The First-Cycle Electrochemical Lithiation of Crystalline Ge: Dopant

Nov 22, 2011 - Chia-Yun Chou , Myungsuk Lee , and Gyeong S. Hwang. The Journal of .... McDowell, Lee, Harris, Korgel, Wang, Nix, and Cui. 2013 13 (2),...
0 downloads 0 Views 1MB Size
Letter pubs.acs.org/JPCL

The First-Cycle Electrochemical Lithiation of Crystalline Ge: Dopant and Orientation Dependence and Comparison with Si Maria K. Y. Chan,† Brandon R. Long,‡ Andrew A. Gewirth,*,‡ and Jeffrey P. Greeley*,† †

Center for Nanoscale Materials, Argonne National Laboratory, Argonne, Illinois 60439, United States Department of Chemistry, University of Illinois at Urbana−Champaign, Urbana, Illinois 61801, United States



ABSTRACT: We use first principles Density Functional Theory (DFT), cyclic voltammetry (CV), and Raman spectroscopy to investigate the first-cycle electrochemical lithiation of Ge in comparison with Si − both high-capacity anode materials for Li ion batteries. DFT shows a significant difference in the dilute solubility of Li in Si and Ge, despite similarities in their chemical and physical properties. We attribute this difference to electronic, as opposed to elastic, effects. CV and Raman data reveal little dopant dependence in the lithiation onset voltages in Ge, unlike in Si, due to a smaller energy difference between dilute Li insertion in p-type Ge and bulk germanide formation than the corresponding difference in Si. Finally, we show that there is no orientation dependence in lithiation onset voltages in Ge. We conclude that approaches other than microstructuring are needed to fabricate effective electrodes able to take advantage of the higher rate capability of Ge compared to that of Si. SECTION: Energy Conversion and Storage

S

ilicon and germanium have the highest gravimetric and volumetric capacities among all currently known anode materials for Li ion batteries.1 The lithiation properties of various forms of Si, including micro- and nanocrystalline particles, amorphous particles and films, and micro- or nanostructures, have been under intense study (see refs 2, 3, and 4 for review). The electrochemical performance of Ge thin films,5,6 micro-7 and nanocrystals,5 and nanowires8 has also been investigated. Si and Ge are isoelectronic, and both exist in the tetrahedrally-bonded diamond structure, with the Ge lattice parameter exceeding that of Si by only 5%. Because of these similarities, one might expect crystalline Si and Ge to have similar electrochemical properties upon lithiation. Indeed, at high temperatures (∼400 °C), both Si9 and Ge10 lithiate to form similar crystalline LixSi/LixGe zintl phases in several voltage steps. At room temperature, however, the formation of most5,6,8,11 such phases is suppressed, and lithiation leads first to amorphization and eventually to the formation of the isostructural crystalline Li15Si412 or Li15Ge45,7,13 phases. Recently, it was discovered that the lithiation of crystalline Si exhibits strong dopant14 and orientation15 dependence, properties that may be useful for electrode design. We used cyclic voltammetry and in situ Raman spectroscopy to investigate the effects of dopants on lithiation in Si wafers, and we found that the voltage at which Li is first inserted is significantly larger in p-type compared to undoped and n-type Si. 14 Density Functional Theory (DFT) calculations showed the effect to be consistent with an electronic explanation.14 Microstructured Si showed strong anisotropy in lithiation, depending on crystalline orientation, a result that was also confirmed with cyclic voltammetry and Raman spectroscopy.15 Comparable anisotropy has also been found in Si nanowires.16,17 © 2011 American Chemical Society

In this work, using a combination of DFT calculations, cyclic voltammetry, and Raman spectroscopy, we compare Li insertion into crystalline Si and Ge, explore differences between the two materials, and investigate whether the aforementioned dopant and orientation effects also apply to Ge. We conclude that Ge is not as sensitive to these effects as is Si, suggesting that alternative strategies are needed to manipulate and enhance the properties of microstructured Ge electrodes. We begin with a DFT investigation of the energetics of dilute Li insertion. A single Li is inserted into computational cells of 8−1000 Si or Ge atoms. The Li insertion energies, defined as

(1)

where SC = (Si, Ge) are shown as a function of Li density in Figure 1a. We see that Einsertion in undoped Ge becomes negative at sufficiently low Li densities but is always positive in undoped Si. We suggest two possible reasons for this difference; namely, lower lattice strain (elastic effect) and smaller band gap (electronic effect). The Ge lattice parameter is larger than that of Si by 5%, and its bulk modulus is smaller by 23%,18 suggesting that the lattice strain induced by the volume expansion with Li insertion in Ge, and the accompanying energetic penalty, could be smaller in Ge than in Si. Alternatively, as discussed in ref 14, in undoped Si (or Ge), the accompanying electron when a Li+ ion is inserted is accommodated in the conduction band, which adds an energetic penalty for Einsertion. Since the conduction band Received: October 27, 2011 Accepted: November 22, 2011 Published: November 22, 2011 3092

dx.doi.org/10.1021/jz201432d | J. Phys. Chem.Lett. 2011, 2, 3092−3095

The Journal of Physical Chemistry Letters

Letter

Table 1. (a) Decomposition of Einsertion into Elastic and Electronic Components and (b) Comparison of Band Gaps and Einsertion in PBE and HSE (All Numbers in eV) (a) Breakdown of Einsertion SC Si

Ge

Figure 1. Li insertion energies for a single Li into computational cells of 8−1000 Si or Ge atoms, (a) without dopants and (b) with a single dopant atom. The density is given by the inverse of the computational cell volume and is equal for Li and dopant atoms in (b).

property

8-atom cells

64-atom cells

Eelastic 0.16 0.16 Eelectronic 0.36 0.18 Einsertion 0.52 0.34 Eelastic 0.24 0.15 Eelectronic −0.02 −0.19 Einsertion 0.22 −0.04 (b) Dependence on Exchange-Correlation Functional 8-atom cells

SC

minimum is lower (relative to the valence band maximum) in Ge than in Si, this energetic penalty would be lower in Ge than in Si. In order to separate the elastic and electronic effects, we also calculate the energy of pure Si/Ge with the lattice distorted to what it would be with Li inserted; upon fully relaxing the Si/Ge computational cells with Li inserted, we remove the Li, freeze the atoms, and calculate the total energy of the distorted Si/Ge lattice. We can then estimate the elastic and electronic contributions to Einsertion according to

Si

Ge

property band gap Einsertion (undoped) Einsertion (p-type) (B-doped) band gap Einsertion (undoped) Einsertion (p-type) (Ga-doped)

64-atom cells

expt

PBE

HSE

PBE

HSE

1.1

0.6 0.53 −0.95

1.3 0.63 −1.10

0.6 0.34 −0.58

1.3 0.42 −0.74

0.7

0.0 0.22 −0.88

0.7 0.35 −0.95

0.0 −0.04 −0.47

0.7 0.05 −0.54

64-atom cells in Table 1b. The differences in Einsertion between PBE and HSE are between 0.07 and 0.16 eV in all cases, indicating that the errors are small. With HSE, the value of Einsertion is increased for the undoped systems and decreased for the p-type system, as expected from the opening of the band gap in HSE, as compared to in PBE. The far lower Einsertion, and hence higher solubility, of Li in Ge as compared to Si may enhance nucleation events during the initial phase of lithiation. This result, together with the higher diffusivity in Ge, mentioned above,21 suggests that Ge may achieve higher lithiation rate capability than Si. We also investigate using DFT the effects of dopants on Li insertion into Ge. Li insertion energies into doped Si and Ge cells with 8−1000 atoms, with a single dopant atom of B/Ga (p-type Si/Ge) or P/Sb (n-type), are calculated and shown in Figure 1b. We attempt Li insertion at both the nearest and furthest site from the dopant atom in the computational cell, and the lower energy configuration is considered. The dopant effects are similar in Ge and Si, with p-type Ge having lower Einsertion than undoped and n-type Ge, which in turn have similar Einsertion to each other. The dopant density dependence of Einsertion is also similar between the two materials, except with a smaller difference between p-type and undoped/n-type Ge, resulting from the smaller band gap in Ge, than in Si. Since the voltage is given by the reaction energy divided by the charge, 22 from Figure 1b, we infer that dilute Li insertion into p-type Ge occurs at energies of −0.6 to −0.4 eV, corresponding to a voltage of 0.4−0.6 V at dopant concentrations of