The 11th International Conference on Inelastic X-ray scattering

High-Energy X-ray Compton Scattering for Non-destructive and Quantum Characterization in Batteries

*Kosuke Suzuki1, H. Sakurai1, N. Tsuji2, Y. Sakurai2, Y. Orikasa3, Y. Uchimoto4, H. Hafiz5,6, B. Barbiellini7,6 and A. Bansil6

* [email protected]

1Faculty of Science and Engineering, Gunma University 2Japan Synchrotron Radiation Research Institute (JASRI), SPring-8 3 Department of Applied Chemistry, Ritsumeikan University 4Graduate School of Human and Environmental Studies, Kyoto University 5Department of Mechanical Engineering, Carnegie Mellon University 6Department of Physics, Northeastern University 7School of Engineering Science, Lappeenranta University Contents

1.Introduction

2.Compton scattering and It’s advantages

3.Experimental results 3-1. Non-destructive characterization for battery Compton scattering experiment is applied to the commercial lithium coin battery and done non-destructive measurement. 3-2. Quantum characterization for materials Compton scattering experiment is applied to positive electrode material.

4.Conclusion Introduction

Although lithium-ion batteries are already widely used in our daily life, further development of high-performance batteries are expected.

Lithium-ion Ø Problems in real battery battery - e （Discharge） ・Inhomogeneous lithium reaction Power source ・Capacity degradation - e （Charge） Non-destructive measurement Discharge Ø Problems in material Li+ ・High-capacity electrode materials High-stability electrode materials Charge ・ Anode Separator Cathode

(Graphite) (LiCoO2, LiMn2O4, Understanding of electrode reaction LiFePO4) We focus on Compton scattering technique to resolve these problems. Compton scattering

Compton scattering is well known as inelastic scattering technique between photons and electrons.

Ø Energy conservation law Incident X -rays !ω1 !ω + E = !ω + E -rays !ω 1 1 2 2 !k1 Scattering X 2 !k2 Ø Momentum conservation law

!k1 + p1 = !k2 + p2 Electron

E1 ,P1 E2 ,P2

Ø Scattered X-ray energy : Photon energy !ω1 , !ω2 2 !k , !k : Photon momentum 1 2 ! k !k⋅p1 E , E : Electron energy 1 2 !ω2 = !ω1 − − : Electron momentum p1 , p2 2m m Compton profile

Ø Scattered X-ray energy 2 ! k !k⋅p !ω = !ω − − 1 2 1 2m m Ø Energy spectrum : I (!ω2 ) I J p (!ω2 ) ∝ ( z )

Ø Compton profile : J (pz ) J p = ρ p dp dp ( z ) ∫∫ ( ) x y Ø Electron momentum density : ρ(p) : Scattered X-ray energy !ω2 2 p = ( px , py , pz ) : Electron momentum ρ p = n ψ r exp −ip⋅r dr n : Occupation number ( ) ∑ j ∫ j ( ) ( ) j j r ψ j ( ) : j-th wavefunction Compton profile can directly compare with theoretical Compton profile. It is enables to discuss electronic structure underlying reduction-oxidation reaction on the electrodes. Advantage of Compton scattering

Compton scattering technique have mainly two advantages.

(1) High-energy (>100keV) X-rays (2) Line-shape of Compton profile

1 /g) 2 2 Li atom Atomic model calc. 10 0.8 Li atom Mn atom Total Incoherent scattering 100 O atom (Compton)

(0) 0.6 J )/

-2 Coherent scattering z

10 Photoelectron p ( 0.4 absorption J -4 10 0.2

10-6 0 100 101 102 103 -6 -4 -2 0 2 4 6

Mass attenuation coefficients (cm p (atomic units) X-ray energy (keV) z Incoherent scattering is enhanced Quantitative analysis will be possible by by using over 100keV X-rays. digitalizing the line-shape of Compton profile NIST (http://www.nist.gov) Biggs et al., Atomic Data and Nuclear Data Tables 16, 201 (1975).

These features are suitable for analyzing lithium-ion batteries. Shape-parameter (S-parameter)

We have developed analysis method using a shape-parameter (S-parameter) to analyze line-shape of Compton profile. 17.5

Compton profile 25 Li Mn O 2.0 2 4 Li Mn O 20 1.1 2 4 17

) Li Mn O -1 0.5 2 4 15 ) (a.u. z 10 16.5 p J( 5

0 16 -10 -5 0 5 10 S-parameter y=0.59x+15.3 p (a.u.) z Hartree-Fock CALC. 15.5 KKR-CPA CALC. DFT CALC. d3 EXP. H = J (pz ) pz 15 H ∫ d2 0 0.5 1 1.5 2 2.5 3 3.5 S = Li composition by ICP d d W = 2 J p p + 4 J p p W ∫ d ( z ) z ∫ d ( z ) z 1 3 Linear relationship between S-parameters and In this study, d =d =5a.u., d =d =1a.u. were decided. 1 4 2 3 lithium concentration is confirmed.

Lithium quantitation is possible by using S-parameter analysis.

K. Suzuki et al., J. Appl. Phys., 119, 025103 (2016). Non-destructive characterization for real batteries

Compton scattering technique is applied to commercial VL2020 coin battery to observe lithiation state Non-destructive measurement

Many experimental techniques are demonstrated to monitor lithium reactions. ・ X-ray absorption near-edge structure (XANES) ・ ・ ・ Test cell C. Rumble et al., J. Electrochem. Soc., 157, A1317 (2010). H. Arai et al., J. Mater. Chem., 1, 10442 (2013). M. Katayama et al., J. Power Sources, 269, 994 (2014). ・ Nuclear magnetic resonance (NMR) ・ ・ ・ Test cell S. Chandrashekar et al., Nature Mat., 11, 311 (2012). ・ Micro X-ray diffraction ・ ・ ・ Test cell J. Liu et al., J. Phys. Chem. Lett., 1, 2120 (2010). ・ Particle induced γ–ray/X-ray emission (RIGE/PIXE) ・ ・ ・ Test cell K. Mima et al., Nucl. Instrum. Methods Phys. Rev. B 290, 79 (2012). ・ Raman micro-spectroscopy ・ ・ ・ Test cell T. Nishi et al., J. Electrochem. Soc., 160, A1785 (2013). ・ Hard X-ray photoemission spectroscopy (HX-PES) ・ ・ ・ Cylindrical cell H. Hori et al., J. Power Sources, 242, 844 (2013). ・ Neutron diffraction ・ ・ ・ Cylindrical cell Xun-Li Wang et al., Scientific Reports, 2, 1 (2012).

These experiments are mainly adopted to the test cells and It is difficult to measure Li directly. We focus on Compton scattering technique. Sample and Experimental setup

SPring-8 BL08W Ø Sample (Panasonic VL2020)

Stainless steal (SUS)

Positive electrode (V2O5) 0.8mm Separator 0.1~0.3mm Negative electrode (LiAl) 0.3mm Detector Spacer 0.1mm

Ø Experiment setup (BL08W SPring-8)

Ge solid-state detector Collimator

Incident slit X-ray X-ray Collimator slit 25 μm (H), Sample 500 μm (W) Φ 500 μm Sample Scattered X-rays Slit

Incident X-rays 115keV z-stage Stage x, y-stage X-ray camera Charge-discharge device Energy spectrum and S-parameter

Energy spectrum was measured in order to distinguish battery components.

Energy spectra S-parameter

Energy spectrum of SUS, electrodes Energy spectrum was converted to S-parameter. and separator was measured. 0.35 2 SUS 0.3 Positive 1.8 electrode 0.25 SUS

Separator SUS

1.6 Spacer 0.2 Negative electrode 1.4 0.15 1.2 0.1 S-parameter

0.05 1 Positive Negative electrode electrode Separator

Normarized intensity (arb. units) 0 -6 -4 -2 0 2 4 6 0.8 p (a.u.) 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 z z position (mm) Energy spectra change with S-parameters can reveal internal the components of the battery. structure of the battery.

K. Suzuki et al., J. Synchrotron. Rad., 24, 1006 (2017). S-parameter distribution

S-parameter distribution during charge-discharge cycle was obtained.

Charge-discharge curve

Vertical position 1.55 SUS 1.55 1.60 1.57 1.55 1.22 1.22 1.52 Positive 1.49 electrode Observed region Positive 1.47 electrode 0.9 0.90 1.44 (mm) z 1.41 1.39 0.57 0.57 Negative Separator 1.36 z position (mm) electrode position

z 1.33 Negative 1.31 X-ray 0.25 0.25 intensity parameter electrode - 1.28 S X-ray size: 25μm(h), 500μm(w) Spacer 1.25 Measurement time: 15 sec/point SUS -0.07 -0.07 1.23 2 1.5 1 1.20 S-parameter 0 2 4 6 8 10 Time (h) K. Suzuki et al., J. Synchrotron. Rad., 24, 1006 (2017). ΔS-parameter distribution

In order to clarify the change of S-parameter at the positive and negative electrodes, ΔS-parameter is taken.

ΔS = St – Save When the battery charged, S : S-parameter at t sec high S-parameter region exist t Charge-discharge Save: averaged S-parameter curve in the separator. on total measurement time

Vertical position 1.55 1.55 SUS 0.06 0.05 0.04 Positive electrode 1.22 1.22 0.03 Positive 0.03 electrode 0.02 Uniform reactions (mm) 0.9 0.90 0.01 parameter

- are occurred.

0.00 S -0.00 0.57 0.57 -0.01

Separator position z position (mm) -0.02

z Negative electrode -0.03 Negative 0.25 0.25 -0.03 electrode The reactions occur

-0.04 Variation of Spacer -0.05 at the surface. -0.07 SUS -0.07 -0.06 2 1.5 1 S-parameter 0 2 4 6 8 1 Time (h) 0

K. Suzuki et al., J. Synchrotron. Rad., 24, 1006 (2017). S-parameter at the separator

Raw data of S-parameters around the separator position. 2 SOC0 (1st) SOC100 SOC0 (2nd) 1.8

1.6 S-parameter 1.4 State of charge: SOC SOC100: fully charged state SOC0: fully discharged state 1.2 0.2 0.3 0.4 0.5 0.6 0.7 z (mm) ・The reproducibility was confirmed between 1st and 2nd full discharged state. ・The S-parameters increase at the separator position when full charged state.

Some lithium-ions remain into the separator. This fact might be induce capacity loss of the battery. Lithium composition

Lithium compositions during charge-discharge cycle are determined. 0.6 0.2 5

O 0.5 Al 2 x

V 0.15 x 0.4

0.3 0.1

0.2 0.05

0.1 Li composition of Li composition of 1st_discharge 1st_charge 2nd_discharg Rest 0 e 0 0 2 4 6 8 10 Time (h) Li composition ICP analysis S-parameter Positive electrode Li composition is agree 0.426 0.470±0.012 SOC0 with that of obtained from Negative electrode ICP analysis. 0.178 0.170 0.006 SOC100 ±

Non-destructive and in-operando analysis are possible by our technique. Quantum characterization for battery materials

High-resolution Compton scattering and magnetic Compton scattering experiments are applied

to LiMn2O4 to reveal electrode reaction LiMn2O4

Spinel LiMn2O4 is a representative positive electrode material for lithium-ion batteries.

Advantage Ø Thermodynamic stability Ø Low cost

Disadvantage Li 8 Charge/discharge curve Mn 16 L. Guohua et al., J. Electrochem. Soc., 143 (1996) 178 O 32

Cell voltage of LiMn2O4

Operation 60% region

Y. S. Meng et al., Energy Environ.Sci., (2009). The capacity decreases 60% at 100 cycles Electrode reaction of LiMn2O4

Charge + - LiMn2O4 xLi +Li(1-x)Mn2O4+xe （+3.5） Discharge （+4）

The volume expansion by Jahn-Teller ion induces capacity loss

¨ LMTO-ASA（H. Berg et al., J. Mater. Chem.9, 2813 (1999) ) ¨ FP-LMTO（G. E. Grechnev et al., Phys. Rev. B65, 174408 (2002)）

On the other hand

¨ DV-Xa（Y. L iu e t al., Solid State Ionics,126, 209 (1999)） ¨ LDDFT（M. K. Aydinol et al., J. Electrochem., 144, 3832 (1997)） O 2p orbital plays important role

The electrode reaction is not fully understood yet. The nature of the electrode reaction is investigated. Previous study Sample

Compton profiles of LixMn2O4 (x = 0.496 and 1.079) were measured. The difference Compton profile are shown with KKR-CPA first-principle calculation result. LixMn2O4 powder

Mn site is almost constant

Interstitial and O sites increase with Li increase

The O 2p orbitals play a dominant role on the electrode reaction of LiMn2O4 positive electrode. Magnetic Compton scattering is applied to study electronic structure of Mn atom. K. Suzuki et al., PRL, 114, 087401 (2015). Magnetic Compton profile (MCP)

Magnetic Compton profiles of Li Mn O x 2 4 I , I (x = 0.41, 0.50 and 0.92) were measured. + -

0.25 1 Li Mn O ⎛ ⎞ 0.41 2 4 I − I µspin 0.2 /Mn) 0.8 R + − A B = = ) Li Mn O ⎜ ⎟ 0.50 2 4 I I N -1 + + − ⎝ ⎠ Li Mn O 0.15 0.92 2 4 0.6 ) (a.u. z p

( 0.1 0.4 mag J 0.05 0.2 SQUID MCP ( )

Magnetic moment ( spin 0 0 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 6 Li concentration, x p (a.u.) z The line-shape of MCPs does Spin magnetic moment increase not show significantly change. with increasing Li concentration. We do comparison with theoretical calculation results to reveal increase of spin magnetic moments in operation region of the battery. Theoretical MCP and magnetic moment

Theoretical magnetic Compton profile was calculated by first-principle DFT calculation using exchange-correlation function of SCAN meta-GGA.

(a) (a) (b) 1 Theory SQUID

/Mn) 0.8 MCP Exp.: Li0.92Mn2O4 B Theo.: Li0.75Mn2O4 0.6

0.4

0.2

Magnetic moment ( 0 0 0.2 0.4 0.6 0.8 1 Li concentration, x

Calculated spin magnetic moment in lithium concentration 0 < x <1 have same trend with experimental spin magnetic moment. H. Hafiz et al., PRB, submitted. Spin-dependent partial density of state 2 3+ Spin-dependent partial density of state associated with the eg and t2g orbitals of Mn and Mn4+ ions are calculated. 1.0 Theory 8.2 Exp.

0.8 (Å) The Mn eg orbital split and avg 8.1 a 2 Mn 3dz bands move to lower energies

0.6 8 0 0.5 1 /Mn) x B 0.4 Theory LiMn2O4 has band gap M ( MCP 0.2 SQUID

Mn 3d 2 bands are partially occupied 0.0 z 0 0.2 0.4 0.6 0.8 1 1.2 x

Li0.75Mn2O4 becomes metallic FIG. 1. Ferrimagnetic phase in Li xMn2O4. Mag- and have ferrimagnetic state netic moments for various lithium concentrations x obtained via Compton scattering (red squares) and SQUID (yellow squares) measurements are compared with the corresponding theoretical predictions (green circles). The gray dashed lines are guides for the eye. The inset shows how the experimental and theoretical values of the cubic lattice constant varies with FIG. 2. Spin-dependent partial-density-of-states. H. Hafiz et al., PRB, submitted. Computed spin-dependent partial-density-of-states (PDOS) x. 3+ 4+ associated with the eg and t2g orbitals of Mn and Mn ions in LixMn2O4 for Li concentrations (a) x =1.0and(b) how the manganese 3d states evolve with Li intercala- x =0.75. See panel (a) for meaning of lines of various colors. Vertical dashed line marks the Fermi energy (EF ). Up and tion and how their magnetism drives performance of the down arrows indicate the contributions of spin up and down battery. PDOS respectively. In an MCS experiment, one measures the magnetic 19 Compton profile (MCP) , Jmag, which can be expressed in terms of a double integral of the spin-dependent elec- from the sample with a scattering angle of 178 were mea- tron momentum density, ⇢mag(p), as sured using a ten-segmented Ge solid-state detector with external magnetic field of 25 kOe. The estimated momen-

Jmag(pz)= ⇢mag(p)dpxdpy. (1) tum resolution is 0.50 a.u. full-width-at-half-maximum. ZZ The spherically averaged profile Jmag(p) was extracted from the di↵erence between two spectra taken under the Here, ⇢mag(p)=⇢ (p) ⇢ (p), where ⇢ (p) and ⇢ (p) " # " # same experimental conditions with alternating directions are the momentum densities of up-spin (majority) and of magnetization of the sample, aligned by an external down-spin (minority) electrons, respectively. magnetic field. The observed spectra were corrected for In this way, MCS allows access to magnetic electrons in the energy-dependent scattering cross section, eciency materials and its potential in this regard was recognized of the detector and absorption of the sample. quite early in the field22,23. The technique has proven Polycrystalline samples of LixMn2O4 (x =0.41, 0.50, especially successful in extracting occupation numbers of 0.92 and 1.08) were prepared by chemical lithium extrac- 24,25 3d Mn orbitals in bilayer manganites and recent MCS tion following the method reported previously21.The studies have revealed fine details of the magnetic orbitals compositions were determined by inductively coupled 26–28 in a number of materials . plasma (ICP) measurements. X-ray powder di↵raction analyses confirmed spinel phases and the increase of the lattice constant with increasing x over the range x = II. MATERIALS AND METHODS 0 1 as observed previously21. Total magnetic moments were obtained from SQUID magnetometer (MPMS5-SW, A. Magnetic Compton experiments Quantum Design, Inc) measurements.

The MCS experiments were carried out at 10 K on beamline BL08W at SPring-8, Japan29,30. Elliptically B. First-principles calculations polarized x-rays emitted from an elliptical multi-pole wig- gler were monochromatized to 183 keV by a bent Si The first-principles calculations were performed us- 620 crystal. Energy spectra of Compton-scattered x-rays ing the pseudopotential projector augmented-plane-wave Magnetic structure 3 Magnetic configuration models are considered.

(a) x=1.0 (b) x=0.75

Mn4+ Mn4+

Mn3+ Mn3+ Li O Li vacancy b

FIG. 3. Schematics of magnetic configurations of LixMn2O4 for lithium concentrations, x =1.0 and x =0.75. Charge-ordering is prominent for the x = 1 phase with an overall AFM configuration (a). As soon as the Li ions are removed, see (b), charge-ordering becomes unstable, and results in partial ordering,Charge spin-flipping,-ordering and reduction become of the averageunstable, magnetic moment perCharge unit cell.- Thisordering induces is ferrimagnetism prominent. in the x =0.75, which is seen also in the partial-densities-of-states4+ of Figs. 2(b). and spin-flipping of Mn is occur.

H. Hafiz et al., PRB, submitted. method31 as implemented in the Vienna Ab-Initio Simu- III. RESULTS AND DISCUSSIONS lation Package (VASP)32,33, with a kinetic energy cuto↵ of 600 eV for the plane-wave basis set. Computations A. Ferrimagnetic (FIM) phase were carried out using both the Generalized Gradient Approximation (GGA)34,35 and the recently constructed Figure 1 highlights evolution of the magnetic moment strongly-constrained-and-appropriately-normed (SCAN) in Li Mn O for Li concentrations varying from zero to meta-GGA11 exchange-correlation functional. A 4 4 4 x 2 4 around 1.2 per formula unit. The AFM state (zero av- -centered k-point mesh was used to sample the Brillouin⇥ ⇥ erage moment) is realized in the end compounds MnO zone of a primitive spinel unit cell (containing eight for- 2 and LiMn O , but the system assumes an FIM state with mula units or 32 oxygen atoms). Equilibrium positions 2 4 a net non-zero moment at other compositions, including of all atoms were calculated via structural optimization, the over-discharged regime of x>1. The moments ob- where the internal degrees of freedom, along with the tained via SQUID measurements are in good accord with shape and volume of the unit cell, were allowed to vary those obtained from MCS experiments. Some di↵erences until the residual forces per atom were less than 0.005 are to be expected since MCS measures only the spin eV/A.˚ We obtained spin momentum densities, ⇢ (p), mag moment while the SQUID couples to the total moment, and the spherical Compton profiles J (p) of the va- mag which includes the orbital component of the magnetic lence electrons using Kohn-Sham orbitals following the moment. Our theoretical predictions are seen to be in method of Makkonen et al.36. This scheme has been re- good agreement with the experimental results. cently used to study the Compton profile of lithium iron 21 37 Notably, Suzuki et al have previously shown by phosphate (LiFePO4) , which is as an exemplar cathode battery material. using first-principles Korringa-Kohn-Rostoker-coherent- potential-approximation (KKR-CPA)38,39 calculations that the e↵ective moment on Mn atoms increases with increasing x if the system is described by a spin-glass- like behavior with randomly oriented Mn moments. The trend of strengthening Mn moment with Li insertion in Fig. 1, which is consistent with muon-spin-rotation experiments40, can thus be understood within the KKR- CPA model, but such a model fundamentally fails to cap- Magnetic orbital

Magnetic orbital of Li0.75Mn2O4 is visualized.

(a) (b) (a) (a) (a) (b)

Mn 3d magnetic electrons in LixMn2O4 (x < 1 ) have mainly t2g symmetry, and they prevent structural distortion promoted by eg character.

The O 2p orbitals play a dominant role on the electrode reaction and Mn 3d orbitals contribute to the stability of the structure in LiMn2O4. Summary

Non-destructive characterization study for the commercial battery and quantum characterization study for the electrode material were shown.

l Non-destructive characterization for VL2020 battery Ø Lithium reactions of the electrodes are revealed nondestructively. Ø Remnant lithium ions are observed in the separator when the battery was charged.

l Quantum characterization for LiMn2O4 Ø The O 2p orbitals play a dominant role on the electrode reaction.

Ø The Mn 3d orbitals contribute to the stability of the structure.

Compton scattering technique is powerful tool for non- destructive and quantum characterizations in the batteries.