600 The TIMMS, KONRATH AND CHIRNSIDE THE DETERMINATION OF [Vol. 83 Determination of Impurities in Carbon Dioxide Gas Chromatography, with Special Reference to Coolant Gas for Nuclear Reactors BY D. G. TIMMS, H. J. KONR.ATH AND R. C. CHIRNSIDE (Research Laboratories of the General Electric Company Limited, Wernbley, England) It has been shown that, with a 25-ml sample, hydrogen, argon, oxygen, nitrogen, methane and carbon monoxide present at levels as low as 5 to 20 p.p.m. in carbon dioxide can be determined by a gas-chromatographic method. Special features of the technique are the use of a molecular sieve as the solid adsorbent, the high speed of analysis and the use of a katharometer of high sensitivity. IT is necessary to determine the impurities in the bulk supplies of carbon dioxide supplied for use as coolant in a gas-cooled reactor and also to detect and determine the level of these impurities and any others that may be present while the reactor is in operation.The main problem is the detection and determination of small concentrations of per- manent gases. These might normally include hydrogen, oxygen, argon, nitrogen, carbon monoxide and methane in the range 10 to 3000 p.p.m. v/v. As argon can become radioactive under irradiation, it may be necessary to detect this at even lower levels, say, 1 p.p.m. At present, the mass spectrometer is in use for these analyses, and, although we have no direct experience of the merits or demerits of the technique for this particular purpose, it seemed likely that a reliable alternative method would be of considerable interest, par- ticularly if it were cheaper and simpler to operate than the mass spectrometer.Classical methods of gas analysis, apart from their other shortcomings, are much too insensitive for the purpose. Ideally, an appar,atus is required by means of which the com- position of the gas can be automatically monitored, the different impurities continuously resolved and measured and the values for the measurement presented directly. It should be reliable and give results that are reproducible and of reasonably high accuracy. In some circumstances the infra-red gas analysers commercially available readily meet certain of these requirements; unfortunately, a number of the gases in which we are particularly interested cannot be determined in this way, since they have no characteristic infra-red absorption spectra.This is true, for example, of argon, nitrogen, oxygen and hydrogen.Nov., 19581 IMPURITIES IK CARBON DIOXIDE BY GAS CHROMATOGRAPHY 601 We have therefore chosen to investigate the feasibility of analysing the coolant gas for impurities by means of gas chromatography. Although by far the greatest developments have been in gas - liquid partition chromatography, introduced by James and Martin in 1952,l gas-adsorption chromatography in some form is of longer standing. Originally, “displacement” techniques were devised by Claesson,2 in which the components of a mixture are first adsorbed and then displaced successively from the solid phase by a more strongly adsorbable gas or vapour introduced into the carrier-gas stream.Later, elution techniques were used by JanAk3 in Czechoslovakia, Ray“ in England and Patton, Lewis and Kaye5 in the U.S.A. In these, the gases are reversibly adsorbed on the solid phase and are separated by virtue of their different adsorption coefficients. The elution technique would seem t o be essential if any rapid separation of the per- manent gases is to be achieved, In a series of paper^,^ JanAk has described the application of this technique to the separation of a large number of gases, and, for example, to the analysis of mine gases. His technique involves the use of carbon dioxide as carrier gas, which is subsequently absorbed in alkali in a nitrometer. The gases that have been separated from the column are measured volumetrically.The chief adsorbents used have been activated charcoal, silica gel and alumina, but these have not been wholly satisfactory for the separation of permanent gases; although charcoal has been used with moderate success, it gives a poor separation of oxygen and nitrogen. In the face of this difficulty and stimulated by some comments of van der Craats during the discussion at the Symposium on Vapour Phase Chromatography in 1956,6 it was decided to investigate the use of molecular sieves as solid adsorbents for our particular problem. We were encouraged to think that they might be of particular value, because Kyryacos and Boord’ had shown in another context that the permanent gases could be separated very effectively on a column with a molecular sieve as adsorbent.EXPERIMENTAL ADSORBENTS- It was mentioned earlier that charcoal had been used by other workers for the separation of the permanent gases, apparently with only moderate success. Our initial experiments were carried out on activated charcoal, but, with a short column, only a partial separation of oxygen and nitrogen could be achieved. We have subsequently confirmed experimentally that hydrogen, oxygen and nitrogen are separated more effectively on molecular sieves than on charcoal. Of the three grades of molecular sieve commercially available-Linde 4A, 5A and 13X -experiment showed that, although both 5A and 13X are effective, grade 5A gives the best separation, and most of the subsequent work has been carried out with this material. blolecular sieves are synthetic zeolites from which the water of constitution has been removed to leave a lattice containing holes of molecular dimensions.Polar compounds are strongly adsorbed and carbon dioxide in particular is de-sorbed only slowly from a molecular sieve. It is not possible, therefore, to effect a rapid direct separation and determination of the different impurities with which we are concerned and which are present in the carbon dioxide in low concentration. This difficulty has been readily overcome by passing the gas first through a soda lime tube to remove the carbon dioxide quantitatively; only the impurity gases pass into the column, from which they can be rapidly eluted and a complete analysis effected in a few minutes. Moisture will de-activate the molecular sieve and must also be removed from the gases entering the column ; this too is absorbed chemically by means of magnesium perchlorate.Any moisture that may get through will give rise to tailing, and this gives an immediate warning of the need to replace the desiccant. The need for the prior removal of carbon dioxide might seem at first sight a disadvantage, but in one important respect it simplifies the subsequent procedure, for if the carbon dioxide is absorbed, it becomes possible to carry out the analysis with carrier gases of ordinary commercial purity. If carbon dioxide were allowed to enter the column, it would displace from the adsorbent any impurity gas that might be present in the carrier gas. The impurities so displaced would then be eluted from the column and would be recorded as apparent impurities in the carbon dioxide.It would be essential therefore to use only carrier gases of very high purity.602 TIMMS, KONRATH AND CHIRNXDE THE DETERMINATION OF [Vol. 83 CARRIER GASES- It was found by experiment that, on a column 6 feet long containing granules of molecular sieve 5A, a satisfactory separation of hydrogen, oxygen and argon (together), nitrogen, methane and carbon monoxide can be effected. In our particular work it was found that the maximum amount of information could be obtained at one time by the use of either h,ydrogen or argon as the carrier gas. However, within the range of experimental conditions we have tried, it was not possible to separate oxygen from argon. When argon is used as the carrier gas, i%ny argon that may be present in the sample is not detectable and does not interfere with the oxygen signal; hydrogen, oxygen, nitrogen, methane and carbon monoxide can, however, be separated and detected or determined.When hydrogen is used as the carrier gas, it is possible to achieve a greater general sensitivity, but argon and oxygen are not separable and they appear together as a single peak. The argon can therefore be determined if first argon and then hydrogen is used as carrier gas, but only as a difference figure. An alternative method that we have investi- gated involves the use of De-oxo catalyst*; the gas remaining after absorption of the carbon dioxide is passed through a short column containing crushed pellets of De-0x0, the oxygen is removed and a signal for argon alone is obtained.It is understood that the determination of argon in coolant gas is of special importance, and, if the argon should be present in very low concentration, the direct method just described would be preferable to the difference method. It could be carried out conveniently on a separate apparatus. APPARATUS- METHOD The apparatus required for gas - solid chromatography consists of- (i) a regulated supply of carrier gas, (ii) a gas-sampling system, (iii) a chromatographic column, (iv) a thermostatically controlled heating jacket for the ( v ) a detector and measuring equipment. To columns column, and Fig. 1. Gas-samplin: system. Scale: 1 to 8 (i) Carrier gas supply-Hydrogen and a.rgon are used for this particular application of the gas-chromatographic technique ; they are obtained from high-pressure gas cylinders.A gas pressure of 40 lb per sq. inch is provided by the usual type of cylinder regulator and it is further reduced and accurately regulated at 3 to 10 lb per sq. inch by a precision gas regulator (Negretti & Zambra or Norgren). The supply to the columns is finally regulated * De-oxo is a platinum catalyst marketed by the Baker Platinum Company, and, as the name implies, oxygen IS removed by catalytic reaction with hydrogen t o form water.Nov., 19581 IMPURITIES IN CARBON DIOXIDE BY GAS CHROMATOGRAPHY 603 by means of fine-adjustment needle valves V,, V, and V, (see Fig. 1). Valves V, and V, serve to equalise the gas flow through each column. The carrier gas is left flowing, even when the apparatus is not in use, in order to prevent atmospheric moisture from entering the columns.(ii) Gas-sampling system-This consists of a gas-sampling valve and a carbon dioxide absorption tube. The sampling valve shown in Fig. 2 provides a means of introducing a sample into the stream of carrier gas with the minimum interruption of flow and with no risk of contamination from the atmosphere. With the gas channels in position A, carrier gas passes directly to the column while the sample loop is swept with sample gas. On rotation of the channels to position B, the carrier gas passes through the sample loop and sweeps the sample into the column. The main flow of sample gas meanwhile passes out to the atmosphere. Fig. 2. Gas-sampling valve.Scale: Full size Since this valve was designed, we have learned that valves operating on a similar principle have been available commercially for some time. The carbon dioxide absorption tube is made from 20 s.w.g. copper tubing, 28 inches long and $g inch external diameter. It is packed for about three-quarters of its length with fresh 14 to 20-mesh soda lime. The exit end of the tube is packed with a mixture of 2 parts by volume of 16 to 36-mesh magnesium perchlorate and 1 part of 16 to 36-mesh crushed firebrick. I t is emphasised that the soda lime should contain an appreciable propor- tion of moisture, about 15 to 20 per cent., as it was found in the course of this work that dry soda lime will not absorb carbon dioxide quantitatively; indeed, if it is very dry, no absorption appears to take place.As a consequence, it is necessary to refill or replace the absorption tubes daily, even if no carbon dioxide has been passed through the apparatus, for the stream604 TIMMS, KONRATH AND CHIRNSIDE : THE DETERMINATION OF [Vol. 83 of carrier gas, which we prefer to keep flowing through the apparatus, slowly removes moisture from the soda lime and eventually makes it ineffective. Soda asbestos has been found to be completely inactive under these particular conditions; it appears to be even more sensitive to the dryirig effect of the carrier gas than is soda lime. The magnesium perchlorate serves to prevent moisture from passing into the column, where it would de-activate the molecular sieve. The crushed firebrick is used to prevent the hydrated perchlorate from blocking the absorption tube.The amount of magnesium perchlorate used is obviously insufficient to absorb all the moisture present in the soda lime, but serves to prevent moisture from entering the column over a period of a t least 1 day under normal conditions of use. (iii) Column-Two interchangeable columns are used, one for the separation and the other to equalise the gas paths. Both consist of 6-foot lengths of &inch external diameter 20 s.w.g. stainless-steel tubing packed with Linde 5A molecular sieve. Copper tubing was used originally, but it was found that, when hydrogen was used as carrier gas, some reaction took place on the copper surfaces and gave rise to anomalous results for oxygen. When heating is provided by means of a vapour jacket, we have used simple U-shaped columns, but coiled columns are used in a more compact form of the apparatus in which a hot-air bath is provided.From gas-sampling system A = Sunvic thermostat, type TS.3 B = Serum cap C = Katharometer D = Column (approxi- mately 5 turns of 4-inch diameter) $ 8 1 I.$ @ n,,,,,,,,, < ,I,,,,,,, <,# l,,,,,,, n l ' r l , l l r r m l l l , l l l , , , , , , , , , , , Fig. 3. Section through heating jacket With the straight column it is customary t o activate the molecular sieve before packing by heating it in air to 350" C for a few hours, but it is possible to activate or regenerate a packed column by heating it to 350" C for about 3 or 4 hours while it is being purged with a stream of dry air. The coiled columns are packed, before they are bent, with granules of 36 to 52-mesh molecular sieve and the ends are plugged with glass-wool; they are then coiled round a 4-inch mandrel and the ends are bent to shape.(iv) Heating jacket-Two types of heating jacket have been used, a simple steam jacket and a thermostatically controlled air jacket. The steam jacket has certain advantages for laboratory use; in particular, the column and t'ne katharometer block may be heated quickly t o the operating temperature (100" C) and temperature fluctuations are negligible.Nov., 19.581 IMPURITIES IN CARBON DIOXIDE BY GAS CHROMATOGRAPHY 605 The air jacket is provided by means of an electrically heated aluminium pot, 6 inches in diameter, with a flanged end 8 inches in diameter.The coiled column and the katharometer block are located in the pot as shown in Fig. 3. -3 oe 0'205" diam. A Filament R, R, = 30-ohm resistors R3 = 10-ohm resistor R4 = 20-ohm 10-watt potentiometer K1, K, = Katharometer filaments G = 2.5-mV recording voltmeter Fig. 4. Construction and mounting of katharometer : ( a ) , section through one channel; ( b ) , section showing double-channel arrangement; (c), bridge circuit; ( d ) , filament mounted on copper - glass seal The pot is heated by means of Nichrome wire wound on glass tape round the outside It was wound originally with a single coil, but this was found to give rise to of the pot.606 TIMMS, KONRATH AND CHIRNSIDI3: THE DETERMINATION OF [Vol. 83 electrical noise. By replacing the single coil with a non-inductive winding consisting of 40 turns of 0.0142-inch diameter Nichrome wire, mains-borne noise has been reduced to a minimum, The current is supplied through a Sunvic hot-wire relay operated by a thermo- stat that dips into the pot.The dissipation of the winding is about 300 watts at 240 volts. The temperature of the pot can be adequately controlled, but the rate of heat transfer to the katharometer block is low and the time taken to reach equilibrium (2 to 3 hours) greatly exceeds that required by the column in the steam jacket (4 hour). This problem could be largely overcome by pre-heating the katharometer block by means of an auxiliary heating coil. A layer of lagging material, eg., cotton-wool, is required over the tops of the heating jackets to minimise thermal e.m.f.s in the leads from the katharometer. (n) Detector-A sensitive katharometer is used as the detector.It is made from a copper block, 2 inches x 21 inches x 4 inches, through which two &inch diameter gas channels are drilled, a reference and a detector channel; these operate under identical conditions. Tungsten filaments having a resistance of about 30 ohms a.t 20" C are mounted on copper - glass seals and are held in the gas channels under slight tension by means of grooved sapphire or ceramic pegs. Details of the construction and mounting of the katharometer and of the associated circuitry are shown in Fig. 4. It has been found convenient to use a recording millivoltmeter that has a range of 0 to 24mV. PROCEDURE- The apparatus is calibrated by the injection of small volumes of the appropriate gas through a serum cap on the inlet to the column and measurement of the height of the peak produced on a recording millivoltmeter in the katharometer circuit, A suitably small volume of gas for calibration purposes may be obtained from the dead space in the end of a syringe.A tuberculin syringe of about 1-ml capacity is flushed with the gas and then, with the needle of the syringe close to the serum cap on the column inlet, the plunger is slowly pushed down to its full extent. The needle is inserted im- mediately into the serum cap, the plunger is withdrawn to the 1-ml mark so as to draw in carrier gas and is then quickly pushed down again to its fullest extent. The gas in the dead space is thus diluted with carrier gas and eFfectively introduced into the column.The syringe is calibrated by weighing it before and after flushing with water. The volume of the dead space in the syringe is about 45 cubic millimetres and a correction must be made for the gas remaining in this space. We have used the following gases for purposes of calibration- The pegs are located symmetrically by means of grub screws. Oxygen and nitrogen-Air. Hydrogen-2 per cent. of hydrogen in air. Methane-Commercial, first passed through an absorption tube to remove carbon dioxide Carbon monoxide-Commercial. With methane and carbon monoxide, corrections have to be made for hydrogen, oxygen and nitrogen present as impurities. and moisture. TABLE I APPROXIMATE AMOUST OF IMPURITY REQ'LTIRED TO GIVE A 1-mm DEFLECTION Amount of impurity with Amount of impurity with Impurity argon as carrier gas, hydrogen as carrier gas, P.P.m.(./V) P,P.m. (v/v) . . 0.5 - 3 Hydrogen . , . . Argon . . .. * . . . Oxygen . . . . . . . . 5 3 Nitrogen . . .. . . 10 3 - Methane . . . . . . 5 7 Carbon monoxide . . .. 20 6 An alternative method of calibration involves a direct comparison of the unknown sample with one of known composition. The reference gas is made up to be similar in com- position to the sample under test. This method. has the advantages that no corrections for temperature and pressure need be applied and the time required for calibration is reduced.NOV., 19581 IMPURITIES IN CARBON DIOXIDE BY GAS CHROMATOGRAPHY 607 The apparatus has been calibrated for these impurity gases with both argon and hydrogen as carrier gases and it has been found that over the range covered there is a linear relationship between peak height and concentration.It will be appreciated that all calculations must include corrections for temperature and pressure unless the calibration is carried out at the same time as an analysis. The sensitivity is determined by a number of factors-katharometer wire temperature, nature of carrier gas, size of sample, etc. Under our conditions of operation and with a 25-ml sample, the approximate amount of each impurity required to give a 1-mm deflection is shown in Table I. RE s u LTS At this stage of the development of nuclear power plant, trials of these proposed methods of analysis have been restricted to samples of commercial-quality carbon dioxide and to pure carbon dioxide to which various impurities have been added in known concentrations.COMMERCIAL CARBON DIOXIDE- Two different rates of withdrawal were used from each cylinder; it will be noted that there were significant differences in the concentrations of the impurities in the two samples from either cylinder. This effect is due to a different distribution of impurity gases between the gas and the liquid phases in the cylinder. At fast rates of withdrawal the composition of the gas would be expected to approach the composition of the liquid phase; at slow rates it may resemble more closely the equilibrium composition of the gas phase. The results of one examination with argon as carrier gas are shown in Table 11.TABLE I1 DETERMINATIOS OF IMPURITIES IN COMMERCIAL CARBOY DIOXIDE FROM CYLINDERS Column temperature, 100" C Argon flow rate, 50 ml per minute Katharometer current, 300 mA Sample volume, 25 ml Samples were taken from small cylinders supplied by two manufacturers. .\mount of impurity found in carbon dioxide from cylinder A at a rate of Amount of impurity found in carbon dioxide from cylinder B a t a rate of Impurity withdrawal of- withdrawal of- I A > f A > 100 ml per minute, 3000 ml per minute, 100 ml per minute, 3000 ml per minute, p.p.m. p.p.m. p.p.m. p,p.m. Hydrogen . , 9.5 11 110 60 Oxygen . . 85 not detected 1500 1000 Nitrogen , . 500 360 6600 4400 The input t o the recorder was attenuated where necessary. SYNTHETIC CARBON DIOXIDE MIXTURE- Impurities were added in known amount to the gas obtained from solid carbon dioxide; the levels of impurity were of the same order as those expected to be present in the coolant from a nuclear power plant.The main analysis was carried out with argon as carrier gas and the results are shown in Table 111. TABLE I11 DETERMINATION OF IMPURITIES IN SYNTHETIC CARBON DIOXIDE MIXTURE Column temperature, 100" C Argon flow rate, 30 ml per minute Katharometer current, 300 mA Sample volume, 25 ml Impurity iZmount of impurity added, Amount of impurity found, Peak height, P.P.m. (v/v) P P.m. (v/.9 mm Hydrogen , . . . 41 40 85 Oxygen . . . . 101 100 21 Nitrogen . . . . 390 400 49 Methane . . .. 175 180 35 Carbon monoxide 1260 1280 69608 TIMMS, KONRATH AND CHIRNSIDE: THE DETERMINATION OF [Vol.83 The analysis of a sample similar in composition was also carried out with hydrogen as carrier gas. For this analysis the apparatus was modified by the insertion of a De-oxo tube between the carbon dioxide absorption tube and the separating column on one side of the apparatus. The De-oxo pellets remove oxygen and so allow the direct determination of argon. Carbon monoxide is also removed during this process, so that the analysis is restricted to the determination of argon, nitrogen and methane. However, by reversing the functions of the two columns, and thus obviating the need to remove the De-oxo catalyst tube, nitrogen and methane, and also carbon monoxide and argon plus oxygen can be determined in the other side of the apparatus. By the use of both columns in this way the full analysis of the mixture, including argon, was carried out.The results are shown in Table IV. The actual chromatograms obtained during this and the analysis in which argon was used as carrier gas are shown in Fig. 5 . Fig. 5, Chromatograms of impurities in a prepared sample of carbon dioxide: ( a ) , argon used as carrier gas: ( b ) , hydrogen used as carrier gas, oxygen removed with De-oxo catalyst TABLE IV DETERMINATION OF IMPURITIES IN SYNTHETIC CARBON DIOXIDE MIXTURE Column temperature, 100" C Hydrogen flow rate, 40 ml per minute Katharometer current, 750 mA Sample volume, 25 ml Impurity Amount of impurity added, Amount of impurity found, Peak height,* p.p.m. (v/v) P.P.m. (v/v) mm Oxygen . . . . 106 100 40 Xitrogen . . . . 405 400 141 hlethane .. . . 173 170 28 Carbon monoxide 1250 1240 241 * The input of the recorder was attenuated in order to fit the peak for carbon monoxide into the Argon( . . . . - 5 7 3.5 chromatogram.Nov., 19581 IMPURITIES IN CARBON DIOXIDE BY GAS CHROMATOGRAPHY 609 DISCUSSION OF THE METHOD The work described represents a preliminary survey of the potential application of the principle of gas chromatography to the analysis of reactor coolant gases. Although the experimental work has been restricted to synthetic mixtures of carbon dioxide with minor concentrations of added impurity gases, the results suggest that the method should be of value for the analysis of coolant gases. It is simpler and more rapid than mass spectrometric measurements and we believe it may be sufficiently sensitive. The time required for a complete analysis is inversely proportional to the rate of flow of carrier gas, but some loss of sensitivity and resolution occurs at high rates of flow. It has been found experimentally that the optimum performance of the apparatus described is obtained at a flow rate of about 30 t o 40 ml per minute. Under these conditions, a complete analysis can be carried out in about 8 minutes. The accuracy with which any particular impurity can be determined is finally set by the volume of sample taken. Most of the analyses have been carried out on 25-ml samples of gas, but this could be increased with advantage for the determination of argon. The disadvantage of a larger sample is the need to absorb larger amounts of carbon dioxide and thus to require more frequent replacement of the absorption tubes. Experiments now in progress with more sensitive methods of detection may provide a better solution to the problem. REFERENCES 1. 2. 3. 4. 5. 6. 7. James, A. T., and Martin, A. J. P., Analyst, 1952, 77, 915. Claesson, S., Ark. Kemi. Min. Geol., 1946, 23A, No. 1. Janak, J., Chem. Listy, 1953, 47, 464, 817, 828, 837, 1184, 1190, 1348 and 1476. Ray, N. H., J . Appl. Chem., 1954, 4, 21 and 82. Patton, H. W., Lewis, S. J., and Kaye, W. J., Anal. Chem., 1955, 27, 170. Desty, D. H., Editor, “Vapour Phase Chromatography” (Proceedings of the Symposium sponsored by the Institute of Petroleum, 1956), Butterworths Scientific Publications, London, 1957. Kyryacos, G., and Boord, C. E., Anal. Chem., 1957, 29, 787. Received September Sth, 1958