Carbide-Derived Carbons - American Chemical Society

Department of Materials Science and Engineering, UniVersity of PennsylVania, ... A. J. Drexel Nanotechnology Institute, Drexel UniVersity, Philadelphi...
0 downloads 0 Views 103KB Size
Langmuir 2006, 22, 8945-8950

8945

Carbide-Derived Carbons: A Comparative Study of Porosity Based on Small-Angle Scattering and Adsorption Isotherms Giovanna Laudisio,† Ranjan K. Dash,‡ Jonathan P. Singer,† Gleb Yushin,‡ Yury Gogotsi,‡ and John E. Fischer*,† Department of Materials Science and Engineering, UniVersity of PennsylVania, Philadelphia, PennsylVania 19104, and Department of Materials Science and Engineering and A. J. Drexel Nanotechnology Institute, Drexel UniVersity, Philadelphia, PennsylVania 19104 ReceiVed March 30, 2006. In Final Form: May 5, 2006 Porous carbons have received much attention recently for potential applications in energy generation and storage, molecular sieving, and environmental remediation. Property optimization for specific applications rests largely on controlling the volume, size, and shape of the pores at the synthetic level. Direct atom-scale experiments which might accurately and reliably measure these quantities are problematic, so indirect methods such as gas sorption are generally employed. Here we apply a second indirect method, small-angle X-ray scattering (SAXS), to study porosity in carbidederived carbons (CDC). The results qualitatively confirm and reinforce model-dependent conclusions drawn from gas sorption isotherms. In particular, both techniques indicate the onset of broad polydispersity under the same processing conditions for particular porous carbon materials.

Carbide-derived carbons (CDCs) are a new class of microporous materials produced by thermochemical etching of carbides1 The selective removal of the metal or metalloid from the crystalline structure of the carbide by etching agents such as chlorine leaves behind an amorphous carbon with specific surface area (SSA) up to 2000 m2/g and up to 80% open pore volume. This topdown approach offers the possibility to tune the porosity by choice of starting carbide, chlorination temperature, and other processing variables.2-5 Total SSA, average pore size, and pore size distribution (PSD) can be tailor-made for specific applications, a capability which makes CDCs unique compared to activated carbons, carbon nanotubes, and other porous carbons. An important application area for which tunable porosity offers a major advantage is the optimization of hydrogen cryosorbers for on-board storage in fuel cell vehicles. Physisorption of hydrogen on carbon nanostructures relies on weak van der Waals interactions, so the pore structure must be engineered to overcome intrinsically small heats of adsorption. In the past, optimization of storage capacity was believed to be a simple matter of maximizing the total pore surface area. However, a systematic study of hydrogen sorption in a variety of CDCs showed convincingly that storage capacity correlates only with the volume of pores smaller than 1 nm and not at all with larger pores6 nor with total SSA as previously reported.7 At atmospheric pressure, small pores are clearly more efficient for hydrogen sorption, whereas mesopores between 2 and 50 nm contribute little to * To whom correspondence should be addressed. E-mail: fischer@ seas.upenn.edu. Fax: 215-573-2128. † University of Pennsylvania. ‡ Drexel University. (1) Yushin, G.; Nikitin, A.; Gogotsi, Y. In Nanomaterials Handbook; Gogotsi, Y., Ed.; Taylor & Francis: Boca Raton, FL, 2006; pp 69-105. (2) Gogotsi, Y.; Nikitin, A.; Ye, H.; Zhou; W., Fischer, J. E.; Yi, B.; Foley, H. F.; Barsoum, M., W. Nat. Mater. 2003, 2, 591-594. (3) Dash, R. K.; Nikitin, A.; Gogotsi, Y. Microporous Mesoporous Mater. 2004, 72, 203-208. (4) Dash, R. K.; Yushin, G.; Gogotsi, Y. Microporous Mesoporous Mater. 2005, 86, 50-57. (5) Dash, R. K.; Chmiola, J.; Yushin, G.; Gogotsi, Y.; Laudisio, G.; Singer, J. P.; Fischer, J. E.; Kucheyev, S. Carbon, in press. (6) Gogotsi, Y.; Dash, R. K.; Ysuhin, G.; Laudisio, G.; Fischer, J. E. J. Am. Chem. Soc. 2005, 127 (46), 16006-7. (7) Zuttel, A. Mater. Today 2003, 6, 24-33.

gravimetric capacity. Hence a precise knowledge of the porosity is a requirement of paramount importance for the manufacture of effective hydrogen storage media. Standard adsorption methods for measuring SSA, PSDs, pore volume, and shapes give values that depend on the probe gas employed and the models used to analyze the data. Porosity analysis is generally performed using N2 or Ar at cryogenic temperature. Slow diffusion of these molecules, especially when small pores are present, requires great care and lengthy procedures to obtain reliable results. This is particularly inconvenient for rapid screening of samples from a new family of materials within which one wishes to study the effects of several process parameters on material performance. CDCs are a good example of such a family. A complete isotherm measurement of one sample at cryogenic temperature and relative pressures from 10-6 to 1 can take 35 h or more. Analysis software provided by instrument manufacturers usually assumes a specific pore shape to extract SSA and PSDs from isotherms, a potentially major limitation when studying CDCs since different pore shapes may result from different choices of precursor crystal structures. Given these limitations, and lacking direct atomistic probes of the pore structure, it is prudent to at least confirm the isotherm results by using a complementary and independent technique. In this context, small-angle X-ray scattering (SAXS) has the additional advantage that a probe molecule is not required. Data collection takes less than 1 h using in-house facilities; synchrotron sources are even faster and would also facilitate in situ experiments. The SAXS signal originates in density differences between matrix and pores; as such, it detects all the pores whereas N2 or Ar sorption detects only open pores larger than 0.4 nm. It is therefore quite possible that a material which shows limited porosity according to the sorption data may merit further processing efforts to open the pores if SAXS indicates a large pore volume. Another advantage is that only a few milligrams of powdered material is required for adequate signal-to-noise. The present work has two goals: to demonstrate that the nature of CDC porosity and its variation with processing, as determined from our previous gas sorption measurements, is borne out by comparative SAXS experiments; and that a simple postprocessing

10.1021/la060860e CCC: $33.50 © 2006 American Chemical Society Published on Web 09/16/2006

8946 Langmuir, Vol. 22, No. 21, 2006

Laudisio et al.

treatment which enhances the hydrogen capacity results from optimization of the PSD by increasing the volume of small pores, as determined by both methods. Experimental Section CDCs were produced form boron carbide (B4C, Alfa Aesar, USA, average particle size 6 µm), titanium carbide (TiC, Alfa Aesar, USA, average particle size 2 µm), zirconium carbide (ZrC, H. C.Stark, Germany, average particle size 8 µm), and silicon carbide (β-SiC, Superior Graphite, USA, average particle size 0.7 µm). The chlorine leaching equipment and methodology is described elsewhere.1,3-5 Precursor powders were loaded into a quartz boat and placed into the 1 in. diameter quartz tube of a resistively heated furnace. The furnace was then heated to the desired temperature (400-1200 °C) under argon (Air Gas, ultrahigh pure grade) purge. Once the desired temperature was reached, chlorine (Air Gas, ultrahigh pure grade) at 10-15 cm3/min was passed through the tube furnace for 3 h and then cooled to room temperature under pure argon purge. In addition to the as-produced chlorinated materials, we also performed a simple post-treatment consisting of annealing in hydrogen at 600 °C for 2 h using the same furnace and procedure. Gas sorption analysis was performed using a Quantachrome Autosorb-1 with argon and hydrogen at 77 K. The argon isotherm at 77 K was used to calculate SSA from the Brunauer-EmmetTeller (BET) equation for volume of gas adsorbed at relative pressures between 0.05 and 0.3 where the BET isotherm is linear.8-11 Volume-weighted pore size distributions (PSDs) and pore volumes were determined using nonlocal density functional theory (NLDFT) included in the Quantachrome data reduction software (version 1.27) which assumes slit pores.9 Weighed pore size of pores was defined as n

∑d V

i i

i)1 n

∑V

Figure 1. Small-angle X-ray scattering (SAXS) profiles recorded from carbide-derived carbons synthesized at different chlorination temperatures, increasing from bottom to top: (A) zirconium carbide precursor and (B) titanium carbide precursor. Both are rock-saltlike fcc crystal structures. The clear distinction between qualitative aspects of the profiles of low-T and high-T materials suggests a major structural rearrangement at sufficiently high temperature for both precursors.

i

i)1

where d is pore size and V is pore volume. SAXS experiments were performed on a multiangle diffractometer equipped with a Cu rotating anode, double-focusing optics, evacuated flight path, and a two-dimensional wire detector.12 Powder samples were dried at 80 °C overnight, sealed in 1 mm-diameter glass capillaries, and measured in transmission for 1 h. Data were collected over a scattering vector Q range of 0.02-1.60 Å-1 (scattering vector Q ) 4π sin θ/λ where θ is the scattering angle and λ is the X-ray wavelength). An intensity profile from an empty capillary was corrected for X-ray absorption and subtracted from the sample profile. This procedure left a small Q-independent residual background which obscures the sample scattering at high Q.

Results and Discussion Our point of departure for presenting, analyzing, and discussing SAXS profiles from CDCs is the ideal behavior of monodisperse pores, for which the scattered intensity I(Q) has a simple analytical form. In the low Q limit, I(Q) is a constant proportional to the total pore volume and the contrast between pores and matrix. For sufficiently large contrast features (large pores, grains, surface asperities...), this important limit may lie below the instrumental Q cutoff. At intermediate Q we have the classic Guinier behavior: (8) Brunauer, S.; Emmett, P.; Teller, E. Adsorption of Gases in Multimolecular Layers. J. Am. Chem. Soc. 1938, 60, 309-19. (9) Ravikovitch, P. I.; Neimark, A. V. Colloids Surf. 2001, 187-88, 11-21. (10) Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area and Porosity; Academic Press: London, 1982; pp 42-54. (11) Lowell, S.; Schields, J. E. Powder Surface Area and Porosity; Chapman and Hall: New York, 1998; pp 17-29. (12) http://www.lrsm.upenn.edu/lrsm/facMAXS.html.

13 I ∼ exp(-Q2R 2/3) where R , the radius of curvature, depends g g on both pore size and shape. A plot of log(I) versus Q2 gives a

straight line with slope -Rg2/3. At still higher Q the pore surfaces give a power law contribution I ∼ Q-n where n ) 4 for smooth surfaces and n < 4 if asperities of order 2π/Q are present. A log-log plot of this simple example would show a progression with increasing Q from a constant value, through a gradually decreasing curve referred to henceforth as a “knee” (the Guinier regime), to a power law with exponent -4. SAXS profiles of CDC show portions of two such progressions; one is welldescribed by monodisperse micropores while the other, at lower Q, has two different interpretations depending systematically on the particular CDC under consideration. Finally, the absolute value of I(Q) is also proportional (at all Q) to the absolute incident intensity and sample density, neither of which are easy to measure. Figure 1A shows on a log-log scale the evolution of SAXS profiles with increasing chlorination temperature for ZrC-CDC. Starting at low Q, all the profiles can be qualitatively described as a progression of two linear (power law) regimes with different exponents, separated by a more-or-less constant plateau, and finally terminating in the aforementioned constant background at the highest Q. The profiles for different chlorination temperatures fall into two groups for this precursor. In the range 300-600 °C, the low-Q exponent is large, the plateau is welldeveloped, and the “knee” between the plateau and the higher-Q linear regime is centered in the range 0.25 < Q < 0.5 Å-1. As described above, the log-log “knee” represents Guinier behavior (13) Guinier, A.; Fournet, G., Small-Angle Scattering of X-rays; John Wiley &Sons: New York, 1955.

Carbide-DeriVed Carbons

Langmuir, Vol. 22, No. 21, 2006 8947

representative of monodisperse contrast elements (i.e., pores) with a rather well-defined radius of gyration Rg or at least a narrow distribution of Rg’s. In contrast, the 3 high-T profiles have smaller low-Q exponents, narrower plateau regions and “knees” at significantly lower Q. As we show below, this qualitative behavior is attributed to the abrupt onset of larger pores and concomitant broadening of the isotherm-derived PSDs; the dashed curves in Figure 1 are model fits describing such behavior (see below). Figure 1B shows a similar but more pronounced progression for TiC-CDC. Here the low Q power law extends over 2-3 decades in intensity for low chlorination temperatures while the plateau no longer exists at 1200 °C. Again the model fits describe this behavior very well (dashed curves) and are consistent with NLDFT-derived pore size distributions, to which we turn next. Figure 2A,B shows the evolution of PSDs with chlorination temperature for ZrC-CDC and TiC-CDC, respectively, corresponding to the SAXS profile evolutions in Figure 1. PSDs are obtained from nonlocal density functional theory applied to argon isotherms measured at T ) 77 K for a range of relative pressures 10-6 < p/p0 < 1 where p0 ) 200 mmHg is the saturated pressure of Ar at 77 K. The graphs give the volume of slit pores having a certain width as a function of pore width. The qualitative aspects of the PSDs are in accord with the qualitative deductions from SAXS profiles. For example, the PSD of ZrC-CDC begins to broaden noticeably toward larger pores at 600 °C, the highest temperature at which reasonably monodisperse Guinier behavior is observed, Figure 1A. At 1000-1200 °C the PSDs show substantial pore volume ranging over 0.5-4 nm pore widths, which corresponds to the broadening and shift to lower Q of the “high-Q” Guinier regime and the reduction in low-Q exponent representing the dominance of polydisperse Porod contributions over the scattering from fractal power grain surfaces observed at lower chlorination temperatures. Similar observations apply to TiC-CDC, for which the polydispersity at 1200 °C is even more pronounced. Next we analyze the SAXS profiles using a model suggested by the qualitative observations. Group 1 materials, chlorinated at low temperatures, include ZrC-CDC from 300 to 600 °C and TiC-CDC for 400-800 °C. Here the two power-law regimes at low- and high-Q ranges are assigned to scattering from 2 different kinds of electron density contrast with Q-dependent intensity given by13

I(Q) ≈ 2πIeF2SQ-4

(1)

where Ie is the intensity scattered by a single electron, F is the electron density difference between carbon and vacuum, and S is the total interfacial area. In our analysis, “vacuum” means either empty pores or the macroscopic volume between powder grains depending on the Q range under discussion. Similarly, S refers either to the pores or the powder grains. For the low-temperature group 1 materials, the PSDs in Figure 2 are dominated by pore widths 1.0 nm, for higher chlorination temperatures. See text.

the second term there is no explicit separation of Guinier and Porod behaviors, while Rg is the usual radius of gyration of the micropores. The first term dominates at low Q, giving rise to the first power-law regime. I(Q) remains relatively constant up to the limit QRg e x6 which defines Rg and demarcates a crossover to the second power-law regime at higher Q. As noted above, an equivalent constant intensity plateau from the powder particles lies below the low Q cutoff of the experiment. The group 1 SAXS profiles can be fitted quite well with eq 3 as indicated by the dashed curves in Figure 1. The low-Q exponents n are collected in Table 1. They are all close to 4, the systematic deviations indicating fractal dimensions Ds ) 6 - n which increase with increasing chlorination temperature. The model also accounts for the different constant plateau Q ranges for different chlorination temperatures, in particular the trend of shifting the plateau range to higher Q for smaller pore dimensions. Finally, the SAXS model Rg values agree well with the average pore size from PSDs, as shown in Figure 3A,B for ZrC-CDC and TiC-CDC, respectively. The maximum difference is less than 0.3 nm. For both CDCs the smallest average pore size is obtained with chlorination temperature 600 °C. SAXS profiles from high-temperature group 2 materials (800-1200 °C for both ZrC-CDC and TiC-CDC), while qualitatively different from those of those of group 1, can also be fitted with eq 3, again shown by the dashed curves in Figure 1. Now the low Q exponents are as small as 2.1 (Table 1) while the Rg values increase in correspondence with the PSD broadening which signals the onset of mesopore formation (Figure 2) and large average pore sizes (Figure 3). For group 2 the “plateau” becomes a very broad shoulder below the power law regime, which shifts to lower Q with increasing chlorination temperature. The smearing of the plateau into the low-Q power law regime is consistent with the development of mesopores as seen in the PSDs. Mesopores, while certainly larger than micropores since they are observed at lower Q, are however still small enough to satisfy QRg e x6 in a Q range accessible to our instrument. Mesopore and micropore scattering are additive; the effect of mesopores on the model described by eq 3 is to create the shoulder and reduce the exponent n, see Table 1. The model accurately reproduces the observation that, as the mesopores increase in average size, the shoulder shifts to lower Q. As a further check on isotherm-derived quantities, we calculate specific surface area from SAXS for comparison with “traditional” values based on gas sorption. We use the method for two-phase systems presented by Guinier13 and subsequently applied to porous

Figure 3. Comparison of SAXS-derived radii of gyration with the NLDFT-based average pore sizes, from refs 4 and 5 for ZrC-CDC and TiC-CDC, respectively: (A) ZrC-CDC and (B) TiC-CDC as functions of chlorination temperature. Both experiments exhibit the same general trend, an increase in characteristic pore dimension at higher temperatures.

carbon materials by Laszlo et al.17 The specific surface area per unit volume, SSA, is given by

SSA ) πφ(1 - φ)

lim[I(Q)Q4]

∫0∞I(Q)Q2 dQ

(4)

where φ is the fraction of the powder sample occupied by carbon, which is generally obtained from total pore volume measured by gas sorption and carbon density, e.g., from He pycnometry. The numerator in eq 4 is evaluated from SAXS profiles by first rearranging eq 2

I(Q)Q4 ) K + cQ4

(5)

and evaluating K from the intercept found by plotting the data as I(Q)Q4 versus Q4, then associating K with the numerator of eq 4. In the denominator integral, I(Q) is corrected for the constant background associated with atomic disorder c and thus becomes

∫0∞ (I(Q) - c)Q2 dQ The integral is infinite, but I(Q) is known only in a certain interval.17 Cutting off the integral below the instrumental low Q limit is acceptable thanks to the Q2 weighting factor. For Q > Qmax, the upper limit of the data range, the remaining contribution to the integral is approximated by

∫Q∞

max

(I(Q) - c)Q2 dQ )

∫Q∞

max

KQ-4Q2 dQ )

K (6) Qmax

The combined error introduced by these approximations is estimated to be less than 10%.17 (17) Laszlo, K.; Czakkel, O.; Josepovits, K.; Rochas, C.; Geissler, E. Langmuir 2005, 21, 8443-8451.

Carbide-DeriVed Carbons

Figure 4. Effect of the starting carbide on (A) SAXS profiles and (B) PSDs. The three equimolar binary compounds TiC, SiC, and ZrC (starting from the bottom) show similar behavior in both experiments: two well-developed power law regimes separated by a plateau in SAXS, and narrow PSDs with small average size (cf. Figure 2A,B). In sharp contrast, both experiments show different behavior for the rhombohedral 4:1 precursor B4C (top).

We applied this method to TiC-CDC chlorinated at 800°C, repeating it twice to partly average out subjective aspects of the procedure. We found 1970 m2/g based on a pycnometric density of 2.302 g/cm3 and average particle size 2 µm from the manufacturer. The corresponding BET result is 21% smaller, 1566 m2/g. The discrepancy could mean that the probe gases do not fill the smallest pores, or that SAXS detects additional closed porosity. A systematic comparison based on a wide range of CDCs, and/or the use of another complementary technique, would help understand the discrepancy. Even at this level of agreement, SAXS can clearly be used as a complementary technique to evaluate specific surface areas of microporous materials. The effect of precursor crystal structure on porosity was also studied with SAXS and gas sorption as a further consistency check. Carbon atoms in the rocksalt ZrC, TiC, and sphalerite SiC precursors have a unique first-neighbor separation of order 0.4 nm. During CDC synthesis, Zr, Ti, or Si atoms leave as volatile chlorides, and one might therefore expect a uniform collapse of the carbon network as the C-C distance shrinks to the 0.14 nm value characteristic of sp2 hybridization, as confirmed by nearedge X-ray absorption fine structure (NEXAFS)5 and many other methods. A uniform collapse might be expected to yield a network of small uniform pores. In contrast, rhombohedral B4C has multiple first-neighbor C-C separations and potentially larger pore volume, so the pore structure should be significantly different with respect to mono/polydispersity and pore shape. Both SAXS and NLDFT PSDs bear out these considerations. Figure 4A,B shows, respectively, the SAXS profiles and PSDs of CDCs from the two different groups. For CDC synthesis in this comparative study, we purposely selected 600 °C for B4C, ZrC, and TiC but 1100 °C for SiC. The precursors with uniform C-C distances

Langmuir, Vol. 22, No. 21, 2006 8949

Figure 5. Effect of H2 annealing (3 h at 600 °C) on (A) SAXS profiles and (B) PSDs for TiC-CDC prepared at 400 and 800 °C. As measured by SAXS, the effect of chlorination temperature is much more dramatic than the H2 anneal, implying that the latter does not induce major structural rearrangement. On the other hand, the PSDs show that H2 annealing increases the pore volume (800 °C) and/or shifts the distribution to smaller pores (400 °C).

all give narrow PSDs (see also Figure 2A,B) and “group 1” SAXS profiles, despite the fact that SiC-CDC is a hightemperature material. Conversely, B4C-CDC gives a very broad PSD and “group 2” SAXS behavior despite the fact that it is a low-temperature material. More work is clearly required to gain an atomistic understanding of these effects; it is, however, gratifying to have another point of agreement between sorption and SAXS data. Post-treatment of CDC, such as activation or surface modification, offers additional potential for controlled tuning of porosity. The effect of H2 annealing on TiC synthesized at 400 and 800 °C was investigated with SAXS and PSDs; the results are shown in Figure 5A,B, respectively. The annealing process has little effect on SAXS profiles, in contrast to the obvious effects of variable synthesis temperatures discussed above. According to SAXS, annealing in hydrogen does not significantly change the carbon structure for a given chlorination temperature, Figure 5A. On the other hand, the PSDs suggest the opening of previously inaccessible microporosity for the 400 °C materials, and widening of pore volume for all pore sizes at 800 °C. This is also reflected in the SAXS profiles by a slight reduction of Rg after annealing the 400 °C material. We conclude that H2 annealing removes byproducts of the chlorination reaction without modifying the carbon structure.

Conclusions We have demonstrated that SAXS is a fast, probe-independent and reliable technique to monitor the structure of CDC materials

8950 Langmuir, Vol. 22, No. 21, 2006

upon change of different processing parameters. Our results are in very good qualitative agreement with adsorption measurements which are commonly used to analyze porous materials, but are dependent on the theory used to interpret the isotherms. A narrow PSD can be obtained from cubic carbides chlorinated at low temperature, while high temperatures result in broadening of the PSDs. Post-treatments such as H2 annealing increases the volume of pores available for adsorption by removing chlorination

Laudisio et al.

byproducts (chlorides, chlorine) without modifying the carbon structure. Acknowledgment. The authors are grateful to E. Geissler for patiently guiding us through the SAXS surface area analysis. This work was supported by the U.S. Department of Energy EERE Program under Grant DE-FC36-04GO14282. LA060860E