The valence of Ti and Ni is close to zero and invariant during hydrogen cycling. None of the metals enter substitutionally or interstitially into the crystalline lattice ...
Spatial Configurations of Ti- and Ni- Species Catalyzing Complex Metal Hydrides: X-Ray Absorption Studies and First-Principles DFT and MD Calculations A. Yu. Ignatov1,6, J. Graetz2, S. Chaudhuri3, T. T. Salguero4, J. J. Vajo4, M. S. Meyer5, F. E. Pinkerton5, and T. A. Tyson6 1
Center for Biophysics at the NSLS, Case Western Reserve Univ., Brookhaven Natl. Lab, Upton, NY 11973 2 Department of Energy, Science, and Technology, Brookhaven Natl. Lab, Upton, NY 11973 3 Department of Chemistry, Brookhaven Natl. Lab, Upton, NY 11973 4 HRL Laboratories, LLC, Malibu, CA 90265 5 Materials and Processes Lab, General Motors Research and Development Center, Warren, MI 48090 6 Department of Physics, New Jersey Inst. of Tech., Newark, NJ 07102
Abstract. We have performed Ti K-edge EXAFS and XANES measurements on 4 and 3 wt% TiCl3-activated NaAlH4 and (LiBH4+0.5MgH2) and Ni K-edge measurements on 3 and 11 wt% NiCl2-activated (LiBH4+0.5MgH2) and (Li3BN2H8) - prospective hydrogen storage materials. The valence of Ti and Ni is close to zero and invariant during hydrogen cycling. None of the metals enter substitutionally or interstitially into the crystalline lattice of the initial or final products. For the Ti- activated NaAlH4 and (LiBH4+0.5MgH2), amorphous TiAl3 and TiB2 alloys are formed, which are almost invariant during cycling. The Ni doped (LiBH4+0.5MgH2) initially forms amorphous Ni3B, which is partly converted to amorphous Mg2NiHy upon hydrogen loading. Local structure around Ti(Ni) atoms is expressed in terms of a cluster expansion and the interatomic distances, coordination numbers and Debye-Waller factors are determined for competitive structural models. For Ti-activated NaAlH4 the models are elaborated by Ti K-edge XANES, which are interpreted in terms of single-electron multiple scattering calculations. Structural properties and phase stability of hypothetical hydrogenated TiAl3 as well as several products of the decomposition reaction are determined from density functional theory calculation. First-principles molecular dynamics simulations of surface diffusion and chemical reactivity imply that the formation of a few monolayers of TiAl3 on the surface may be responsible for the significant increase in the reaction rate. Keywords: Hydrogen storage materials, alanates, borohydrides, nanocatalysts, EXAFS, XANES, DFT, and MD PACS: 71.20. Ps, 78.70 En, 61.10 Ht, 78.70 Dm
INTRODUCTION Much research has focused on studying complex metal hydrides suitable for reversible hydrogen storage at moderate temperatures and hydrogen pressures [1-7, 10-12]. Bogdanovic et al. [1] demonstrated that adding ~2 wt% of Ti increased the rate of desorption and absorption of hydrogen in sodium alanate (~5.6 wt%): NaAlH4 ↔ 1/3Na3AlH6+2/3Al+H2 ↔ NaH+Al+3/2H2
(1)
Knowledge of spatial configurations of the metal dopant atoms and their changes during cycling is necessary to develop a microscopic understanding of energy landscapes and engineering new catalysts. Xray diffraction, inelastic neutron scattering, and other long-range order techniques have failed to identify the
structure of the local species. First-principles simulations [4] were not conclusive either due to vast parameter space of size, shape, and composition. Being local, element specific, and sensitive to impurities, x-ray absorption spectroscopy (XAS) is well suited to identify the local configuration of the active species. Preliminary XAS studies [5,6] determined that Ti is coordinated by Al and is zerovalent. In this paper, we report on a combined experimental (EXAFS and XANES) and theoretical (DFT and MD) study of Ti-activated alanate. EXAFS results on Ti- and Ni-activated LiBH4+0.5MgH2 [Eq. (2), H capacity ~7.8 wt%] and Ni-activated Li3BN2H8 [Eq. (3), capacity ~11 wt% [7]] are presented. LiBH4+0.5MgH2 ↔ LiH + 0.5MgB2+2H2 Li3BN2H8 → Li3BN2+4H2
(2) (3)
EXPERIMENTAL The preparation of 4 wt% Ti-doped NaAlH4 was described elsewhere [5]. 3 wt% TiCl3-activated LiBH4+0.5MgH2 was cycled four times. Results on samples after the second hydrogenation and fourth desorption are presented. 3 wt% NiCl2-activated LiBH4+0.5MgH2 was cycled two times. 11wt% NiCl2activated Li3BN2H8 was quenched at the halfway (- 5 wt%) and complete hydrogen desorption (- 11wt%). Ti and Ni K-edge spectra were collected at beamlines X19a and X9b at the NSLS using a Si(111) double crystal monochromator. A reflecting mirror was used to suppress the higher order harmonics. Spectra were acquired in fluorescence yield using a PIPS detector or a 13-element Ge detector (Canberra) with an energy resolution of 260 eV for the diluted samples. Due to low Ti/Ni concentrations, the spectra were not corrected for “self-absorption.” From 4 to 12 spectra were collected to assure reproducibility. All samples were kept at ~40 K. The standard procedures for experimental data reduction implemented in EDA [8] and VIPER [9] were used to obtain k2χ(k) EXAFS. The resulting fits are shown in Fig. 1 and the obtained structural parameters are summarized in Table 1.
(2S) models. Though the agreement factor, Ra, of the 2S model is less than that of the 1S model, the corresponding reduced agreement factors are opposite: Ra(2S)/ν2 > Ra(1S)/ν1 meaning that the 2S model is statistically unfavorable for the given experimental data set. Analysis of χ(k) spectra indicates that ~5±3 H atoms may be located at 1.78±0.04 Å. The issues will be elaborated by XANES and DFT presented below. In Ti-activated LiBH4+0.5MgH2 the Ti-species formed after cycling resembles amorphous TiB2 (aTiB2) with reduced Ti-B and Ti-Ti coordination. This suggests either that LiBH4 and a-TiB2 compete for boron, or repeated cycling promotes nano-dispersion. As a whole, similar to Ti-activated NaAlH4, the aTiB2 is almost invariant during the cycling. Ni-based species in LiBH4+0.5MgH2 do change upon the cycling. Amorphous Ni3B (a-Ni3B) seen in the desorbed sample is partly converted to amorphous MgNi2Hy (a-MgNi2Hy) in the hydrogenated phase. TABLE 1. Local structure parameters of Ti- and Ni-based species obtained from EXAFS fit. S0 =0.82, 0.93, and 0.96 for NaAlH4, LiBH4, and both Ni-activated samples, as derived from the fit of TiAl3, TiB2, and Ni2B standards (not listed). Ra=1/(Mν) Σ (χ exp -χ calc)2, Rν=Ra /ν where M is the number of data points and ν is the number of degrees of freedom in the fit. Typical accuracy for N and σ2 is ~10%.
8
10 k (Å-1)
12
4wt% of Ti in NaAlH4, Eq. (1)
Des., Fig. 1(a)
6
Ti-Al Ti-Al Ti-Ti
2.73±.02 2.90±.02 3.81±.04
3.5* 7.0* 1.8
4.2 8.9 5.2
2S, 1.8
Ti-Al
2.82±.01
10
18.2
1.4
Des
4
Rνx10
Ti-B Ti-Ti
2.38±.02 3.02±.01
7.6 4.0
9.5 8.8
Ti-B Ti-Ti
2.38±.02 3.03±.01
10.3 4.7
8.2 8.3
Ni-B Ni-Ni Ni-Ni
2.09±.03 2.51±.02 2.74±.02
3.0 7.3 3.2
15.1 8.9 9.1
60% of the Des (Ni-B,Ni fixed as above) + 40% of Hyd 8.2 6.2 Ni-Mg 2.79±.01 3.8 4.5 Ni-Ni 4.61±.02
14
FIGURE 1. Comparison of the BFT data (triangles) with models, Table 1. Models having lowest Ra are shown by solid lines. Alternative models are shown by dots.
In Ti-activated NaAlH4, the Ti species is similar to amorphous TiAl3 with local structure about the Ti given by a cluster expansion of Ti-Hx-Al10-Ti2-... The structure is nearly invariant during H cycling. A Ti-Al contribution was fit with singe-shell (1S) and two-shell
Ni-B Des., Fig . 1(d)
-2
2
Hyd
2LiNH2+LiBH4
3
σ210 ,Å
Des
(d)
R, Å
N
Halfway
0
Pair
Hyd, 1(c)
11wt% of Ni in
3wt% of Ti in LiBH4 Eq. (2)
3wt% of Ni in LiBH4+0.5MgH2
2 (c)
3wt% of Ni in LiBH4 +0.5MgH2
3wt% of Ti in LiBH4
(b)
11wt% of Ni in 2LiNH2+LiBH4 , Eq. (3)
4
4wt% of Ti in NaAlH4
(a)
Ni-(B,Ni)
k2 χ(k)
6
Ni-(B,Ni) Ti-(B,Ti) Ti-Al pairs Ni-(Mg,Ni)
RESULTS AND DISCUSSION
2.08±.03
2.4
8.2
2.50±.02
7.8
8.1
Ni-Ni
2.70±.02
2.8
7.7
Ni-B
Ni-Ni
2.10±.03
2.8
7.8
Ni-Ni
2.49±.02
5.0
5.9
Ni-B Ni-Ni Ni-Ni
2.09±.03 2.49±.02 2.71±.02
2.6 7.6 2.9
12.8 12.5 14.1
3
2S, 1.9#
2.3#
*N1=2N2. # Ni-B parameters were constrained for the Rν estimate.
Absorption (arb. units)
a-Ni3B is formed in 11wt% Ni-activated Li3BN2H8 during ball milling. When the sample is halfway decomposed the Ni-Ni shell becomes more ordered and upper shells are formed in the fully desorbed sample that appears to be only slightly different from the crystalline Ni3B. Special care needs to be taken not to underestimate the Ni-Ni occupation numbers that result from non-Gaussian Ni-Ni PDF of Ni3B. (a)
|
|
(b)
(c) (d)
1.0
(b) Computed: 1/2 ML TiAl3 at Al surface (solid) TiAl3 (dots). I4/mmm, 118 atoms
0.5
(c) Computed: 0 (dots), 5 (solid), and 8 H atoms (dashed line) in the T-position (d) Exp.: TiAl3 (dots) vs. 0.8TiAl3+0.2TiH2 (solid)
4960
5000
5040 Energy (eV)
5080
FIGURE 2. Three scenarios [(b), (c), and (d)] reproducing observed trends of changes in Ti K-edge XANES (a) as one goes from a reference TiAl3 (dots) to the Ti-based species (solid). Signatures of I4/mmm (that will not appear in cubic, Pm-3m) marked with bars.
Ab-initio XANES calculations have been performed [10]. XANES spectra are interpreted in terms of single-electron multiple scattering calculations for a large cluster of atoms. Close to zero Ti valence is validated. Simulations confirmed the EXAFS results: Ti does not enter substitutionally or interstitially into the perfect or fragmented lattice. XANES narrowed down the choice of structural models put forward by EXAFS. Experimental trends in the data [Fig 2(a)] are reproduced reasonably well assuming that 0.5ML thick TiAl3 clusters are formed at the Al surface [Fig 2(b)]; or ~ 5-8 H atoms occupy the T-position [Fig 2(c)]; or TiAl3 and TiH2 phases are mixed together in the ratio ~4:1 [Fig 2 (d)]. Structural properties and phase stability of hydrided Ti-Al alloys, NaAlH4 and Na3AlH6, and several products of the decomposition reaction were determined at 0 K within the LDA approximation to DFT using the LAPW method [10]. The calculations reveal that partial decomposition of NaAlH4 accompanied by formation of TiAl alloys is preferred to Ti substitution for Na, in good agreement with our XAS finding. Of possible Ti products, TiAl3 is the most favorable thermodynamically, while the hydrides
(TiAl3Hx, x=1,8) have positive formation energy ruling out the scenario shown in Fig. 2(c). TiAl3 in H2 (gas) environment is only ~ 10 kJ/(mol H2) more stable than TiH2 and Al (fcc) implying a possible coexistence of TiAl3 and TiH2 at elevated temperatures and in presence of Al and H2. Understanding the local structure of the Ti-based species guided a recent theoretical study [11]. Trends of hydrogen chemisorption for different a-TiAl3 surface models were investigated. With 0.25 ML coverage and surface (sub-surface) Ti sites, the overall reaction is exothermic (endothermic) Ech~ -0.6(+0.2) eV, respectively. Hydrogen chemisorption stabilizes the Ti atoms on the Al(001) surface more than the subsurface Ti-atoms: it “pumps” Ti atoms to the surface preventing formation of crystalline TiAl3 even after 100 cycles [6]. Car-Parinello MD simulations are used to study the stability and the diffusion rate of different AlnH3n clusters. AlH3 is the most probable cluster. The manifold increase in the rate of hydrogenation reaction is attributed to three processes [11]: (i) catalytic chemisorption of molecular hydrogen on a-TiAl3 (ii) formation and (iii) diffusion of AlH3 that transports both Al and H into the NaH lattice to form first Na3AlH6 and then NaAlH4. Therefore vacancy formation in NaAlH4 may provide an alternative mechanism of the enhanced desorption kinetics [12]. In conclusion, this work illustrates a potential of combined approach relying on EXAFS, XANES, and first-principles DFT and MD calculations in order to reveal atomic-level structure of Ti-based species in NaAlH4. For the first time we determined the local structure of Ti- and Ni-species in new promising hydrogen storage materials: LiBH4+0.5MgH2 and Li3BN2H8. Contrary to the Ti- species, the Ni-based species are not invariant during the hydrogen cycling: a-Ni3B is partly converted to a-Mg2NiHy upon hydrogenation. This work was supported by U.S. DOE Contract No. DE-AC02-98CH10886.
REFERENCES 1. 2. 3. 4. 5. 6.
B. Bogdanovic et al., J. Alloys. Comp. 253/254, 1 (1997). V. Balema, et al., J. Alloys. Comp. 329, 108 (2001). D. Sun, et al., J. Alloys. Comp. 337, L8 (2002). J. Iniguez, et al. Phys. Rev. B 70, 054103 (2004). J. Graetz, et al., Appl. Phys. Lett. 85, 500 (2004). M. Federhoff, et al., Phys. Chem. & Chem. Phys. 6, 4369 (2004). 7. F. E. Pinkerton et al., J. Phys. Chem. B 109, 6 (2005). 8. A. Kuzmin, Physica B 208-209, 175 (1995). 9. K. Klementiev, VIPER, www.desy.de/~klmn/viper.html 10. A. Yu. Ignatov, et al., (submitted). 11. S. Chaudhuri, et al., J. Am. Chem. Soc. 128, 11404 (2006). 12. C. M. Araujo, Phys. Rev. B 72, 165101 (2005).
