Anal. Chem. l W Q , 62,2313-2317
2313
Simultaneous Determination of Bruneaur-Emmett-Teller and Inverse Gas Chromatography Surface Areas of Solids Hengchang Song a n d Jon F.P a r c h e r *
Department of Chemistry, University of Mississippi, University, Mississippi 38677
Inverse gas chromatography, IGC, and mass spectrometric tracer pulse chromatography, MSTPC, were used concurrently to provide independent, but complementary, methods for the determination of adsorbent surface areas. Nitrogen at 77 K was used as the adsorbate in both procedures. The IGC technique involved the use of argon as the probe solute, whereas ”N, was used as the probe in the MSTPC method. The IGC method identifies the point of completion of a monolayer of adsorbed nitrogen and hence the surface area from the amount of adsorbed nitrogen requlred to change the primary retention mechanism for argon from gas-solid adsorption to gas-liquid partition in the adsorbed film of “liquid” nitrogen. The MSTPC method was used to determine the amount of nitrogen adsorbed. The experimentally measured adsorption isotherm data for N, at 77 K were fit to the classical BET equation to determine the surface area of different solids. The total surface area of the adsorbent in a packed GC, SFC, or HPLC column can be determined in situ, Le., without unpacking or disturbing the column in any way. The two independent surface area measurements were carried out simultaneously on a given adsorbent to provide both theoretical and modelless values for the monolayer capacity. It was demonstrated that the IGC, MSTPC, and volumetrically measured surface areas (where available) agreed with an average relative deviation of < I O % for a series of adsorbents including graphitized carbon blacks, silica, chemically bonded sllicas, glass beads, and diatomaceous earth chromatographic adsorbents. I n the case of the glass bead adsorbents, it was shown that the BET plot was not linear, whereas the IGC plots were linear and gave a more reliable surface area 8stimate. The method IGC was also used to determine the Henry’s law constant of 0.10 atm for argon in liquid nitrogen at 77 K.
INTRODUCTION Volumetric, gravimetric, and chromatographic methods have all been used successfully to determine the surface area of solids from the adsorption isotherms of nitrogen a t 77 K by the classical BET (I) technique. Although the BET method is popular and useful for the measurement of surface areas, the theory does not represent an accurate adsorption model. There are several discrepancies in the model that have been discussed extensively in the literature (2-4). The major deficiencies in the model are the assumptions of (i) a planar, homogeneous surface and (ii) negligible solute-solute interactions in the adsorbed phase. The primary datum from the BET method is the monolayer capacity of an adsorbent, and this can be determined relatively accurately. However, this must be related to the surface area by a “size” parameter for the adsorbate. The accepted standard is nitrogen with an assumed size of 16 A2 at 77 K, although there is some controversy concerning the validity of this figure for all adsorbents. Because of the uncertainties inherent in the BET method, several other procedures have been developed that do not 0003-2700/90/0362-2313$02.50/0
depend upon a specific theoretical model for the estimation of the monolayer capacity of a solid adsorbent. One such technique was developed by Serpinet (5) in 1976. The method utilized inverse gas chromatography, IGC, to determine the monolayer capacity of hydroxylated adsorbents for nonvolatile, high molecular weight “adsorbates”. The method was based on the assumption that the dominant retention mechanism for a probe solute would change from adsorption to partition at the point of formation of a monolayer of stationary liquid phase as the liquid loading was varied from submonolayer to multilayer surface coverage. The monolayer capacity was determined graphically from the breakpoint in a plot of the retention volume of the probe solute versus the reciprocal of the weight of stationary liquid phase in each column. The size parameter, which could vary with orientation for the high molecular weight adsorbate, had to be determined from adsorption on well-characterized adsorbents. Another potential problem with this method is that the surface areas previously measured with nonvolatile, high molecular weight adsorbates have often been observed (6, 7)to be significantly lower than the surface area from BET measurements with nitrogen on the same adsorbent. Thus, the method must be limited to meso- and macroporous adsorbents with pores sizes (usually >60 A) larger than the molecular size of the “adsorbates”. One variant of this method eliminates the need for multiple columns but requires retention volume measurements for 8-10 temperatures above and below the melting transition of the liquid phase as well as prior knowledge of the value of the partition coefficient of the probe solute in the stationary liquid phase. Although the modelless IGC methods avoid the theoretical problems of the BET method, the IGC methods also have the disadvantages and limitations already cited. An obvious solution to both the theoretical and practical problems of the BET and IGC methods would be the development of an IGC method using a s m d inert adsorbate (stationary liquid phase), such as nitrogen at 77 K. Recently, mass spectrometric tracer pulse chromatography with labeled nitrogen as the probe was used to determine the BET surface areas of diverse adsorbates from nitrogen adsorption isotherms (8). It was observed in that investigation that the retention volume of argon varied with the amount of nitrogen adsorbed. It is the purpose of the present work to introduce an IGC method for the measurement of surface areas using argon as the probe solute with nitrogen a t 77 K as the adsorbate. In this case, the BET and IGC measurements could be carried out simultaneously to yield both a theoretical value and a modelless value for the monolayer capacity of any adsorbent for nitrogen.
EXPERIMENTAL SECTION The instrumentation and experimental procedures have been described in a previous publication (8). The experimental procedure was tracer pulse chromatography with a mass specific detector. Mixtures of nitrogen and helium with differing compositions were used as the carrier gas. A gaseous mixture of neon, was injected with a gas sampling argon, and labeled nitrogen, 15N2, valve. The retention time of neon was used for the deadtime, to;the retention time of the labeled nitrogen was used to determine 0 1990 American Chemical Society
2314
ANALYTICAL CHEMISTRY, VOL. 62, NO. 21, NOVEMBER 1, 1990
Table I. Description of Chromatographic Columns and Adsorbents particle adsorbent material
size, pm
Carbopack A Carbopack B Carbopack C
180-250 180-250 150-180 5 5 5
silica
C18-bondedsilica diol-bonded silica cyano-bonded silica glass beads Chromosorb P
packing weight,
column dimens, mm
mg
100 x 3 60 X 3
150.0 44.1 831.3 49.1 15.4 40.3 59.6 9320 304.2
400 X 3 50 X 3 50 X 3 50 X 3 50 X 3 400 X 6.4 300 X 3
5 180-250 180-250
401
0
P d 9'
1
04
P/P
OB
08
150
200
250
300
Nitrogen Adsorbed ( k m o l / g )
P
I
100
Figure 2. Retention volume of argon on Chromosorb P with preadsorbed nitrogen at 77 K: (M) experimental data; (- - -) Henry's law line for argon in N,.
,n'
02
50
1
argon on the exposed solid surface. The retention volume decreased as the amount of adsorbed nitrogen increased because the nitrogen effectively blocked the surface for argon adsorption. At surface coverages higher than 150 pmol/g, the retention mechanism was primarily partition of argon into the adsorbed, liquidlike layer of nitrogen, Le., simply gasliquid partition chromatography with adsorbed nitrogen as the stationary liquid phase. The retention volume of argon increased directly with the amount of nitrogen adsorbed. Thus, the overall retention equation would have the form
(Nitrogen)
Figure 1. Adsorption isotherm of nitrogen on Chromosorb P at 77 K: (0)adsorption:
(A)desorption.
the amount of nitrogen adsorbed, r; and the retention time of argon was used to calculate the retention volume of the IGC probe solute. The surface area of the adsorbent was calculated from the linear form of the BET equation
c-1
P
r(P- P ) = ( E k )
+
P
cT,( p")
(1)
where P and Po are the pressure of nitrogen in the column and the vapor pressure of liquid nitrogen at the column temperature, rm is the monolayer capacity, and C is an empirical parameter that is related to the heat of adsorption and the shape of the isotherm in the region of monolayer formation. The adsorbents used in the investigation are listed in Table I.
RESULTS AND DISCUSSION Macroporous Adsorbent. The experimentally measured adsorption isotherm of nitrogen on Chromosorb P at 77 K is shown in Figure 1. The isotherm showed little or no hysteresis even at the highest pressures. The BET plot for this adsorbent was linear over the range 0.02 5 P I P 5 0.40 and gave a calculated surface area of 5.4 m2/g. Figure 2 shows the IGC retention volume data for argon on Chromosorb P as a function of the amount of nitrogen adsorbed, which was determined from the retention volume of the MSTPC probe, 15N2. The plot shows two distinct regions corresponding to the two predominant retention mechanisms, e.g., adsorption and partition. From the BET plots for nitrogen, it was calculated that a monolayer of nitrogen was formed at about 56 pmol/g (marked with an arrow in the figure). At surface coverages of less than 56 pmol/g, the retention mechanism was predominantly adsorption of
where K, is the partition coefficient for solute i in liquid nitrogen, u is the molar volume of liquid nitrogen, KA is the adsorption coefficient of solute i on the adsorbent, 8 is the fractional surface coverage of the solid adsorbent by nitrogen, and A, is the surface area of the adsorbent. The dashed line in the figure represents the calculated linear regression line for the data at the highest surface coverage (>ZOO pmollg). The value of K4u = 64 L/mol was determined from the slope of this line. No previous value for this solubility "parameter" was found in the literature. This type of solubility measurement would be difficult to achieve with classical volumetric or gravimetric procedures because the volatility of the solvent, N2,was greater than that of the solute, Ar. In fact, 77 K is below the melting point of solid argon. This measurement can be performed chromatographically because of the inherent capability to operate in the "infinite dilution" region of the partition isotherm of argon in liquid nitrogen. Several authors (9-13) have suggested that one way to determine the separate contributions to a complex retention mechanism was to plot the data in the form of reciprocals. That is, eq 2 can be rearranged to give
where rm= r/0 is the monolayer capacity required for the surface area calculations. The full equation would only hold for the range r 5 rm,i.e. KA = 0 for r 2 rm.The value of rm can be determined from a plot of ( V R / f - K g J vs l/r for values of r 5 rm.Such a plot is shown in Figure 3 for the Chromosorb P adsorbent. As is commonly observed, the plot was not perfectly linear at high values of l/r;however, linear
ANALYTICAL CHEMISTRY, VOL. 62, NO. 21, NOVEMBER 1, 1990
2315
2.5
z i h 4
3
=.E Y
U ~
I
-0.5 0
I
0.01
I
0.02
l/r
0.03
1
0.04
-0 5
1 0.05
(g/fimol)
0
1
I
1
0 0005
0.001
0.0015
0 002
1/ r (g/ mol) Figure 5. Reciprocal plot for argon on silica.
Figure 3. Reciprocal plot (eq 3) for argon on Chromosorb P: (W) experimental data; (-) hear regression results. The arrow shows the monolayer capacity calculated from the BET equation.
4000 1
20000
E
v
3000
1\ I
O L
0
1000
zoo0
3000
4000
5000
Amount of Nitrogen Adsorbed (umol/g) Figure 6. Retention volume of the IGC probe, argon, on silica. 0
04
02
P/P
OB
OB
1
(Nitrogen)
Flgure 4. Adsorption isotherms of nitrogen on 5-pm silica at 77 K: (0) experimental data: (- -) volumetric data ( 74).
-
regression of the data in the range 0.5 I{ V&- - K,u] I2.0 gave a value for rmof 58.3 ymollg a t ( VRt/r - K,y) = 0 and a surface area of 5.6 m2/g, which agreed with the BET value of 5.4 m2/g. The value of rm was determined from the intercept on the abscissa, rather than the ratio of the slope and ordinate intercept, in order to emphasize the graphic interpretation and empiricism of the analysis. Mesoporous Adsorbents. The full adsorption isotherm of nitrogen on a 5 pm silica stationary phase over a range of relative pressures from 0.01 to 0.95 is shown in Figure 4, along with independently measured (14) volumetric data. The BET surface area was determined to be 148 m2/g, and the volumetric measurements (14) gave values of 151 and 155 m2/g for duplicate analyses. A significant hysteresis loop, which is characteristic of mesoporous adsorbents, was observed a t the higher pressures. The MSTPC measurements were not reliable in this region due to the pressure drop across the column. A plot of the data in the form for IGC analysis is shown in Figure 5. Again, the plot was linear in the range 0.5 II VRl/r - K,u) I 2.0, and the calculated surface area was 141 m2/g, which is in agreement with the BET values. The IGC method provides an independent means for surface area measurement, which agrees with both chromatographic and
static BET results well within the f10% figure commonly cited for the uncertainty of the BET method (2). The data plotted in the more common form of VR(argon) vs r(nitrogen) on silica are illustrated in Figure 6 and show a maximum in the retention volume at very low surface coverage. Such maxima have been observed previously with both volatile (1516) and nonvolatile (17,18) adsorbates on graphitized carbon black adsorbents. The postulated cause of the maximum is enhanced adsorption and retention at low surface coverages due to solute-adsorbate-adsorbent (threebody) interactions, which are completely neglected in the BET model as well as the retention model represented by eq 2. Three chemically bonded stationary phases, e.g., diol-, cyano-, and octadecyl-bonded silica, were studied in the same manner. Figure 7 shows the retention volume data for argon on all four silica-based SFC adsorbents plotted in the form of the retention volume of argon as a function of the total amount of nitrogen adsorbed. The observed coincidence of the curves at high surface coverage confirms the supposition that the primary retention mechanism for the nitrogen-coated adsorbents was partition of argon into the adsorbed layer of condensed nitrogen. Thus, the retention volume of argon should be proportional to the amount of nitrogen adsorbed and independent of the type of adsorbent for multilayer adsorption of nitrogen. The straight line in Figure 7 had a measured slope of 66 L/mol, which gave a calculated Henry's law constant of 0.10 atm.
2316
ANALYTICAL CHEMISTRY, VOL. 62, NO. 21, NOVEMBER 1, 1990
200
00
0 2
04
06
08
1
P/P" (Nitrogen) 0
200
400
800
Amount of Nitrogen Adsorbed ( p m o l ) Figure 7. Retention volume of argon on sillca-based adsorbents: (0) silica: C,,-silica; (A)dol-silica; (0) CyanMilica; (-) linear regression for 1300 umol adsorbed.
m)
Figure 9. Adsorption isotherm of nitrogen on graphitized carbon black (Carbopack C) at 77 K: (0)experimental data: (-) Avgul and Kiselev (79). 20 ,
t y"
L
1001
-5
:0
20
30
40
50
Nitrogen Adsorbed ( k m o l / m z )
-
2
Figure 10. IGC plot for argon of Carbopack C with preadsorbed nitrogen.
7
B
0
I
! i t
2
t
15'-
E
I
v
3
;
1-
I
E
i
l L
i
I-
$005
0006
0007
bo08
0009
( E / m") Figure 8. Retention volume (A) and reciprocal plot (B) of the IGC probe, argon, on Carbopack A. i/i-
The diol-bonded silica gave a chromatographic BET surface area of 241 m2/g and an IGC surface area of 222 m2/g. For the cyano-bonded silica the results were 112 and 122 m2/g for the BET and IGC methods, respectively. Similar results, e.g., 94 and 106 m2/g, were observed for the octadecyl-bonded silica. In this case, the same material had been determined by volumetric analysis to have a BET surface area of 89 and 98 m2/g for duplicate analyses (14). Nonporous Adsorbents. Similar experiments were carried out for three graphitized carbon black adsorbents, e.g., Carbopacks A, B, and C. The results for the retention volume of argon on Carbopack A and the IGC analysis are shown in Figure 8 for six relative pressures of nitrogen in the range 0.02 IP / p I0.20. In this range the BET plot was linear with a calculated monolayer capacity of 124 pmol/g and surface area of 12.0 m2/g (8). The retention volume of the argon probe
decreased linearly with surface coverage for the lowest surface coverages, as shown in Figure SA. This is in agreement with the retention volume equation for IGC probe i on a liquidmodified solid adsorbent. Extrapolation of the linear portion of the plot at low 8 to VRl- K,p r = 0 gives the monolayer capacity. In this case, the calculated value was 129 pmol/g, which gave a surface area of 12.4 m2/g. The reciprocal plot is shown as Figure 8B, and the surface area measured from regression of the data in this form gives the same value of 12.4 m2/g. The surface area for one batch of Carbopack B was also determined by the same method. The measured BET surface area was 80 m2/g, whereas the ICG linear extrapolation produced a value of 82 m2/g. Carbopack C was investigated more extensively due to the popularity of this adsorbent. The adsorption isotherm of nitrogen is shown in Figure 9 along with previously measured data published by Avgul and Kiselev (19). The very sharp "knee" in the isotherm indicates the clean formation of a distinct monolayer of nitrogen at 9-10 pmol/m2. The BET analysis gave a surface area of 9.1 m2/g with a large C value of 340, also indicative of the sharp knee. The IGC analysis gave a surface area of 10.0 m2/g, which agreed within 10% of the BET area. The IGC plot for this adsorbent is shown in Figure 10 and clearly demonstrates a third type of retention mechanism in addition to adsorption on uncovered support and partition into the bulk liquid nitrogen. In the range 10-30 pmol/m2 of nitrogen adsorbed, i.e., in the region of formation of the second and third layers of adsorbed nitrogen, the retention volume of argon was constant or diminished slightly with
ANALYTICAL CHEMISTRY, VOL. 62, NO. 21, NOVEMBER 1, 1990
0’3
7
7 I
, .
R
‘;2
0.2
W
t
5-1 2:
0.1
0
I
I
0.05
0.1
0.15
02
P/P” (Nitrogen) Figure 11. BET plot for nitrogen on glass beads at 77 K.
increasing amount of nitrogen adsorbed. This indicates that the solubility of argon in (or the adsorption of argon on) the second and third adsorbed layers was slightly greater t h a n in the bulk liquid nitrogen. That is, another term would need to be added to eq 2 with a different K value for the presumably oriented second and third layers. The contribution of this mechanism would decrease with the amount of nitrogen adsorbed because the structured character of these layers was progressively destroyed until all of the adsorbed nitrogen acted as bulk liquid at high loadings. Thus,this exceptionally simple system, e.g., nitrogen/argon/carbon a t 77 K, displays very complex retention mechanisms including cooperative and competitive gas-solid adsorption, gas-liquid partition, and an uncertain mechanism involving the interaction of argon with a very thin film of adsorbed nitrogen only two to three molecular diameters thick. The simple BET equation is not adequate for this system, and a full retention volume equation would require at least four terms, two increasing with r
(cooperatiue adsorption and partition) and the other two decreasing with r (competitive adsorption). This type of complex retention mechanism has been observed for other “simple” systems involving volatile solutes and nonvolatile “adsorbates” on silica (20) and alumina (21). The nitrogen isotherm was also measured for a batch of glass beads with a nominal surface area of 0.014.1 m2/g. The BET plot for these beads is shown in Figure 11. The plot was not linear even over the nearly ideal range of 0.02 IP / P I0.2, and the calculated surface areas ranged from 0.07 m2/g at the high-pressure range to 0.12 m2/g at the low-pressure range. On the other hand, the IGC plot (eq 2) was linear over the range 0.2 I{VR,/r- K,p) I2.0 and gave a calculated surface area of 0.14 m2/g. This is an example where the complementary IGC method provided an independent surface area measurement in a case where the classical BET analysis was unsatisfactory.
CONCLUSIONS MSTPC and IGC methods can be combined to give two independent estimates of the amount of nitrogen or other adsorbate required to form a monolayer on a solid surface and thus the surface area of the solid. The MSTPC method produces a multipoint isotherm, which can be regressed to the BET model or used to determine the so-called “point B” of the isotherm. The IGC method measures the amount of
2317
adsorbate that must be adsorbed to change the retention mechanism for an IGC probe, such as argon, from adsorption on uncoated adsorbent surface to partition in “liquid” adsorbate when multilayer adsorption occurs. One method depends upon a theoretical model; the other method is em-‘ pirical. The methods are independent, but complementary. In most cases, both methods give the same surface area within experimental uncertainty. However, in the case where one method is unreliable, the other method will usually provide a better estimate of the adsorbent surface area. The advantages of this dual method over other chromatographic techniques for the measurement of solid surface areas are as follows: The primary experimental data are retention times not peak areas. No detector calibration or blanks are required. The experimental systems remain isothermal and isocratic (for a data point). The baseline and flow rate disturbances associated with adsorption (cooling) and desorption (heating) cycles and breakthrough curves are avoided. The surface areas of adsorbents in packed HPLC, SFC, or GC columns can be measured in situ. A large pressure drop across a packed column will degrade the accuracy of the BET method but has little or no effect on the IGC measurements. Any combination of adsorbates and probes could be used; however, nitrogen, argon, and krypton are the most convenient probes and adsorbates. Both methods, IGC and BET, measure only the monolayer capacity. An accurate size parameter for the probe adsorbate is required for the determination of surface areas. The methods are fast, simple, and accurate. However, a mass-specific detection system and labeled isotopic solutes are required for the MSTPC procedure. Registry No. Ar, 7440-37-1;Nz, 7727-37-9.
LITERATURE CITED Bruneaur, S.; Emmett, P. H.; Teller, E. J. Am. Chem. Soc. 1938, 60, 309. Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area and Porosity; Academic Press: New York, 1982. Ross, S.; Oliver, J. P. On Physical Adsorption; Interscience: New York, 1964. Steele, V. A. The Interaction of Gases with SolM Surfaces; Pergamon Press: New York, 1974. Serpinet, J. J. Chromatogr. 1976, 779, 483. Dyer, A.; Leigh, D.; Sharples, W. E. J. Chromatogr. 1976, 778, 319. Tremaine, P. R.; Wikjord, A. G.; LeBlanc, J. C. Can. J. Chem. 1982, 6 0 , 2859. Song, H.; Strubinger, J. R.; Parcher, J. F. J. Chromatogr. 1990, 578, 319. Conder, J. R.; Locke, D. C.; Purnell, J. H. J. Chromatogr. 1969, 73, 700. Conder, J. R.; Young, C. L. Physicochemical Measurements by Gas Chromatography; Wiley-Interscience: New York, 1979. Komaita, T.; Naito, K.; Takei, S. J. Chromatogr. 1975, 774, 1. Naito, K.; Takei, S. J. Chromatogr. 1980, 790, 21. Moriguchi, S.; Takei, S. J. Chromatogr. 1985, 350, 15. Micromeritics, Inc., Atlanta, GA. Hyver, K. J.; Parcher, J. F. Anal. Chem. 1984, 5 6 , 274. Parcher, J. F.; Johnson, D. M. J. Chromatogr. Sci. 1985, 2 3 , 459. Bruner, F.; Ciccioli, P.; Crescentini, G.; Pistolesi, M. T. Anal. Chem. 1973, 45, 1851. Di Corcia, A.; Liberti, A. I n Advances in Chromatography;Glddlngs, J. C., et al., Eds.; Marcel Dekker: New York, 1976; Chapter 7. Avgul, N. N.; Kiselev, A. V. I n Chemistry and Physics of Carbon; Walker, P. L., Ed.; Marcel Dekker: New York, 1970. Naito, K.; Sagara, T.; Takei, S. J. Chromatogr. 1990, 503, 25. Naito, K.; Ohwada, N.; Moriguchi, S.; Takei, S. J. Chromatogr. 1985, 330, 193.
RECEIVED for review April 9, 1990. Accepted July 30, 1990. Acknowledgment is made to the National Science Foundation for support of this work.