42 RILEY : SIMULTANEOUS DETERMINATION OF WATER AND Simultaneous Determination of Water [Vol. 83 and Carbon Dioxide in Rocks and Minerals BY J. P. RILEY (Department of Oceanography, The University of Liverpool) A method is described for the simultaneous determination of water and carbon dioxide in minerals and rocks, The samples are inserted in a combustion tube at 1100" to 1200" C ; the resultant water and carbon dioxide are removed with a current of nitrogen, absorbed and determined gravimetrically. About twelve determinations can be made per day, since it is not necessary to allow the furnace to cool between analyses. Blank values for the apparatus are low, amounting to about 0.1 to 0.2 mg per hour. Water is completely removed from even the most resistant minerals, such as stauro- lite, within 30 minutes at 1200" C.The method showed a standard deviation of &0-04 per cent. with rocks containing (about 5 per cent. of water when a 0.5-g sample was used. A comparison has been made between a number of methods used for the determination of water in rocks and it was concluded that both the methods of Penfield and of Shapiro and Brannock tend to give low results owing to incomplete removal and absorption of water. A KNOWLEDGE of the water content of rocks and minerals is of considerable importance to petrologists and mineralogists. The determination of water in rock samples by "loss on ignition" is still apparently being carried out at the present time by some petrologists,l even though it has been known for a long time to be unreliable. High results are caused by loss of volatile constituents other than water, such its sulphur or carbon dioxide.Low resultsJanuary, 19581 CARBON DIOXIDE IN ROCKS AND MINERALS 43 are due to gain in weight through oxidation of ferrous oxide to ferric oxide; this oxidation is usually not complete. It is probable that the majority of water determinations at the present time are carried out by various modifications of the Penfield2 method. In this procedure, the sample is heated in a bulb blown at one end of a hard-glass tube, and the liberated water is condensed further along the tube. The portion containing the water is weighed and then weighed again after being thoroughly dried. There are two principal sources of error in the Penfield method, both of which cause the results to be low.It is difficult to ensure that all the liberated water is condensed, particularly under conditions of low relative humidity. Many common minerals retain some water up to at least 900°C, and therefore it may be almost impossible to remove all the water from the sample by heating in a glass tube. This difficulty may be minimised by carrying out the heating in a tube of fused silica, but the removal of the heated end of the tube is more difficult. Shapiro and Brannock3 have recently described a modification of the Penfield method, in which the sample is heated in a Pyrex-glass boiling-tube and the liberated water is retained in a weighed strip of filter-paper, which is then re-weighed. An empirical correction factor is used in the calculation of the results.Probably the most accurate method for the determination of water in rocks is that in which the sample is heated in a combustion tube in a current of dry air and the expelled water is collected in a weighed absorption tube, the increase in weight of the absorption tube being determined. The procedure, as described by grave^,^ is rather time-consuming, since only three determinations can be carried out per day. The temperature (about 1000" C) used in expelling the water is not sufficiently high to remove all the water from the more refractory minerals, without the use of a flux, such as sodium tungstate, which requires the use of a platinum boat. The blank value is high, being between 0.7 and 3 mg. In the method described by Groves, much time is spent in waiting for the combustion tube to cool between determinations.To speed up the process, the sample contained in an alumina boat was placed in the cold end of the silica combustion tube and then, after the end of the tube had been closed, it was inserted into the furnace by means of a stainless-steel rod sliding through a polytetrafluoroethylene washer. In order to ensure that all the water was expelled from the sample, the combustion tube was heated by means of an electric furnace to 1100" C, or to 1200" C if resistant minerals were present (the temperature was measured inside the combustion tube). At 1100" C all the water is removed from the majority of rocks in 20 minutes, and even refractory minerals, such as topaz, epidote and staurolite, give up all their water in about the same time at 1200" C.The simultaneous determination of water and carbon dioxide in rocks and minerals has been described by GoochK,6 and by Dittrich and Eitel,' but in each method elaborate platinum apparatus is required. When an attempt was made to determine the carbon dioxide and water simultaneously by heating at 1100" to 1200" C in a combustion tube and absorbing and weighing the resultant carbon dioxide and water, very high blank values for carbon dioxide were found. These were attributed to the formation of acidic oxides of nitrogen from the air used for sweeping out the apparatus. In order to reduce this error, nitrogen from a cylinder was used instead of air, and next to the main combustion tube was placed a further silica tube packed with short lengths of copper wire heated to 700" to 750" C to reduce oxides of nitrogen.Interference from sulphur compounds was prevented by a layer of silver pumice placed after the copper wire, and by a bubbler containing a saturated solution of chromium trioxide in phosphoric acid. With this apparatus extremely low blank values were found for both water and carbon dioxide (0.1 and 0-2 mg per hour, respectively). METHOD APPARATUS- The heated section is then drawn off. The apparatus used is shown diagramatically in Fig. 1. High-temperature furnace-The high-temperature furnace consists of an alumina tube, 22 cm long and having an external diameter of 4.2 cm, with 84 turns of 0.91-mm Kanthal A1 resistance wire on a 15-cm portion of the tube, the turns being more closely spaced at the ends of the tube than in the centre.The resistance wire is bedded in aluminous cement and the space between it and the 10-cm diameter silica outer jacket is packed with pure alumina. The furnace tube and the outer jacket were held in position by end plates made of Sindanyo.Scale: 2 2 inches = I foot A = B = C = D = E = F = G = Fig. 1. Insertion device Silica combustion tube Silica tube containing copper wire and silver pumice Absorption tube containing magnesium perchlorate Absorption tube containi ng soda-asbestos Bubbler containing chromium trioxide Side-arm containing magnesium perchlorate Apparatus for determining carbon dioxide and waterJanuary, 19581 CARBON DIOXIDE I N ROCKS AND MINERALS 45 Co++er furnace-The low-temperature furnace consists of a translucent silica tube, 9 cm long and having an external diameter of 2.3 cm, with 16 turns of 1-58-mm x 0.25-mm Nichrome tape on it.The whole is contained in a 6.3-cm diameter brass tube and the inter- vening space is packed with kieselguhr. In operation, the two furnaces are run in series with one another. The high-temperature furnace is regulated to the desired temperature with a rheostat or Variac auto-transformer. The total energy dissipated at 1100" C is about 950 watts. Variation of the temperature of the copper furnace over the range 700" to 750" C is of no consequence. Gas-purification train-Nitrogen gas (not of "oxygen-free" quality) from a cylinder is passed via a multistage regulator and needle valve, through a bubble counter containing concentrated sulphuric acid.It is purified by passage through tubes containing soda lime, fused calcium chloride and, finally, anhydrous magnesium perchlorate. Insertion device-In order to avoid the necessity for allowing the combustion tube to cool between determinations, samples were pushed into the hot zone of the closed combustion tube by means of an insertion device, A. This consists of a brass stuffing box made airtight with a polytetrafluoroethylene washer, through which slides a stainless-steel pusher bar 35 cm long and 6 mm in diameter, with a 12-mm diameter head. The stuffing box is fitted with a brass tube for the introduction of nitrogen into the combustion tube and is connected to the latter by means of a rubber bung. Combustion tube-The combustion tube, B, consists of a translucent silica tube, 45 cm long and having an internal diameter of 1.8 cm, fused at one end to a 3-cm length of silica tubing having an external diameter of 5 mm.The combustion tube is supported in the high- temperature furnace so that the 5-mm silica tube is about 10 cm outside it. Although the silica tube is being heated for long periods above the temperature recommended by the manufacturers (1000" C), no deterioration, except for slight sagging and a small amount of surface devitrification, has been noted even after several months' use. Attempts to use an impervious alumina combustion tube led to very variable blank values. Copper tube-This tube, C, whose function is to remove sulphur compounds and oxides of nitrogen, consists of a translucent silica tube, 10 cm long and having an internal diameter of 1 cm, packed with alternate layers of copper wire and silver pumice (prepared by evaporat- ing 14-mesh pumice with strong silver nitrate solution and igniting strongly) held in place by asbestos plugs.Both ends of the tube are fused to 3-cm lengths of silica tubing having an external diameter of 5 mm. At least once a week a current of coal gas should be passed for 15 minutes through the heated tube to reduce any copper oxide to the metal. The life of the tube is about 3 months with rocks of low sulphur content. Absorption tubes-The absorption tubes are 14cm long and have a 1-cm bore; a tube having an external diameter of 5 mm is sealed on at one end and a B10 socket at the other.They are closed by means of B10 cones fused close to the joint to the 5-mm tubing. The narrow tubing at each end is constricted at X to a capillary having a bore of about 1 mm. If samples containing much carbonate are to be analysed, an absorption tube of larger diameter (1.6 cm) should be used for the collection of carbon dioxide. The water absorption tube, D, is filled with anhydrous magnesium perchlorate. The carbon dioxide absorption tube, E, is packed with an 8-cm layer of soda asbestos and a 2-cm layer of anhydrous magnesium perchlorate. The stoppers of both absorption tubes are cemented in position with hard black wax. Connection of the absorption tubes is made with aged rubber tubing. Chromium trioxide bubbler-With samples containing more than 0-5 per cent.of sulphur, the bubbler, F, filled with a saturated solution of chromium trioxide in 85 per cent. phosphoric acid, should be interposed between the water absorption tube, D, and the carbon dioxide absorption tube, E. Its side-arm, G, is packed with magnesium perchlorate. PROCEDURE- For the analysis of the majority of rock samples, the main furnace should be run at 1100" C; if difficultly decomposable minerals such as staurolite are present, the temperature should be raised to 1200" C. The flow of nitrogen should be regulated to about 3 litres per hour. Each day before use, allow nitrogen to pass through the apparatus and absorption tubes for about 20 minutes. Remove the absorption tubes, wipe them carefully and weigh them after 5 minutes.46 RILEY : SIMULTANEOUS DETERMINATION OF WATER AND [Vol.83 Weigh 0.5 to 1.5 g of sample into a previously ignited 2-ml alumina boat lined with a piece of nickel foil. If much fluoride or sulphur is present, cover the sample with a layer of freshly ignited magnesium oxide. Insert the boat into the end of the combustion tube, replace the insertion device, and allow nitrogen to sweep air out of the apparatus for 5 minutes. Connect the weighed absorption tubes, and then push the sample into the furnace bymeans of the stainless-steel rod. When large amounts of readily decomposable carbonates, such as siderite and magnesite, are present, the boat should not be pushed immediately into the hot part of the furnace, since this leads to such rapid evolution of carbon dioxide that it is not completely absorbed.Such samples should be allowed to decompose in the cooler region just outside the furnace, and only after decomposition is nearly complete should the sample be subjected to the full temperature. After a heating period of 30 to 40 minutes, remove the absorption tubes, wipe them and weigh them after 5 minutes. Carry out a blank determination in the same manner, without the sample, before the first determination and at the end of work. The normal blank values for the water and carbon dioxide are 0.1 and 0.2 mg per hour, respectively. When higher blank values are found for carbon dioxide, the copper tube is exhausted and should be treated a s described on p. 45. COMPARISON OF METHODS FOR THE DETERMINATION OF WATER IN ROCKS AND MINERALS In order to compare the precision of different methods for the determination of water, nine samples of metamorphic rocks (ground to pass an 80-mesh sieve) were examined by the proposed method, by the Penfield method2 (heating for 20 minutes in a tube of Pyrex glass), by the Penfield method, but with lead oxide used as flux, and by the method of Shapiro and Brannock.3 “Loss on ignition” was determined by heating the sample in an uncovered platinum crucible at 1000” C for 30 minutes.The results are shown in Table I, each figure being the mean of at least two determinations. The ferrous iron content of the residue remaining after the determination of “loss on ignition” is very variable, ranging from 0.16 to 3.0 per cent. Considerable errors can be caused by correcting the “loss on ignition” figure on the assumption of complete oxidation of ferrous iron.Even when correction is made only for the amount of ferrous iron actually oxidised, results tend to be high, owing to loss of carbon dioxide and sulphur. This is par- ticularly noticeable with samples Nos. 1502c, 1698c and 1698~1, which contained appreciable amounts of pyrite. When sulphides and carbonates are absent, “loss on ignition” results are in fair agreement with those by other methods. In general, it will be seen that the results by the proposed method are appreciably higher than those by the other methods (except for “loss on ignition”); this is attributed to more efficient removal and collection of water. This difference suggests that the water contents given in many earlier analyses of rocks and minerals (particularly amphiboles and micas) may be significantly low.The figures obtained by the methods of Shapiro and Brannock and of Penfield (without flux) are in fair agreement with one another. The comparatively low temperature (less than 800” C) and short period of heating used makes the complete removal of water by these procedures unlikely. This is the reason for the conspicuous failure of both methods with the moderately resistant hornblendite (sample No. 1971). When lead oxide was used as a flux in the Penfield method, the results were higher, but they were still lower than those by the proposed method. This error is caused by a fundamental flaw in the Penfield method and its modifications, i.e., failure to condense all the liberated water.Shapiro and Brannock have attempted to correct for this loss by the use of an empirical correction factor, but the results in Table I suggest that their correction factor is too low. A study of the reproducibility of the methods with the garnet - sillimanite - biotite schist (sample No. 2096) showed standard deviations of 0.04, 0.13 and 0.10 per cent., for the proposed method, the Penfield method, with flux, and the method of Shapiro and Brannock, respectively. EXAMINATION OF REFRACTORY MINERALS Several minerals, such as talc, topaz and epidote, do not give up all their water at tempera- tures below 1100” C. The rate of loss of water fmm these minerals at 1200” C was therefore studied, and it was found that with every mineral, loss of water was complete after heating for 30 minutes.In order t o test the method, replicate determinations of the water content were carried out on the minerals, which had been ground to pass an 80-mesh sieve. CheckTABLE I COMPARISON OF METHODS FOR THE DETERMINATION OF WATER IN ROCKS Sample No.* 672 734 1153 1502c 1698c 1698u 1714 1971 2096 Rock type Biotite - garnet schist . . .. .. Garnet - sillimanite - sericite schist Garnet - cordierite hornfels . . .. .. Garnet - cordierite hornfelsg .. .. Cordierite - garnet - biotite hornfels . . Biotite-rich cordierite hornfels . . .. Biotite - sericite schist . . .. .. Hornblendite . . .. .. .. .. Garnet - sillimanite - biotite schist .. .. Loss on ignition A r \ Un- Corrected Residual Penfield FeO, corrected, At, BS, FeO, method, % % % % % % 9.42 10.29 11-91 9-58 8.92 11.13 8-70 7.15 9.82 2-35 4-48 2-70 4.17 3-80 3-08 3.58 1.19 3.57 3.39 5-62 4-02 5-84 4.79 4.33 4.65 1.98 4-65 3.24 5-41 3.69 5.59 4.50 4.16 4.63 1.96 4.40 1-32 1.93 3.01 2.22 2-29 1.50 0.18 0.16 2.28 2-32 5-19 2.87 4.38 3.92 3.25 3.93 1-84 4.03 Penfield method with PbO, 2.88 5.54 2.83 4-37 3.94 3-65 4-22 2.42 4.08 % Shapiro method, 2.48 4-91 2.94 4-47 3.86 3-41 3.99 1-44 3.80 % * - Y io Proposed method, % 2-98 5.47 3.27 4.67 0 4-10 * 3.70 4.50 0 2-57 ‘ 4.32 v * Sample number refers to the catalogue number of the sample in the collection of the Geology Department, University of Liverpool.t On the assumption of partial oxidation of ferrous iron. All samples originated from Cashel, Co. Galway, Eire. U M t On the assumption of total oxidation of ferrous iron.6 Containing 0.58 per ce& of sulphur. SIMULTANEOUS Magnesite TABLE I11 DETERMINATION OF WATER AND CARBON DIOXIDE IN CARBONATE MINERALS z Calcite Strontianite Dolomite Siderite Scapoli te M Carbon Carbon Carbon Carbon Carbon Carbon dioxide Water dioxide Water dioxide Water dioxide Water dioxide Water dioxide Water v, % Chemical* . . 50.20 50.40 5040 50.55 50.40 i Mean 50.44 BY simultaneous determination Method found, found, found, found, found, found, found, found, found, found, found, found, % % % % % % % % % % % - 43-80 - 29.03 - 47.25 - 38.27 - 5.23t - 0.84 43.80 0.51 29.14 0.35 47-15 0.40 38.37 0.25 5.25 0-58 0.8 1 43.80 0.54 29-15 0.44 47.23 0-40 38-30 0-23 5.37 0.63 0.75 43.70 0.57 29.07 0.50 47.30 0.42 38.20 0.20 5.48 0.66 0.81 43.65 0.48 29.03 - 47.40 0.50 38.40 0.25 5.41 0.63 0.80 43-74 0.53 29.10 043 47-27 0.43 38.32 0.24 5-38 0.63 Method described by Groves.* t Result given by Manchot and Lorenz’s method.g48 RILEY : SIMVLTANEOUS DETERMINATION OF WATER AND [Vol.83 determinations were also made as described by Groves,* with sodium tungstate as flux. results, which are shown in Table 11, indicate that the recoveries of water are satisfactory. The TABLE I1 WATER RECOVERED FROM RESISTANT SILICATES Water found in A Method Epidote, Staturolite, Topaz, Talc, Phlogopite, % % % % % Sodium tungstate flux . . 1-96 1.83 1.69 4-82 2.72 1.96 1.85 1-64 4.86 2.77 1.96 1.85 1.65 4.87 2-78 r I Proposed method (1200° C) 1.97 1.86 1-67 4.89 2.77 2.00 1-87 1.67 4.90 - 2.01 1-88 1-68 - - - - - - 1.89 I Mean 1.98 1.87 1-66 4-58 2.77 SIMULTANEOUS DETERMINATION OF WATER AND CARBON DIOXIDE In order to test the proposed method for the determination of carbon dioxide, several carbonate minerals were examined.The results are shown in Table I11 (p. 47), together with the carbon dioxide contents determined chemically.s In each case the agreement is satisfactory. The temperature at which the dissociation pressure of the carbon dioxide of carbonates becomes 1 atmosphere differs widely with the metallic radicle present, e g . , ferrous carbonate at 490" C; magnesium carbonate at 540" Clo; calcium carbonate at 900" Cll; strontium carbonate at 1289" C12; barium carbonate 1360" C. Samples that have low dissociation tem- peratures should be allowed to decompose initially for 10 minutes in the hot region just outside the furnace.If such samples are subjected immediately to the full temperature of the furnace, there is serious risk of loss of carbon dioxide by incomplete absorption, owing to its too rapid evolution. The vapour pressure of carbon dioxide over strontianite (SrCO,) at 1200" C is sufficiently high to ensure the complete removal of carbon dioxide in 30 minutes. On the other hand, the vapour pressure of carbon dioxide over barium carbonate is so low at this temperature that only partial removal of the carbon dioxide (70 to 80 per cent.) occurs in 30 minutes. Complete recovery can be obtained if the heating period is prolonged to 3 hours. The carbon dioxide content of scapolite is difficult to determine chemically, since the mineral is not completely attacked by phosphoric acid and must be decomposed with hydro- fluoric acid.g Its carbon dioxide is strongly retained at 1100" C, and a temperature of 1200" C must be used for the decomposition; under these conditions the results obtained for this mineral agree well with those found chemically by using hydrofluoric acid.As a further test of the method, replicate analyses of four metamorphic rocks, low in carbonate, were carried out for water and carbon (dioxide. The results are shown in Table IV. TABLE IV SIMULTANEOUS DETERMINATION OF WATIER AND CARBON DIOXIDE IN ROCKS Sample No. Carbon dioxide found, % Water found, % 1564 0.46, 0.48, 0.50, 0 4 3 3-72, 3.71, 3.68, 3.69 1760 0.71, 0.75, 0.70, 0.76 4.14, 4-13, 4.17, 4.15 1431 0.17, 0.18, 0-15, Os.16 6.08, 6.12, 6.07, 6.11 1460A 0.27, 0.32, 0.33, 0.31 4-85, 4.58, 4.60, 4-50 Sample No.Rock type: 1564 1760 1431 1460A Cordierite - spinel - plagioclase hornfels. Cordierite - plagioclase - biotite hornfels. Cordierite - biotite - spinel hornfels. Magnetite - spinel - biotite hornfels. The sample number refers to the catalogue number of the sample in the collection of the Geology Department, University of Liverpool. All samples originated from Cashel, Co. Galway, Eire. From the results it is seen that satisfactory reproducibility was achieved for both water and carbon dioxide.January, 19581 CARBON DIOXIDE I N ROCKS AND MINERALS 49 In order to check the efficiency of the silver pumice and chromium trioxide trap for the prevention of interference by sulphur, 30-mg samples of sulphur and chalcopyrite were analysed ; with neither substance was the gain in weight of the soda-asbestos tube greater than 0.3 mg.If samples to be analysed are known to contain fluorine, sulphur or sulphides, it is advisable to cover the samples with a layer of magnesium oxide to prolong the life of the silver pumice. A more serious error in the determination of carbon dioxide results from the presence of graphite or organic carbon in the samples. Such materials are completely burnt to carbon dioxide. The carbonate-carbon in these materials should be determined by the conventional method in which phosphoric acid is used.* The organic carbon content of muds may be found by determining the carbonate-carbon by the conventional method and total carbon by combustion. I thank Mr. P. Sinhaseni for assistance in the preparation of Fig. 1, and also Professor R. M. Shackleton and Dr. B. Leake for providing the rock and mineral samples used in the investigation. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. REFERENCES Fairbairn, H. W., “A Co-operative Investigation of Precision and Accuracy in Chemical, Spectro- chemical and Modal Analysis of Silicate Rocks,” U.S. Geol. Survey, Bull. 980, 1951, pp. 21-22. Penfield, S. L., Amer. J . Sci., 1894, 48, 30. Shapiro, L., and Brannock, W. W., AnaE. Chem., 1955, 27, 1796, Groves, A. W., “Silicate Analysis,” Second Edition, George Allen & Unwin Ltd., London, 1951, Gooch, F. A., Amer. Chem. J., 1880, 2, 247. -, Chem. News, 1880, 42, 326. Dittrich, M., and Eitel, W., 2. anorg. Chern., 1912, 75, 373. Groves, A. W., 09. cit., p. 112. Manchot, W., and Lorenz, L., 2. anorg. Chem., 1924, 134, 316. Remy, H., “Lehrbuch der anorganischen Chemie,” Band I, S 225, Akad. Verlag, Leipzig, 1931. Dutoit, W., J . Chim. Phys., 1926, 24, 110. Tamman, G., and Pape, W., 2. anorg. Chem., 1923, 127, 50. p. 97. Received July 15th, 1957