5039
2009, 113, 5039–5042 Published on Web 03/09/2009
Exchange of Hydrogen Atoms Between BH4 in LiBH4 David T. Shane, Robert C. Bowman, Jr., and Mark S. Conradi* Washington UniVersity, Department of Physics-1105, Saint Louis, Missouri 63130, and RCB Hydrides LLC, Franklin, Ohio 45005 ReceiVed: January 26, 2009; ReVised Manuscript ReceiVed: February 18, 2009
We report here boron-11 and hydrogen NMR measurements of the lifetime of BH4 units in molten LiBH4, with respect to hydrogen exchange. A well-resolved 11B pattern of B-H J-couplings yields a lower limit of 16 ms on the BH4 lifetime at 285 °C. Carr-Purcell-Meiboom-Gill multiple spin-echo measurements, both on 11B and hydrogen, find decay time constants of ∼2 s. Most or all of these decays can be attributed to boron and hydrogen T1 relaxation. Thus, any hydrogen exchange in molten LiBH4 is slower than our ∼1 s time scale. In solid LiBH4, H-exchange is expected to be even slower. This confirms that the diffusion of hydrogen in solid LiBH4 observed recently results from motion of intact BH4 units. Introduction Because of its high hydrogen content (18% w/w), LiBH4 is a potential hydrogen storage material. Typically, this material is envisioned cycling between LiBH4 and LiH (plus elemental boron), still yielding a high storage capacity for hydrogen of 13.8% (w/w).1-3 However, LiBH4 suffers from very slow dehydriding and rehydriding kinetics,4-6 as well as a high dehydriding reaction enthalpy.1,7 Both of these factors result in required operating temperatures2,7 of 400-600 °C, excessive for most applications. Partly to understand the reaction kinetics of LiBH4, recent reports have addressed atomic motions in solid LiBH4. NMR line-narrowing and T1 relaxation8-10 have been reported as have isotopic exchange measurements.11 The isotopic exchange work studied LiBH4 solid at typically 250 °C in contact with D2 gas. Raman spectroscopy found11 isotopically mixed borohydride units (BH4, BH3D, BH2D2, etc.) in partially exchanged material. This was taken as evidence that individual H (or D) atoms are mobile, by means of exchange of H (or D) between nondiffusing (or only slowly diffusing) borohydride units. A recent NMR study8 of solid LiBH4 found thermally activated H line narrowing in the 170-200 °C region, implying a motion rate of hydrogen in reasonable accord with the isotope exchange work. Crucially, the extent of 11B NMR line narrowing rules out the possibility of stationary boron atoms. Thus, H-exchange between BH4 units cannot be the only mechanism of H-transport in the solid; diffusion of BH4 units is responsible for at least some of the 11B and hydrogen line narrowing. Here, we address whether H-exchange is important at all in the diffusion of H in the molten and solid states. In particular, we set a lower limit on the exchange lifetime of BH4 units in molten LiBH4, using NMR observation of B-H J-couplings. We then reason that any such exchange in the solid will be at least as slow as in the melt. Thus, diffusion of intact BH4 units * To whom correspondence should be addressed. Address: Washington University, Department of Physics-1105, One Brookings Drive, Saint Louis, MO 63130, Office phone: 314-935-6418. Laboratory phone: 314-935-6292. Fax: 314-935-6219. E-mail:
[email protected].
10.1021/jp900768t CCC: $40.75
is essentially the only mechanism of boron and hydrogen transport in both liquid and solid LiBH4. Experimental Methods Powder LiBH4 was obtained from Aldrich with specified nominal purity of 90%. The material was stored and handled in a nitrogen-atmosphere glovebag. The powder was loaded into alumina ceramic tubes (Omega) of 6.3 mm outer diameter. Ceramic was used to avoid the reported reactivity12 of glass with molten LiBH4. Indeed, we had one borosilicate tube with liquid LiBH4 rupture and the contents burn. NMR measurements were performed at 27.28 MHz. For boron-11, this corresponds to a field of 2.0 T; a Varian flux stabilizer (inductive pick-up regulator) and 80 MHz 19F NMR were used for short-term and long-term stabilization, respectively, of the Varian iron-core electromagnet. For hydrogen NMR measurements at the same frequency, the field was decreased to about 0.64 T and operated without NMR stabilization. We note that the low operating frequency here allows the B-H J-couplings (∼81 Hz) to be evident at only modest field uniformity (in ppm units). Initial field shimming was performed with a liquid sample of BF3 · 2H2O (Aldrich) in a separate ceramic tube. The 11B signal of molten LiBH4 was strong enough to allow real-time final field shimming. Typical 11B line widths were ∼12 Hz fwhm. The NMR probe held the sample tube in a solenoidal rf coil with a nearby type-T thermocouple, all inside a vacuuminsulated dewar-tube open at both ends. Temperature-regulated air flowed past the sample at 50 L STP per minute. For good field uniformity, the bottom of the sample tube and the meniscus of the molten sample were located below and above the rf coil, respectively. This reduced the field distortions from the magnetic susceptibility13 of the tube and sample. The sample tube was joined by a gastight O-ring connection to a pressure transducer and gas apparatus. The sample was briefly evacuated and then backfilled with 1 atm of H2 gas to suppress boiling.14 Without it, we observed instances of bubble formation, removing some of the sample from the rf coil region. 2009 American Chemical Society
5040 J. Phys. Chem. C, Vol. 113, No. 13, 2009
Letters
Figure 1. NMR spectrum of 11B in liquid LiBH4 at 285 °C at 27.28 MHz. A five-line pattern due to B-H J-couplings with four equivalent hydrogen spins is present, demonstrating slow exchange (or no exchange) of hydrogen atoms between BH4 units.
To minimize the system pressure and reduce the possibility of bubble formation, all of the measurements were made in the 285-300 °C temperature window, slightly above the 268 °C melting point.7 Results and Discussion The 11B NMR spectrum of molten LiBH4 at 285 °C appears in Figure 1. A five line pattern with intensity ratio 1:4:6:4:1 appears, as expected13,15 for boron coupled to four equivalent hydrogen nuclear spins (spins one-half). That is, the five resonance lines are due to BH4 with H spin orientations + + + +, + + + -, + + - -, + - - -, and - - - -, respectively (and permutations thereof). The existence of 5 sharp resonances clearly places this system in the slow-exchange or no-exchange limit.15,16 If the exchange were rapid, the five lines would collapse into a single resonance, due to time-averaging.16 We note the melting transition7 is obvious, with the line width decreasing markedly and J-couplings becoming evident. The J-coupling strength is 81 ( 1 Hz, in agreement with a report of LiBH4 in alkaline methanol solution at room temperature.17 The spectrum of Figure 1 displays lines with a line width fwhm (full-width at half of maximum) of 9.5 Hz. If all of the width were due to H-exchange, the line width would obey15,16
2π(∆f)fwhm /2 ) (∆ω)hwhm ) 1/2τ*
(1)
where τ* is the BH4-unit lifetime against exchange of any of its four hydrogen atoms. The term (∆ω)hwhm is in radians/s. We note the origin of the factor of 2 on the right side of eq 1: half of the H-exchanging events are spin-invisible, exchanging one H for another of the same spin-state. Because the measured line width has an unknown contribution from field inhomogeneity, the observed line width establishes a lower limit of 16 ms on the exchange lifetime τ*, through eq 1. The CPMG (Carr-Purcell-Meiboom-Gill) pulse sequence15,16 (90y-t-180x-2t-180x-2t-etc) was used to remove the effects of field nonuniformity and to place a more restrictive (larger) lower limit on the BH4 lifetime τ*. We used t values that were short enough to avoid echo decay due to diffusion through field
Figure 2. CPMG spin-echo train from 11B in molten LiBH4 at 285 °C with pulse spacing t ) 15 ms. The decay time constant TC is 2.7 s, demonstrating that H-exchange is very slow here.
inhomogeneities16 and yet long enough to be in the strong collision limit.16 Consider the effect on 11B spin precession phase φ due to changes in the spin state (m ) 1/2 or -1/2) of a hydrogen nuclear spin, whether through a hydrogen T1 process or H-exchange event. A spin changing event causes a sudden change in the 11B frequency of ∆f ) ( 81 Hz. This results in a dephasing of the 11B at the time of the next echo by angle φ = 2π∆ft, with the exact answer depending on the timing of the event relative to the pulse sequence. That is, an event occurring at or near a 180° pulse generates maximum dephasing (about 2π∆ft) while an event occurring at or near an echo peak results in nearly zero dephasing. The important point is that, for ∆ft J 1, the dephasing angle is usually large, justifying the strong collision approximation.16 That is, any BH4 unit that suffers a hydrogen spin-changing event is essentially removed from contributing to all further echoes. CPMG echo trains on 11B at 285 °C with t ) 15 ms display an exponential decay with decay time constant TC of 2.7 s, as displayed in Figure 2. As this value of t is slightly longer than 1/(81 Hz), it is well into the strong collision regime. The decay is slow enough that we ask whether it can be explained entirely by 11B true T2 processes (in a liquid, we expect18 T 2B ) T 1B) and T1H-driven changes of H spin-state, without any hydrogen exchange. Because each boron spin is J-coupled to four hydrogen spins, sudden frequency changes of 11B from hydrogen T1 processes occur stochastically with rate 4W, where W is the probability per time of an individual hydrogen making a latticedriven spin transition. As shown elsewhere16 for spins one-half, 1/T 1H is equal to 2W. Combining the two sources of decay, we predict in the strong collision limit the echo envelope to decay at rate 1/TC,
1/TC ) 1/T B1 + 4W ) 1/T B1 + 2/T H1
(2)
The relaxation times T 1B and T 1H were measured at 285 °C and found to be 8.2 and 9.0 s ((10%), respectively. Using these values in eq 2, the predicted echo decay time constant TC in the absence of exchange is 2.9 s, in excellent agreement with the observed value of 2.7 s. Thus, it appears that all or most of the CPMG decay results from boron T2 and hydrogen T1 processes; any H-exchange is slow on the time scale of seconds.
Letters
J. Phys. Chem. C, Vol. 113, No. 13, 2009 5041
In the above, the T 1H was determined at 27.28 MHz so at a lower field (about 0.64 T) than the 11B CPMG measurements (2.0 T). However, in a liquid (fast correlation limit) relaxing primarily by dipole interactions, T1 is expected to be independent of field.16,18 For small enough values of the CPMG pulse spacing t, the strong collision limit should no longer apply. In the limit of very small t, H spin-changing events should not contribute to the echo train decay.19 For t values of 1, 2, 5, 15, and 25 ms, we measure at 285 °C echo decay time constants TC of 4.0, 4.0, 2.9, 2.7, and 2.3 s. Thus, the variation of TC with pulse spacing t follows the expected trend. At t ) 1 ms, we would expect the echo train to be insensitive19 to hydrogen spin-state changes (∆ft , 1) and TC ) T 1B. Clearly, the echo envelope decays faster than T1B, suggesting there may be an additional and unknown weak source of decay. We also performed hydrogen CPMG measurements at 285 °C. With t ) 15 ms, the decay time constant was 2.8 s. Each hydrogen spin is J-coupled to a single 11B; here we neglect the smaller fraction of boron-10 present, 19%. For dipole-dipole driven relaxation of spin-3/2 11B, the lattice induces transitions at rates 3W, 4W, and 3W for the transitions 3/2-1/2, 1/2-(-1/ 2), and -1/2-(-3/2), respectively. It is easy to show using a master rate equation16 for the populations of the four levels that 1/T1 ) 2W. Thus, the probability per time of a 11B spin making a transition is 3W (in m ) ( 3/2) and 7W (in m ) ( 1/2). Because the spin populations are essentially equal, the average rate of transitions is 5W. Thus, the hydrogen CPMG echo decay rate in the strong collision limit is
1/TC ) 1/T H1 + 5W ) 1/T H1 + 2.5/T B1
(3)
With the values of T1 for hydrogen and boron listed above, the prediction for TC in the absence of H-exchange is 2.4 s, in reasonable accord with the measured value. This confirms our assertion that all or most of the observed decay can be understood without invoking H-exchange; H-exchange must be slow on the time scale of seconds. The establishment of a pool of hydride anions (H-) would be expected to increase the rate of H-exchange and decrease the lifetime τ* of BH4 units, following H* + BH4 T H + BH3H*. Thus we performed 11B NMR on a sample of LiBH4 containing approximately 3% (w/w) of LiH. The melting point of LiH is much higher than our temperatures and we do not know the solubility of LiH in molten LiBH4. Nevertheless, this sample should have a much larger concentration of hydride anions in the melt than for nominally pure LiBH4. We believe any undissolved LiH sinks to the bottom of the liquid and is of no further concern. The 11B spectrum of this sample displayed the familiar 1:4: 6:4:1 J-multiplet of Figure 1, placing the system in the slowexchange regime. The 11B CPMG decay at t ) 15 ms was exponential with time constant TC ) 2.3 s, very close to that of the LiBH4 sample without added LiH. We conclude that H-exchange in molten LiBH4 remains slow, even in the presence of hydride anions (albeit of unknown concentration). To probe the exchange lifetime τ* to much longer times, we prepared a well-stirred powder mixture of nominally equal amounts of LiBH4 and LiBD4 (Isotec). The sample was then heated to 285 °C and examined by 11B NMR. The aim was to see the spectrum evolve in time from a 50-50 superposition of the spectra of BH4 and BD4 (deuterium is spin-one with a smaller moment than the proton) to a sum of signals from BH4,
Figure 3. Spectra of 11B. Heavy curve is data from the molten LiBH4-LiBD4 mixture at 285 °C. The light curve is simulation assuming full isotopic mixing. The dotted curve simulates the case of no isotopic mixing. The simulations assume a 60:40 H:D ratio and use 15 Hz fwhm Lorentzian broadening.
BH3D, BH2D2, BHD3, and BD4, in statistical proportions. The 11 B spectra of each isotopomer is given in ref 20. However, by the time the sample was completely melted (30 min after reaching temperature), the sample appeared to be already fully statistically equilibrated: (1) The spectrum showed minimal changes after an additional 60 m in the molten state, and (2) the experimental spectrum was fit well by a simulated spectrum with the statistical distribution of isotopomers (BH4, BH3D,..., BD4). The spectrum of molten LiBH4 after 1.5 h at 285 °C is displayed in Figure 3 as the heavy curve. A simulated spectrum assuming full isotopic mixing (so BH4, BH3D,..., BD4) appears as the light curve. The H:D ratio was taken to be 6:4 (our LiBD4 was found to have substantial H content) and Lorentzian line broadening of 15 Hz fwhm was used to represent the field inhomogeneity. The dotted curve in Figure 3 simulates the case on no isotopic mixing, so only BH4 and BD4 are present. Clearly, full isotopic mixing does a much better job of describing the measured spectrum. Evidently, the actual exchange lifetime τ* is between the time scales of seconds (from the CPMG data) and ∼30 m. All of the measurements above were performed on molten lithium borohydride. Because the liquid is more disordered than the solid and appears at higher temperatures, we believe that hydrogen exchange in solid LiBH4 should be at least as slow as in the melt. Thus, H-exchange in solid LiBH4 is slower than one event per second (per BH4 unit) at all temperatures. A previous NMR study of atomic motions in solid LiBH4 reported8 diffusion of H nuclear spins with mean time between jumps of about 3 µs at 250 °C. The jump time inferred from the macroscopic H,D-exchange measurements11 was approximately 0.4 µs, in reasonable agreement. These are so much faster than the limits placed above on H-exchange that virtually none of the observed H diffusion can result from H-exchange. Instead, the dominant motion must be diffusion of intact BH4 units. This agrees with the observation of nearly complete 11B line narrowing in the same temperature range as the hydrogen line narrowing.8 A previous study reported isotopic scrambling in solid LiBH4 partially exchanged with D2 gas.11 Clearly, according to the present results, this cannot be due to diffusion of individual H
5042 J. Phys. Chem. C, Vol. 113, No. 13, 2009 or D anions via H,D-exchange, as suggested. Instead, we propose that diffusion of intact borohydride anions occurs to distribute the isotopes (as BH4 and BD4) through the solid. Any local isotope scrambling process, such as slow H,D-exchange as allowed by the present results, will succeed in generating all the isotopomers (BH3D, etc), as observed by Raman spectroscopy.11 Conclusions Boron-11 measurements in molten LiBH4 show a wellresolved pattern of J-coupling multiplets. The narrow width of about 9.5 Hz fwhm yields a lower limit to the lifetime of BH4 against H-exchange of 16 ms at 285 °C. Multiple echo experiments on 11B and on hydrogen using the CPMG sequence show decay time constants of about 2 s. Most or all of the observed decay can be assigned to the 11B T 1B and to the T1H of the hydrogen nuclear spins. Thus, in liquid LiBH4, exchange of hydrogen atoms between BH4 units is slow on the time scale of seconds. The liquid has greater disorder than the solid. Hence, we expect that any H-exchange in solid LiBH4 is at least as slow as in the melt. Thus, the previously reported diffusion of hydrogen in hot solid LiBH4 (170-250 °C), with hopping times of 10 µs and shorter, has essentially no contribution from H-exchange. Instead, the H are carried by diffusing intact BH4 units. Acknowledgment. The authors appreciate helpful discussions with T. J. Udovic, J. G. Kulleck, and J. J. Vajo concerning boiling, foaming, and bubbles in molten LiBH4. Research funding from the U.S. Department of Energy, Basic Energy Science, under Grant DE-FG02-05ER46256 is acknowledged. Robert Corey is thanked for 11B measurements at the outset of the present work.
Letters References and Notes (1) Vajo, J. J.; Skeith, S. L.; Mertens, F. J. Phys. Chem. B 2005, 109, 3719–3722. (2) Orimo, S.; Nakamori, Y.; Kitahara, G.; Miwa, K.; Ohba, N.; Towata, S.; Zu¨ttel, A. J. Alloys Compd. 2005, 404-406, 427–430. (3) Nakamori, Y.; Orimo, S.-I. J. Alloys Compd. 2004, 370, 271–275. (4) Barkhordarian, G.; Klassen, T.; Dornheim, M.; Bormann, R. J. Alloys Compd. 2007, 440, L18–L21. (5) Gross, A. F.; Vajo, J. J.; Van Atta, S. L.; Olson, G. L. J. Phys. Chem. C 2008, 112, 5651–5657. (6) Fang, Z.-Z.; Kang, X.-D.; Wang, P.; Cheng, H.-M. J. Phys. Chem. C 2008, 112, 17023–17029. (7) Mauron, P.; Buchter, F.; Friedrichs, O.; Remhof, A.; Bielmann, M.; Zwicky, C. N.; Zu¨ttel, A. J. Phys. Chem. B 2008, 112, 906–910. (8) Corey, R. L.; Shane, D. T.; Bowman Jr, R. C.; Conradi, M. S. J. Phys. Chem. C 2008, 112, 18706–18710. (9) Matsuo, M.; Nakamori, Y.; Orimo, S.-I.; Maekawa, H.; Takamura, H. Appl. Phys. Lett. 2007, 91, 224103. (10) Skripov, A. V.; Soloninin, A. V.; Filinchuk, Y.; Chernyshov, D. J. Phys. Chem C 2008, 112, 18701–18705. (11) Borgschulte, A.; Zu¨ttel, A.; Hug, P.; Racu, A.-M.; Schoenes, J. J. Phys. Chem. A 2008, 112, 4749–4753. (12) Zhang, Y.; Zhang, W.-S.; Fan, M.-Q.; Liu, S.-S.; Chu, H.-L.; Zhang, Y.-H.; Gao, X.-Y.; Sun, L.-X. J. Phys. Chem. C 2008, 112, 4005–4010. (13) Pople, J. A.; Schneider, W. G.; Bernstein, H. J. High-Resolution Nuclear Magnetic Resonance; McGraw-Hill: New York, 1959. (14) Kostka, J.; Lohstroh, W.; Fichtner, M.; Hahn, H. J. Phys. Chem. C 2007, 111, 14026–14029. (15) Levitt, M. H. Spin Dynamics; Wiley: Chichester, U.K., 2008. (16) Slichter, C. P. Principles of Magnetic Resonance; Springer; New York, 1980. (17) Than, C.; Morimoto, H.; Anders, H.; Williams, P. G. J. Labelled Compounds Radio. 1996, 38, 693–711. (18) Abragam, A. The Principles of Nuclear Magnetism; Oxford: London, 1961. (19) Ansermet, J. Ph.; Slichter, C. P.; Sinfelt, J. H. J. Chem. Phys. 1988, 88, 5963–5971. (20) Smith, B. E.; James, B. D.; Peachey, R. M. Inorg. Chem. 1977, 16, 2057–2062.
JP900768T