mined nature are obtained. -44s predicted by Figure 4,the scattering ratio for 100% samples reverses as one crosses the copper edge, Practice. A number of experiments have been performed, in which the choice of scattered ware length was arbitrary and mostly a matter of convenience. To date, the selection of the most useful scattered wave length has been based on an empirical approach. One other serious limitation of this method is the decreasing intensity of scattered radiation a t longer wave lengths. Nevertheless. several of the experiments performed indicate that this technique can be a powerful tool as an alternative to the internal standard method. Figure 6 gibes data for a multicomponent ore system. This system behaves as a two-component system. iron vs. light elements such as silicon. The change of scattering and Xi Kcu intensity a t constant nickel concentration are shown US. iron concentration. The curves are approximately parallel and hyperbolic. This would indicate that a ratio of Ni R a to scattering at 0.6 A. should be almost independent of iron concentration. Figure 7 shows this to be the case, the absolute nickel measurement being very sensitive to iron concentration, while the ratio is almost independent of iron concentration. Table I shows the results of a direct comparison with the internal standard method, strontium having been added to these materials as a standard for yttrium. Here, the scattering at 0.6 -4.
Table 1. Effect on Yttrium Measurement of Substitution of Holmium Oxide and Dysprosium Oxide for Samarium Oxide (37.9%) in Rare Earth Mixtures
Change in Measurement Measurement I KLY T Ka/Sr KLY T K(u/O.6.2. scattering
Hoz03 -8 4 +51 +4 3
Dy203 -10.5 1-10 - U 3
is as good as internal standard for yttrium as the added strontium. This scattering was not a satisfactory standard for other elements in these samples. Figure 8 shows the results obtained for lead determination in a variety of minerals. In this set of samples, the principal constituent varied from mostly light elements to 60% zinc to 40% iron, vet the use of a line-to-scattering ratio provided practically complete independence of sample composition. The above examples are merely typical ones; in other experiments. too numerous to list here, scattered radiation has been found to be a useful internal standard. SUMMARY
The use of scattered radiation as an internal standard has been shown t o correct for instrumental variables and for absorption effects. Depending upon the problem on hand, this correction has been complete or partial. Additional investigations will better define
tbe limits within which this technique is significantly superior to others. LITERATURE CITED
(1) Adler, I., Bselrod, J. M., ASAL. CHEM.26, 931-2 (1954). (2) ildler, I., ilxelrod, J. M., Spectrochim. Acta 7, 91-9 (1955).
(3) .Lipplied Research Laboratories, Spectrographer’s A’ews Letter 7, S o . :3 (1954). (4) Ihid., 9, No. 4 (1956). (5) Beattie, H. J., Brissey, 11. M., ASAL. CHEX 26, 980 (1954). (6) Campbell, IT. J., Leon, M.,Pitts-
burgh Conference on dnalyt~ical Chemistry and lpplied Spectroscopy, 1957. ( i )Claisse, F., Dept’. of Mines, Laboratories Branch, Quebec, P. R. 327
(1956). (8) Compton, A. H., dllison, Y. K.,
“X-Rays in Theory and Esperiment,” 2nd ed., pp. 138-40, Van Nostrand. New 1 ork. 1935. Kemp, J. R., ’ - 4 s ~ (‘HEX. ~. 28, 1838-43 (1956).
Xemp, J. TV., Andermann, Geo., Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, 1956; SpectrochiiTL. Acta 8, 114 9 (1956). (11) Liebhafsky, H. A, Winslow, E. H., ANAL.CHEM.28, 586-8 (1956). f l 2 ) Mack, 31. A-orelco R ~ p t r .3, 3i-9 (1956). (13) Salmon, hl. L., Blackledge, .J. P., Ibid.. 3. 68 (1956). , ’ J., ‘Specirochirn. Actn 7, 6 (1955).
V, H., Potter, F. R., Pittu-
31k“ (1955) ’(abstract).
RE(EIVEDfor review July 26, 1957. 4ccepted Fehruary 24, 1958.
Low Pressure Electric Discharge Detectors R. C. PITKETHLY Research Station, The British Petroleum Co., Ltd., Sunbury-on-Thomes, Middlesex, England
b Stable and extremely sensitive detectors for gas-liquid chromatography have been developed from the simple low pressure electric discharge device described by Harley and Pretorius. Although other methods of detecting changes in the characteristics of the discharge can be envisaged, the work described was limited to direct current operation and measurement of the running voltage. For stobility and linearity of response, a homogeneous, nonsputtering cathode of large surface area and relatively small anode-tocathode spacing is desirable. With modified neon lamps as detectors, paraffin hydrocarbons were readily detected at concentrations of 1 in
10,000,000 and peak height response to propane and the butanes was linear in the range from IO-” to mole.
T
of electric discharges in a gas are greatly affected by the nature of the gas. In theory this fact should permit the detection of changes in gas composition, and high sensitivity is to be expected in certain cases. For example, Emeleus (1) states that the striking voltage in an inert gas a t several millimeters pressure can be lowered 300 to 400 volts by the presence of mercury a t a partial pressure of 0.01 mni. Howewr, no successful applicaHE CHARACTERISTICS
tion of this principle to analysis has been reported until recently. Harley and Pretorius (4) briefly reported the use, as a detector for gas chromatography, of a simple discharge tube forming one arm of a Wheatstone bridge supplied with 900 volts direct current. The device permitted analysis of hydwcarbon samples in quantities of the order of 10-lO mole and detection of 10-12 mole. This sensitivity is approximately 10,000 times greater than that of detectors presently used in gas chromatography and is, therefore, of great interest. The preliminary investigation reported here was undertaken to determine whether stable detectors of this type could he developed. VOL. 30, NO. 8, AUGUST 1958
1309
t
TO A T U O S P C L ~ E OR TO =L@%MET Z
AILV
+TO
4 u E
~ A L - A S T ( 3 L TER-), Ah2
LlLLLh
PAL D
-Figure 2. Bridge-impedance electric discharge detector RI, Rls.
changer
400V
circuit
for
47K, incorporated in neon lamp
Rz, R11. 40K, wire wound R3,
Rs.
R4, Rg.
6 ohms, wire wound 18 ohms, wire wound
6 8 ohms, wire wound 15K, wire wound R7, Rlz. 20K, wire wound VRI, VR2. 20K, wire wound V I , Vz. Modified neon lamp detectors Va. 6SN7 HT supply, Groyonics, Type 2 1 1 Recorder, Brown-Elektronic potentiometric, 1 0-mv. full-scale defiection
R,, Rlo.
Re.,
CHMM ATOGRADIY COLUMNS )METER. G U M 0
Figure 1. detector
Test equipment for modified neon lamp
METHODS OF DETECTING CHANGES DISCHARGE CHARACTERISTICS
R11.
IN
lead to delayed recovery after a peak of vapor concentration has passed.
Several possible approaches may be made to the problem of detecting the response of the discharge to changes in the gas composition. These are rn follows.
Although alternating current excitation with external electrodes was theoretically attractive and oscilloscope trials indicated the feasibility of striking voltage measurements, the direct current running voltage method was selected for the present work on account of the simplicity and availability of the equipment required.
With direct current excited discharge, measurement of the running voltage is straightforward and balanced bridge circuits are readily set up. If the discharge tube is incorporated in a saw-toothed oscillator circuit, the voltage across the discharge tube will oscillate between the striking and running voltages. Measurement of the peak voltage then gives a value for the striking voltage averaged over a period dependent on the time constant of the measuring circuit. This may be more sensitive to changes in gas composition than the running voltage. Alternatively, with alternating current coupling to the voltage-measuring circuit, the difference between striking and running voltage is obtained. Alternating current excitation can be used and, especially with radiofrequencies, it becomes possible to employ electrodes external to the tube and to use higher gas pressures in the discli~trge. The response of the detector should then he controlled to a greater extent by the gas composition, and the effect? associated with the electrodes and malls should be correspwdingly reduced. Another possible technique is t o use radio-frequency excitation and take advantage of the rectifying properties of the discharge to pick off a direct currwt voltage from a pair of probes, following the technique of Lion (6). A probable disadvantage of this arrangement is that the probes are not subject to the full cleaning effect of the discharge; slow desorption of vapors may 13 10 *
ANALYTICAL CHEMISTRY
EXPERIMENTAL
Materials. The hydIocarbons employed in this work were samples of purity greater than 99 mole yo,available as the result of preparative work carried out over a number of years. Calor gas is a commercial product intended for domestic uses. The sample used here contained 2 i mole % ’ propane, 26 mole % isobutane, and 47 mole % n-butane. Kieselguhr (Celite 545, Johns-Manville Corp.) mas prepared for use as B stationary phase support according to the procedure described by James and Martin (5) but omitting the acid treatment The silicone oil (MS 550, Hopkins 8: Williams, Ltd.) was diluted with five times its volume of %pentane before addition t o the prepared kieselguhr and the mixture was stirred continuously during evaporation of the solvent. Apparatus. Preliminary tests had previously been carried out with discharge tubes based on Harley and Pretorius’ design but with longitudinal gas flow. Although certain of the early observations are mentioned later in the discussion of discharge tube design, the following descriptions refer only t o the testing of a modified neon lamp as a detector. The gas-liquid chromatography columns used were
conventional and intended merely to provide peaks of known but very low vapor concentration; no attempt was made at this stage to exploit the characteristics of the detector t o improve column efficiency. DETECTOR CELLS.Gas inlet and outlet tubes were sealed to two Philips SBC 200/260-volt neon indicator lamps and an internal seal was incorporated (Figure 1) so that vapors held up in the sealing tip in the stem of the lamp would be s m p t away from the electrodes. The inlet tubes were sealed to 3 . 5 foot lengths of 0.25-mm. capillary and the outlet tubes were connected to the pumping system, which included a ballast volume of 3 liters to reduce pressure fluctuations. Electrical connections to the tubes were made with standard small bayonet sockets and flexible leads. The electrodes and supporting wires in these lamps were of iron; one was a circular disk 10 mm. in diameter and the other consisted of a strip, 2 mm wide, bent into a ring of approximately the same diameter. The spacing between the electrodes (1 to 1.5 mm.) was not critical. A current-limiting resistor (ca. 50,000 ohms) connected in one lead was incorporated in the base; this affected the output but suitablc operating conditions can be selected as described below. In neon tubcs rated for 100-volt supplies, the series resistor is of lox value (100 ohms) but their use was avoided because the electrodes usually have a thick and relatively fragile oxide coating. Voltage-current characteristics of such a tube, obtained by measuring potential drops a t various applied voltages while operating in the apparatus shown diagranimatically in Figures 1 and 2, are given in Figure 3. It is evident that the operating conditions giving the lowest sensitivity to current and pres-
sure variations are 1- to 1.5-ma. current and 3- to 5-mm. pressure. CHROMATOGRAPHY COLUMNS.These were 1 meter long and 4 mm. in internal diameter, packed in the normal way with Celite carrying 20% of its weight of silicone hfS 550 oil. Both columns were supplied from the same controlled nitrogen stream. One was provided
I
with a sample introduction device (Figure 1). The columns operated a t atmospheric pressure a t the inlet and a t an appreciably reduced pressure a t the outlet, determined by the gas flow rate and the pressure drop in the capillaries connecting the column outlets to the detector lamps.
I
6
I
200
220
I 240
I 280
I
260
J
I
I
220
200
240
260
P O T E N T I A L ACROSS DISCHARGE
280
I 300
- VOLTS
Figure 3. Voltage-current-pressure characteristics of modified neon tube detector
---
1RUN N o \
Figure 4.
Nitrogen effluent from n-hexatriacontone-Celite Nitrogen effluent from silicone.Celite column at 25'
column at 21
C.
C
I2
_________ Typical chart records showing reproducibility of sample injection Calor gas, 0.4770by volume in nitrogen
ELECTROKIC EQUIPMENT. A simple balanced pair of cathode followers using a 6SK7 valve was employed initially to feed a 2-ma. Evershed and Vignoles recorder, but for sensitivity measurements the circuit was modified and a 10mv. Brown-Elektronic potentiometric recorder was used. It was intended that detectors of metal construction with grounded outer electrode should also be tested; therefore, it was more convenient in the modified circuit t o ground the positive high voltage line and take the output from low value resistors in the anode loads (Figure 2). Three ranges of sensitivity were provided to give full-scale deflections on the 10mv. recorder with approximately 2.5, 10, or 40 volts input; the impedance changer-recorder combination was cabbrated by measuring grid-to-grid voltages with a valve voltmeter and noting the recorder reading a t various settings of the zero shift potentiometers. Sample Dilution Techniques. The introduction of microgram and smaller quantities of hydrocarbons into the chromatograph columns by use of dilute solutions in cetane and silicone MS 550 oil, which had been degassed a t 50' C. in high vacuum, was attempted. A 10-cm. trap column packed with Celite-silicone oil WRS connected at the inlet of the analyzing column and 10-pl. samples of the dilute solutions (0.002 to 0.1% by weight) were added by micropipet. Cetane vapor soon penetrated the column, upsetting the balance completely, and the silicone oil contained sufficient volatile matter to give two large peaks a t retention volumes near to n-heptane and a subsequent succession of low, wide peaks which seriously affected the baseline stability. This method was abandoned in favor of dilution with nitrogen using a 1-liter flask fitted with a serum cap and containing a quantity of Dixon gauze rings. Measured amounts of hydrocarbons, gaseous or liquid, were injected from calibrated syringes into the nitrogenfilled flask and mixed by vigorous shaking. A thousandfold dilution was thus obtainable a t one step; up to 10 1-ml. samples could be withdrawn from the flask with less than 1% reduction in the pressure. Procedure. The nitrogen flow n-as started and vented t o the atmosphere a t the sample introduction point After the system was swept out, a soap film flow meter was attached to the vent; the flow rate t o the atmosphere was adjusted t o a convenient value (BO to 100 ml. per minute) and measured. The vacuum pump was then started and a proportion of the nitrogen stream was drawn through the columns and discharge tubes; the change in nitrogen flow rate to tl!e atmosphere gave the total flow In both columns. The distribution hctween the columns was found by running the columns separately with tlie connection to the vacuum punip throttled to give the same pressure (3 mm.) in the discharge tubes. Volatile materials were removed from the columns by heating them in a curVOL. 30,
NO. 8, AUGUST 1958 *
131 1
rent of air a t 60" C. while nitrogen was flowing. The power supply (400 volts, stabilized) to the discharge tubes and impedence changer was switched on and if the tubes did not light, the high voltage line was temporarily connected to the 550-volt unregulated supply. Failure to fire under these conditions indicated the presence of high vapor concentrations. I n this case sweeping out was continued. When stable discharges were established, the connections to the tubes were reversed, if necessary, to obtain similar distribution of luminosity and mininum difference in running voltage between the two tubes. The zero shift on the analyzing tube was set to maximum to give maximum response, and zero adjustments were made on the reference side. When approximate balance was achieved, the recorder was connected and the apparatus was allowed to run until equilibrium was reached. Samples of diluted gas and vapor mixtures were injected from a 1-mi. graduated syringe with a 5-inch needle into the gas stream entering the analyzing column. The heights and areas of the resulting peaks were obtained from the recorder chart. Examples of chart records show reproducibility of sample injection (Figure 4) and baseline noise and response to small amounts of Calor gas (Figure 5). ,4ir peaks of variable and sometimes appreciable size were obtained. The air responsible for these peaks was introduced either into the gas mixture by leakage through the serum cap of the dilution flask or into the column during injection of the sample-e.g., in the tip of the hypodermic needle. In either case the amount of air was small compared with the sample volume and did not significantly affect the amount of hydrocarbon introduced. However, when estimating the height of a propane peak which followed a large air peak, allowance was made for the tail of the latter.
Sensitivity to Hydrocarbon Vapors. One set of results presented in Table I and Figure 6 shows the response of the system t o several paraffin hydro-
Table I.
Component Propane Isobutane n-Butane
a c
Sample Volumt.,
d.
rn v. 100
-
0-
BASE-LINE
NOISE O N 2 5 - V O L T
13 1 2
ANALYTICAL CHEMISTRY
Figure 5. Baseline noise and response to small amounts of Calor gas
F S D RANGE
AIR 41R
, s L '
RUN N o 3 5
o Q x 10-6m e
-JJ
RUN N o 3 4 4.4 x 10-6 m e
carbons in the range from propane to octane. In Figure 7 data for propane, isohutane, and n-butane show the near-linearity of the relationship between peak height and quantity of vapor in the range from 10- to lop9 liter. On this log-log plot the observed slopes were 0.95, whereas a linear response requires a slope of unity. The ultimate sensitivity of the detector for CBand Cd hydrocarbons was approximately 2.5 X lo-' ml. mole), the response then being of the same order as the base-line noise (& 10 mv.), as may be seen in Figure 5. This corresponded to a peak vapor concentration of the order of 1 in 50 million by volume, Concentrations of l in 10,000,000are readily detectable. DISCUSSION
Factors Mecting Design and Operation of Discharge Tubes. Preliminary tests with discharge tubes consisting of platinum electrodes sealed into glass tubing confirmed the ex-
I
RUN Ne. 26 4 , 7 x 10-5 m e
430
d l
P
/iGL-
>
'
1 -
E
,
rnde
01
6
-~ ~ - _ _ _ NJMBEP
OF - A R e O \
a ATOMS
Figure 6. Sensitivity to paraffin hydrocarbon vapors in nitrogen
tremely high sensitivity and fast response of the low pressure discharge as a vapor detector, but the tube designs based on Harley and Pretorius' note (4) were not entirely satisfactory for the follo-sing reasons: Sensitivity to pressure variations Production of a discouraging amount of electrical noise which was possibly
Response of Neon Lamp Detector to Paraffin Hydrocarbon Vapors
(Xitrogen carrier gas, 35 ml. per minute) Peak Peak Area, Weight Height, VoltY 1Iicromole Voltsa Min. I
26.8 0.0210 0.92 0.505* 23 8 0.0202 1.17 0.486* > 10 0.0368 2.14 0.879* 3.13 0.0043 0 31 0,00049" Isopentane 23.6 2.61 0.0362 0.0041@ n-Pentane 17.5 3.08 0.00467c 0.0358 n-Hexane 7.80 0.0320 3.20 0.00467c n-Heptane 2.92 : i 28 0.0288 0 . 0O46'ic n-Octane Measured output (equals 0.56 X change in voltage across discharge). Volume of gas a t 20" C., 760 mm. Volume of liquid at 20" C.
~~~
7
r
5.5 7.4
...
2.22 18.5 27.4 30.6 31.4
Retention Volume, M1. 38 59 82 155 211 612 1895 5240
Response Volt-liter Volt-liter pmole per Y 9.17 12.8
0.208 0.221
18 17 26 33 38
0:iso
1 9 7 5 2
0.248 0.311 0.334 0.335
associated with sputtering of the cathode metal and a tendency to instability of the discharge near the anode Nonlinearity resulting from discontinuities in response associated with changes in the number of striations in the positive column of the discharge Consideration of the characteristics of discharges through gases led to considerable improvements in the design and operation of these detectors. The theoretical treatment of the phenomena occurring in discharges has made 7), but great progress in recent years (I is far from complete. However, much empirical information of direct application to this problem has been accumulated. In a low pressure discharge the potential drop is affected by the work function of the cathode, the ionization potential of the gas, and the mean free path of electrons in the gas. As all these factors can be influenced by the nature of the gas, the functioning of the detector depends on their combined effect. However, in a discharge operating in the “normal” regime-i.e., when the negative glow does not completely cover the cathode surface-a large fraction of the potential drop occurs in the cathode (Crookes) dark space. This so-called normal cathode fall of potential is then a t a minimum,
VOLUME OF CAS
is independent of the current flowing, and may also be independent of the gas pressure over a limited range. (See, for example, the voltage-current-pressure characteristics in Figure 3). Therefore, there are distinct advantages, as far as stability is concerned, in running the discharge in the normal regime. Furthermore, the probability of irregular and nonlinear response will be reduced if the normal regime can be maintained during peaks of vapor concentration. This may not be a sufficient condition for linearity, as the other factors mentioned above are involved, but it is almost certainly a necessary one. For work with hydrocarbons and other vapors which cause an increase in the potential drop, this suggests that the unperturbed standing current should be chosen near the upper limit of the normal regime. In circuits using a current-limiting resistor, the fall in current accompanying a peak of vapor concentration which can be tolerated before crossing the lower limit is then a t a maximum. If large currents are employed, sufficient cathode area must be provided to keep the current density in the constant zone. A development of this argument indicates that use of a constant current generator-e.g., a pentode in place of a
(IITERSX
10-9)
Figure 7. Response of modified neon lamp detector to propane, isobutane, and n-butane Nitrogen carrier gor, 35 ml. p e r minute 0 lrobutane n-Butone
A Propane
current-limiting resistor-may raise the upper limit of the concentration range over which the detector is effective. The normal cathode dark space a t an aluminum electrode in nitrogen a t 1mm. pressure is of the order of 0.3 mm. thick and proportionately less a t higher pressures. The electrode spacing can therefore be reduced to about 1 mm. without affecting appreciably the negative glow and dark space. The positive column and the irregularities of response associated with it can thus be eliminated and noise emanating from instability of the discharge near the anode may be reduced. Finally, considering the cathode material, the main requirement is that the work function of the surface should be stable and reproducible after temporary contamination by vapors. Secondary considerations are: a low work function to permit operation a t comparatively low applied voltages, and resistance to sputtering to avoid insulation troubles and background noise. Very low work functions are obtxinable with oxide-coated cathodes, but these were rejected on account of their sensitivity to contaminants. A homogeneous cathode material is preferred because formation of temporarily active areas and changes of surface composition are less likely to occur. Of the common metals, magnesium, aluminum, and iron are favored on account of their low work function and relative freedom from sputtering. Although cathodes of pure aluminum, a “machinable” aluminum alloy, stainless steel, and wrought iron were tested, the most satisfactory results in terms of voltage drop and stability were obtained with the iron electrodes used in neon lamps. Stainless steel cathodes appeared to be stable but discoloration of the metal indicated some surface changes. The characteristics of modern voltage stabilizer tubes indicate that zirconium and molybdenum mould probably be suitable. Modified Neon Lamps as Vapor Detectors. The above considerations led to the testing of small neon lamps as vapor detectors. These lamps had suitably proportioned electrodes, closely spaced and with a reasonably large cathode surface. of a metal having a low work function, stability, and freedom from sputtering. Furthermore, they were readily adapted by glass-blowing for use in a gas flow system (Figure I). These detectors operated in the normal regime with currents between 0.5 and 2.5 ma., a t voltages across the discharge of 240 to 300 volts, and were not greatly sensitive to pressure variations in the range from 3 to 5 mm. The presence of a resistor in the neon lamps and the desirability of working with an amilable 400-volt stabilized VOL. 30, N O . 8, AUGUST 1958
1313
power supply limited the standing curtent which could be used. The value selected (ca. 1.5 ma.) represented a compromise between high current (within the normal regime) and loss of output when the external resistor was reduced in value. With discharges running a t about 250 volts it became possible to supply the tubes and the impedance changer required to feed the recorder from the same stable 400-volt source instead of from two separate power packs. This was not only more economical but gave greater stability. A disadvantage of the lower supply voltage was the increased probability of extinction of the discharge by very high vapor concentrations and subsequent difficulty in relighting. The use of closely spaced electrodes improved the performance but did not eliminate irregular response to large peaks. In this case discontinuities were associated with erratic movements of the glow and departure from the normal regime. The use of pentodes, mentioned above but untried as yet, may improve the behavior in this respect. The data on sensitivity to paraffin hydrocarbons (Table I and Figure 6) and baseline noise (Figure 5) show a satisfactory signal-to-noise ratio for concentrations down to 1 in 10,000,000 and indicate that the detector is capable of the analysis of mixtures in quantities between the microgram and millimicrogram levels. Calibration is required but the linearity of response to paraffins appears to be good at levels below a few micrograms. With regard to the variation of response with molecular weight of hydrocarbons (Figure 6), little can he said as yet. Although the ionization potential of the gas or vapor is usually considered to be an important factor governing the potential drop in a discharge, it is evi-
dent from the data so far obtained that it is certainly not controlling: A progressive rise in molar response was observed between Ca and CS, whereas the ionization potentials show little change beyond C4. Corresponding data on the tendency to electron attachment of hydrocarbons and their effect on the work function of metals is meager, but these factors are no doubt important in determining the response of this type of detector. An instrument consisting essentially of a vibrating reed electrometer has been used to measure changes in contact potential of metal surfaces resulting from the presence of vapors in a gas stream (3,8),but failed as a detector for elution gas chromatography because of slow desorption of the vapors. It is interesting that no serious difficulty of this type has been encountered with the discharge tubes, although some tailing of peaks has been observed; no carbonaceous deposits accumulated in several weeks of daily use. A possible explanation for the difference between the two detectors is that the discharge itself tends to clean the surface and assists removal of adsorbed materials. FUTURE DEVELOPMENTS
The three most striking characteristics of this detector are its simplicity, its high output level, and its great sensitivity. Future developments which take advantage of these points can be foreseen. Firstly, it should be possible to design sensitive detectors which are robust both in the detector and in the ancillary equipment. Metal equipment with st& bility and sensitivity comparable to the glass apparatus described has been constructed. Secondly, the low concentration levels a t which the detector operates will
permit most interesting developments in gas chromatography, such as the use of new stationary phases with very low absorptive capacities. Increased column efficiencies and the possibility of analyzing materials of low volatility a t r e l a tively low temperatures can be expected. Results already indicate that mass transfer effects in the stationary phase are much reduced. For example, the columns used in this work maintained their normal efficiency although they were operated under conditions of low pressure and high gas velocity which, with existing detectors, would have been inefficient. A direct application of practical use is the analysis of hydrocarbons present in low concentration in hydrogen or other gas streams. ACKNOWLEDGMENT
The author thanks the chairman and directors of the British Petroleum Co. for permission to publish this paper. LITERATURE CITED
(1) Emeleus, K. G., ‘‘Condu$ion of Electricity through Gases, p. 14, Methuen & Co.. Ltd.. London. 1951.
Francis, V. J., Jenkins, H. G., Repts. Progr. in Phys. 7, 230 (1940). Griffiths, J., James, D., Phillips, G., Analyst 77, 897 (1952).
Harley, J., Pretorius, V., Nature 178, 1244 (1956).
James, A. T., Martin, .4. J. P., Biochem. J . 50,679 (1952). Lion, K. S., Rev. Sci. Instr. 27, No. 4, 222 (1956).
Lunt, R. W., von Engel, A,, Repts. Progr. in Phys. 8,338 (1941). Phillim. G.. J. Sci. Instr. 28, 342 RECEIVED for review August 26, 1957. Accepted February 8, 1958. Division of Analytical Chemistry, 132nd meeting, ACS, New York, N. Y., September 1957.
Determination of Germanium by the HeteropoIy Blue Method ELWOOD R. SHAW and JAMES
F. CORWIN
Department of Chemistry, Antioch College, Yellow Springs, Ohio ,The cause of the up to 100% variation in absorbancies of identical samples of germanium dioxide, analyzed by the heteropoly blue-molybdate complex method, was traced to different interpretations of the literature recommended time interval between the addition of the molybdate and the reductant. The optimum interval was selected by a time study, and
1 3 14
ANALYTICAL CHEMISTRY
the best acid and molybdate concentrations were determined. Different operators obtained results which agree within 5%.
N
methods of colorimetric analysis using organic reagents have been deireloped, but the most widely used and convenient method for routine analysis is the molybdenum blue EW
method. Several modifications of the method reported in the literature have been summarized by Krause and Johnson (7‘). Most publications refer to the instability of the yellow molybdogermanic acid complex; however, no specific study was made of its rates of formation and decomposition under the conditions of the recommended procedures. The literature (1, 8) rec-