A N A L Y T I C A L CHEMISTRY, V O L . 46, N O . 8, J U L Y 1974
bent-in-a-can sampler (2) showed it to be easier to use and to require less sampling time. While we have used this sampler for both inshore and offshore field work for the past year under a variety of sea surface conditions, the purpose of this report is to discuss laboratory studies performed to determine the selectivity of the sampler for hydrocarbons and related compounds as a function of molecular structure, and the efficiency of recovery for petroleum films of varied thicknesses. EXPERIMENTAL Sampling Apparatus and Procedure. The surface film sampler (Figure 1) consists of a disk of 2-mm Teflon which is attached to a 4-mm marine aluminum backing by means of 24 aluminum countersunk bolts. The original prototype tested by one of us (RM) utilized a “heat cured” epoxy coated wooden backing Estes, P. G Mikolaj. R. R. Thaman, and L. W . Senger, Proceedings 1973 Conference on Prevention and Control of Oil Spills, American Petroleum Institute,Washington, D.C., 1973.
(2) J. E .