Chemisorption, physisorption and hysteresis during hydrogen storage in carbon nanotubes

Chemisorption, physisorption and hysteresis during hydrogen storage in carbon nanotubes

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7 Available online at S...

815KB Sizes 1 Downloads 62 Views

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7

Available online at

ScienceDirect journal homepage:

Chemisorption, physisorption and hysteresis during hydrogen storage in carbon nanotubes Seyed Hamed Barghi, Theodore T. Tsotsis, Muhammad Sahimi* Mork Family Department of Chemical Engineering & Materials Science, University of Southern California, Los Angeles, CA 90089-1211, United States

article info


Article history:

The question of chemisorption versus physisorption during hydrogen storage in carbon

Received 9 July 2013

nanotubes (CNTs) is addressed experimentally. We utilize a powerful measurement tech-

Received in revised form

nique based on a magnetic suspension balance coupled with a residual gas analyzer, and

26 October 2013

report new data for hydrogen sorption at pressures of up to 100 bar at 25  C. The measured

Accepted 30 October 2013

sorption capacity is less than 0.2 wt.%, and there is hysteresis in the sorption isotherms

Available online 2 December 2013

when multi-walled CNTs are exposed to hydrogen after pretreatment at elevated temperatures. The cause of the hysteresis is then studied, and is shown to be due to a com-


bination of weak sorption e physisorption e and strong sorption e chemisorption e in the

Hydrogen storage

CNTs. Analysis of the experimental data enables us to calculate separately the individual

Carbon nanotubes

hydrogen physisorption and chemisorption isotherms in CNTs that, to our knowledge, are


reported for the first time here. The maximum measured hydrogen physisorption and


chemisorption are 0.13 wt.% and 0.058 wt.%, respectively.


Copyright ª 2013, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.



Hydrogen is considered a promising sustainable source of energy, due to its high energy density, and the fact that it can be produced from a variety of renewable sources, including biomass (via gasification) and water electrolysis (e.g., via renewable solar energy). Hydrogen combustion, in addition, does not emit green-house gases to the atmosphere [1], which justifies its classification as a clean source of energy. To be able to use hydrogen as an energy source, its safe and efficient storage is a key requirement, but has remained a most challenging issue. Compressing hydrogen at room temperature to high pressures of up to 350e700 bar, or its liquefaction at cryogenic temperatures and lower pressures, were the initial two methods used for hydrogen storage. The

high costs and the safety risks involved are, however, key drawbacks of these two conventional storage techniques and have hindered their commercial applications so far [2]. Adsorbing hydrogen on a solid adsorbent at moderate pressures and ambient temperatures is another potential method for hydrogen storage. The method is safer and requires less expensive storage equipment than the high-pressure compression and cryogenic systems [3]. Synthesis of adsorbent materials capable of adsorbing large amounts of hydrogen at such conditions is, however, the main issue that still needs to be addressed, in order to commercialize hydrogen storage systems based on solid adsorbents. Metal hydrides and carbon-based materials are the two major types of adsorbents that have been considered to date, and have been the subject of many studies in the field of hydrogen storage over the past 10 years [4e8].

* Corresponding author. Tel.: þ1 (213) 740 2064; fax: þ1 (213) 740 8053. E-mail address: [email protected] (M. Sahimi). 0360-3199/$ e see front matter Copyright ª 2013, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7

Among the various types of carbon-based adsorbents proposed for hydrogen storage, carbon nanotubes (CNTs) have received considerable attention in recent years, due to their high surface area, nanometer size pores with a narrow pore size distribution, and low mass density [9e11]. To investigate the performance of CNTs for hydrogen storage, it is necessary to accurately determine the equilibrium hydrogen adsorption/desorption isotherms on such materials. In particular, the kinetics of adsorption are highly important to determining the charging time of any practical hydrogen storage system, as well as the method by which such charging may be implemented. Moreover, in order to be able to ensure that the storage system is capable of delivering the adsorbed hydrogen to the vehicle’s engine with an appropriate rate, the kinetics of hydrogen desorption from the CNTs must also be measured. To date, three techniques have been used to measure hydrogen adsorption in CNTs, namely, the Sievert method, thermal desorption spectroscopy (TDS), and the gravimetric method [12]. The Sievert method is a commonly-used volumetric technique based on changes in the pressure of the measurement vessel caused by adsorption/desorption of the adsorptive gas. Its simplicity and the fact that it may better approximate the conditions under which commercial hydrogen storage systems operate are, reportedly, the key advantages of the method. On the other hand, since in the Sievert method the change in the pressure of the adsorption vessel is the main means via which adsorption and/or desorption amounts are measured, omnipresent leaks during the experiments can (and often do) lead to erroneous results [13]. In addition, since vessel fluid dynamics interfere with measurements at short times, it is difficult to accurately measure fast sorption kinetics and, thus, to accurately differentiate by this method between weak e physisorption e and strong e chemisorption e isotherms, as we are able to do with the technique used in this study. Another drawback of the Sievert method is the fact that atmospheric pressure is the lowest pressure at which desorption experiments may be conveniently carried out [14]. The TDS method is based on measuring the hydrogen that is adsorbed in CNTs via its desorption in high vacuum using mass spectrometry. The high sensitivity of the technique, allowing it to detect adsorption even by a small amount of CNTs (as low as 1 mg), is the main advantage of the method [15]. On the other hand, it is not possible to use this technique for identifying adsorption hysteresis that may occur, or to differentiate among the various types of sorption that may take place simultaneously. In addition, since hydrogen desorption happens under high vacuum, it is difficult to gauge from the isotherms measured by this method whether the material studied can function as a practical storage medium in the range of pressures (of up to 100 bar) suggested by the U.S. Department of Energy [16]. The gravimetric method is the third technique utilized for the analysis of hydrogen adsorption. In this method, solid adsorbents are exposed to the adsorptive gases at various pressures, while the gravimetric balance monitors the weight change of the adsorbent that is caused by


adsorption. The advantage of this method, compared to the TDS, is that one can independently determine both the adsorption and desorption isotherms, as we do in this study and, thus, detect the presence of hysteresis, In addition, one is able to study fast dynamics and, thus, to discriminate among various types of sorption phenomena occurring either simultaneously or at distinct time scales. Compared to the Sievert method, the advantage of the gravimetric method is that it is not affected by the often inevitable gas leaks that develop in the measurement cell [12]. Though the technique cannot, by itself, distinguish the various gases adsorbed (a key requirement for studying multi-component adsorption), coupling the gravimetric balance equipment to a sensitive mass analyzer (as is the case with the experimental system used in this study) overcomes this main drawback [13]. The hydrogen uptake capacities of carbon nanomaterials published in the literature are remarkably scattered. Geng et al. [17] reported, for example, that CNTs are capable of adsorbing 0.1 wt.% hydrogen at 293 K and 10 MPa, whereas Chambers et al. [18] claimed that tubular graphite nanofibers adsorb 11.26 wt.% hydrogen at 298 K and 11.35 MPa. Other experimental data, which are in variance with the two aforementioned studies, have been reported as well [19]. Such inconsistent results have been a matter of controversy in the area of hydrogen storage using carbon nanomaterials. To resolve the controversy, Tibbetts et al. [20] examined hydrogen sorption in nine carbon materials, including graphite particles, activated carbon, graphitized PYROGRAF vapor-grown carbon fibers (VGCF), CO and air-etched PYROGRAF fibers, Showa-Denko VGCF, carbon filaments, and nanotubes from the MER Corp. and Rice University. Sorption experiments were carried out at temperatures between 80  C and 500  C and at pressures of up to 11 MPa. The maximum hydrogen uptake for the nine carbon materials at room temperature was 0.1 wt.%. According to Tibbetts et al., the hydrogen adsorption capacity of a number of the carbon materials at room temperature is so low that, without doing careful calibration, it is impossible to even detect it with a reliable accuracy. These findings can potentially cast doubt on the prior experimental work that has claimed hydrogen uptake capacities higher than 1 wt.% for carbon materials at room temperature. In another work, Zuttel et al. [21] investigated the hydrogen storage capacity of a number of carbon materials, including multi-walled CNTs (MWCNTs) fabricated by the pyrolysis of acetylene. Their experiments yielded hydrogen storage capacities, at room temperature and at a pressure of 10 MPa, not greater than 0.6% (on a per mass basis). Zuttel et al. suggested that the early reports of extraordinary large hydrogen adsorption capacities of CNTs must, therefore, be viewed with skepticism. In addition to the aforementioned experimental studies, molecular simulations have also indicated relatively low hydrogen adsorption capacities for CNTs. Dodziuk and Dolgonos [22] used molecular mechanics calculations and molecular dynamics simulations, for example, to study the ability of individual armchair, zigzag, chiral, and bundles of nanotubes to function as hydrogen storage media, suing the consistent-valence force field (CVFF) and the extensible systematic force field (ESFF) in their simulations. Their results suggested that high hydrogen storage in CNTs cannot be achieved through physisorption alone.


i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7

Hydrogen adsorption/desorption hysteresis in CNTs has been reported previously in studies that used the conventional Sievert method to measure the sorption isotherms [14,23,24]. It has been hypothesized that the reason for observing the hysteresis may be the formation of some type of chemical bonding between hydrogen and the metal catalyst residues or the amorphous carbon in CNTs. Previous studies have not, however, been able to perform quantitative experiments in order to confirm the hypothesis. For example, one cannot exclude conclusively that the observed differences between the adsorption and desorption isotherm branches are not simply due to the experimental difficulties associated with leaks in the Sievert measurement vessel [12]. Another possible scenario is that the hysteresis may have been a result of chemisorption of hydrogen on CNTs, which has been reported [25] as an adsorption mechanism in CNTs. Thus, it is crucial to be able to clearly distinguish between hydrogen chemisorption from physisorption occurring in CNTs. In this paper, we specifically address the question of hysteresis during hydrogen sorption in CNTs, as well as whether chemisorption or physisorption occurs in such nanostructured materials. Another motivation for the present work is the need for high-precision experiments, so that one can potentially identify the reason for the scattered hydrogen adsorption uptake data in the literature. We report here new measurements of hydrogen sorption in CNTs, using a precise measurement technique that enables us to measure separately the contributions of chemisorption and physisorption to the overall storage capacity of CNTs. In contrast, most measurement techniques used in the past [26] cannot readily differentiate between physisorption and chemisorptions. To carry out the experiments, we use a high-pressure gravimetric instrument, coupled with a highly-sensitive mass analyzer system for measuring hydrogen sorption isotherms of MWCNTs. To our knowledge, the technique we use has not been utilized before in such studies, and there have been no previous studies that could unequivocally distinguish between chemisorption and physisorption of hydrogen occurring on MWCNTs at room temperature. This has been made possible in this study by the combination of the gravimetric and mass spectrometry techniques because, (i) Hydrogen adsorption and desorption isotherms are measured accurately and simultaneously with the same experimental method. This is not possible with the TDS method, whereas the Sievert method is sensitive to system leaks [27e30]. Therefore, since the two methods are distinct with their own inherent experimental errors, a quantitative comparison between the results obtained with these two methods is not reliable. (ii) With our experimental technique, the presence of potential impurities in the adsorptive gas or the desorption of volatile impurities from the solid sample can be detected with the residual gas analyzer that is connected to the gravimetric equipment. This then distinguishes the method from the gravimetric method used alone, or from the conventional Sievert method.




Characterization of carbon nanotubes

The hydrogen used in the experiments was ultrahigh pure grade (99.999%) from Gilmore. The MWCNTs with a purity of 97.46 wt.% were purchased from Nanostructure and Amorphous Materials, Inc. (the manufacturer reports the following other elements as impurities: Al, 0.19 wt.%; Cl, 1.02 wt.%; Co, 1.09 wt.%, and S, 0.24 wt.%). The MWCNTs are produced by natural gas catalytic decomposition over a Co-based catalyst. The inner and outer diameters of the nanotubes are approximately 5 and 8 nm. Specific surface area, average pore diameter, and pore volume of the MWCNTs sample were determined by us via nitrogen adsorption at its normal boiling temperature using an ASAP 2010 BET instrument. The crystalline structure of MWCNTs was examined with the aid of a Rigaku XRD equipment.


Monitoring the gas composition

The composition of hydrogen gas used in the experiments was monitored with a residual gas analyzer (RGA200), manufactured by Stanford Research Systems (SRS), Inc. The RGA system was equipped with an electron multiplier that makes it possible to measure partial pressures of species as low as 1014 Torr. The effect of the background noise on the mass analysis results was corrected by running the RGA200 system without injecting any gases. In the first step, the composition of the ultrahigh pure hydrogen was determined prior to injecting the gas into the system, in order to ensure that there were no unexpected impurities in the hydrogen feed. In the next step the gas stream leaving the adsorption vessel was analyzed to check that there was no source of contamination in the system, and that MWCNTs do not contain any unstable contaminations. The aim of this step was to guarantee that the data represent solely hydrogen adsorption, and that possible adsorption of other gasses or desorption of impurities from MWCNs do not cause experimental errors.


Effect of the drift

The gravimetric hydrogen uptake of MWCNTs was measured using a magnetic suspension balance (MSB), manufactured by Rubotherm, by monitoring the changes in the weight of the MWCNTs sample due to hydrogen adsorption. Therefore, in order to make definitive estimates of the change in the sample’s weight during the experiments, it is necessary to consider the drift of the MSB over time, which is the error in the weight measurement by the MSB that may happen during the experiments. Since no adsorption or desorption would occur when there is no sample in the container, measuring over time the weight of the empty stainless steel sample container is the most accurate way of observing the effect of the drift on the accuracy of the data.


Stability of the MWCNT sample

Before starting the experiments, it is important to remove the potential adsorbed impurities (e.g., water, CO2, and various

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7


would lead to misleading estimates of the hydrogen uptake capacity of MWCNTs.


Determining the volume of the MWCNTs sample

Buoyancy forces have a small but non-negligible effect on the sample’s weight, particularly at the higher pressures. To account for their impact, one must first measure the sample’s true solid volume, otherwise, known as the skeletal volume that excludes the pore volume. For gravimetric measurements, this is typically accomplished by measuring the sample’s (apparent) weight, mP , at various pressures of He and by correlating to the sample’s true weight (and also its apparent weight under vacuum conditions) m0 according to the simple Archimedes formula m0 ¼ mP þ Vs rg

Fig. 1 e (a) Analysis of the gas streams entering and leaving the adsorption vessel using the residual gas analyzer. (b) Weight of the empty sample container and the fresh sample’s weight vs. time under dynamic vacuum at pressures of 10L5 bar. (c) Plot of Eqn. 1 for the MWCNTs for various pressures of Helium gas.

hydrocarbons) from the MWCNTs surface. This was done by heating-up the sample at 120  C for 5 h under dynamic vacuum. The sample’s weight was subsequently recorded under vacuum at 25  C. The goal was to ensure that there is no desorption of possible adsorbed impurities from the MWCNTs sample during the hydrogen adsorption experiments. This step is important because desorption of any impurities from the sample during the hydrogen adsorption experiments


where Vs is the sample’s true (skeletal) volume, and rg is the density of He. He was used as a test gas because it is considered to be inert, non-adsorbing and the lightest among the noble gases. In the experiments, its bulk gas density, rg , was measured directly by weighing a reference stainless steel insert of known volume at various pressures. Plotting ðm0  mP Þ vs. the gas density, rg , yields a straight line [Fig. 1(c)] with its slope being the sample’s true volume,Vs . Once the skeletal volume is known, and the hydrogen density is measured at various pressures, the true sample weight during adsorption/desorption is calculated by adding to the apparent weight the buoyancy correction term. The measured hydrogen density is 0.007625 g/cm3 at 100 bar, 0.000801 g/cm3 at 10 bar and 0.0000813 g/cm3 at 1 bar (the calculated densities using the Peng Robinson equation of state [31] at 100, 10, and 1 bar are 0.00783 g/cm3, 0.00081 g/ cm3, and 0.0000813 g/cm3, respectively) and, thus, the total buoyancy term corresponds to 0.354 wt.% at 100 bar, 0.0372 wt.% at 10 bar, and 0.00378 wt.% at 1 bar. During the sorption calculations the assumption was made, due to the small amounts of hydrogen adsorbed (particularly the strongly adsorbed hydrogen that occupies less than 5% of the BET surface area), that no significant changes in the sample’s skeletal volume take place. Relaxing that assumption and assuming a 5% volume change would correspond to an error of 0.0177% at 100 bar, 0.00186% at 10 bar, and 0.000189% at 1 bar, and would, in no way, change any of the conclusions about adsorption hysteresis and the presence of strong and weak adsorption occurring in MWCNT.


Hydrogen adsorption isotherms

The hydrogen uptake of the MWCNTs sample was determined by weighing the sample at various hydrogen pressures in the measurement vessel. The effect of the buoyancy forces on the results was taken into account in the calculations using Eq. (1). After reaching the equilibrium state at each pressure, a backpressure controller increased the system’s pressure to the next higher pressure. After hydrogen adsorption was measured at 100 bar, the system’s pressure was decreased step by step to measure the desorption isotherm, which represents the first hydrogen adsorption/desorption cycle.


i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7

At the end of the first hydrogen sorption cycle, the hydrogen uptake of the MWCNTs sample was measured again, representing the second hydrogen adsorption/desorption cycle. The difference between the first and the second cycles was in the initial preparation of the MWCNTs samples. For the first cycle, the MWCNTs were first degased at elevated temperature of 120  C. For the second cycle the MWCNTs were not degased at the elevated temperatures and, therefore, the hydrogen remaining on the surface from the first cycle was still adsorbed on the MWCNTs at the beginning of the second cycle.


Results and discussion

The surface area and pore structure characteristics of the MWCNTs sample were measured using the BET technique. The specific surface area is 441.3 m2/g, the average tube diameter is 7.4 nm, while the pore volume is 0.82 cm3/g. The Xray diffraction (XRD) analysis of the MWCNTs produces two sharp peaks for the carbon layers that are characteristic of the C(002) and C(100), which are the typical XRD peaks for carbon reported for MWCNTs, but detects no other crystalline phases [32]. The mass-spectrometric analyzer (RGA200, coupled with an electron multiplier), was used to analyze the compositions (in the range of 1e64 amu) of the gas streams entering and leaving the sorption measurement system. The results of the analysis are reported in Fig. 1(a) and they show only the presence of molecular hydrogen (a sharp peak at 2 amu); there are no other impurities (analytical detection limit of <0.1 ppmv) in the gas phase that could have resulted in experimental uncertainties. We note that there was no drift in the measured weight of the empty container, even after the measurement had been carried out for 3 days [see Fig. 1(b)]. This means that the drift of the magnetic suspension balance during the hydrogen adsorption experiments is negligible. Fig. 1(b) also shows the MWCNTs sample’s weight at 25  C and under dynamic vacuum as a function of time (prior to that the sample was degased at 120  C for 6 h to remove water and other impurities that might potentially had adsorbed during exposure to laboratory conditions). The sample’s weight remains constant under dynamic vacuum even after 2 days, indicating the absence of volatile impurities on the surface of the MWCNTs. Coupled to the findings that no other species are detected in the gas phase [see Fig. 1(a)], one concludes that any subsequent weight changes detected with this particular MWCNT sample is only a result of hydrogen adsorption/desorption, and not due to instrument drift or potential instability of the MWCNTs themselves. Fig. 1(c) illustrates the results for the measurement of the sample’s weight at various helium pressures, corresponding to various gas densities. The linear relationship between the sample’s weight and the helium gas density confirms the validity of Eq. (1) for determining the MWCNTs sample’s true volume. The true density of the MWCNTs sample, determined using helium gas, was 2.15 g/cm3, which is virtually identical to what has been previously reported (w2.1 g/cm3) by others for such materials [33]. Fig. 2 shows the hydrogen adsorption in the MWCNTs as a function of time at a pressure of 5 bar. For this experiment, the

MWCNTs were exposed to a hydrogen flow for the first time after the sample was degased in vacuum at 120  C for 6 h; this also represents the first hydrogen sorption step on the way to generate the adsorption isotherm. Fig. 2 demonstrates that it takes around 4e5 h for the system to reach equilibrium, which is in line with the equilibration times reported in the literature [24] for such materials. Careful examination of Fig. 2 reveals, furthermore, that a considerable part of the hydrogen adsorbs almost instantaneously on the MWCNTs, and that w90% of the total amount is adsorbed within the first hour. This fast “charging” time is an important consideration in terms of the eventual practical application of these materials. After the sample’s weight had equilibrated (upon raising the pressure from vacuum to 5 bar), the pressure was raised again in a step-wise manner in order to generate the adsorption isotherm, which is indicated as the 1st hydrogen adsorption cycle in Fig. 3. Subsequently, the pressure was lowered in a step-wise manner in order to obtain the desorption branch of the cycle, which is also shown in Fig. 3. When the pressure reached back to 5 bar, the weight of the sample was allowed to equilibrate and, subsequently, the vacuum pump was turned on. The change in the sample weight is also shown in Fig. 2. Interestingly, even after the weight of the sample levels off, it does not return to its initial weight, and in fact the sample retains almost 35% of the total hydrogen adsorbed on the CNTs at the end of the first adsorption isotherm run. Keeping the sample under 105 bar of dynamic vacuum for an additional two and a half days (the data not shown here) did not change the sample’s weight any further beyond what is shown in Fig. 2. The data in Fig. 3 clearly show that there is a substantial hysteresis between the adsorption and the desorption branches of the first cycle. Subsequently, the pressure of the sample was raised from 105 bar to 5 bar and the adsorption and desorption isotherms were again generated, indicated as the 2nd adsorption/

Fig. 2 e (a) Hydrogen adsorption at the beginning of the 1st adsorption/desorption cycle (step from 0 to 5 bar); (b) Hydrogen desorption at the end of 1st hydrogen adsorption/desorption cycle (from 5 to 0 bar). (c) Desorption of chemisorbed hydrogen under vacuum at 120  C after the end of the 3rd hydrogen adsorption/desorption cycle.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7

desorption cycle in Fig. 3. Interestingly, no hysteresis was observed between the adsorption and desorption branches of the 2nd adsorption/desorption cycle, which are completely indistinguishable from each other, as well as being very close to the desorption branch from the first adsorption/desorption cycle. Upon completion of the second adsorption/desorption cycle, a third cycle was carried out and, again, no hysteresis was observed (the results are not shown here because they are very much indistinguishable from those of the 2nd cycle). It is clear from the data in Fig. 3 that evacuation of the sample at 25  C does not return its weight to the original value. Since the weight difference involved is rather small (0.25 mg or w0.07% of the original MWCNT sample weight of w300 mg e however, as previously noted, this is a large fraction w35% of the total amount of hydrogen adsorbed during the first adsorption cycle), the possibility exists that it may have resulted from dust particles in the gas atmosphere or in the apparatus chamber being deposited on the MWCNTs sample. To exclude such a possibility, upon the completion of the 3rd adsorption/desorption cycle the temperature of the sample was raised (under 105 bar of dynamic vacuum) to 120  C. The sample weight started decreasing (see Fig. 2) and after w6 h it returned to its original weight. During the same period the mass analyzer did not indicate the presence of any other gas species (in the range 1e64 amu) other than hydrogen. In our view, the data in Fig. 3 indicate two different types of hydrogen species on the surface, weakly adsorbed, or physisorbed, hydrogen and strongly adsorbed, or chemisorbed, hydrogen, the latter being defined here as the adsorbed species that will not desorb from the surface under dynamic vacuum at 25  C for a period of more than two days (see Fig. 2) and which are, thus, unlikely to desorb from the same MWCNTs during the normal charging and discharging cycles as well. In Fig. 4 we plot the weight change during the 1st adsorption/desorption cycle as we transition from 20 bar to 40 bar, as well as the corresponding weight change profile as we transition back from 40 to 20 bar during the desorption branch of the cycle. Shown in the same figure are the rates of change in the

Fig. 3 e The 1st and 2nd hydrogen adsorption/desorption cycles (note that for the 2nd cycle the adsorption and desorption branches are indistinguishable).


sample’s weight, dW/dt. As already noted above, during the adsorption step one observes a clearly sharp rise in the weight, followed by a more gradual increase. Interestingly, in the rate of change in the weight profiles one observes two branches with two very distinct slopes. For the desorption step again most of the weight loss (98%) occurs in the first 7 min. If one is to assume that the strongly adsorbed species, defined here as those that do not desorb under dynamic vacuum of 105 bar, are unlikely to desorb either as one transitions from 40 to 20 bar in pressure, then the desorbed amount can be fully attributed to weakly-adsorbed or physisorbed hydrogen. Since no hysteresis exists during adsorption/desorption from the part of the MWCNT surface that is not covered by the strongly adsorbed hydrogen species (see Fig. 3), one can then assume that the amount that is desorbed during the desorption branch of the first cycle is equal to the amount that is physisorbed during the adsorption part of the cycle. This way one can generate an adsorption isotherm (for the first experimental cycle) for the weakly-adsorbed hydrogen that is shown in Fig. 5. By subtracting the physisorbed isotherm from the total 1st cycle adsorption isotherm one can, in addition, generate the chemisorption isotherm, which is also shown in Fig. 5. The data suggest that for pressures of up to 20 bar the physisorption and chemisorption have relatively the same magnitude. But, above 20 bar physisorption becomes the primary adsorption mechanism as the strongly-adsorbing sites on the surface of the MWCNT seem to saturate rather quickly. The assumption that the amount desorbed at each pressure step during the desorption branch of the 1st cycle is only due to weakly-adsorbed hydrogen and by subtracting that amount from the total adsorption isotherm one can correctly generate the chemisorption isotherm, entails the fundamental assumption that there are two distinctly different types of sites on the MWCNT, one on which exclusively weak adsorption occurs, and another on which preferential strong adsorption takes place. If that is indeed the case, and one subtracts from the total adsorption isotherm during the 2nd (and 3rd) cycle the amount that remains irreversibly adsorbed under dynamic vacuum at 25  C, one should be able to generate the physisorption isotherm of Fig. 5. To prove that this is indeed the case, all three (the physisorption isotherm of Fig. 5 and the

Fig. 4 e Hydrogen uptake (w) and absolute value of hydrogen uptake change rate (dw/dt).


i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7

MWCNT surface, the fraction of the BET surface area of MWCNT covered by adsorbed hydrogen at the end of adsorption branch is rather small, w9.2% for the physisorbed hydrogen and w4% for the chemisorbed hydrogen, since it is unclear, at this point, what kind of surface species are involved in these two different types of adsorption, or whether it even makes good sense to compare the surface area based on liquid nitrogen adsorption to the area occupied by hydrogen, these estimates are based on the simple idea that the adsorbed hydrogen molecule occupies an area on the BET surface of MWCNTequivalent to a circle with a diameter equal to its Lennard-Jones diameter. This is a likely to be a good assumption for the physisorbed hydrogen, but less so for the chemisorbed one, though any differences that may exist are unlikely to be substantial. Such small coverages are in line with prior modeling investigations [34]. Fig. 5 e The calculated sorption isotherms in the MWCNTs at 25  C.

estimated physisorption isotherms for cycles 2 and 3) are shown in Fig. 6. The three lines nearly coincide, thus lending credence to the above hypothesis. This, then, means that hydrogen sorption hysteresis during the first cycle is a consequence of hydrogen chemisorption on the MWCNTs. The chemisorbed hydrogen atoms do not leave the MWCNTs when the system’s pressure decreases during the desorption experiments. Therefore, the accumulation of chemisorbed hydrogen at various pressures during adsorption gives rise to the observed hysteresis during desorption. For practical applications, these results validate the idea that only physisorption is involved during the charging and discharging parts of the cycle. Despite the fast rates of physisorption and the apparent strong chemisorption affinities between hydrogen and



Precise measurements by a new technique have demonstrated that hydrogen adsorption in multi-walled carbon nanotubes at room temperature is a combination of reversible physisorption and irreversible chemisorption. The adsorption measurements unveil and confirm that it is the chemisorption part that gives rise to hysteresis in hydrogen adsorption/ desorption isotherms. For instance, while the chemisorbed hydrogen was still on the surface of the MWCNTs, the sample was once more exposed to hydrogen and the adsorption/ desorption isotherms were measured again. No hysteresis was detected for the new cycle, implying that no chemisorption happened during the second cycle. The equilibrium values of the new hydrogenation cycle are different from those of the first cycle. Our calculations confirmed that, by subtracting the chemisorbed values from the equilibrium data of the first cycle, the adsorption and desorption branches of the first hydrogenation cycle and the equilibrium data of the second cycle take on the same values. This supports the conclusion that the reversible hydrogen adsorption on MWCNTs during practical conditions is completely due to physisorption. The technique used in this work is a practical method for determining separately hydrogen physisorption and chemisorption in carbon nanotubes and other nanostructured porous materials.

Acknowledgments SB is grateful for the support from a Provost’s Fellowship at the University of Southern California. The support of the U.S. Department of Energy (DE-SC0003586) and the National Science Foundation (CBET-0854427) is also gratefully acknowledged.

Fig. 6 e The amount of hydrogen physisorption on the MWCNTs during the first sorption cycle and the equilibrium hydrogen values during the second and third hydrogen sorption cycles (the amount of the strongly sorbed hydrogen after the 1st cycle is subtracted).


[1] Schlapbach L, Zu¨ttel A. Hydrogen-storage materials for mobile applications. Nature 2001;414:353e8.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 3 9 0 e1 3 9 7

[2] Corgnale C, Hardy BJ, Tamburello DA, Garrison SL, Anton DL. Acceptability envelope for metal hydride-based hydrogen storage systems. Int J Hydrogen Energy 2012;37:2812e24. [3] Sakintunaa B, Lamari-Darkrim F, Hirscher M. Metal hydride materials for solid hydrogen storage: a review. Int J Hydrogen Energy 2007;32:1121e40. [4] Rosi NL, Eckert J, Eddaoudi M, Vodak DT, Kim J, O’Keeffe M, et al. Hydrogen storage in microporous metal-organic frameworks. Science 2003;300:1127e9. [5] Dimitrakakis GK, Tylianakis E, Froudakis GE. Pillared graphene: a new 3-D network nanostructure for enhanced hydrogen storage. Nano Lett 2008;8(10):3166e70. [6] Park SJ, Kim BJ, Lee YS, Cho MJ. Influence of copper electroplating on high pressure hydrogen-storage behaviors of activated carbon fibers. Int J Hydrogen Energy 2008;33:1706e10. [7] Langhammer C, Zoric I, Kasemo B. Hydrogen storage in Pd nanodisks characterized with a novel nanoplasmonic sensing scheme. Nano Lett 2007;7(10):3122e7. [8] Berseth PA, Harter AG, Zidan R, Blomqvist A, Arau´jo CM, Scheicher RH, et al. Carbon nanomaterials as catalysts for hydrogen uptake and release in NaAlH4. Nano Lett 2009;9(4):1501e5. [9] Lee SY, Park SJ. Effect of temperature on activated carbon nanotubes for hydrogen storage behaviors. Int J Hydrogen Energy 2010;35:6757e62. [10] Zhu H, Cao A, Li X, Xu C, Mao Z, Ruan D, et al. Hydrogen adsorption in bundles of well-aligned carbon nanotubes at room temperature. Appl Surf Sci 2001;178:50e5. [11] Ci L, Zhu H, Wei B, Xu C, Wu D. Annealing amorphous carbon nanotubes for their application in hydrogen storage. Appl Surf Si 2003;205:39e43. [12] Hirscher M, Becher M, Haluska M, Quintel A, Skakalova V, Choi YM, et al. Hydrogen storage in carbon nanostructures. J Alloys Compd 2002;330e332:654e8. [13] Panella B, Hirscher M, Roth S. Hydrogen adsorption in different carbon nanostructures. Carbon 2005;43:2209e14. [14] Hou PX, Xu ST, Ying Z, Yang QH, Liu C, Cheng HM. Hydrogen adsorption/desorption behavior of multi-walled carbon nanotubes with different diameters. Carbon 2003;41:2471e6. [15] Sudan P, Zuttel A, Mauron P, Emmenegger Ch, Wenger P, Schlapbach L. Physisorption of hydrogen in single-walled carbon nanotubes. Carbon 2003;41:2377e83. [16] DOE hydrogen and fuel cells program plan drafthttp://www1.; 2011. [17] Geng HZ, Kim TH, Lim SC, Jeong HK, Jin MH, Jo YW, et al. Hydrogen storage in microwave-treated multi-walled carbon nanotubes. Int J Hydrogen Energy 2010;35:2073e82. [18] Chambers A, Park C, Terry R, Baker K, Rodriguez NM. Hydrogen storage in graphite nanofibers. J Phys Chem B 1998;102:4253e6. [19] Luxembourg D, Flamanta G, B^eche E, Sans JL, Giral J, Goetz V. Hydrogen storage capacity at high pressure of raw and






[25] [26]










purified single-wall carbon nanotubes produced with a solar reactor. Int J Hydrogen Energy 2007;32:1016e23. Tibbetts GG, Meisner GP, Olk CH. Hydrogen storage capacity of carbon nanotubes, filaments, and vapor-grown fibers. Carbon 2001;39:2291e301. Zu¨ttel A, Nu¨tzenadel C, Sudan P, Mauron P, Emmenegger C, Rentsch S, et al. Hydrogen sorption by carbon nanotubes and other cabon nanostructures. J Alloys Compd 2002;330:676e82. Dodziuk H, Dolgonos G. Molecular modeling study of hydrogen storage in carbon nanotubes. Chem Phys Lett 2002;356:79e83. Tarasov BP, Maehlen JP, Lototsky MV, Muradyan VE, Yartys VA. Hydrogen sorption properties of arc-generated single-wall carbon nanotubes. J Alloy Compd 2003;356:510e4. Liu C, Fan YY, Liu M, Cong HT, Cheng HM, Dresselhaus MS. Hydrogen storage in single-walled carbon nanotubes at room temperature. Science 1999;286:1127e9. Bulyarskii SV, Basaev AS. Chemisorption of hydrogen by carbon nanotubes. Technol Phys 2009;54:1612e7. Orinakova R, Orinak A. Recent applications of carbon nanotubes in hydrogen production and storage. Fuel 2011;90:3123e40. Hirscher M, Becher M, Haluska M, Dettlaff-Weglikowska U, Quintel A, Duesberg GS, et al. Hydrogen storage in sonicated carbon materials. Appl Phys A 2001;72:129e32. Panella B, Hirscher M, Ludescher B. Low-temperature thermal-desorption mass spectroscopy applied to investigate the hydrogen adsorption on porous materials. Microporous Mesoporous Mater 2007;103:230e4. Bianco S, Giorcelli M, Musso S, Castellino M, Agresti F, Khandelwal A, et al. Hydrogen adsorption in several types of carbon nanotubes. J Nanosci Nanotechnol 2010;10:3860e6. Rather S, Naik M, Hwang SW, Kimb AR, Nahm KS. Room temperature hydrogen uptake of carbon nanotubes promoted by silver metal catalyst. J Alloys Compd 2009;475:17e21. Valderrama JO. Interaction parameter for hydrogencontaining mixtures in the Peng Robinson equation of state. Fluid Phase Equilibria 1986;31:209e19. Stamatin I, Morozan A, Dumitru A, Ciupina V, Prodan G, Niewolski J, et al. The synthesis of multi-walled carbon nanotubes (mwnts) by catalytic pyrolysis of the phenolformaldehyde resins. Physica E 2007;37:44e8. Lehman JH, Terrones M, Mansfield E, Hurst KE, Meunier V. Evaluating the characteristics of multiwall carbon nanotubes. Carbon 2011;49:2581e602. Ng TY, Rena YX, Liew KM. Adsorption of hydrogen atoms onto the exterior wall of carbon nanotubes and their thermodynamics properties. Int J Hydrogen Energy 2010;35:4543e53.