Table V. Comparison of Results Obtained on Known Blends of Voranol CP3001 and Voranol CP4000
Primary Hydroxyl, yo of Total Hydroxyl Voranol Trityl CP3001 Nitrite ChloAcetic % Method ride Anhydride 0 10 20
80 90 100
0. 3. 9.
34. 43. 46.
0. 5. 7.
34. 37. 41.
A comparison of three methods is shown in Table V. Again good agreement is found except a t the higher concentrations of primary hydroxyl.
for the privilege of including their data.
INTERFERENCES
(1) Bashkirov, A. N., Lodizik, S. A., Kamzolkin, V. V., Trudy Inst. Nefti, Akad. A'auk SSSR 12, 297 (1958). (2) Critchfield, F. E., Hutchinson, J. A., ANAL.CHEM.32. 862 (1960).
,4ny alcohol or glycol present in the polyglycol will interfer with the determination. Ketones, which also would cause high results, are not likely t o be present in polypropylene glycols. The same is true of propenyl unsaturation, which, however, can be eliminated by hydrogenation if it should be present. ACKNOWLEDGMENT
40. f 4
method indicates the uncertainty in drawing the secondary hydroxyl rate line. The higher percentage is probably most nearly correct. However, great care must be taken in interpreting results by this method when the amount of primary hydroxyl is high.
The results by the trityl chloride rate method reported in Table IV were obtained by J. C. Ambuhl, Central Laboratory, The Dow Chemical Co., Freeport, Texas. The results by the acetic anhydride rate method in Tables I11 and IV were obtained by R. A. Hummel of the Special Services Laboratory. The author thanks these chemists
LITERATURE CITED
(3) Hanna, J. G., 'Siggia, S., 'J.Polymer Scz. 56, No. 164, 297 (1962). (4) Hendrickson, J. G., Texas Div., The Dow Chemical Co., FreeDort, Texas,
private communication.
-
( 5 ) Morton, R. A., Stubbs, A. L., -4nalyst 71, 348 (1946). (6) Pernarowski, M., Knevel, A. hf., Christian, J. E., J . Phann. Sci. 50, 943 (1961). (7) Pernarowski, M., Knevel, A. M., Christian, J. E., Ibid., 946 (1961). (8) Schmulyakovskii, Y . E., Khim. i Tekhnol. Topliv i Masel 4, 46 (1949). (9) Schmulyakovskii, Y . E., Zhur. Prikl. Khim. 32, 2513 (1959). (10) Siggia, S., Hanna, J. G., ANAL.CHEM. 33, 896 (1961).
RECEIVEDfor review April 9, 1962. Accepted May 25, 1962.
An Efficient Dynamic Method for Surface Area Determinations H. W. DAESCHNER and F. H. STROSS Shell Development Co ., Emeryville, Calif,
b The measurement of surface area of solids b y the sorption of nitrogen from a flowing mixture of nitrogen and helium first proposed by Nelsen and Eggertsen has been evaluated and further developed. Variables, sources of error, and means for controlling them are discussed. The response of a thermal conductivity detector is shown as a function of the composition of the gas mixtures used in the method. By suitable manifolding of sample cells and use of a special calibration procedure, a complete determination of surface area (three relative pressures and BET computation) can b e made in about 40 minutes. The experimental procedures are described. The revised method has a repeatability of better than 370, and is applicable over a range of surface areas from 0.04 sq. meter per gram to the highest normally encountered.
A
to the problem of determining the surface area of solids was described by Nelsen and Eggertsen (6). Like most methods in common use (3), their method involves NEW APPROACH
1 150
ANALYTICAL CHEMISTRY
the measurement of the amount of gas adsorbed at a temperature near the boiling point of the gas. The novel feature in their technique is to pass a known nitrogen-helium mixture continuously a t atmospheric pressure through the sample, to measure, by thermal conductivity devices, changes in the nitrogen concentration of the gas leaving the sample, and t o show these changes on the chart of a recording potentiometer. The sample cell is immersed in liquid nitrogen; the sample thus cooled adsorbs nitrogen and temporarily reduces the nitrogen concentration in the gas effluent. Upon warming the sample, the adsorbed nitrogen is liberated, temporarily increasing the nitrogen concentration in the effluent. The adsorption and desorption processes are represented as peaks on the recorder chart; the areas of these peaks are direct functions of the amount of nitrogen adsorbed and desorbed by the sample. Improvements leading to the design of an apparatus for routine analysis were discussed by Lee and Stross ( 4 ) ; the commercial version of this instrument was described by Ettre ( 2 ) . This apparatus featured a nitrogen injector valve, which provided convenient means
for frequent calibration of the instrument with amounts of nitrogen comparable to the amount sorbed during an actual determination. The present paper describes the results obtained in evaluating the response of the thermal conductivity detector as a function of gas composition. It discusses errors that can arise from the flow changes consequent to adsorption or desorption of the measuring gas, and describes a method for avoiding such errors. It also outlines a scheme for calibrating the instrument to make frequent use of the injector valve unnecessary. EXPERIMENTAL
The flow scheme shown in Figure 1 is similar in principle to that described by Nelsen and Eggertsen. Helium and nitrogen are introduced from cylinders equipped with pressure regulators and pass through flow controllers. After blending, the gases' pass through the apparatus as follows: a cold trap to remove any condensibles; B mixing chamber; a thermal conductivity cell (reference) ; the sample cell; a nitrogen injector valve; a second mixing chamber; a second thermal conductivity cell (measuring); and finally through a
-6CAPIL-
Figure 1 .
soap-film flow meter. Critical factors which needed to be given special attention in developing the present apparatus were: precise flow control of helium and nitrogen and repeatable adjustment and measurement of their flow rates; thorough mixing of the gas streams to give a constant known composition of the gas entering the sample cell; unrestricted flow to provide near atmospheric pressure throughout the lines and sample cell; sufficient volume between sample cell and the measuring thermal conductivity cell to allow for re-establishment of the original flow rates through the detector during measurement of the desorption peak; and a simple device for calibration. -4 mixing chamber, consisting of a 6-inch length of l/&ch brass tubing, has been inserted after the sample cell. The purpose of this chamber is to allow the gas flow rate, disturbed by the evolution of desorbing gas, to become re-established before the desorbed gas reaches the detector and to mix the sample gas with the carrier gas; this should result in a symmetrical and less sharp peak. The combined volume of the mixing chamber, lines, and injection valve is between 45 and 50 ml. An additional method to ensure reestablishment of the normal gas flow rate consists in venting the carrier just ahead of the sample cell during desorption. This method is described further below. The venting valve may be attached to a sample manifold as shown in Figure 2. The manifold is formed from a series of 4-way valves to which sample cells or auxiliary calibration cells may be connected. In practice, a number of samples may then be equilibrated a t the relative pressure of nitrogen provided by the carrier gas a t the same time. After adsorption has been completed, each sample is desorbed in turn in a manner such that the nitrogen desorbed bypasses the other cells through the 4-way stopcocks. Use of the manifold avoids the necessity of sweeping the lines between determinations, or each time a tube is changed; this results in a considerable saving of time and gaseous reagents. The gas mixtures can be obtained by suitable mixing chambers and flow control valves from the pure components; alternatively, a number of gas
Flow diagram
mixtures equal to the number of points one wishes to obtain on the isotherm may be prepared ahead of time and used as required (7). DISCUSSION
The primary data used for the computation of the specific surface of solids by the dynamic method are the areas of the curves produced by a thermal conductivity detector as a result of the changes in concentration of the heliumnitrogen gas mixture on cooling the samples to temperatures a t which substantial sorption takes place, and on reheating these samples to the former temperatures. The areas of the recorded peaks are not necessarily directly proportional to the amount of nitrogen sorbed, for two reasons in particular: A, the relationship between gas concentration and detector response in terms of pen deflection is not linear, and B, if the evolution of gas, as it desorbs, is fast enough to cause a significant change in total flow rate of the gas T O THERMAL CONDUCTIVITY C E L L
Figure 2.
through the detector, the areas of the sorptogram being recorded on a time axis will reflect the change in gas flow. A. In the first description of the method by Nelsen (6),linearity of response over the entire nitrogen-helium composition range was assumed as a first approximation. To determine the relation more accurately, the base lines on the recorder were noted for different helium-nitrogen compositions passing through the sample detector while pure helium was passed through the reference detector cell. The shift, plotted against nitrogen concentration in the gas mixture, is shown in Figure 3. During adsorption and desorption of nitrogen, however, the composition of the gas mixture passing the detector vanes over a certain range. The magnitude of the detector response during the recording of a peak, because of this effect, will then depend on the gas composition a t the start of the peak (Le., the composition of the carrier gas) and the maximum departure from this composition that This was obtained during a peak. change in composition is a function of the rate of sorption as well as of the amount sorbed. Experiments were made to get an idea of the magnitude of the changes of peak area caused by the change in response of the thermal conductivity detector to the change in composition of the gases. Table I shows the change in peak area with a change in composition of the carrier gas on injecting the same volume of nitrogen under otherwise similar conditions. The change in area response is greater, the greater the volume injected, as one would expect. The deviation from linearity in the helium-nitrogen system is not large over FROM T H E R M A L CONDUCTIVITY C E L L
Manifold for sample cells and carrier gas vent
A, 8, and C. Circle seal 4-way valves (Model P1-418, Pasadena, Calif.) to which are attachedl A Calibration tube fltted with Eck and Krebs 4-way glass stopcock, B. Colibration tube with T-bore stopcocks, and C. Sample tuba
VOL 34, NO. 9, AUGUST 1962
1 151
Figure 4.
Figure 3. Variation of recoraer pen position with composition of carrier gas Thermal conductivity detestor
reasonably narromz limits, and assump-
used in the prdent apparatus, with which known amounts of nitrogen can be injected into the instrument which are comparable in volume with the nitrogen sorbed by a sample during an analysis. This valve, which can handle four different volumes of gas, is shown in Figure 4. Occasionally, for unknown Table I. Calibration with Typical Volumes of Nitrogen: Change of Peak Area with Carrier Gas Composition
(Constant flow rate and attenuation) Per Cent Detector Na ?i Volume of N2 Response Garner Injected, ml. Peak Area 0 8 17 25 0 8 17 25 0 8 17 25
1152
19.16
14.12
10.34
116.9 113.4 112.8 108.6 80.1 77.2 77.2 73.5 53.5 54.2 53.0 51.7
ANALYTICAL CHEMISTRY
samples or for pore volume determina. tions, these four volumes may not cover the expected range; therefore, addi-' tional tubes similar to those illustrated in Figure 2 are used for calibration in these cases. From a family of curves relating the volumes of injected nitrogen to peak areas at different relative pressures, the corresponding volumes sorbed by the sample during an analysis can then be derived by interpolation or extrapolation. The details of this procedure are described under calibration. B. The second source of error is the change in flow rate resulting from the desorption step. I n case of samples having very large surface areas it may be impossible to control the desorption sufficiently well to prevent a substantial change in the gas flow rate. The peak area representing the amount of sorbed gas is independent of total gas flow rate when plotted against total volume (sample plus carrier gas) eluted since the start of the experiment; it varies, however, with flow rate when plotted as a function of time, as is normally the case in conventional sorptograms ( I ) . This results from the relation between the parameters in question, and has no connection with flow sensitivity of detectors. Flow sensitivity of a detector is expressed by a shift in the position of the base line with a change in gas flow rate. In the apparatus under discussion,
Nitrogen injector valve
the sample detectors are of the pretzel type and depend only on normal convection flow of gas past the heated detecting element. The base line of a gas or gas mixture of constant composition stayed constant as the gas flow was varied over a wide range. In even more critical experiments, the flow rate through the sample detector was varied from 1 to 40 ml. per minute, while that of the gas passing the reference detector was held constant a t 30 ml. per minute. This experiment was repeated several times in both directions, with gases of the same as well BS of different compositions passing through the two detectors. There was no base line drift as long as the compositions of the gases were kept constant during any one experiment. T o confirm the validity of these considerations, and to ensure the proper functioning of the apparatus parts involved, a sample cell was filled with nitrogen and swept with helium-nitrogen gas mixture at different flow rates. The areas were proportional to the flow rate, consistent with theory; the peak shapes changed slightly with flow rate because of the changes in mixing characteristics resulting from the variation of the gas flow (6). Table I1 shows the results of the experiments; the product, carrier gas flow rate times peak area, shows only a slight trend, because of the nonlinearity of the detector response. After taking this trend into account, the residual relative error is substantially less than 1%. Both factors discussed under A and B are operative when nitrogen is desorbed from the sample at different rates into the flowing carrier gas. In experiments on the same sample, in which all conditions except the rate of desorption are held constant, a change in this rate will affect the peak area obtained when the rate of desorption is high enough signifi-
Table 11. Change of Peak Area of Injected Standard Volume with Carrier Gas Flow Rate
11002
i
B
HELIUM - A - PURE N2 IN m L I U M A 17%
(2.0 ml. of nitrogen into pure helium) He Gaa Peak Area F!ow Rate (Integrator Flow Rate X (ml./min.) Units) Peak Area 13.9 147 2043 14.3 144.5 2066 20.7 99 2049 21.6 97.5 2106 22.1 94.2 2081 25.7 83 2133 Table 111. Change of Peak Area with Rate of Injection of Calibration Gas (10 ml. of nitrogen into helium containing
8% nitrogen) Rate of Injection, ml./sec. Peak Area 7.5 12.5 17.5 30 60
V O L U M E OF MTROCEN, M L (CALIBRATION CELL)
Figure 5. Change of peak area with calibration volume of nitrogen injected into carrier gas stream of different compositions
cantly to affect the total gas flow rate, and when the different rates of desorption give rise to substantial differences in maximum concentration of sorbate in carrier gas. The results given in Table I11 show how the same volume of nitrogen gas released a t different rates into the same carrier gas affects the peak area observed. It is difficult to evaluate the relative contributions from the two effects, but this is not necessary in view of the procedure developed further below. The foregoing experiments also imply that the adverse influences are more effective, the larger the volume desorbed. In the case of catalysts having a large surface area, for instance, the smallest samples that can conveniently be weighed by conventional methods will still yield gas volumes of the order of 20 ml. STP. If this volume is released over a period of one minute into a carrier gas stream flowing at, say, 20 ml. per min., the gas flow rate is doubled, and the corresponding area on the chromat6gram is halved. It is clear that a sizable error can easily be introduced even if care is taken to carry out the desorption slowly to minimize the effect. The problem of error from these sources, however, can be eliminated altogether by diverting the carrier gas stream from the sample cell during the desorption period by a 3-way stopcock located just ahead of the sample cell, as
shown in Figure 2. The usual operating procedure thenis modified as follows: After the adsorption of nitrogen has been brought about by cooling the sample cell with liquid nitrogen] the base line representing the carrier gas mixture is allowed to re-establish itself. When the base line is constant, the carrier gas is diverted directly to the atmosphere by stopcock D (Figure 2). The nitrogen then is desorbed from the sample by removing the Dewar containing the liquid nitrogen from the sample cell, and the nitrogen evolved is permitted to expand toward the detectors over a period of 15 to 20 seconds. Then the carrier gas is directed over the sample cell again t o cause it to sweep the nitrogen evolved through the sample detector cell. In this manner, the nitrogen peak is brought into the detector a t the carrier gas flow rate established a t the beginning of the experiment. Two conditions must be met by the instrument if the procedure just described is to attain its purpose : the volume provided between sample cell and detector must be larger than the largest amount of nitrogen desorbed during an analysis, to keep the desorbate gas from expanding into the detector cell while the carrier gas is diverted. Such expansion would give rise to precisely the conditions we wish to avoid-i.e., the gas flow rate would be changing while the desorption peak is appearing on the
118 ~~.
120 121 124 128
chromatogram; and, the detector should be insensitive to flow changes-i.e., its base line should be stable when the rate of flow is changed. Several advantages result from this modification: A constant] even flow rate of gas through the detector is produced. From the discussion above it is clear that this is necessary to obtain the conditions essential for a precise calibration procedure. Regular peaks having repeatable peak heights are obtained; this reduces the error possible under conditions of irregular desorption. Changes in flow rate regulator settings caused by the sudden flow upsets are avoided. Such changes have been observed when using excessively sensitive or worn regulators. RESULTS
The apparatus and method described here were used following two procedures. In the case of completely unknown samples, one makes determinations a t several relative pressures and calculates the isotherm by the Brunauer-EmmettTeller (BET) equation. Different samples of similar type, however, can be more rapidly determined by a comparative procedure a t a single relative pressure. This procedure is of considerable value for routine refinery use where large numbers of samples of uniform type are analyzed daily. Results obtained by the sorptometer are compared with results obtained by the conventional pressure-volume method in Table IV. In all cases, adsorptions were measured a t three nitrogen (or krypton) pressures, and the results were calculated by the BET equation. To improve sensitivity, the pressure volume method commonly uses VOL 34, NO. 9, AUGUST 1962
11 53
Table IV. Comparison of Surface Areas by the Dynamic and Static Methods Surface Area, sq. m./g.
Sample Si02-A1,08 Catalyst A
Dynamic \112
Pressurevolume (static) 106 109
( :;
Si02-Al10a 66 Catalyst B Silicon carbide (powdered) 0 145" 0.143 (I