- Email: [email protected]
Elucidating the porous structure of activated carbon fibers using direct and indirect methods
Carbon Vol. 34, No. 10, pp. 1191&200,1996 Copyright 0 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved OOIX-6223/96 $15.00 + 0.00
Pergamon 80008-6223( 96)00065-6
ELUCIDATING THE POROUS STRUCTURE OF ACTIVATED CARBON FIBERS USING DIRECT AND INDIRECT METHODS M. A. DALEY,~ D. TANDON,~ J. ECONOMY~ and E. J. HIPPOS “1304 W. Green Street, Department of Materials Science and Engineering, University of Illinois at C-U, Urbana, IL 61801, U.S.A. bDepartment of Mechanical Engineering and Energy Processes, Southern Illinois University at Carbondale, Carbondale, IL 62901, U.S.A. (Received 3 January 1996; accepted in revised form 23 April 1996) Abstract-The porous structure of a series of commercial activated carbon fibers (ACFs) (with increasing surface area) was studied using both direct and indirect methods. The porous structure was characterized directly using scanning tunneling microscopy (STM) techniques and indirectly using the Dubinin-
Radushkevich-Stoeckli
(D-R-S)
eauation applied to the nitrogen adsorption
isotherm at 77 K. The
porous structure was imaged at ‘boih the fib&-surface and fiber cross-se&on and compared to analyses using indirect methods in terms of pore size, pore shape and pore size distribution. An average pore size was calculated from the fiber cross-section using section analysis in the standard STM software package and image analysis software. The pore size distribution of the ACFs was determined by applying the D-R-S equation to the nitrogen adsorption isotherm at 77 K. Additionally, a series of phenolic fibers activated under inert conditions was shown to form a stable porous structure. Comparison of this system to the ACFs using STM provided an important insight into the mechanism of pore generation in the ACFs. Copyright 0 1996 Elsevier Science Ltd
Key Words-Activated (D-R-S)
equation,
carbon fibers (ACFs), micropores,
scanning
tunneling
microscopy
mesopores,
Dubinin-Radushkevich-Stoeckli
(STM).
1. INTRODUCTION
Activated carbons have been used for many years to control contaminants in air and water [l]. Until recently, little progress has been made in relating the porous structure of activated carbon to its adsorption properties [ 2,3]. Traditionally, activated carbons have been characterized using the B.E.T. method applied to the adsorption isotherm to calculate surface area. Surface area was presumed to be a measure of a carbon’s ability to adsorb contaminants. It was believed that higher surface area activated carbons were more effective for the adsorption of contaminants than lower surface area activated carbons. Dubinin and then Foster proposed that surface area is not a true indicator of the adsorption characteristics of activated carbon because it does not contain information about the pore size, pore shape and pore surface chemistry [ 3,4]. A number of researchers have attempted to characterize the micropore size distribution of various carbons to understand their role in adsorption; but, they have been limited to the use of indirect techniques. These include the use of fundamental theory as applied to adsorption isotherms [4-91, the use of molecular probes [ 10-121, small angle X-ray scattering [13-171, and neutron scattering [18,19]. Although each of these techniques provides some insight into the microporous structure of the carbon, they require several assumptions or theories to interpret the results. Recently the microporosity of activated carbon has 1191
been directly imaged using scanning tunneling microscopy (STM) techniques [2,20] and transmission electron microscopy [21,22]. Using these techniques, it is possible to directly image the microporosity of activated carbon. In this paper we will demonstrate the use of STM for characterizing the microporous structure of a number of different activated carbon fibers (ACFs) at both the fiber surface and fiber crosssection and then proceed to address the following questions: What is the nature of the microporous structure of the ACFs? What is the origin of micropores in the ACFs? What is the average pore width as measured by STM compared to that measured using the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K? How does the micropore size affect the adsorption characteristics of the ACFs? These and a number of related questions addressed in the following sections.
will be
2. EXPERIMENTAL 2.1 Sample preparation The samples used in this study consisted of a series of activated carbon fibers with increasing surface area (increasing degree of burnoff, decreasing yield). These fibers were commercially prepared by Nippon Kynol by carbonizing and activating a phenolic fiber precursor (Kynol@) in steam/carbon dioxide. The prepara-
1192
M. A.
tion of these systems is extensively described in the literature [23,24]. The fibers used in this study were ACC-5092-10 (ACFlO), ACC-5092-15 (ACF15), and ACC-5092-25 (ACF25) and had B.E.T. surface areas (S.A.s) of 900 m’/g, 1200 m’/g and 2000 m’/g, respectively. The B.E.T. S.A. measurements were determined using the standard Brunnauer, Emmet and Teller (B.E.T.) method applied to an experimental adsorption isotherm over a relative pressure range of 0.01 to 0.25. The adsorption isotherms were measured using a Micromeritics (Norcross, Georgia) ASAP 2400 using nitrogen at 77 K. Another series of high surface area fibers was produced by activating phenolic fiber precursors under inert conditions in ultra high purity nitrogen at temperatures of 500 to 800°C each for a period of 30 minutes. The porosity in this new family of activated fibers was due solely to volatilization of degradation by-products. The sample chamber was purged with ultra-high purity nitrogen for 2 hours prior to reaction and the reaction vessel was cooled under ultra-high purity nitrogen to prevent the interaction of the sample with air. The B.E.T. S.A. measurements were performed by the method mentioned previously and resulted in carbon fabrics whose surface area increased with activation temperature, e.g. 35 m2/g to 650 m2/g for 30 minutes at activation temperatures of 500 and 800°C respectively. This family is designated as nitrogen activated phenolic fibers (NAPFs). It should be noted that a phenolic derived carbon fiber with very low surface area designated as ACFOS can only be obtained by heating to 2100°C. In our previous paper in this journal [Z], the temperature was erroneously given as 700°C. 2.2 STA4 techniques A commercial scanning tunneling microscope (STM) Nanoscope II (Digital Instruments Inc.) was employed for the study of the carbon fibers under ambient conditions. A piezo-electric head type “A” was used for scan areas no larger than 600 nm x 600 nm. The tips were made of Pt(SO)/Ir(20) and were commercially supplied by Digital Instruments Inc. (Nanotipm). The STM was secured on a suspended platform which was contained within a Styrofoam housing to minimize vibrations. The constant tunneling current mode was used for all studies. The magnitude of the tunneling current ranged between 2 and 3 nA and the magnitude of the bias voltage ranged from 20 to 100 mV. The surface relief is expressed with increasing shades of grey where lighter regions are “hills” and darker regions are “valleys”. The surface of the fibers was explored by placing the as-received fibers directly under the STM tip. Several methods were attempted for preparing crosssectional sections of the fibers. The best method involved packing a large number of oriented fibers into a rigid piece of polyethylene tubing. The tubing was then cut to expose the cross-sections of the fibers.
DALEY et al
Due to the dense packing of the fibers, the crosssections remained immobilized during scanning. The pore width and length as well as the pore size distribution were measured by section analysis using the standard STM software and image analysis programs (NIH image). It becomes important to provide a definition for a pore so that it will not be confused with surface roughness or cracks. Generally as the pore width increases, the apparent pore depth that is measurable increases. This is not to say that a small pore is not very deep, but, it becomes increasingly difficult to measure the “true pore depth” as the pore becomes smaller due to the interaction of the STM tip with the sides of the pore walls. This identifies one limitation of the STM technique, namely that it is not possible to measure the true pore depth or interconnectivity between pores. For the purpose of this paper, where the ACFs have a range of surface areas, we arbitrarily define a pore as having steep walls and an apparent pore depth of at least 0.2 nm. The apparent pore depth adresses our limitation to probe down into the pore as the pore becomes narrower. This arbitrary value was considered to be adequate after evaluating carbons with little or no porosity so as to not confuse porosity with surface roughness. In some cases this definition of a pore is limiting because it fails to identify the large number of pores less than 5 A in width because they fail to meet the criterion of a pore. Fortunately for this study it was found that pores less than 5 A in width do not contribute significantly to the total pore volume. Using the sectional analysis in the STM software package, sections of the STM image were chosen at random in both the horizontal and vertical directions and the pore width was measured by applying the arbitrary definition of a pore. Pore size distributions were measured using image analysis; however, this technique becomes complicated because of the appearance of interconnected pores due to the large amount of porosity. As a result, the measured pore size distributions may not be representative of the observed structure. Therefore, the pores must be assigned by discriminating between pores using discrete pore depth intervals. It is important that pores are not counted more than once. To avoid confusion, the pore size distributions presented in this paper were measured using the section analysis method in the standard STM software package.
2.3 Measurement
of adsorption
isotherms
A Coulter Omnisorb 100 (Hialeah, Florida) was used for the volumetric measurement of the nitrogen adsorption isotherm at 77 K. All samples were degassed for 36 hours under vacuum at 200°C. The nitrogen adsorption experiments were performed in static mode using the mass flow controller which was programmed to supply a fixed dose of nitrogen to the sample container. After equilibration, the pressure was recorded and the process was continued at higher
Elucidating the porous structure of activated carbon fibers using direct and indirect methods
1193
dose pressures until the saturation pressure was reached. The pressure difference between the previous pressure measurement and the next equilibrated pressure along with the dose volume was used to calculate the amount of nitrogen adsorbed by the sample. The pore size distribution was calculated from the nitrogen adsorption isotherm using the D-R-S equation [4].
3. RESULTS
3.1 STA4 results Over 800 images of the surface and cross-section of the commercial ACFs and NAPFs have been obtained using STM. The images and data which are presented are representative of this large body of data. 3.1.1 NAPFs. These STM images give important information concerning the initiation of microporosity in ACFs, since the NAPFs are prepared in the absence of external etching agents, and the microporosity can only arise from degradation and volatilization processes. Figure 1 is a 32 nm x 32 nm image of the fiber surface of the NAPF (800°C). Small mesopores and micropores are observed on the surface ranging in width up to 5 nm. Figure 2 is a 32 nm x 32 nm image of the crosssection of the NAPF (SOO’C). Micropores are observed across the fiber cross-section and measure less than 1.1 nm in width, with the majority of the micropores being in the pore range of less than 0.6 nm. Attempts to image the NAPFs with surface areas of only 35 m’/g (500°C) were complicated by (1) the low conductivity of the fiber and (2) the relatively low concentration of microporosity. 3.1.2 ACFs. These STM images convey information about the microporous structure of ACFs with surface areas of 900,120O and 2000 m’/g, respectively. Figure 3 is a 100 nm x 100 nm image of the fiber surface of ACFlO. At this length scale, discrete domains of mesopores and micropores were observed. Traditionally micropores have been classified as pores less than 2 nm in width and mesopores are between
Fig. 1. 32 nm x 32 nm STM image of the surface of NAPF.
Fig. 2. 32 x 32 nm STM image of the cross-section of NAPF.
Fig. 3. 100 x 100 nm STM image of the surface of ACFlO. 2 and 50 nm in width. The mesopores were elongated and measured up to 30 nm in length and 20 nm in width. Micropores were also observed in the mesoporous domains. Within these domains, mesopores varied in width across their length suggesting that these mesopores were generated by widening of micropores until the pore walls between adjacent micropores were burned out. The result was a mesopore of varying width. Figure 4 is a 32 nm x 32 nm image of the surface of ACFlO. At this higher magnification, a large number of elongated mesopores were observed in addition to a smaller number of ellipsoidally-shaped micropores. The mesopores ranged from 2 to 11 nm in length and from 2 to 6 nm in width while the micropores ranged in length and width from 1 to 1.9 nm. In this image, more mesoporosity was observed than microporosity. Surprisingly, the porosity was aligned parallel to the fiber direction. This feature may arise from ordering of polymeric chains at the fiber surface during the original melt spinning of the phenolic fiber precursor. Figure 5 is a 40 nm x 40 nm image of the crosssection of ACFlO. Ellipsoidally-shaped micropores were observed which ranged in size from several
1194
Fig. 4. 32 x 32 nm STM image of the surface
Fig. 5.40 x 40 nm STM image of the cross-section
M. A. DALE
of ACFlO.
of ACFIO.
angstroms to as large as 2.5 nm in width. Micropores were randomly distributed and homogeneous throughout the bulk of the fiber. In three-dimensions, the pores presumably were unoriented and formed a highly interconnected microporous network. Support for this interpretation was obtained from X-ray flat plate photos which only showed amorphous halos. It seems reasonable that the microporous structure must be interconnected from the large pore volumes of the ACFs [24]. Density measurements confirm that the porosity of the fibers must be interconnected [ 251. Figure 6 is a 61 nm x 61 nm image of the surface of ACF15. On this length scale, elongated micropores were observed parallel to the fiber axis. The micropores ranged in length from 1 to 20 nm and in width from 0.9 to 2 nm. This oriented microstructure represents less than 20% of the surface of ACFlS. Again, this oriented microporous structure may result from the shear associated with melt spinning of the original fiber precursor. Figure 7 is a 160 nm x 160 nm image of the surface of ACFlS. A highly mesoporous structure was
Fig. 6. 60 x 60 nm STM image of the surface of ACF15.
Fig. 7. 160 x 160 nm STM image of the surface
of ACF15.
observed at this length scale where the mesopores were ellipsoidally-shaped and were randomly distributed with respect to the fiber axis. This image is representative of more than 80% of the fiber surface. Figure 8 is a 60 nm x 60 nm image of the crosssection of ACF15. The micropores were randomly
Fig. 8.60 x 60 nm STM image of the cross-section
of ACF15.
Elucidating
the porous
structure
of activated
distributed in the bulk and were ellipsoidally-shaped. The micropores were homogeneous in the crosssection when the fiber was scanned across its entire diameter. It was not possible to resolve the interface between the fiber surface and cross-section. However, this region must be less than 60 nm since no difference in the bulk structure was observed when the fiber was scanned to the fiber edge. The pore size ranged from several angstroms to 4.2 nm in width. Figure 9 is a 100 nm x 100 nm image of the surface of ACF25. Large mesopores were observed on the surface of ACF25. These mesopores were ellipsoidally-shaped and ranged in size from 5 to 200 nm. Elongated micropores were observed on the surface of the ACF25. These micropores ranged in width from less than 0.5 to 2 nm and were as large as 30 nm in length. Figure 10 is a 32 nm x 32 nm image of the surface of ACF25. Both micropores and mesopores were observed. The degree of etching is much less in this image than that shown in Fig. 9 suggesting that the degree of gasification at the surface occurs in many different stages.
carbon
fibers using direct and indirect
of ACF25.
1195
Figure 11 is a 60 nm x 60 nm image of the crosssection of ACF25. Ellipsoidally-shaped micropores and small mesopores were readily observed. The pore size ranged from several angstroms to 46 A in width. The pores were randomly distributed and homogeneous across the fiber cross-section. Pore size distributions, Figs 12-14, were measured from the STM images of the cross-sections of ACFlO, ACF15 and ACF25 using section analysis and image analysis. The average pore size (width) for ACFlO is
Fig. 11. 60 x 60 nm
Fig. 9. 100 x 100 nm STM image of the surface
methods
0
STM
image ACF25.
of the
cross-section
20
5
of
25
PO:: Width’;& Fig. 12. Measured pore size distribution in the bulk for ACFlO as imaged at the fiber cross-section using STM.
16.5
23 1
Pore Width
Fig. 10. 32 x 32 nm STM image of the surface
of ACF25.
297
363
42.9
49.5
(A)
Fig. 13. Measured pore size distribution in the bulk for ACFlS as imaged at the fiber cross-section using STM.
M. A. DALEY et al. evolution of volatile species. The stability of these incipient micropores is directly related to the stability and persistence of the cross-linked structure in the precursor fiber. The transition from the surface porosity to the bulk porosity must result in an interface consisting of narrowing pores or wedge-shaped pores.
3.2 Adsorption results 3.3
9.9
16.5
23.1
29.7
Pore Width
36.3
42.9
49.5
(A)
Fig. 14. Measured pore size distribution in the bulk for ACF25 as imaged at the fiber cross-section using STM. 0.94 nm and ranged from
The nitrogen adsorption isotherms at 77 K for the ACFs are shown in Fig. 15 for a partial pressure range of 1O-5 to 0.12. All isotherms are Type I and exhibit microporous character. From the hysteresis loop of the adsorption isotherm from partial pressures of 10m5 to 1, we can conclude that the ACF25 had some mesoporosity. By applying standard theories to the adsorption isotherm, the pore size may be determined. The Dubinin-Radushkevich-Stoeckli (D-R-S) eqn (1) is one method of calculating the pore size distribution [2]. It is based on the equation
x exp ((- 1*(M2)*((~~-~2)2)/(2*b2)))
(1)
where Wz is the total micropore volume, x, is the pore half-width assuming a slit-shaped pore and corresponds to the maximum in the distribution curve, and 6 is the variance in B when a Gaussian distribution is assumed. Using the DubininRadushkevich (D-R) equation, the micropore volume, Wz, was calculated as the intercept on a log W vs log2(P,/P) plot, as depicted in Fig. 16. In this case, the micropore volume has been calculated for ACFlO as 0.392 mL/g of carbon fabric, for ACF15 as 0.481 mL/g of carbon fabric, and for ACF25 as 0.762 mL/g of carbon fabric. The value for x, and 6 may be determined from the experimental adsorption isotherm using the following relationship, eqn (2): Bj = - (ln( Wj)- ln( Wj_ I))/(( T/B)‘*( log2 x
~p~lpj~~10~2~p~lpj-~~~~
(2)
where Wis the amount of nitrogen adsorbed in mL/g of fabric, P/P, is the partial pressure, Tis the adsorption temperature which in this case is 77 K, and B is the adsorbent-adsorbate interaction parameter which has been discussed elsewhere and is equal to 0.43 [26]. Bj may be related to x using the following eqns (3) and (4): B.=M*x? J I
(3)
where M=(0.01915/k)2
(4)
and the determination of k has been described elsewhere and in this case is equal to 13 [4,16]. From the distribution of the xjs, 6 may be calculated. From our analysis, 6 was calculated as 2.71*10-’ for ACFlO, 2.9425*10-’ for ACF15, and 7.8247*10-’ for ACF25. The values of x, as calculated using the D-R equation are 0.583 nm for ACFlO, 0.648 nm for
Elucidating
the porous
structure
of activated
Nitrogen
carbon
adsorption
fibers using direct and indirect
isotherms
methods
1197
at T=77K
,O°C
P
B 6
300
2 3
200
.
.~~oo***~~o~o~~*
s >
0’
adsorption
.
.
.
.
.
.
A ACF25 . ACFLS 0 ACFlO
100
Fig. 15. Nitrogen
.
O.b2
isotherms
~ y=2.4041 + -0.010273x y=2.4925 + -0.012684x ......-.- y=2.6928 + -0.023025x
O.b4
0.66 P/PO
for the ACFs at T=77
O.b8
K over a range
o.io
’ 0.1
of partial
pressures
from zero to 0.12.
R=0.99657 R=0.99729 R=0.99402
2.1 2.6 2.5 g2.4 + 2.3 0
5
IO
15
20
25
30
35
40
Pore width (A) 2.1 0
5
10
15
20
log* (PO/P)
Fig. 17. Pore sire distribution for the ACFs as calculated from the nitrogen adsorption data using the DubininRadushkevisch-Stoeckli equation.
Fig. 16. Nitrogen adsorption data for the ACFs in the form of a Dubinin-Radushkevisch Plot. 4. DISCUSSION
4.1 Nature of the microporous structure ACFlS, and 0.873 nm for ACF25, respectively. Using these parameters, the D-R-S equation can be solved yielding the pore size distribution as a change in micropore volume per change in micropore halfwidth as a function of the micropore half width. In Fig. 17, the pore size distributions are given for ACFlO, ACF 15 and ACF25. As calculated from the D-R-S equation, the ACFlO had an average pore width of 1.166 nm and ranged from 0.5 to 1.7 nm, the ACF15 had an average pore width of 1.296 nm and ranged from 0.7 to 1.8 nm, and the ACF25 had an average pore width of 1.746 nm and ranged from 0.6 to 2.7 nm. From this distribution, it is apparent that the pore size increased as the surface area increased and that the pore size distribution became broader with increasing surface area.
The structure of any activated carbon depends on the precursor and etching conditions. Generally, the pore surface of microporous carbons consists of highly convoluted carbon sheets which are often crooked and the structure is highly disorganized [22]. Even though the structure is highly disorganized, it has been proposed to contain “slit-shaped” micropores [27]. These micropores are basically the gaps between the crooked or crumpled aromatic carbon sheets. Using STM, both elongated and ellipsoidal micropores and mesopores can be observed at the ACF surface. At the fiber cross-section, ellipsoidally-shaped micropores or small mesopores have been identified. The pores at the fiber cross-section were randomly distributed and homogeneous throughout the bulk. This evidence suggests that the structure of the carbon consists of many elongated
1198
M. A. DALEY~~~.
tubes which wind and twist throughout the carbon fiber creating its microporous character. These tubes can be viewed as an interconnected porous network consisting of pores of varying size. These pores are not isolated entities but must be interconnected due to the large pore volumes of the ACFs as shown from adsorption experiments and density measurements. Unfortunately, we are not able to directly image the interconnectivity within the porous network. When the porous tube is sliced or viewed from the edge, an elongated pore is observed. From these data, we infer that the structure of the ACFs has both an elongated pore (tube length) with a limiting pore diameter denoted by the cross-section of the tube which when viewed at the cross-section resembles an ellipsoidally-shaped micropore. The structure at the fiber surface consisted of large mesopores and some micropores whereas the structure at the fiber cross-section was far more uniform and consisted of micropores and small mesopores. Since the pore structure of the activated carbon fibers is continuous, the larger mesopores at the surface must narrow and empty into the micropores and smaller mesopores in the bulk of the fiber resulting in a narrowing pore at the surface-bulk interface. It was not possible to directly image the interface between the fiber surface and the fiber cross-section. From scans across the fiber diameter, it has been shown that the surface to bulk interface is strictly less than 60 nm in width compared to a 12 pm fiber.
4.2 Micropore generation jiber precursors
in ACFs jiom phenolic
From the work on the NAPFs, it is apparent that high surface area activated carbon fibers may be produced under inert conditions presumably from the evolution of gaseous by-products. This does not appear to be a general phenomenon for fibers that char, but rather is limited to those systems that develop a stable cross-linked structure during charring. Thus the micropores begin to develop at temperatures as low as 500°C (30 minutes) producing surface areas of 35 m’/g and up to 650 m’/g at SOO’C (30 minutes). From the images of the NAPFs, one observes the presence of a microporous structure where the measured pore width in the bulk is less than 1.1 nm and where most of the micropores are less than 0.6 nm in width. As might be expected, the surface of the NAPFs is etched to a lesser degree than those of the corresponding ACFs as observed from the STM images. The weight loss of the ACFs is much larger than that of the NAPFs and the fiber diameter of the ACFs generally decreases from 15 to 9-12 pm whereas only modest change is noted in the fiber diameter of the NAPFs. The difference in surface features between NAPFs and ACFs provides an insight into the two mechanisms of pore formation operative at the fiber surface. Thus, the ACFs are aggressively attacked at the fiber surface due to oxidative etching whereas the
only etching which occurs at the surface of the NAPFs is due to the evolution of gaseous by-products. Despite the differences in etching of the fibers at the surface, the porosity in the bulk of the fibers as observed from all of the fiber cross-sections remains remarkably homogeneous. The NAPFs obviously have very small micropores which continue to widen when exposed to an oxidative environment. Presumably, the homogeneous distribution of micropores which are initially generated at - 500°C result from the evolution of gases evolving from a very stable, cross-linked phenolic precursor.
4.3 Factors afectingpore
structure
Both the carbon precursor and treatment method affect the shape, size and distribution of porosity in a carbon. Laine and Yumes studied the effects of various treatments on the pore size distribution of activated carbon from coconut shell. When the carbon was activated with carbon dioxide at 800°C both wide mesopores and narrow micropores were observed. Catalyst (K,PO,) addition increased the micropore diameter and decreased the number and diameter of the macropores [28]. Kasaoka et al. have carried out an elegant study using various dyes in the liquid phase as molecular probes of the micropores and mesopores in activated carbon fibers (Kynol). They showed that the higher surface area activated carbon fibers were more effective at adsorbing larger dye molecules due to a larger critical pore diameter than activated carbon fibers with lower surface areas. Foster also showed that the average pore size increases with decreasing yield of the activated carbon fiber using the Dubinin-Radushkevich equation applied to the nitrogen adsorption isotherm at 77 K [26]. The images from the STM studies confirm these results and verify that the average pore size increases with increasing surface area. Measured and calculated pore size distributions reveal that the pore size distribution becomes broader as the activation yield decreases. Based on the increase in pore size with increasing extent of reaction, one can conclude that the activation of activated carbon fibers in the bulk occurs in three stages: (1) generation of small micropores, (2) widening of small micropores creating larger micropores, and (3) generation of small mesopores by continued widening of larger micropores until burnout of adjacent walls between micropores. 4.4 Comparison
of direct and indirect techniques to measure pore size
The average pore width as measured by STM was 0.94nm (ACFlO), 1.61 nm (ACF15) and 1.94 nm (ACF25) while the average pore size as calculated using the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K was 1.17 nm (ACFlO), 1.3 nm (ACF15) and 1.75 nm (ACF25). The average pore size as calculated from the D-R-S equation may not be directly compared with that directly
Elucidating the porous structure of activated carbon fibers using direct and indirect methods measured using STM since the first is a volumetric average and the second is a number average. However, the general trends may be thoroughly examined and compared. The D-R-S equation assumes a slit-shaped micropore and a Gaussian distribution of pores about the average pore-half width. STM techniques reveal both elongated and ellipsoidally-shaped pores and a non-Gaussian distribution about the average pore width. Despite the differences between the assumptions in the D-R-S equation and the features observed using STM, the measured and calculated pore width are surprisingly similar. The calculated and measured average pore size increase with increasing activation. Furthermore, the pore size distributions become broader with increasing activation. 4.5 Implications of pore size on adsorption Marsh [S] proposed a relationship between pore size and adsorption. Using the DubininRadushkevich equation, he demonstrated that as % burnoff increased the pore half-width increased. These results were correlated to the experimental adsorption isotherm and it was proposed that activated carbons with smaller pore size were more effective at removing contaminants at low concentrations due to a higher overlap in potential than activated carbons with larger pore sizes. Foster [26] extended this analysis to a series of activated carbon fibers. He observed a cross-over regime for the adsorption of butane by the ACFs. At low concentrations the ACFs with smaller pore size (lower surface area) were more effective at adsorbing butane whereas at higher concentrations of butane the ACFs with larger pore size (higher surface area) were more effective. The higher surface area ACFs are more efficient than the lower surface area ACFs because they have a larger pore volume and the pores of the lower surface area ACF are filled. The assumptions made to explain the crossover regime namely that lower surface area ACFs have a small pore size and that higher surface area ACFs have a larger pore size have been confirmed using STM.
ACFs can best be described as consisting of many elongated tubes which wind and twist throughout the carbon fiber creating its microporous character. These pores are not isolated entities but must be interconnected due to the large pore volumes of the ACFs. Based on the studies of the NAPFs, the generation of micropores begins at temperatures as low as 500°C due to the evolution of volatiles. To our surprise, surface areas of up to 650 m’/g were obtained by the reaction of the phenolic fiber under inert conditions of 800°C for 30 minutes. It seems reasonable to conclude that this process contributes significantly to pore generation in the ACFs. The average micropore half-width is observed to increase with increasing surface area using both STM techniques and the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K. Also, the pore size distribution becomes broader with increasing surface area suggesting that pore widening and pore generation occur cooperatively or that pores are generated over a broad activation range and then pore widening occurs for some pores at the expense of smaller pores. The average micropore width as measured using STM techniques was 0.94 nm (ACFlO), 1.6071 nm (ACFlS) and 1.94 nm (ACF25) while the average pore size obtained indirectly using the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K was 1.17 nm (ACFlO), 1.30 nm (ACFlS) and 1.746 nm (ACF25). The calculated and measured average pore size increased with increasing activation. The pore size distribution became broader with increasing activation. The crossover regime observed by Marsh and Foster was confirmed to be the result of increasing pore size with increasing surface area as proposed by the original researchers.
REFERENCES 1. A. L. Myers and G. Belfort, Proceedings of the Engineer-
2.
5. CONCLUSIONS STM revealed ellipsoidal micropores and ellipsoidal and elongated mesopores at the fiber surface and ellipsoidal micropores and small mesopores at the fiber cross-section. The interface between the fiber surface and the bulk could not be directly observed, but presumably results in a narrowing of mesopores into smaller mesopores or micropores resulting in wedge shaped pores. The surface structure is less than 60 nm in thickness and thus represents a small portion of the total fiber. No evidence was found in the ACFs to support the view of slit-shaped pores. The micropores and small mesopores at the fiber cross-section are randomly distributed and homogeneous across the fiber diameter. The structure of the
1199
3. 4. 5. 6. 7,
ing Foundation Conference held at Schloss Elmau, Bavaria, West Germany, May 6-11, (1983). J. Economv. M. A. Dalev, E. Hiuno and D. Tandon, carbon 33,-j, 344 (1995): I_ K. L. Foster, R. G. Fuerman, J. Economy, S. M. Larson and M. J. Rood, Chemistry of Materials 4, 5, 1068 (1992). M. M. Dubinin and H. F. Stoeckli, J. Colloid and Interface Science 75, 1, 34-42 (1980). H. Marsh and B. Rand, J. Colloid and Interface Science 33, 101-l 15 (1970). E. Bekyarova and D. Mehandjiev, J. Colloid and Interface Science 160, 115 (1993). E. M. Freeman, T. Siemienjewskam, H. Marsh and B. Rand, Carbon 8, 7-17 (1970). J. P. Olivier, W. B. Conklin and M. V. Szombathrly, Determination of Pore Size Distribution from Density Functional Theory: A comparison of Nitrogen and Argon Results. Micromeritics Instrument Corporation, Inc.,
Norcross, GA 30093 U.S.A. J. P. Olivier, W. B. Conklin
and
M. V. Szombathrly,
Determination of Pore Size Distribution from’ Density Functional Theory: A comparison of Nitrogen and Argon
1200
10.
11. 12. 13.
14. 15. 16. 17. 18.
M. A. DALEY et al.
Results. Micromeritics Instrument Corporation, Inc., Norcross, GA 30093 U.S.A. S. Kasaoka, Y. Sakata, E. Tanaka and R. Naitoh, international Chemical Engineering 29, 1, 101-114 (1989). S. Kasaoka, Y. Sakata, E. Tanaka and R. Naitoh, International Chemical Engineering 29, 134-142 (1989). J. J. Kipling and R. B. Wilson, J. Appl. Chem. 10, 109%113 (1960). A. Guinier, G. Fourret, C. Walker and K. Yudowitch. Small Angle Scattering of X-rays. Wiley, New York (1955). M. D. Foster and K. F. Jensen Carbon 29, 2, 271-282 (1991). P. W. Schmidt, J. Appl. Cryst. 24, 414-435 (1991). M. M. Dubinin, G. M. Plavnik and E. D. Zaverina, Carbon 2, 261-268 (1964). N. Setoyama, M. Ruike, T. Kasu, T. Suzuki and K. Kaneko, Langmuir 9, 2612-2617 (1993). J. Kieffer, J. Appl. Phys. 72, 12 (1992).
19. J. D. F. Ramsay, In Neutron Scattering from Porous Solids from Characterization of Porous Solids pp 23-34. Elsevier, Amsterdam (1988). 20. W. P. Hoffman, M. B. Fernandez and M. B. Rao, Carbon 32, 1383 (1994). 21. M. G. Dobb, H. Guo and D. J. Johnson, In Carbon 1994 Preprints of Extended Abstracts. Granada Spain (July, 1994). 22. H. Marsh, D. Crawford, T. M. O’Grady and A. Wennerberg, Carbon 20, $419-426 (1982). 23. James Economy and Rodger A. Clark, Patent Application Number 710,292, filed March 4, 1968. 24. Joseph S. Hayes, Jr., In American Kynol. Reprinted from Kirk-Othmer: Encyclopedia of Chemical Technology, Vol. 16, Third Edition, pp. 125-138. 25. R. Y. Lin and J. Economy, Applied Polymer Symposium. 21, 143-152 (1973). 26. Ken Foster, Ph.D. thesis, University of Illinois, U.S.A. (1993). 27. H. F. Stoekli, Carbon 28, 1 (1990). 28. J. Laine and S. Yumes, Carbon 30, 601 (1992)
Pergamon 80008-6223( 96)00065-6
ELUCIDATING THE POROUS STRUCTURE OF ACTIVATED CARBON FIBERS USING DIRECT AND INDIRECT METHODS M. A. DALEY,~ D. TANDON,~ J. ECONOMY~ and E. J. HIPPOS “1304 W. Green Street, Department of Materials Science and Engineering, University of Illinois at C-U, Urbana, IL 61801, U.S.A. bDepartment of Mechanical Engineering and Energy Processes, Southern Illinois University at Carbondale, Carbondale, IL 62901, U.S.A. (Received 3 January 1996; accepted in revised form 23 April 1996) Abstract-The porous structure of a series of commercial activated carbon fibers (ACFs) (with increasing surface area) was studied using both direct and indirect methods. The porous structure was characterized directly using scanning tunneling microscopy (STM) techniques and indirectly using the Dubinin-
Radushkevich-Stoeckli
(D-R-S)
eauation applied to the nitrogen adsorption
isotherm at 77 K. The
porous structure was imaged at ‘boih the fib&-surface and fiber cross-se&on and compared to analyses using indirect methods in terms of pore size, pore shape and pore size distribution. An average pore size was calculated from the fiber cross-section using section analysis in the standard STM software package and image analysis software. The pore size distribution of the ACFs was determined by applying the D-R-S equation to the nitrogen adsorption isotherm at 77 K. Additionally, a series of phenolic fibers activated under inert conditions was shown to form a stable porous structure. Comparison of this system to the ACFs using STM provided an important insight into the mechanism of pore generation in the ACFs. Copyright 0 1996 Elsevier Science Ltd
Key Words-Activated (D-R-S)
equation,
carbon fibers (ACFs), micropores,
scanning
tunneling
microscopy
mesopores,
Dubinin-Radushkevich-Stoeckli
(STM).
1. INTRODUCTION
Activated carbons have been used for many years to control contaminants in air and water [l]. Until recently, little progress has been made in relating the porous structure of activated carbon to its adsorption properties [ 2,3]. Traditionally, activated carbons have been characterized using the B.E.T. method applied to the adsorption isotherm to calculate surface area. Surface area was presumed to be a measure of a carbon’s ability to adsorb contaminants. It was believed that higher surface area activated carbons were more effective for the adsorption of contaminants than lower surface area activated carbons. Dubinin and then Foster proposed that surface area is not a true indicator of the adsorption characteristics of activated carbon because it does not contain information about the pore size, pore shape and pore surface chemistry [ 3,4]. A number of researchers have attempted to characterize the micropore size distribution of various carbons to understand their role in adsorption; but, they have been limited to the use of indirect techniques. These include the use of fundamental theory as applied to adsorption isotherms [4-91, the use of molecular probes [ 10-121, small angle X-ray scattering [13-171, and neutron scattering [18,19]. Although each of these techniques provides some insight into the microporous structure of the carbon, they require several assumptions or theories to interpret the results. Recently the microporosity of activated carbon has 1191
been directly imaged using scanning tunneling microscopy (STM) techniques [2,20] and transmission electron microscopy [21,22]. Using these techniques, it is possible to directly image the microporosity of activated carbon. In this paper we will demonstrate the use of STM for characterizing the microporous structure of a number of different activated carbon fibers (ACFs) at both the fiber surface and fiber crosssection and then proceed to address the following questions: What is the nature of the microporous structure of the ACFs? What is the origin of micropores in the ACFs? What is the average pore width as measured by STM compared to that measured using the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K? How does the micropore size affect the adsorption characteristics of the ACFs? These and a number of related questions addressed in the following sections.
will be
2. EXPERIMENTAL 2.1 Sample preparation The samples used in this study consisted of a series of activated carbon fibers with increasing surface area (increasing degree of burnoff, decreasing yield). These fibers were commercially prepared by Nippon Kynol by carbonizing and activating a phenolic fiber precursor (Kynol@) in steam/carbon dioxide. The prepara-
1192
M. A.
tion of these systems is extensively described in the literature [23,24]. The fibers used in this study were ACC-5092-10 (ACFlO), ACC-5092-15 (ACF15), and ACC-5092-25 (ACF25) and had B.E.T. surface areas (S.A.s) of 900 m’/g, 1200 m’/g and 2000 m’/g, respectively. The B.E.T. S.A. measurements were determined using the standard Brunnauer, Emmet and Teller (B.E.T.) method applied to an experimental adsorption isotherm over a relative pressure range of 0.01 to 0.25. The adsorption isotherms were measured using a Micromeritics (Norcross, Georgia) ASAP 2400 using nitrogen at 77 K. Another series of high surface area fibers was produced by activating phenolic fiber precursors under inert conditions in ultra high purity nitrogen at temperatures of 500 to 800°C each for a period of 30 minutes. The porosity in this new family of activated fibers was due solely to volatilization of degradation by-products. The sample chamber was purged with ultra-high purity nitrogen for 2 hours prior to reaction and the reaction vessel was cooled under ultra-high purity nitrogen to prevent the interaction of the sample with air. The B.E.T. S.A. measurements were performed by the method mentioned previously and resulted in carbon fabrics whose surface area increased with activation temperature, e.g. 35 m2/g to 650 m2/g for 30 minutes at activation temperatures of 500 and 800°C respectively. This family is designated as nitrogen activated phenolic fibers (NAPFs). It should be noted that a phenolic derived carbon fiber with very low surface area designated as ACFOS can only be obtained by heating to 2100°C. In our previous paper in this journal [Z], the temperature was erroneously given as 700°C. 2.2 STA4 techniques A commercial scanning tunneling microscope (STM) Nanoscope II (Digital Instruments Inc.) was employed for the study of the carbon fibers under ambient conditions. A piezo-electric head type “A” was used for scan areas no larger than 600 nm x 600 nm. The tips were made of Pt(SO)/Ir(20) and were commercially supplied by Digital Instruments Inc. (Nanotipm). The STM was secured on a suspended platform which was contained within a Styrofoam housing to minimize vibrations. The constant tunneling current mode was used for all studies. The magnitude of the tunneling current ranged between 2 and 3 nA and the magnitude of the bias voltage ranged from 20 to 100 mV. The surface relief is expressed with increasing shades of grey where lighter regions are “hills” and darker regions are “valleys”. The surface of the fibers was explored by placing the as-received fibers directly under the STM tip. Several methods were attempted for preparing crosssectional sections of the fibers. The best method involved packing a large number of oriented fibers into a rigid piece of polyethylene tubing. The tubing was then cut to expose the cross-sections of the fibers.
DALEY et al
Due to the dense packing of the fibers, the crosssections remained immobilized during scanning. The pore width and length as well as the pore size distribution were measured by section analysis using the standard STM software and image analysis programs (NIH image). It becomes important to provide a definition for a pore so that it will not be confused with surface roughness or cracks. Generally as the pore width increases, the apparent pore depth that is measurable increases. This is not to say that a small pore is not very deep, but, it becomes increasingly difficult to measure the “true pore depth” as the pore becomes smaller due to the interaction of the STM tip with the sides of the pore walls. This identifies one limitation of the STM technique, namely that it is not possible to measure the true pore depth or interconnectivity between pores. For the purpose of this paper, where the ACFs have a range of surface areas, we arbitrarily define a pore as having steep walls and an apparent pore depth of at least 0.2 nm. The apparent pore depth adresses our limitation to probe down into the pore as the pore becomes narrower. This arbitrary value was considered to be adequate after evaluating carbons with little or no porosity so as to not confuse porosity with surface roughness. In some cases this definition of a pore is limiting because it fails to identify the large number of pores less than 5 A in width because they fail to meet the criterion of a pore. Fortunately for this study it was found that pores less than 5 A in width do not contribute significantly to the total pore volume. Using the sectional analysis in the STM software package, sections of the STM image were chosen at random in both the horizontal and vertical directions and the pore width was measured by applying the arbitrary definition of a pore. Pore size distributions were measured using image analysis; however, this technique becomes complicated because of the appearance of interconnected pores due to the large amount of porosity. As a result, the measured pore size distributions may not be representative of the observed structure. Therefore, the pores must be assigned by discriminating between pores using discrete pore depth intervals. It is important that pores are not counted more than once. To avoid confusion, the pore size distributions presented in this paper were measured using the section analysis method in the standard STM software package.
2.3 Measurement
of adsorption
isotherms
A Coulter Omnisorb 100 (Hialeah, Florida) was used for the volumetric measurement of the nitrogen adsorption isotherm at 77 K. All samples were degassed for 36 hours under vacuum at 200°C. The nitrogen adsorption experiments were performed in static mode using the mass flow controller which was programmed to supply a fixed dose of nitrogen to the sample container. After equilibration, the pressure was recorded and the process was continued at higher
Elucidating the porous structure of activated carbon fibers using direct and indirect methods
1193
dose pressures until the saturation pressure was reached. The pressure difference between the previous pressure measurement and the next equilibrated pressure along with the dose volume was used to calculate the amount of nitrogen adsorbed by the sample. The pore size distribution was calculated from the nitrogen adsorption isotherm using the D-R-S equation [4].
3. RESULTS
3.1 STA4 results Over 800 images of the surface and cross-section of the commercial ACFs and NAPFs have been obtained using STM. The images and data which are presented are representative of this large body of data. 3.1.1 NAPFs. These STM images give important information concerning the initiation of microporosity in ACFs, since the NAPFs are prepared in the absence of external etching agents, and the microporosity can only arise from degradation and volatilization processes. Figure 1 is a 32 nm x 32 nm image of the fiber surface of the NAPF (800°C). Small mesopores and micropores are observed on the surface ranging in width up to 5 nm. Figure 2 is a 32 nm x 32 nm image of the crosssection of the NAPF (SOO’C). Micropores are observed across the fiber cross-section and measure less than 1.1 nm in width, with the majority of the micropores being in the pore range of less than 0.6 nm. Attempts to image the NAPFs with surface areas of only 35 m’/g (500°C) were complicated by (1) the low conductivity of the fiber and (2) the relatively low concentration of microporosity. 3.1.2 ACFs. These STM images convey information about the microporous structure of ACFs with surface areas of 900,120O and 2000 m’/g, respectively. Figure 3 is a 100 nm x 100 nm image of the fiber surface of ACFlO. At this length scale, discrete domains of mesopores and micropores were observed. Traditionally micropores have been classified as pores less than 2 nm in width and mesopores are between
Fig. 1. 32 nm x 32 nm STM image of the surface of NAPF.
Fig. 2. 32 x 32 nm STM image of the cross-section of NAPF.
Fig. 3. 100 x 100 nm STM image of the surface of ACFlO. 2 and 50 nm in width. The mesopores were elongated and measured up to 30 nm in length and 20 nm in width. Micropores were also observed in the mesoporous domains. Within these domains, mesopores varied in width across their length suggesting that these mesopores were generated by widening of micropores until the pore walls between adjacent micropores were burned out. The result was a mesopore of varying width. Figure 4 is a 32 nm x 32 nm image of the surface of ACFlO. At this higher magnification, a large number of elongated mesopores were observed in addition to a smaller number of ellipsoidally-shaped micropores. The mesopores ranged from 2 to 11 nm in length and from 2 to 6 nm in width while the micropores ranged in length and width from 1 to 1.9 nm. In this image, more mesoporosity was observed than microporosity. Surprisingly, the porosity was aligned parallel to the fiber direction. This feature may arise from ordering of polymeric chains at the fiber surface during the original melt spinning of the phenolic fiber precursor. Figure 5 is a 40 nm x 40 nm image of the crosssection of ACFlO. Ellipsoidally-shaped micropores were observed which ranged in size from several
1194
Fig. 4. 32 x 32 nm STM image of the surface
Fig. 5.40 x 40 nm STM image of the cross-section
M. A. DALE
of ACFlO.
of ACFIO.
angstroms to as large as 2.5 nm in width. Micropores were randomly distributed and homogeneous throughout the bulk of the fiber. In three-dimensions, the pores presumably were unoriented and formed a highly interconnected microporous network. Support for this interpretation was obtained from X-ray flat plate photos which only showed amorphous halos. It seems reasonable that the microporous structure must be interconnected from the large pore volumes of the ACFs [24]. Density measurements confirm that the porosity of the fibers must be interconnected [ 251. Figure 6 is a 61 nm x 61 nm image of the surface of ACF15. On this length scale, elongated micropores were observed parallel to the fiber axis. The micropores ranged in length from 1 to 20 nm and in width from 0.9 to 2 nm. This oriented microstructure represents less than 20% of the surface of ACFlS. Again, this oriented microporous structure may result from the shear associated with melt spinning of the original fiber precursor. Figure 7 is a 160 nm x 160 nm image of the surface of ACFlS. A highly mesoporous structure was
Fig. 6. 60 x 60 nm STM image of the surface of ACF15.
Fig. 7. 160 x 160 nm STM image of the surface
of ACF15.
observed at this length scale where the mesopores were ellipsoidally-shaped and were randomly distributed with respect to the fiber axis. This image is representative of more than 80% of the fiber surface. Figure 8 is a 60 nm x 60 nm image of the crosssection of ACF15. The micropores were randomly
Fig. 8.60 x 60 nm STM image of the cross-section
of ACF15.
Elucidating
the porous
structure
of activated
distributed in the bulk and were ellipsoidally-shaped. The micropores were homogeneous in the crosssection when the fiber was scanned across its entire diameter. It was not possible to resolve the interface between the fiber surface and cross-section. However, this region must be less than 60 nm since no difference in the bulk structure was observed when the fiber was scanned to the fiber edge. The pore size ranged from several angstroms to 4.2 nm in width. Figure 9 is a 100 nm x 100 nm image of the surface of ACF25. Large mesopores were observed on the surface of ACF25. These mesopores were ellipsoidally-shaped and ranged in size from 5 to 200 nm. Elongated micropores were observed on the surface of the ACF25. These micropores ranged in width from less than 0.5 to 2 nm and were as large as 30 nm in length. Figure 10 is a 32 nm x 32 nm image of the surface of ACF25. Both micropores and mesopores were observed. The degree of etching is much less in this image than that shown in Fig. 9 suggesting that the degree of gasification at the surface occurs in many different stages.
carbon
fibers using direct and indirect
of ACF25.
1195
Figure 11 is a 60 nm x 60 nm image of the crosssection of ACF25. Ellipsoidally-shaped micropores and small mesopores were readily observed. The pore size ranged from several angstroms to 46 A in width. The pores were randomly distributed and homogeneous across the fiber cross-section. Pore size distributions, Figs 12-14, were measured from the STM images of the cross-sections of ACFlO, ACF15 and ACF25 using section analysis and image analysis. The average pore size (width) for ACFlO is
Fig. 11. 60 x 60 nm
Fig. 9. 100 x 100 nm STM image of the surface
methods
0
STM
image ACF25.
of the
cross-section
20
5
of
25
PO:: Width’;& Fig. 12. Measured pore size distribution in the bulk for ACFlO as imaged at the fiber cross-section using STM.
16.5
23 1
Pore Width
Fig. 10. 32 x 32 nm STM image of the surface
of ACF25.
297
363
42.9
49.5
(A)
Fig. 13. Measured pore size distribution in the bulk for ACFlS as imaged at the fiber cross-section using STM.
M. A. DALEY et al. evolution of volatile species. The stability of these incipient micropores is directly related to the stability and persistence of the cross-linked structure in the precursor fiber. The transition from the surface porosity to the bulk porosity must result in an interface consisting of narrowing pores or wedge-shaped pores.
3.2 Adsorption results 3.3
9.9
16.5
23.1
29.7
Pore Width
36.3
42.9
49.5
(A)
Fig. 14. Measured pore size distribution in the bulk for ACF25 as imaged at the fiber cross-section using STM. 0.94 nm and ranged from
The nitrogen adsorption isotherms at 77 K for the ACFs are shown in Fig. 15 for a partial pressure range of 1O-5 to 0.12. All isotherms are Type I and exhibit microporous character. From the hysteresis loop of the adsorption isotherm from partial pressures of 10m5 to 1, we can conclude that the ACF25 had some mesoporosity. By applying standard theories to the adsorption isotherm, the pore size may be determined. The Dubinin-Radushkevich-Stoeckli (D-R-S) eqn (1) is one method of calculating the pore size distribution [2]. It is based on the equation
x exp ((- 1*(M2)*((~~-~2)2)/(2*b2)))
(1)
where Wz is the total micropore volume, x, is the pore half-width assuming a slit-shaped pore and corresponds to the maximum in the distribution curve, and 6 is the variance in B when a Gaussian distribution is assumed. Using the DubininRadushkevich (D-R) equation, the micropore volume, Wz, was calculated as the intercept on a log W vs log2(P,/P) plot, as depicted in Fig. 16. In this case, the micropore volume has been calculated for ACFlO as 0.392 mL/g of carbon fabric, for ACF15 as 0.481 mL/g of carbon fabric, and for ACF25 as 0.762 mL/g of carbon fabric. The value for x, and 6 may be determined from the experimental adsorption isotherm using the following relationship, eqn (2): Bj = - (ln( Wj)- ln( Wj_ I))/(( T/B)‘*( log2 x
~p~lpj~~10~2~p~lpj-~~~~
(2)
where Wis the amount of nitrogen adsorbed in mL/g of fabric, P/P, is the partial pressure, Tis the adsorption temperature which in this case is 77 K, and B is the adsorbent-adsorbate interaction parameter which has been discussed elsewhere and is equal to 0.43 [26]. Bj may be related to x using the following eqns (3) and (4): B.=M*x? J I
(3)
where M=(0.01915/k)2
(4)
and the determination of k has been described elsewhere and in this case is equal to 13 [4,16]. From the distribution of the xjs, 6 may be calculated. From our analysis, 6 was calculated as 2.71*10-’ for ACFlO, 2.9425*10-’ for ACF15, and 7.8247*10-’ for ACF25. The values of x, as calculated using the D-R equation are 0.583 nm for ACFlO, 0.648 nm for
Elucidating
the porous
structure
of activated
Nitrogen
carbon
adsorption
fibers using direct and indirect
isotherms
methods
1197
at T=77K
,O°C
P
B 6
300
2 3
200
.
.~~oo***~~o~o~~*
s >
0’
adsorption
.
.
.
.
.
.
A ACF25 . ACFLS 0 ACFlO
100
Fig. 15. Nitrogen
.
O.b2
isotherms
~ y=2.4041 + -0.010273x y=2.4925 + -0.012684x ......-.- y=2.6928 + -0.023025x
O.b4
0.66 P/PO
for the ACFs at T=77
O.b8
K over a range
o.io
’ 0.1
of partial
pressures
from zero to 0.12.
R=0.99657 R=0.99729 R=0.99402
2.1 2.6 2.5 g2.4 + 2.3 0
5
IO
15
20
25
30
35
40
Pore width (A) 2.1 0
5
10
15
20
log* (PO/P)
Fig. 17. Pore sire distribution for the ACFs as calculated from the nitrogen adsorption data using the DubininRadushkevisch-Stoeckli equation.
Fig. 16. Nitrogen adsorption data for the ACFs in the form of a Dubinin-Radushkevisch Plot. 4. DISCUSSION
4.1 Nature of the microporous structure ACFlS, and 0.873 nm for ACF25, respectively. Using these parameters, the D-R-S equation can be solved yielding the pore size distribution as a change in micropore volume per change in micropore halfwidth as a function of the micropore half width. In Fig. 17, the pore size distributions are given for ACFlO, ACF 15 and ACF25. As calculated from the D-R-S equation, the ACFlO had an average pore width of 1.166 nm and ranged from 0.5 to 1.7 nm, the ACF15 had an average pore width of 1.296 nm and ranged from 0.7 to 1.8 nm, and the ACF25 had an average pore width of 1.746 nm and ranged from 0.6 to 2.7 nm. From this distribution, it is apparent that the pore size increased as the surface area increased and that the pore size distribution became broader with increasing surface area.
The structure of any activated carbon depends on the precursor and etching conditions. Generally, the pore surface of microporous carbons consists of highly convoluted carbon sheets which are often crooked and the structure is highly disorganized [22]. Even though the structure is highly disorganized, it has been proposed to contain “slit-shaped” micropores [27]. These micropores are basically the gaps between the crooked or crumpled aromatic carbon sheets. Using STM, both elongated and ellipsoidal micropores and mesopores can be observed at the ACF surface. At the fiber cross-section, ellipsoidally-shaped micropores or small mesopores have been identified. The pores at the fiber cross-section were randomly distributed and homogeneous throughout the bulk. This evidence suggests that the structure of the carbon consists of many elongated
1198
M. A. DALEY~~~.
tubes which wind and twist throughout the carbon fiber creating its microporous character. These tubes can be viewed as an interconnected porous network consisting of pores of varying size. These pores are not isolated entities but must be interconnected due to the large pore volumes of the ACFs as shown from adsorption experiments and density measurements. Unfortunately, we are not able to directly image the interconnectivity within the porous network. When the porous tube is sliced or viewed from the edge, an elongated pore is observed. From these data, we infer that the structure of the ACFs has both an elongated pore (tube length) with a limiting pore diameter denoted by the cross-section of the tube which when viewed at the cross-section resembles an ellipsoidally-shaped micropore. The structure at the fiber surface consisted of large mesopores and some micropores whereas the structure at the fiber cross-section was far more uniform and consisted of micropores and small mesopores. Since the pore structure of the activated carbon fibers is continuous, the larger mesopores at the surface must narrow and empty into the micropores and smaller mesopores in the bulk of the fiber resulting in a narrowing pore at the surface-bulk interface. It was not possible to directly image the interface between the fiber surface and the fiber cross-section. From scans across the fiber diameter, it has been shown that the surface to bulk interface is strictly less than 60 nm in width compared to a 12 pm fiber.
4.2 Micropore generation jiber precursors
in ACFs jiom phenolic
From the work on the NAPFs, it is apparent that high surface area activated carbon fibers may be produced under inert conditions presumably from the evolution of gaseous by-products. This does not appear to be a general phenomenon for fibers that char, but rather is limited to those systems that develop a stable cross-linked structure during charring. Thus the micropores begin to develop at temperatures as low as 500°C (30 minutes) producing surface areas of 35 m’/g and up to 650 m’/g at SOO’C (30 minutes). From the images of the NAPFs, one observes the presence of a microporous structure where the measured pore width in the bulk is less than 1.1 nm and where most of the micropores are less than 0.6 nm in width. As might be expected, the surface of the NAPFs is etched to a lesser degree than those of the corresponding ACFs as observed from the STM images. The weight loss of the ACFs is much larger than that of the NAPFs and the fiber diameter of the ACFs generally decreases from 15 to 9-12 pm whereas only modest change is noted in the fiber diameter of the NAPFs. The difference in surface features between NAPFs and ACFs provides an insight into the two mechanisms of pore formation operative at the fiber surface. Thus, the ACFs are aggressively attacked at the fiber surface due to oxidative etching whereas the
only etching which occurs at the surface of the NAPFs is due to the evolution of gaseous by-products. Despite the differences in etching of the fibers at the surface, the porosity in the bulk of the fibers as observed from all of the fiber cross-sections remains remarkably homogeneous. The NAPFs obviously have very small micropores which continue to widen when exposed to an oxidative environment. Presumably, the homogeneous distribution of micropores which are initially generated at - 500°C result from the evolution of gases evolving from a very stable, cross-linked phenolic precursor.
4.3 Factors afectingpore
structure
Both the carbon precursor and treatment method affect the shape, size and distribution of porosity in a carbon. Laine and Yumes studied the effects of various treatments on the pore size distribution of activated carbon from coconut shell. When the carbon was activated with carbon dioxide at 800°C both wide mesopores and narrow micropores were observed. Catalyst (K,PO,) addition increased the micropore diameter and decreased the number and diameter of the macropores [28]. Kasaoka et al. have carried out an elegant study using various dyes in the liquid phase as molecular probes of the micropores and mesopores in activated carbon fibers (Kynol). They showed that the higher surface area activated carbon fibers were more effective at adsorbing larger dye molecules due to a larger critical pore diameter than activated carbon fibers with lower surface areas. Foster also showed that the average pore size increases with decreasing yield of the activated carbon fiber using the Dubinin-Radushkevich equation applied to the nitrogen adsorption isotherm at 77 K [26]. The images from the STM studies confirm these results and verify that the average pore size increases with increasing surface area. Measured and calculated pore size distributions reveal that the pore size distribution becomes broader as the activation yield decreases. Based on the increase in pore size with increasing extent of reaction, one can conclude that the activation of activated carbon fibers in the bulk occurs in three stages: (1) generation of small micropores, (2) widening of small micropores creating larger micropores, and (3) generation of small mesopores by continued widening of larger micropores until burnout of adjacent walls between micropores. 4.4 Comparison
of direct and indirect techniques to measure pore size
The average pore width as measured by STM was 0.94nm (ACFlO), 1.61 nm (ACF15) and 1.94 nm (ACF25) while the average pore size as calculated using the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K was 1.17 nm (ACFlO), 1.3 nm (ACF15) and 1.75 nm (ACF25). The average pore size as calculated from the D-R-S equation may not be directly compared with that directly
Elucidating the porous structure of activated carbon fibers using direct and indirect methods measured using STM since the first is a volumetric average and the second is a number average. However, the general trends may be thoroughly examined and compared. The D-R-S equation assumes a slit-shaped micropore and a Gaussian distribution of pores about the average pore-half width. STM techniques reveal both elongated and ellipsoidally-shaped pores and a non-Gaussian distribution about the average pore width. Despite the differences between the assumptions in the D-R-S equation and the features observed using STM, the measured and calculated pore width are surprisingly similar. The calculated and measured average pore size increase with increasing activation. Furthermore, the pore size distributions become broader with increasing activation. 4.5 Implications of pore size on adsorption Marsh [S] proposed a relationship between pore size and adsorption. Using the DubininRadushkevich equation, he demonstrated that as % burnoff increased the pore half-width increased. These results were correlated to the experimental adsorption isotherm and it was proposed that activated carbons with smaller pore size were more effective at removing contaminants at low concentrations due to a higher overlap in potential than activated carbons with larger pore sizes. Foster [26] extended this analysis to a series of activated carbon fibers. He observed a cross-over regime for the adsorption of butane by the ACFs. At low concentrations the ACFs with smaller pore size (lower surface area) were more effective at adsorbing butane whereas at higher concentrations of butane the ACFs with larger pore size (higher surface area) were more effective. The higher surface area ACFs are more efficient than the lower surface area ACFs because they have a larger pore volume and the pores of the lower surface area ACF are filled. The assumptions made to explain the crossover regime namely that lower surface area ACFs have a small pore size and that higher surface area ACFs have a larger pore size have been confirmed using STM.
ACFs can best be described as consisting of many elongated tubes which wind and twist throughout the carbon fiber creating its microporous character. These pores are not isolated entities but must be interconnected due to the large pore volumes of the ACFs. Based on the studies of the NAPFs, the generation of micropores begins at temperatures as low as 500°C due to the evolution of volatiles. To our surprise, surface areas of up to 650 m’/g were obtained by the reaction of the phenolic fiber under inert conditions of 800°C for 30 minutes. It seems reasonable to conclude that this process contributes significantly to pore generation in the ACFs. The average micropore half-width is observed to increase with increasing surface area using both STM techniques and the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K. Also, the pore size distribution becomes broader with increasing surface area suggesting that pore widening and pore generation occur cooperatively or that pores are generated over a broad activation range and then pore widening occurs for some pores at the expense of smaller pores. The average micropore width as measured using STM techniques was 0.94 nm (ACFlO), 1.6071 nm (ACFlS) and 1.94 nm (ACF25) while the average pore size obtained indirectly using the D-R-S equation applied to the nitrogen adsorption isotherm at 77 K was 1.17 nm (ACFlO), 1.30 nm (ACFlS) and 1.746 nm (ACF25). The calculated and measured average pore size increased with increasing activation. The pore size distribution became broader with increasing activation. The crossover regime observed by Marsh and Foster was confirmed to be the result of increasing pore size with increasing surface area as proposed by the original researchers.
REFERENCES 1. A. L. Myers and G. Belfort, Proceedings of the Engineer-
2.
5. CONCLUSIONS STM revealed ellipsoidal micropores and ellipsoidal and elongated mesopores at the fiber surface and ellipsoidal micropores and small mesopores at the fiber cross-section. The interface between the fiber surface and the bulk could not be directly observed, but presumably results in a narrowing of mesopores into smaller mesopores or micropores resulting in wedge shaped pores. The surface structure is less than 60 nm in thickness and thus represents a small portion of the total fiber. No evidence was found in the ACFs to support the view of slit-shaped pores. The micropores and small mesopores at the fiber cross-section are randomly distributed and homogeneous across the fiber diameter. The structure of the
1199
3. 4. 5. 6. 7,
ing Foundation Conference held at Schloss Elmau, Bavaria, West Germany, May 6-11, (1983). J. Economv. M. A. Dalev, E. Hiuno and D. Tandon, carbon 33,-j, 344 (1995): I_ K. L. Foster, R. G. Fuerman, J. Economy, S. M. Larson and M. J. Rood, Chemistry of Materials 4, 5, 1068 (1992). M. M. Dubinin and H. F. Stoeckli, J. Colloid and Interface Science 75, 1, 34-42 (1980). H. Marsh and B. Rand, J. Colloid and Interface Science 33, 101-l 15 (1970). E. Bekyarova and D. Mehandjiev, J. Colloid and Interface Science 160, 115 (1993). E. M. Freeman, T. Siemienjewskam, H. Marsh and B. Rand, Carbon 8, 7-17 (1970). J. P. Olivier, W. B. Conklin and M. V. Szombathrly, Determination of Pore Size Distribution from Density Functional Theory: A comparison of Nitrogen and Argon Results. Micromeritics Instrument Corporation, Inc.,
Norcross, GA 30093 U.S.A. J. P. Olivier, W. B. Conklin
and
M. V. Szombathrly,
Determination of Pore Size Distribution from’ Density Functional Theory: A comparison of Nitrogen and Argon
1200
10.
11. 12. 13.
14. 15. 16. 17. 18.
M. A. DALEY et al.
Results. Micromeritics Instrument Corporation, Inc., Norcross, GA 30093 U.S.A. S. Kasaoka, Y. Sakata, E. Tanaka and R. Naitoh, international Chemical Engineering 29, 1, 101-114 (1989). S. Kasaoka, Y. Sakata, E. Tanaka and R. Naitoh, International Chemical Engineering 29, 134-142 (1989). J. J. Kipling and R. B. Wilson, J. Appl. Chem. 10, 109%113 (1960). A. Guinier, G. Fourret, C. Walker and K. Yudowitch. Small Angle Scattering of X-rays. Wiley, New York (1955). M. D. Foster and K. F. Jensen Carbon 29, 2, 271-282 (1991). P. W. Schmidt, J. Appl. Cryst. 24, 414-435 (1991). M. M. Dubinin, G. M. Plavnik and E. D. Zaverina, Carbon 2, 261-268 (1964). N. Setoyama, M. Ruike, T. Kasu, T. Suzuki and K. Kaneko, Langmuir 9, 2612-2617 (1993). J. Kieffer, J. Appl. Phys. 72, 12 (1992).
19. J. D. F. Ramsay, In Neutron Scattering from Porous Solids from Characterization of Porous Solids pp 23-34. Elsevier, Amsterdam (1988). 20. W. P. Hoffman, M. B. Fernandez and M. B. Rao, Carbon 32, 1383 (1994). 21. M. G. Dobb, H. Guo and D. J. Johnson, In Carbon 1994 Preprints of Extended Abstracts. Granada Spain (July, 1994). 22. H. Marsh, D. Crawford, T. M. O’Grady and A. Wennerberg, Carbon 20, $419-426 (1982). 23. James Economy and Rodger A. Clark, Patent Application Number 710,292, filed March 4, 1968. 24. Joseph S. Hayes, Jr., In American Kynol. Reprinted from Kirk-Othmer: Encyclopedia of Chemical Technology, Vol. 16, Third Edition, pp. 125-138. 25. R. Y. Lin and J. Economy, Applied Polymer Symposium. 21, 143-152 (1973). 26. Ken Foster, Ph.D. thesis, University of Illinois, U.S.A. (1993). 27. H. F. Stoekli, Carbon 28, 1 (1990). 28. J. Laine and S. Yumes, Carbon 30, 601 (1992)