J. Phys. Chem. C 2009, 113, 13933–13939
13933
Reverse Spillover of Hydrogen on Carbon-Based Nanomaterials: Evidence of Recombination Using Isotopic Exchange Anthony J. Lachawiec, Jr. and Ralph T. Yang* Department of Chemical Engineering, UniVersity of Michigan, Ann Arbor, Michigan 48109 ReceiVed: April 17, 2009; ReVised Manuscript ReceiVed: June 25, 2009
The hydrogen spillover phenomenon at ambient temperature is of interest in the development of adsorbents for storage applications. The mechanism of hydrogen spillover is studied using equilibrium dosing of the deuterium hydride molecule. Temperature programmed desorption spectra for both primary and secondary spillover materials, with the deuterium peak appearing first, have supported partitioning of the desorption behavior into reverse spillover and recombination directly from the atomic receptor. Desorption kinetics are faster relative to adsorption for hydrogen and deuterium on all carbon-based adsorbents in the study. Diffusion time constants are computed using a surface diffusion model based on Fick’s second law. Kinetic data are studied for pressures from atmospheric to 100 atm and equilibrium adsorbed amounts to 5.5 mmol/g. Values follow an inverse relationship with amount and are 2 orders of magnitude higher (10-4 s-1) than those reported for nitrogen physisorption on carbon adsorbents. Adsorption rates measured using volumetric techniques do not account for recombination, while desorption rates include both reverse spillover and recombination. Subtraction of the two rates yields the rate of recombination for hydrogen or deuterium atoms. The highest rate of recombination is observed for secondary spillover (1.3 × 1012 atoms H/cm2 · s), and the maximum is nearly 1 order of magnitude higher than recombination for primary spillover. Recombination rates decrease with final adsorbed amount, supporting literature simulation of stable hydrogen clusters adsorbed to graphene sheets. The maximum deuterium atom recombination rate is greater than hydrogen, indicating that it is less stable on the receptor. Introduction Spillover of hydrogen atoms from a transition metal particle source to a high capacity receptor has recently been studied as a mechanism for enhanced storage of hydrogen in fuel cell applications. The phenomenon has been shown to increase both the capacity and improve the kinetic response of carbon-based adsorbent materials.1-10 The diffusion of hydrogen atoms on the surface of these materials is an important parameter to understand and optimize. Rapid charge and discharge kinetics of adsorbents and capacity are equally important when considering materials for practical application. The requirements for feasible storage compounds to be used for on-board transportation application have been delineated by the Department of Energy.11 Hydrogen spillover kinetics are influenced by the physical contact of a transition metal source with a receptor and the rate of diffusion of atomic hydrogen on the receptor surface. In two studies, Neikam and Vannice observed spillover of atomic hydrogen from platinum to Y-zeolite exchanged with ceria in the presence of perylene.12,13 The perylene molecules were shown to act as a bridge that reduced the activation energy to transport hydrogen from platinum to zeolites. Boudart et al. observed a similar behavior at elevated temperatures for spillover from platinum to carbon after pretreatments were found to form physical carbon bridges between the two components, otherwise considered carbon contamination of platinum.14 Bridges and metal-support contact are important in that they promote an increased spillover rate compared to those without such * To whom correspondence should be addressed. E-mail: yang@ umich.edu.
enhancements even when such catalysts are studied at elevated temperatures.15 Adsorption and desorption requires forward and reverse spillover. Hydrogen atoms must be mobile to and from the transition metal source at ambient conditions. Although atomic hydrogen is chemisorbed to platinum, Pliskin and Eischens observed weak forms of atomic hydrogen bound to platinum using infrared spectroscopy.16 Volumetric studies of platinum catalysts17 and, more recently, simulation results have supported this observation, demonstrating that atomic hydrogen is mobile on the platinum surface and migration from the source to the receptor is possible at ambient conditions with a small energy barrier.18 Once they have transferred to the carbon surface, physisorbed hydrogen atoms can recombine or reach stable conditions at sites with defects, high degrees of curvature, or surface modificationsssuch as functional groups on active carbon.19 Catalytic research has led to the theory of reverse spillover for atomic hydrogen desorption. Reverse spillover is the migration of adsorbed atoms from a receptor to a source where they recombine for desorption as molecules.20,21 Taylor and coworkers22 proposed reverse spillover as the mechanism to explain hydrogenation of ethylene on a alumina supported platinum catalyst. In this study, along with a follow-up by Altham and Webb,23 presorbed tritium traced the hydrogenation of ethylene and propylene, demonstrating that hydrogen migrated to the platinum particle to participate in the hydrogenation. Reverse spillover was determined to be rate controlling for the dehydrogenation of isopropanol over platinum supported on titanium oxide, due to long diffusion distances for low metal loading.24 Asaoka and co-workers found increased rates of
10.1021/jp903555y CCC: $40.75 2009 American Chemical Society Published on Web 07/14/2009
13934
J. Phys. Chem. C, Vol. 113, No. 31, 2009
hydrogen desorption relative to adsorption for a catalyst containing cobalt supported on active carbon, increasing as the metal content rose to 5 wt %.25 Sermon and Bond found that reverse spillover proceeded at a slower rate relative to the forward process for systems containing platinum doped oxides of molybdenum and tungsten, as well as silica, in the range of 273-373 K.26 Reverse spillover of oxygen was observed on an alumina-supported platinum catalyst by researchers studying ammonia oxidation at 433 K.27 The rate of reverse spillover on copper and cobalt supported by active charcoal was shown to be nearly three times faster than forward spillover for hydrogenation reactions at 673 K.25 Fujimoto and co-workers observed atomic hydrogen transfer, reaction, and desorption as molecular hydrogen via reverse spillover from active carbon to various supported transition metal particles (excluding platinum group metals) during dehydrogenation of paraffins in the temperature range 673-773 K.28,29 Le Van Mao et al. observed enhanced aromatization performance for zeolite-supported cocatalysts, including gallium and zinc oxides, which was attributable to reverse hydrogen spillover.30 Isotopic tracer experiments have been used to study the radial diffusion mechanism of forward and reverse spillover.31-33 Prior work performed by our laboratory has proven such a mechanism for carbon-based materials at ambient temperature.34 Recombination of atomic deuterium and atomic hydrogen to form deuterium hydride occurs readily on single-crystal platinum.35-37 Our work has consistently demonstrated faster hydrogen desorption rates relative to adsorption rates for both primary5 and secondary38 spillover. The implication is that the kinetics of this process are faster relative to forward spillover if desorption proceeds through the metal particle in a reverse spillover mechanism. An alternative explanation is that during desorption, hydrogen and deuterium atoms recombine and directly desorb without requiring the transition metal particle pathway. The current work employs isotope equilibration to trace desorption and elucidate its mechanism for carbon-based materials where spillover is active. Primary and secondary spillover are studied using transition metal directly doped on a carbon-based receptor and a commercial catalyst bridged to high capacity receptors. Temperature programmed desorption (TPD) results indicate that recombination of hydrogen atoms is a plausible explanation for faster desorption rate. Time constants are computed for atomic hydrogen and deuterium diffusion on carbon-based adsorbents from time-dependent adsorption data. Differences in the rates of adsorption and desorption, as measured by volumetric techniques, are used to calculate the rate of atomic hydrogen recombination for primary and secondary spillover adsorbents. Experimental Methods The materials used in this study were designed to study both primary and secondary spillover. The primary spillover adsorbent was a templated carbon receptor doped to 6 wt % loading with platinum (6 wt % Pt/TC) using ultrasound-assisted impregnation.5,34 Two materials were used to study secondary spillover. The source of spillover atoms for both materials was a 5 wt % Pt/C commercially available catalyst (Strem Chemicals, Inc., 78-1600) while IRMOF-839,40 and AX-2141,42 were used as receptors. The components were subjected to bridge building techniques using simple sugar precursors. Synthesis and characterization of the materials has been extensively reviewed in previous publications.43-45 Table 1 lists the BET surface area and pore volume of the composite materials. The apparatus TPD experiments consisted of a dosing manifold and an analysis manifold connected to a mass
Lachawiec and Yang
Figure 1. Configuration of TPD apparatus.
TABLE 1: Material Characterization for Primary and Secondary Spillover Adsorbents
6 wt % Pt/TC AX-21/PtC/Bridge (8:1:1)
BET SA (m2/g)
total pore volume (mL/g)
micropore volume (mL/g)
2730 ( 10 2162 ( 17
1.4 ( 0.2 1.1 ( 0.1
0.9 ( 0.1 0.6 ( 0.1
spectrometer (Figure 1). Analysis of evolved gases was performed with an AeroVac 1200 Magnetic Sector mass spectrometer (VTI, Inc.), operated at an accelerating voltage of 70 eV. An electron multiplier, operating at 1000 V, was used to increase the sensitivity to low current signals. A thin 316 SS tube was used to deliver evolved gas directly to the inlet of a molecular leak valve (Varian, Inc.). A sheath around the sampler could be dynamically pumped with a mechanical vacuum pump to remove residual gas if the pressure rose above the maximum allowable for the mass spectrometer inlet. This was not required, as the manifold remained below 300 mbar during all experiments. Heating was accomplished using an external heater constructed from nichrome wire. The sample holder for TPD was a single ended tubular chamber constructed from 316 stainless steel. All connections were made with VCR fittings to ensure leak integrity. The sample was placed at the end of a 30 cm length of tube to enable suspension in a temperature bath. A filter disk (0.5 µm) was placed at the outlet VCR fitting to prevent particle intrusion at the manifolds. A bellows valve was installed at one end to allow transfer of the sample under vacuum at cryogenic conditions from the dosing to the analysis manifold. Dosing was completed on a Micromeritics ASAP 2000 instrument, which enabled high accuracy pressure measurements. Freezing spillover to 77 K was required prior to TPD and provided ample time to transfer the sample between dosing and analysis manifolds. The analysis manifold was outfitted with isolation valves to maintain roughing vacuum in the line (1 × 10-3 mbar) and prevent moisture intrusion while idle. Since moisture contamination can cause errors in measurement of hydrogen fragments, attention was given to the purity of all gases used in this study. Ultrahigh-purity helium (99.995%) and hydrogen (99.995%) were passed through molecular sieve 5A beds to ensure the gases were dry and free of contaminants. Deuterium (UHP, 99.97%) and deuterium hydride (96 atom % D, Sigma-Aldrich, 488690) were used without further purification. Baseline TPD results were obtained for a blank chamber, undoped templated carbon, and platinum black. Results for TPD of the blank chamber and undoped templated carbon have been reported elsewhere.34 Platinum black (98% min., Strem Chemicals, Inc., 78-1420) was used to determine the exchange behavior of the metal alone. The platinum black amount was established to equal the surface area of platinum crystallites supported on
Reverse Spillover of Hydrogen on Nanomaterials templated carbon. The specific surface area was calculated from platinum particle geometry established from TEM and XRD studies,34 and the BET surface area of platinum black was measured as 37 m2/g. In order to observe the reverse spillover mechanism, deuterium hydride was used as the dosing species. The sample was degassed in situ for 8 h at 623 K (523 K, IRMOF-8 receptor) and equilibrated for at least 2 h at room temperature under vacuum (1.3 × 10-7 mbar). The sample was equilibrated at the desired temperature under a dose of 0.4 bar HD for 1 h to ensure an equilibrium distribution of hydrogen and deuterium atoms on the surface of the adsorbent. The time was selected on the basis of kinetics of spillover at pressures less than 100 kPa. Additional experiments performed with dose times of 2 and 8 h confirmed the equilibration period. The sample was cooled to 77 K in order to prevent desorption via reverse spillover. TPD experiments were performed at a heating rate of 15 K/min after transfer to the analysis manifold and evacuating the gas phase. For templated carbon and 6 wt % Pt/TC, the upper limit of TPD was 650 K to prevent Pt sintering46,47 and carbon gasification.48 TPD was limited to 523 K for the IRMOF-8 bridged PtC material to prevent its decomposition. The controlled temperature was measured with an external surface thermocouple, and the sample temperature was calculated using a calibrated offset. The sample was treated at 650 K (523 K, IRMOF-8 receptor) and 1.3 × 10-7 mbar for 3 h between each TPD experiment to remove atomic hydrogen and deuterium that might have remained from prior exposure. Repeat experiments confirmed that atoms remaining after this treatment were sufficiently anchored to prevent interaction with subsequent doses. Recombination rates were computed from rate of adsorption data collected using a custom high-pressure volumetric instrument. The instrument and data analysis techniques have been described in detail elsewhere.49 The fractional completion was measured for uniform steps (∼0.5 mmol/g) in adsorption amounts along the isotherm to facilitate comparison between adsorption and desorption, and hence, recombination rates. Fractional completion was converted to moles adsorbed (or desorbed) and differentiated with respect to time to yield molar adsorption and desorption rates. The results were used to compute the rate of hydrogen and deuterium recombination for these steps along the adsorption isotherm as a function of time to complete equilibration. Results and Discussion Deuterium hydride was equilibrated on the surface of 6 wt % Pt/TC for 1 h at 298 K. After the sample was cooled to 77 K and the gas phase was pumped off, TPD was performed and the result is shown in Figure 2. Physisorbed deuterium hydride below 200 K is subtracted from this spectrum. As the temperature increases, HD and D2 peaks begin to form. The D2 peak is substantially smaller relative to the HD peak and its maximum occurs at a lower temperature. For uniformly distributed hydrogen and deuterium atoms at equilibrium, a single peak of deuterium hydride would be expected. The result implies that there is some partitioning of species on the carbon surface after spillover. If the difference in bond energies for species containing deuterium are compared with those containing hydrogen, the former are always more energetic.50 While these bond energies are not exact for the physisorbed species that are predicted by simulation,18,51 a similar trend is likely to hold. Deuterium has a higher probability to cluster together since the interaction energy is slightly greater
J. Phys. Chem. C, Vol. 113, No. 31, 2009 13935
Figure 2. TPD result for HD equilibration on 6 wt % Pt/TC at 298 K.
Figure 3. TPD result for HD equilibration on IRMOF-8/PtC/Bridge (8:1:1) at 298 K.
relative to hydrogen. On desorption, deuterium atoms would then be more likely to recombine with each other as they diffuse on the surface by hopping. This would cause them to desorb from the surface prematurely, since the strength of the D2 bond is larger than that of H2. This behavior is a plausible explanation for the consistently smaller deuterium peaks in the TPD results. HD equilibration was also measured for IRMOF-8/PtC/Bridge material to observe the response for a secondary spillover adsorbent. The results are shown in Figure 3. A desorption spectrum similar to Figure 2 is observed, with a shift of peak temperatures to lower temperatures. The spillover phenomenon occurs as atoms hop along the surface due to a concentration gradient.38,52,53 This type of surface diffusion process is best represented by a twodimensional, radial coordinate system. Figure 4 depicts the situation for a secondary spillover source and receptor system. With the techniques described in the experimental section, bridges are formed between the support material of the source (PR ) primary receptor) and the secondary receptor. There is no simple analytical solution for the boundary condition of a step increase in concentration at the center of the geometry; however, one exists for a step increase at the outer boundary. Yang et al. describe the inaccuracies in the diffusion time constant caused by implementing this solution.38 The values calculated from experimental data are predicted
13936
J. Phys. Chem. C, Vol. 113, No. 31, 2009
Lachawiec and Yang
Figure 4. Schematic of spillover geometry.
lower than the actual values for adsorption as the surface is assumed to load in the opposite direction. Concentration and temperature behavior is correct and is useful for comparing relative rates as these conditions vary. Fick’s second law for a two-dimensional, radial geometry becomes
∂C ∂C 1∂ rD ) ∂t r ∂r s,e ∂r
(
)
Figure 5. Adsorption fraction complete for primary spillover at 298 K for final equilibrium amount: 0.5, O; 3.5, ∆; 4.5 mmol/g, 0. Step size was uniformly 0.5 mmol/g.
(1)
This equation may be solved by the method of separation of variables. Defining as a solution
C ) u(r) exp(-R2τ)
(2)
assuming a constant surface diffusion coefficient (Ds,e) with dimensionless time, τ ) (Ds,et)/Rs2, and satisfying Bessel’s equation of order zero.
d2u 1 du + + R2u ) 0 2 r dr dr
(3)
For constant concentration at the outer boundary of spillover, r ) Rs and with a constant initial concentration on the surface, C(r,0) ) Ci, the solution is ∞
(
)
Ds,e 1 F(t) ) 1 - 4 exp - 2 R2nt 2 Rs n)1 Rn
∑
(4)
with Rn values equal to the positive roots of the Bessel function of the first kind, order zero (Jo(Rn) ) 0). Note that the diffusion time constant appears in this relationship as a result of substituting the relationship for dimensionless time. The advantage of this solution is that it is valid over the entire adsorption event, compared to the short time solution employed previously.38 This enables asymptotic behavior of the data at equilibrium to be adequately captured when calculating the diffusion time constant. Figure 5 is a plot of data, reduced according to methods described in the Experimental Methods section, and the solution to the two-dimensional Fick diffusion relationship for several equilibrium amounts on the 6 wt % Pt/TC equilibrium adsorption isotherm at 298 K. As data indicate, the rate of adsorption for primary spillover decreases with adsorbed amount. Desorption behavior is shown in Figure 6. It is interesting to note that the rate of desorption is faster than the rate of adsorption for all points on the isotherm. The fitting parameter for eq 4 is the
Figure 6. Desorption fraction complete for primary spillover at 298 K for final equilibrium amount: 0.5, O; 3.5, ∆; 4.5 mmol/g, 0. Step size was uniformly 0.5 mmol/g.
diffusion time constant. Values for this parameter are plotted in Figure 7 for adsorption and desorption as a function of adsorbed amount. No literature values of D/R2 were readily available for hydrogen; however, the values are 2 orders of magnitude higher than those reported for nitrogen on molecular sieve carbon (7.0 × 10-6 s-1).54 The parameter is lower compared to pore diffusion in physisorption modeling.49 Since the approximate diffusion distance is known from TEM images,32 the atomic hydrogen diffusion coefficient may be roughly estimated from the time constant. On average, platinum sources are 300 Å apart, so the maximum diffusion distance, Rs, is 150 Å. Using the values of the diffusion time constant in Figure 7, the diffusion coefficient for hydrogen atoms on carbon is estimated as 2 × 10-15-5 × 10-16 cm2/s. These values are 1 order of magnitude higher compared to a prediction from literature using an Arrhenius relationship to estimate Deff, H (298 K) ) 4 × 10-17 cm2/s,55-57 indicating that hydrogen atoms may have not reached the distance assumed above. A similar analysis was performed for the selected secondary spillover adsorbent. The results of Fick’s diffusion model applied to AX-21/PtC/Bridge (8:1:1) are reported for adsorption and desorption in Figures 8 and 9, respectively. Adsorption kinetics are slower for the secondary spillover adsorbent relative to the
Reverse Spillover of Hydrogen on Nanomaterials
J. Phys. Chem. C, Vol. 113, No. 31, 2009 13937
Figure 7. Diffusion time constants for primary spillover of H on 6 wt % Pt/TC at 298 K.
Figure 10. Diffusion time constants for secondary spillover of H on AX-21/Pt/C/Bridge (8:1:1) at 298 K.
Figure 8. Adsorption fraction complete for secondary spillover at 298 K for final equilibrium amount: 0.5, O; 2, ∆; 4, 0; 4.5 mmol/g, ]. Step size was uniformly 0.5 mmol/g.
Figure 11. Diffusion time constants for primary spillover of D on 6 wt % Pt/TC at 298 K.
Figure 9. Desorption fraction complete for secondary spillover at 298 K for final equilibrium amount: 0.5, O; 2, ∆; 4, 0; 4.5 mmol/g, ]. Step size was uniformly 0.5 mmol/g.
primary spillover adsorbent. In both cases, desorption is faster relative to adsorption with substantial differences at high equilibrium amounts. The diffusion time constant is plotted versus adsorbed amount in Figure 10.
The reachable distance of hydrogen atoms is not as well characterized in this case. If the range of diffusion coefficients calculated above is assumed to apply to secondary spillover, the maximum reachable spillover distance is ∼800 Å. Considering that most adsorbent particles are several thousand angstroms in diameter, this diffusion distance confirms the assumption that sources should be highly dispersed on receptors to maximize the extent of spillover. The diffusion time constant was estimated for deuterium adsorption and desorption on 6 wt % Pt/TC at 298 K. The result is shown in Figure 11. Comparing the results to those for primary spillover of hydrogen atoms (Figure 7), deuterium adsorption rates are slower with subsequent lower diffusion time constants, due to kinetic effects.58 An interesting behavior occurs for desorption as the adsorbed amount decreases. The expected trend is not followed and deuterium diffusion time constants are larger relative to hydrogen at nearly all desorption conditions in this study. Recombination during desorption influences the magnitude of the difference. The earlier assumption that all of the spillover atoms desorb through a reverse spillover mechanism must be modified, as a certain fraction of these atoms seem to undergo recombination on the receptor during the desorption step. Deuterium appears to be tracing recombination on the surface that would explain
13938
J. Phys. Chem. C, Vol. 113, No. 31, 2009
Figure 12. Rates of hydrogen recombination for primary spillover during equilibration time for 0-0.5 (dashes), 3-3.5 (solid), and 4-4.5 mmol/g step (solid-dashes).
faster desorption rates. In a system of uniform hydrogen concentration, clusters of hydrogen atoms could form at highenergy defect sites or on surface functional groups. As noted, the simulation work of Yakobson and co-workers demonstrated stable clusters of six hydrogen atoms that spilled over onto a model graphene sheet.51 Recombination of these clusters without the need to fully reverse back to the source particle favors the observation of increased desorption kinetics. If the theory of atomic hydrogen cluster stability is accepted, then there is likely to be some recombination during adsorption until equilibrium is reached. This serves to slow down the net adsorption rate. The calculation technique using volumetric adsorption data as a function of time to describe spillover kinetics does not couple the net uptake of hydrogen atoms to recombination. However, desorption is a measure of both recombination and reverse spillover. The difference in the observed adsorption and desorption rates should give an estimate of the rate of recombination. Figure 12 indicates the result for subtraction of the rate of adsorption from that of desorption for primary spillover of hydrogen at 298 K. The rate of recombination is computed on a unit surface area basis using the BET values in Table 1. Recombination rates are presented for three final equilibrium amounts. The highest rate of 9.5 × 1011 atoms H/cm2 · s is found for the lowest amount and less than 1 min into the spillover event. Recombination rates decrease as the quantity of adsorbed atoms increases, a phenomenon that is in agreement with a higher probability of stable hydrogen atom clusters found on a model graphene receptor.51 The recombination rate behavior with adsorbed amount is somewhat unexpected if the mechanism follows a classical rate law. One would expect recombination to increase with adsorbed amount in that case. The observations here point toward increasingly stable configurations of spillover hydrogen atoms as the adsorbed amount increases. The computational binding energy reported by Yakobson and co-workers for hydrogen atoms in aromatic clusters demonstrated an increasing trend with cluster size (for 6-24H clusters).51 Our experimental results support this simulation in that a decreasing recombination rate for increased adsorbed amount tends to indicate that the probability of newly spiltover hydrogen atoms joining a stable cluster is greater when there are larger quantities of atoms present on the surface. As the cluster size grows, the configu-
Lachawiec and Yang
Figure 13. Rates of hydrogen recombination for secondary spillover during equilibration time for 1.5-2 (dashes) and 3.5-4 mmol/g step (solid).
ration further stabilizes and supports the observed elevated capacity at room temperature and moderate pressure due to spillover on carbon-based receptors. The recombination rate for hydrogen atoms on a secondary spillover adsorbent is presented in Figure 13 for AX-21/PtC/ Bridge (8:1:1). The recombination rate for secondary spillover is nearly 1 order of magnitude larger compared to primary spillover for equilibrium adsorbed amount less than 2 mmol/g. The difference between primary and secondary spillover adsorbents decreases as the amount increases; however, recombination rates are always greater for secondary spillover. Such behavior indicates stability of adsorbed configurations. Atoms must travel farther on secondary receptors to achieve a stable state. It must be noted that as the adsorbed amount increases, hydrogen atoms are spread out over an increasingly large portion of the surface. The local concentration of the receptor surface is not directly proportional to the overall adsorption amount as reported from equilibrium isotherm measurements. Recombination is significantly altered by the receptor and the differences are most evident for low adsorption amount. The recombination rate for deuterium atoms was computed for two different equilibrium adsorbed amounts, and the results are presented in Figure 14. Recombination rates are greater for deuterium relative to hydrogen at low amount adsorbed (