Patent Application
20210408580
The invention is directed to the field of solid state batteries with alkali metal sulfide solid state electrolytes.
Solid-state lithium ion conductors, the key component to enabling all solid-state lithium ion batteries, are one of the most pursued research objectives in the battery field. The intense interest in solid-state electrolytes, and solid-state batteries more generally, stems principally from improved safety, the ability to enable new electrode materials and better low-temperature performance. Safety improvements are expected for solid-state battery cells as the currently used liquid-electrolytes are typically highly-flammable organic solvents. Replacing these electrolytes with non-flammable solids would eliminate the most problematic aspect of battery safety. Moreover, solid-electrolytes are compatible with several high energy density electrode materials that cannot be implemented with liquid-electrolyte based configurations. Solid-electrolytes also maintain better low temperature operation than liquid-electrolytes, which experience substantial ionic conductivity drops at low temperatures. Such low temperature performance is critical in the burgeoning electric-vehicles market.
Of the currently studied solid-electrolytes, sulfides remain one of the highest-performance and most promising families. Sulfide glass solid-electrolytes and glass-ceramic solid-electrolytes, where crystalline phases have precipitated within a glassy matrix, have demonstrated ionic conductivities on the order of 0.1-1 mS cm−1 and above 1 mS cm−1, respectively. The ceramic-sulfide electrolytes, most notably Li10GeP2S12 (LGPS) and Li10SiP2S12 (LSPS), are particularly promising as they maintain exceptionally high ionic conductivities. LGPS was one of the first solid-electrolytes to reach ionic conductivities comparable to liquid-electrolytes at 12 mS cm−1, only to be displaced by LSPS, which achieved an astonishingly high ionic conductivity of 25 mS cm−1. Despite these promising conductivities, the ceramic-sulfide family is plagued by a narrow stability window. That is, LGPS and LSPS both tend to reduce at voltages below approximately 1.7 V vs lithium metal or oxidize above approximately 2.1 V. This limited stability window has proven a major barrier for battery cells that need to operate in a voltage range of approximately 0-4 V.
Thus, there is a need for improved solid state batteries incorporating solid state electrolytes with controllable structural properties and surface chemistry.
We have developed rechargeable solid state batteries using solid state electrolytes with improved cycling performance. The rechargeable solid state batteries disclosed herein are advantageous as the solid state electrolytes have superior voltage stability and excellent battery cycle performance.
Batteries of the invention may be stabilized against the formation of lithium dendrites and/or can operate at high current density for an extended number of cycles.
In one aspect, the invention features a rechargeable battery including a first electrode, a second electrode, and a solid state electrolyte disposed therebetween. The solid state electrolyte includes a sulfide that includes an alkali metal, such as lithium. In certain embodiments, the solid state electrolyte is under a volumetric constraint sufficient to stabilize the solid state electrolyte during electrochemical cycling. In particular embodiments, the volumetric constraint exerts a pressure of about 70 to about 1,000 MPa, e.g., about 100-250 MPa, on the solid state electrolyte, e.g., to enforce mechanical constriction on the microstructure of solid electrolyte on the order of 15 GPa. In certain embodiments, the volumetric constraint provides a voltage stability window of between 1 and 10 V, e.g., 1-8V, 5.0-8 V, or greater than 5.7 V, or even greater than 10V.
In some embodiments, the solid state electrolyte has a core shell morphology. In certain embodiments the alkali metal is Li, Na, K, Rb, or Cs, e.g., Li. In some embodiments, the solid state electrolyte includes SiPS, GePS, SnPS, PSI, or PS. In some embodiments, the solid state electrolyte is Li10SiP2S12, Li10GeP2S12, or Li9.54Si1.74P1.44S11.7ClO0.3. In some embodiments, the first electrode is the cathode, which can include LiCoO2, LiNi0.5Mn1.5O4, V Li2CoPO4F, LiNiPO4, Li2Ni(PO4)F, LiMnF4, LiFeF4, or LiCo0.5Mn1.5O4. In certain embodiments, the second electrode is anode and can include lithium metal, lithiated graphite, or Li4Ti5O12. In particular embodiments, the volumetric constraint provides a mechanical constriction on the solid state electrolyte between about 1 to about 100 GPa, e.g., about 15 GPa.
In another aspect, the invention features a rechargeable battery including a first electrode, a second electrode, and a solid state electrolyte disposed therebetween, wherein the second electrode is an anode comprising an alkali metal and graphite. In some embodiments, the battery is under a pressure of about 70-1000 MPa, e.g., about 100-250 MPa. In particular embodiments, the alkali metal and graphite form a composite. In some embodiments, the alkali metal is Li, Na, K, Rb, or Cs, e.g., Li. In some embodiments, the solid state electrolyte includes SiPS, GePS, SnPS, PSI, or PS. In certain embodiments, the solid state electrolyte is Li10SiP2S12, Li10GeP2S12, or Li9.54Si1.74P1.44S11.7Cl0.3. In particular embodiments, the first electrode is the cathode and can include LiCoO2, LiNi0.5Mn1.5O4, V Li2CoPO4F, LiNiPO4, Li2Ni(PO4)F, LiMnF4, LiFeF4, or LiCo0.5Mn1.5O4. In some embodiments, the battery is under an external stress that provides a mechanical constriction on the solid state electrolyte between about 1 to about 100 GPa, e.g., about 15 GPa.
In another aspect, the invention features a rechargeable battery including a first electrode, a second electrode, and a solid state electrolyte disposed therebetween, wherein the solid state electrolyte may include a sulfide including an alkali metal; and the battery is under isovolumetric constraint. In some embodiments, the isovolumetric constraint is provided by compressing the solid state electrolyte under a pressure of about 3-1000 MPa, e.g., about 100-250 MPa. In certain embodiments, the alkali metal is Li, Na, K, Rb, or Cs, e.g., Li. In some embodiments, the solid state electrolyte includes SiPS, GePS, SnPS, PSI, or PS. In certain embodiments, the solid state electrolyte is Li10SiP2S12, Li10GeP2S12, or Li9.54Si1.74P1.44S11.7Cl0.3. In particular embodiments, the first electrode is the cathode and can include LiCoO2, LiNi0.5Mn1.5O4, V Li2CoPO4F, LiNiPO4, Li2Ni(PO4)F, LiMnF4, LiFeF4, or LiCo0.5Mn1.5O4. In some embodiments, the isovolumetric constraint provides a mechanical constriction on the solid state electrolyte between about 1 to about 100 GPa, e.g., about 15 GPa. In another aspect, the invention features a rechargeable battery having a first electrode, a second electrode, and a solid state electrolyte disposed therebetween. The solid state electrolyte includes a sulfide that includes an alkali metal, and optionally has a core-shell morphology. The first electrode includes an interfacially stabilizing coating material. In certain embodiments, the first and second electrodes independently include an interfacially stabilizing coating material. In certain embodiments, one of the first and second electrodes includes a lithium-graphite composite.
In some embodiments, the first electrode comprises a material as described herein, e.g., in Table 1. In some embodiments, the coating material of the first electrode is a coating material as described herein, e.g., LiNbO3, AlF3, MgF2, Al2O3, SiO2, graphite, or in Table 2. In certain embodiments, the alkali metal is Li, Na, K, Rb, or Cs, e.g., Li. In some embodiments the solid state electrolyte includes SiPS, GePS, SnPS, PSI, or PS. In certain embodiments, the solid state electrolyte is Li10SiP2S12, Li10GeP2S12, or Li9.54Si1.74P1.44S11.7Cl0.3. In some embodiments, the first electrode is the cathode and can include LiCoO2, LiNi0.5Mn1.5O4, V Li2CoPO4F, LiNiPO4, Li2Ni(PO4)F, LiMnF4, LiFeF4, or LiCo0.5Mn1.5O4. In some embodiments, the battery is under an external stress that provides a mechanical constriction on the solid state electrolyte between about 1 to about 100 GPa, e.g., about 15 GPa. In certain embodiments, the battery is under a pressure of about 70-1000 MPa, e.g., about 100-250 MPa.
In another aspect, the invention features a method of storing energy by applying a voltage across the first and second electrodes and charging the rechargeable battery of the invention. In another aspect, the invention provides a method of providing energy by connecting a load to the first and second electrodes and allowing the rechargeable battery of the invention to discharge.
, •, ▪, ▾,
stand for LCO(PDF #44-0145), LSPS(ICSD #252037), SiO2(PDF #48-0476), Li3PO4(PDF #45-0747), Cubic Co4S3(PDF #02-1338), Monoclinic Co4S3(PDF #02-1458) respectively. In (B), ▴,
, •, ▪,
stand for SnO2(PDF #41-1445), LSPS(ICSD #252037), SiO2(PDF #34-1382), P2S5(PDF #50-0813), and Li2S(PDF #23-0369) respectively. In (C), ▴,
,
: stand for LTO(PDF #49-0207), LSPS(ICSD #252037) and Li1.95Ti2.05S4 (PDF #40-0878) respectively. In (D), ▴,
stand for SiO2(PDF #27-0605) and LSPS(ICSD #252037) respectively. The shaded regions in (A-D) highlight where significant phase change happened after heating to 500° C. The interfacial chemical compatibility decreases from (A) to (D), corresponding well with the predicted interfacial decay energies of 200, 97, 75, and 0 meV/atom for LCO, SnO2, LTO and SiO2, respectively. (E, F) CV results for Li2S and SnO2. The shaded regions predict if the curve in that region will be dominantly oxidation, reduction, neutral.
LGPS; +Li; * Graphite; x LiS2; ∇ GeS2;
GeLi5P3. (D) The structure of Li/Graphite anode in LGPS based all-solid-state battery; (E) SEM image of the cross section of Li/Graphite anode; (F) FIB-SEM of the interface of Li and Graphite.
The invention provides rechargeable batteries including a solid state electrolyte (SSE) containing an alkali metal and a sulfide disposed between two electrodes. The solid state electrolytes may have a core-shell morphology, imparting increased stability under voltage cycling conditions. These batteries of the invention are advantageous as they may be all-solid-state batteries, e.g., no liquid electrolytes are necessary, and can achieve higher voltages with minimal electrolyte degradation.
Core-shell morphologies in which a core of ceramic-sulfide solid-electrolyte is encased in a rigid amorphous shell have been shown to improve the stability window. The mechanism behind this stabilization is believed to be tied to the tendency of ceramic-sulfides to expand during decay by up to more than 20%. Applying a volume constraining mechanism, this expansion is resisted which in turn inhibits decay. We have generalized this theory and provide experimental evidence using post-synthesis creation of a core-shell morphology of LGPS to show improved stability. Based on the decay morphology, the magnitude of stabilization will vary. A mean-field solution to a generalized strain model is shown to be the lower limit on the strain induced stability. The second decay morphology explored, nucleated decay, is shown to provide a greater capability for stabilization. Moreover, experimental evidence suggests the decay is in fact the later (nucleated) morphology, leading to significant potential for ceramic-sulfide full cell batteries.
Further developments of the theory underpinning the enhanced stability and performance of core-shell electrolytes have revealed that the strain stabilization mechanism is not limited to the materials level but can also be applied on the battery cell level through external stress or volume constriction. The strain provided by the core-shell structure stabilizes the solid electrolyte through a local energy barrier, which prevents the global decomposition from happening. Such stabilization effect provided by local energy barrier can also be created by applying an external stress or volume constriction from the battery cell, where up to 5.7 V voltage stability window on LGPS can be obtained as shown in
In solid state batteries, lithium dendrites form when the applied current density is higher than a critical value. The critical current density is often reported as 1-2 mA cm−2 at an external pressure of around 10 MPa. In the present invention, a decomposition pathway of the solid state electrolyte, e.g., LGPS, at the anode interface is modified by mechanical constriction, and the growth of lithium dendrite is inhibited, leading to excellent rate and cycling performances. No short-circuit or lithium dendrite formation is observed after the batteries are cycled at a current density up to 10 mA cm−2.
Solid State Electrolytes
A rechargeable battery of the invention includes a solid electrolyte material and an alkali metal atom incorporated within the solid electrolyte material. In particular, solid state electrolytes for use in batteries of the invention may have a core-shell morphology, with the core and shell typically having different atomic compositions.
Suitable solid state electrolyte materials include sulfide solid electrolytes, e.g., SixPySz, e.g., SiP2S12 such as Li10SiP2S12, or β/γ-PS4. Other solid state electrolytes include, but are not limited to, germanium solid electrolytes, e.g., GeaPbSc, e.g., GeP2S12 such as Li10GeP2S12, tin solid electrolytes, e.g., SndPeSf, e.g., SnP2S12, iodine solid electrolytes, e.g., P2S8I crystals, glass electrolytes, e.g., alkali metal-sulfide-P2S5 electrolytes or alkali metal-sulfide-P2S5-alkali metal-halide electrolytes, or glass-ceramic electrolytes, e.g., alkali metal-PgSh-i electrolytes. Another material includes Li9.54Si1.74P1.44S11.7Cl0.3. Other solid state electrolyte materials are known in the art. The solid state electrolyte material may be in various forms, such as a powder, particle, or solid sheet. An exemplary form is a powder.
Alkali metals useful for the solid state electrolytes for use in batteries of the invention include Li, Na, K, Rb, and Cs, e.g., Li. Examples of Li-containing solid electrolytes include, but are not limited to, lithium glasses, e.g., xLi2S(1−x)P2S5, e.g., 2Li2S—P2S5, and xLi2S-(1-x)P2S5—LiI, and lithium glass-ceramic electrolytes, e.g., Li7P3S11-z.
Electrode Materials
Electrode materials can be chosen to have optimum properties for ion transport. Electrodes for use in a solid state electrolyte battery include metals, e.g., transition metals, e.g., Au, alkali metals, e.g., Li, or crystalline compounds, e.g., lithium titanate such as Li4Ti5O12 (LTO). An anode may also include a graphite composite, e.g., lithiated graphite. Other materials for use as electrodes in solid state electrolyte batteries are known in the art. The electrodes may be a solid piece of the material, or alternatively, may be deposited on an appropriate substrate, e.g., a fluoropolymer or carbon. For example, liquefied polytetrafluoroethylene (PTFE) has been used as the binder when making solutions of electrode materials for deposition onto a substrate. Other binders are known in the art. The electrode material can be used without any additives. Alternatively, the electrode material may have additives to enhance its physical and/or ion conducting properties. For example, the electrode materials may have an additive that modifies the surface area exposed to the solid electrolyte, such as carbon. Other additives are known in the art.
High voltage cathodes of 4 volt LiCoO2 (LCO, shown in
Electrode Coatings
In some cases, the electrode materials may further include a coating on their surface to act as an interfacial layer between the base electrode material and the solid state electrolyte. In particular, the coatings are configured to improve the interface stability between the electrode, e.g., the cathode, and the solid electrolyte for superior cycling performance. For example, coating materials for electrodes of the invention include, but are not limited to graphite, LiNbO3, AlF3, MgF2, Al2O3, and SiO2, in particular LiNbO3 or graphite.
Based on a new high-throughput analysis schema to efficiently implement computational search to very large datasets, a library of different materials was searched to find those coating materials that can best stabilize the interface between sulfide solid-electrolytes and typical electrode materials, using Li10SiP2S12 as an example to predict over 1,000 coating materials for cathodes and over 2,000 coating materials for anodes with both the required chemical and electrochemical stability. These are generally applicable for LGPS. Table 2 provides the predicted effective coating materials.
External Stress
Strain stabilization mechanism for enhancing electrolyte stability is not limited to the materials level but can also be applied on the battery cell level through external stress or volume constriction. In certain embodiments, the external stress is a volumetric constraint applied to all or a portion, e.g., the solid state electrolyte, of the rechargeable battery, e.g., delivered by a mechanical press. The external stress can be applied by a housing, e.g., made of metal. In some cases, the volumetric constraint can be from about 70 MPa to about 1,000 MPa, e.g., about 70 MPa to about 150 MPa, about 100 MPa to about 300 MPa, about 200 MPa to about 400 MPa, about 300 MPa to about 500 MPa, about 400 MPa to about 600 MPa, about 500 MPa to about 700 MPa, about 600 MPa to about 800 MPa, about 700 MPa to about 900 MPa, or about 800 MPa to about 1,000 MPa, e.g., about 70 MPa, about 75 MPa, about 80 MPa, about 85 MPa, about 90 MPa, about 95 MPa, about 100 MPa, about 150 MPa, about 200 MPa, about 250 MPa, about 300 MPa, about 350 MPa, about 400 MPa, about 450 MPa, about 500 MPa, about 550 MPa, about 600 MPa, about 650 MPa, about 700 MPa, about 750 MPa, about 800 MPa about 850 MPa, about 900 MPa, about 950 MPa, or about 1,000 MPa. In the present invention, “about” means±10%.
The solid state electrolyte may also be compressed prior to inclusion in the battery. For example, the solid state electrolyte may be compressed with a force between about 70 MPa to about 1,000 MPa, e.g., about 70 MPa to about 150 MPa, about 100 MPa to about 300 MPa, about 200 MPa to about 400 MPa, about 300 MPa to about 500 MPa, about 400 MPa to about 600 MPa, about 500 MPa to about 700 MPa, about 600 MPa to about 800 MPa, about 700 MPa to about 900 MPa, or about 800 MPa to about 1,000 MPa, e.g., about 70 MPa, about 75 MPa, about 80 MPa, about 85 MPa, about 90 MPa, about 95 MPa, about 100 MPa, about 150 MPa, about 200 MPa, about 250 MPa, about 300 MPa, about 350 MPa, about 400 MPa, about 450 MPa, about 500 MPa, about 550 MPa, about 600 MPa, about 650 MPa, about 700 MPa, about 750 MPa, about 800 MPa about 850 MPa, about 900 MPa, about 950 MPa, or about 1,000 MPa. Once pressed, the solid state electrolyte can then be employed in a battery. Such a battery may also be subjected to external stress to enforce a mechanical constriction on the solid state electrolyte, e.g., at the microstructure level, i.e., to provide an isovolumetric constraint. The mechanical constriction on the solid state electrolyte may be from 1 to 100 GPa, e.g., 5 to 50 GPa, such as about 15 GPa. The external stress required to maintain the mechanical constriction may be from about 1 MPa to about 1,000 MPa, e.g., about 1 MPa to about 50 MPa, about 1 MPa to about 250 MPa, about 3 MPa to about 30 MPa, about 30 MPa to about 50 MPa, about 70 MPa to about 150 MPa, about 100 MPa to about 300 MPa, about 200 MPa to about 400 MPa, about 300 MPa to about 500 MPa, about 400 MPa to about 600 MPa, about 500 MPa to about 700 MPa, about 600 MPa to about 800 MPa, about 700 MPa to about 900 MPa, or about 800 MPa to about 1,000 MPa, e.g., about 70 MPa, about 75 MPa, about 80 MPa, about 85 MPa, about 90 MPa, about 95 MPa, about 100 MPa, about 150 MPa, about 200 MPa, about 250 MPa, about 300 MPa, about 350 MPa, about 400 MPa, about 450 MPa, about 500 MPa, about 550 MPa, about 600 MPa, about 650 MPa, about 700 MPa, about 750 MPa, about 800 MPa about 850 MPa, about 900 MPa, about 950 MPa, or about 1,000 MPa. The external stress employed may change depending on the voltage of the battery. For example, a battery operating at 6V may employ an external stress of about 3 MPa to about 30 MPa, and a battery operating at 10V may employ an external stress of about 200 MPa. The invention also provides a method of producing a battery using compression of the solid state electrolyte prior to inclusion in the battery, e.g., with subsequent application of external stress.
Methods
Batteries of the invention may be charged and discharged for a desired number of cycles, e.g., 1 to 10,000 or more. For example, batteries may be cycled 10 to 750 times or at least 50, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1,000, 1,500, 2,000, 3,000, 4,000, or 5,000 times. In embodiments, the voltage of the battery ranges from about 1 to about 20V, e.g., about 1-10V, about 5-10V, or about 5-8V. Batteries of the invention may also be cycled at any appropriate current density e.g., 1 mA cm−2 to 20 mA cm−2, e.g., about 1-10 mA cm−2, about 3-10 mA cm−2, or about 5-10 mA cm-2.
The cyclic voltammograms (CV) of Li/LGPS/LGPS+C were measured under different pressures between open circuit voltage (OCV) to 6 V at a scan rate of 0.1 mVs−1 on a Solartron electrochemical potentiostat (1470E), using lithium (coated by Li2HPO4) as reference electrode. A liquid battery using LGPS/C thin film as cathode, lithium as anode and, 1 M LiPF6 in EC/DMC as electrolyte was also assembled for comparison. The ratio of LGPS to C is 10:1 in both solid and liquid CV tests.
The cathode and anode thin films used in all-solid-state battery were prepared by mixing LTO/LCO/LNMO, LGPS, Polytetrafluoroethylene (PTFE) and carbon black with different weight ratios. The ratios of active materials/LGPS/C are 30/60/10, 70/27/3, 70/30/0 for LTO, LCO and LNMO thin film electrodes, respectively. This mixture of powder was then hand-grinded in a mortar for 30 minutes and rolled into a thin film inside an argon-filled glove box with 3% PTFE added. Solid electrolytes used in all-solid-state Li ion batteries were prepared by mixing LGPS and PTFE with a weight ratio of 97:3, then hand-grinding the mixed powder in a mortar for 30 minutes and finally rolling it into a thin film inside an argon-filled glove box. To assemble an all-solid-state Li ion battery cell, the prepared composite cathode (LCO or LNMO) thin film, LGPS thin film (<100 μm), and anode (LTO) thin film were used as cathode, solid electrolyte, and the anode, respectively. The three thin films of cathode, electrolyte and anode were cold-pressed together at 420 MPa, and the pressure was kept at 210 MPa by using a pressurized cell during battery cycling test. The charge and discharge behavior was tested using an ArbinBT2000 workstation (Arbin Instruments, TX, USA) at room temperature. The specific capacity was calculated based on the amount of LTO.
Theory—The Physical Picture
The mechanism by which strain can expand the LGPS stability window is depicted in
G
0(xD)=(1−xD)GLGPS+xDGD (1)
The lowest Gibbs energy state is xD=1 (all decomposed) and the initial state is xD=0 (pristine LGPS). Accordingly, the reaction energy is ΔG0=G0(1)−G° (0)=GD−GLGPS. This system is inherently unstable. That is, ∂x
Next, consider the application of a mechanical system that constrains the LGPS particle. Given that LGPS tends to expand during decay, any mechanical constraint will require that decomposition induce strain in the surrounding neighborhood. Such a constraining system could be either materials-level (i.e. a core-shell microstructure) or systems-level (i.e. a pressurized battery cell) or a combination of the two. Ultimately, this mechanical system can only induce a finite strain before fracturing. The energy needed to fracture the system is denoted Gfracture.
Prior to the fracturing of the constraining mechanism, any decomposition of the LGPS must lead to an increase in strain energy. The green line in
If the constraint induced strain Gibbs (Gstrain) is sufficiently steep, the slope of the total Gibbs at xD<xf will be positive (as depicted in
Two Work Differentials
The presence of Gstrain as a function of xD stems from the nature of LGPS to expand upon decomposition. Depending on the set of decomposed products, as determined by the applied voltage, this volume expansion can exceed 20-50%. As such, the process of LGPS decomposition is one that can include significant “stress-free” strain—that is, strain that is the result of decomposition and not an applied stress. Proper thermodynamic analysis of such decay pathways requires careful consideration of the multiple work differentials, which are reasonably neglected for other systems.
The general approach to showing the equivalency of these two differential work expressions is as follows. The solid-like stress and strain tensors are separated into the compression and distortion terms via the use of deviatoric tensors as defined in equation 2. The pressure is generalized in terms of the stress matrix p≡⅓tr(σ)=−⅓σii and volume strain ϵ≡(V−Vref/Vref.
Using these definitions, the solid-like work can be separated into one term that only includes compression and one term that only includes deformation.
δW=Vrefσijδϵij=Vref(σijdδϵijd−pδϵ) (3)
In the fluid limit, where there is no shape change, equation 3 reduces to δW=−Vrefpδϵ=pδV assuming that δVref=0, giving back the fluid-like work differential. In most mechanical systems, this assumption is valid as the undeformed reference volume does not change. However, it fails in describing LGPS decomposition because the undeformed volume changes with respect to xD and, hence, δVref≠0.
V
ref(xD)=(1−xD)VLGPS+xDVD (4)
Instead, proper thermodynamic analysis of LGPS decomposition requires consideration of both work terms. The fluid term−pδVref indicates the work needed to compress the reference volume (i.e., change xD) in the presence of a stress tensor a and the solid term represents the work needed to deform the new reference state Vrefσijδϵij. Considering this, the full energy differential is given by equation 5.
δE=TδS+μαNα−pδVref+Vrefσijδϵij (5)
Transforming to the Gibbs energy G=E−TS+pVref−Vrefσijϵij=μαNα, yields the differential form:
δG=−SdT+μαδNα+Vδp−Vrefϵijδσij (6)
Note that the transformation used frequently in solid mechanics, G=E−TS−Vrefσijϵij=μαNα−pVref, is sufficient so long as Vref is constant and, hence, −pVref can be set as the zero point.
At constant temperature, equation 6 gives the differential form of G′(xD) of
δx
∂x
In the following discussion we consider two limiting cases for Gstrain as a function of xD, which provides a range of values for which LGPS can be stabilized. The first case is that of a LGPS particle that decomposes hydrostatically and is a mean field approximation. The fraction of decomposed LGPS is assumed to be uniform throughout the particle (xD({right arrow over (r)})=xD for all {right arrow over (r)}). The second limiting case is that of spherically symmetric nucleation, where LGPS is completely decomposed within a spherical region of radius Ri (xD({right arrow over (r)})=1: r≤Ri) and pristine outside this region (xD({right arrow over (r)})=0: r>Ri). As is shown below, the hydrostatic case yields a lower limit for ∂xDGstrain whereas the nucleation model shows how this value could, in practice, be much higher.
Hydrostatic Limit/Mean Field Theory
The local stress σ({right arrow over (r)}) experienced by a subsection of an LGPS particle is directly a function of the decomposition profile xD({right arrow over (r)}) as well as the mechanical properties of the particle and, if applied, the mechanically constraining system. In the hydrostatic approximation, the local stress is said to be compressive and equal everywhere within the particle (σij({right arrow over (r)})=−pδij). In the mean field approximation, the same is said for the decomposed fraction xD({right arrow over (r)})=xD. Given the one-to-one relation between σ({right arrow over (r)}) and xD({right arrow over (r)}), these two approximations are equivalent.
We restrict focus to the limit as xD→0 to evaluate the metastability of LGPS about the pristine state. If ∂x
δx
∂x
The reference volume is the volume in the unconstrained system, Vref=(1−xD)VLGPS+xDVD. Combining equation 7 and equation 9 with the metastability condition ∂x
ϵRXNKeff>(GLGPS0−GD0)VLGPS−1 (10)
Equation 9 is solved for in
Spherical Nucleation Limit
The maximally localized (i.e. highest local pressure) decomposition mechanism is that of spherical nucleation as shown in
To fit the decomposed reference state of radius RD into the void of radius Ri, both the decomposed sphere and the remaining LGPS must become strained as shown in
In terms of the displacement vector of the decomposed and pristine materials, {right arrow over (u)}D({right arrow over (r)}) and {right arrow over (u)}P({right arrow over (r)}), and the radial stress components, σrrD({right arrow over (r)}) and σrrP({right arrow over (r)}), the boundary conditions are:
For a spherically symmetric stress in an isotropic material, the displacement vector is known to be of the form u(r)=Ar+Br−2, where the vector notation has been removed as displacement is only a function of distance from the center. The strain Gibbs for a compressed sphere under condition 2, defining p0=−σrr(d)(RD), gives the compressive term σx
(4/3πRo3)−1∂xDGstrain=p0(2+ϵRXN+¾pSp−1) (11)
Passivation Layer Theory
Electrolytes, either liquid or solid, are likely to react with electrodes where the electrode potential is outside of the electrolyte stability window. To address this, it is suggested that electrolytes be chosen such that they form a passivating solid-electrolyte-interface (SEI) that is at least kinetically stable at the electrode potential. Many works on the topic of improving sulfide electrolytes have speculated that by forming electronically insulating layers on the surface of sulfide electrolytes such passivation layers can be formed. In this section, we discuss the role of such passivation layers and provide a quantitative analysis of the mechanism by which we believe an electronically insulating surface layer improves stability.
In
δG=μLi
Applying conservation δNa=−δNc, δa=−δnc gives the well-known equilibrium conditions:
Or, in other words, the electrochemical potential (η=μ+zeϕ) of both the electrons and the lithium ions must be constant everywhere within the cell. As a result, the lithium metal potential (μLi=ηLi
Like equation 13, equation 14 leads to the condition that the lithium metal potential remains constant throughout the cell.
A speculated mechanism for passivation layer stabilization of sulfide electrolytes is depicted in
The authors believe that while an electronically insulating passivation layer is a key design parameter, the above theory is missing a critical role of effective electron conduction that occurs due to the ‘lithium holes’ that are created when a lithium ion migrates out of the insulated region, leaving behind the corresponding electron. The differential Gibbs energy of this system is represented by adding a solid-electrolyte term to equation 12 (denoted by superscript SE).
δG=μLi
The electron and lithium conservation constraints are now:
Constraints 1 and 2 represent the tethering of the electron and lithium density in the case of an insulated particle. Unlike the system governed by equation 12, the Fermi level of the solid-electrolyte is not fixed by an external voltage. The result is that by lowering the number of atoms within the solid-electrolyte by extracting lithium ions, and hence increasing the number of electrons per atom within the insulated region, the number of electrons per atom and the Fermi level increase. In effect, this represents the conduction of electrons by way of lithium-holes. Solving equation 15 for the equilibrium points given the above constraints lead to those of equation 14 between the anode/cathode as well as the following relation between the anode and solid-electrolyte.
The total voltage experienced within the SE can be represented as ϕSE−ϕ0SE−VS where ϕ0SE is the voltage in the absence of lithium extraction from the SE (the original voltage as depicted in
The ultimate result of this voltage relaxation within the electronically insulated region is depicted in
Intrinsically, this has no impact on the solid-electrolyte stability. However, in the limit of very low capacitances, as is expected, only a small fraction of the lithium ions would need to migrate to the anode for ϕ0SE−C−1eNSE≈0. Hence the electronically insulating shell traps the bulk of the lithium ions locally which maintains the high reaction strain needed for mechanical stabilization.
Results and Discussion
Electrochemical Stability
The impact of mechanical constriction on the stability of LGPS was studied by comparing decay metrics between LGPS and the same LGPS with an added core-shell morphology that provides a constriction mechanism. To minimize chemical changes, the constricting core-shell morphology was created using post-synthesis ultrasonication. This core-shell LGPS (“ultra-LGPS” hereafter) was achieved by high-frequency ultrasonication that results in the conversion of the outer layer of LGPS to an amorphous material. Bright-field transmission electron microscopy (TEM) images of the LGPS particles before (
The electrochemical stabilities of non-constricted LGPS and constricted ultra-LGPS were evaluated using cyclic-voltammetry (CV) measurements of Li/LGPS/LGPS+C/Ta (
In contrast, the decomposition of ultra-LGPS was largely suppressed, manifested by only one minor oxidation peak at a higher voltage (3V) during charging, and almost no reduction peak during discharging (
These stability advantages of ultra-LGPS over LGPS were found to be even more prominent when implemented in an all-solid-state half-cell battery. The cycling performance was measured for Li4T5O12 (LTO) mixed with carbon and either ultra-LGPS or LGPS as a cathode, ultra-LGPS or LGPS as a separator, and lithium metal as the anode. The cycling performance of each configuration was taken at low (0.02C), medium (0.1C), and high (0.8C) current rates. The results, depicted in
To isolate the decomposition of LGPS in the LTO cathode composite, the solid-electrolyte layers were replaced by a glass fiber separator.
In each of these results, those ultra-LGPS particles with core-shell morphologies have outperformed the stability of LGPS counterparts. As discussed in ref22, core-shell designs are proposed to stabilize ceramic-sulfide solid-electrolytes via the volume constraint placed on the core by the shell. This experimental electrochemical stability data agrees with this theory. Sulfur deficient shells, as seen in the case of ultra-LGPS, are expected to lower the effective compressibility of the system and hence increase the volume constraint22. The solid-state half-cell (solid-state cathode+glass fiber/liquid electrolyte+lithium metal anode) performance in the voltage range of 1-2.2 V vs lithium demonstrates that ultra-LGPS has, in practice, improved stability over LGPS in the cases of both LGPS oxidation and reduction. Additionally, the Coulombic efficiency of ultra-LGPS is also higher than that of LGPS, indicating an improved efficiency of charge transfer in the system, and less charge participation in unwanted side reactions.
Decomposition Mechanism
To better understand the mechanism by which LGPS decomposes, TEM analyses were performed to study the microstructure of LTO/[ultra-]LGPS interfaces after cycling. An FIB sample (
Since the composite cathode layer is composed of LTO, LGPS and C, there will be minor LTO/LGPS interfaces (hereafter “LTO/LGPS secondary interface”) that are ubiquitous within the cathode layer.
As comparison,
These results suggest that the nucleation limit is a more faithful representation of the true decay process than the hydrostatic limit. The sulfur rich particles formed in LGPS have a length scale on the order of Ri≈20 nm. In ultra-LGPS, the shell thickness is also roughly l≈20 nm. Hence if we consider the formation of such a sulfur particle near the core-shell boundary in ultra-LGPS, the minimum distance from the center of the sulfur rich particle to the exterior of the shell is Ro=Ri+l≈40 nm. In this case R≈8Ri3 which satisfies the condition Ri<<Ro needed to apply the nucleated model. In summary, we know that the LGPS decays via a mechanism that leads to nucleation of sulfur rich particles on the surface. We also know that applying a shell layer with a thickness such that l≈Ri inhibits such decay. These results suggest that the pristine core-shell state is at least metastable with respect to the decay towards the state with nucleated decay just below the core-shell interface.
Conclusions
In summary, we have developed a generalized strain model to show how mechanical constriction, given the nature of LGPS to expand upon decay, can lead to metastability in a significantly expanded voltage range. The precise level to which constriction expands the voltage window is depended on the morphology of the decay. We performed a theoretical analysis of two limits of the decay morphology, the minimally and maximally localized cases. The minimally localized case consisted of a mean field theory where every part of the particle decays simultaneously, whereas the maximally localized case consisted of a nucleated decay. It was demonstrated that, while the maximally localized case was best, both cases had the potential for greatly expanding the stability window. We also developed a theory for the role of an electrically insulating passivation layer in such a stain-stabilized system. This model suggests that such passivation layers aid in stability by keeping lithium ions localized within the particle, maximizing the reaction strain.
Experimental results for the stability performance of LGPS before and after the adding of a constricting shell supports this theory. After the formation of shell via ultrasonication, LGPS demonstrated remarkably improved performance cyclic voltammetry, solid-state battery cycling, and solid-state half-cell cycling. Because the shell was applied in a post-synthesis approach, chemical differences between the core-shell and pure LGPS samples, which might otherwise affect stability, were kept to a minimum. The core-shell is believed to be an instance of mechanically constrained LGPS as during any decomposition, the LGPS core will seek to expand whereas the shell will remain fixed. In order words, the shell provides a quasi-isovolumetric constraint on the core dependent on the biaxial modulus of the shell and the particle geometry.
Analysis of the decay morphology found in LGPS particles but not in ultra-LGPS particle suggests that the nucleated decay limit more accurately reflects the true thermodynamics. It was found that, in LGPS, nucleated sulfur-rich decay centers were embedded in the surface of the LGPS particles after cycling. Further, these nucleated decay centers were not found in the cycled ultra-LGPS. The ultra-LGPS maintained a shell thickness comparable to the decay cites in LGPS (approximately 20 nm), which was predicted to be sufficient for the high level of stabilization afforded by the nucleated model. These results, combined with the improved stability of ultra-LGPS, indicate that not only is strain-stabilization occurring, but that the magnitude at which it is occurring is dominated by maximally localized decay mechanism. This is a promising result as such nucleated decay has been shown to provide a larger value of ∂x
Methods
Sample Preparation
LGPS powder was purchased from MSE Supplies company. Ultra-LGPS was synthesized by soaking LGPS powder into organic electrolytes, such as dimethyl carbonate (DMC) and diethyl carbonate (DEC), and then sonicated for 70h in Q125 Sonicator from Qsonica company, a microprocessor based, programmable ultrasonic processor
Electrochemistry
The cyclic voltammograms (CV) of Li/LGPS/LGPS+C/Ta and Li/ultra-LGPS/ultra-LGPS/Ta cells were measured between 0.5 to 5 V at a scan rate of 0.1 mVs−1 on a Solartron electrochemical potentiostat (1470E), using lithium as reference electrode. The electrochemical impedance spectrums of Li/LGPS/LGPS+C/Ta and Li/ultra-LGPS/ultra-LGPS/Ta cells were measured at room temperature both before and after CV tests, by applying a 50 mV amplitude AC potential in a frequency range of 1 MHz to 0.1 Hz. The composite cathode used were prepared by mixing LTO, (ultra-)LGPS, polyvinylidene fluoride (PVDF) and carbon black with a weight ratio of 30:60:5:5. This mixture of powders was then hand-grinded in a mortar for 30 minutes and rolled into a thin film inside an argon-filled glove box. SEs were prepared by mixing (ultra-)LGPS and PVDF with a weight ratio of 95:5, then hand-grinding the mixed powder in a mortar for 30 minutes and finally rolling it into a thin film inside an argon-filled glove box. To assemble a solid-state cell, the prepared composite cathode thin film, (ultra-)LGPS thin film, and Li metal foil were used as cathode, solid electrolyte, and the counter electrode, respectively. The thin films of composite cathode and (ultra-)LGPS were cold-pressed together before assembling into the battery. A piece of glass fiber separator was inserted between (ultra-)LGPS thin film and Li metal foil to avoid interfacial reaction between these two phases. Only 1 drop of 1 M LiPF6 in ethylene carbonate (EC) and dimethyl carbonate (DMC) solution (1:1) was carefully applied onto the glass fiber to allow lithium ion conduction through the separator. Swagelok-type cells were assembled inside an argon-filled glove box. Assembling process of an (ultra-)LGPS battery is the same with that of an (ultra-)LGPS solid-state battery, except that the (ultra-)LGPS δE layer is removed. The charge/discharge behavior was tested using an ArbinBT2000 workstation (Arbin Instruments, TX, USA) at room temperature. The specific capacity was calculated based on the amount of LTO (30 wt %) in the cathode film.
Characterization
For FIB sample preparation, the cold-pressed thin film of composite cathode and (ultra-)LGPS after 1 charge-discharge cycle in (ultra)LGPS solid-state battery was taken out inside an argon-filled glove box. It was then mounted onto a SEM stub and sealed into a plastic bag inside the same glove box. FIB sample preparation was conducted on an FEI Helios 660 dual-beam system. The prepared FIB sample was then immediately transferred into JOEL 2010F for TEM and STEM EDS/EELS characterization.
Density Functional Theory Calculations
In order to allow comparability with the Material Project crystal database, all DFT calculations were performed using the Material Project criteria. All calculations were performed in VASP using the recommended Projector Augmented Wave (PAW) pseudopotentials. An energy cutoff of 520 eV with k-point mesh of 1000/atom was used. Compressibility values were found by discretely evaluating the average compressibility of the material between 0 GPa and 1 GPa. Enthalpies were calculated at various pressures by applying external stresses to the stress tensor during relaxation and self-consistent field calculations
Like liquid counterparts, the key performance metrics for solid-electrolytes are stability and ionic conductivity. For lithium systems, two very promising families of solid-electrolytes are garnet-type oxides and ceramic sulfides. These families are represented, respectively, by the high-performance electrolytes of LLZO oxide and LSPS sulfide. Oxides tend to maintain good stability in a wide range of voltages but often have lower ionic conductivity (<1 mS cm−1)1. Conversely, the sulfides can reach excellent ionic conductivities (25 mS cm−1)6,20 but tend to decompose when exposed to the conditions needed for battery operation.
Instabilities in solid-electrolytes can arise from either intrinsic material-level bulk decompositions or surface/interfacial reactions when in contact with other materials. At the materials-level, solid-electrolytes tend to be chemically stable (i.e. minimal spontaneous decomposition) but are sensitive to electrochemical reactions with the lithium ion reservoir formed by a battery cell. The voltage stability window defines the range of the lithium chemical potential within which the solid-electrolyte will not electrochemically decompose. The lower limit of the voltage window represents the onset of reduction, or the consumption of lithium ions and the corresponding electrons, whereas the upper limit represents the onset of oxidation, or the production of lithium ions and electrons. The voltage window affects the bulk of any solid-electrolyte particle as the applied voltage is experienced throughout. While interfacial reactions occur between the solid-electrolyte and a second ‘coating’ material at the point of contact, these reactions can either be two-bodied chemical reactions, where only the solid-electrolyte and the coating material are reactants, or three-bodied electrochemical reactions, in which the solid-electrolyte, coating material and the lithium ion reservoir all participate. The two types of reactions are state-of-charge or voltage independent and dependent, respectively, as determined by the participation of the lithium ion reservoir.
Prior studies have revealed that the most common lithium ion electrode materials, such as LiCoO2 (LCO) and LiFePO4 (LFPO), form unstable interfaces with most solid electrolytes, particularly the high performance ceramic sulfides. Successful implementation of ceramic sulfides in solid-state batteries may employ suitable coating materials that can mitigate these interfacial instabilities. These coating materials may be both intrinsically electrochemically stable and form electrochemically stable interfaces with the ceramic sulfide in the full voltage range of operation. In addition, if different solid-electrolytes are to be used in different cell components for maximum material-level stability, then the coating materials may also change to maintain chemically stable interfaces.
In short, the choice of a coating material depends on both the type of solid-electrolyte and the intended use of operation voltage (anode film, separator, cathode film, etc.). Pseudo-binary computational methods can approximately solve for the stability of a given interface, but are computationally expensive and have not yet been developed in very-large scale. A major performance bottleneck for high-throughput analysis of interfacial stability has been the cost to construct and evaluate many high-dimensional convex hulls. In the case of material phase stability, the dimensionality of the problem is governed by the number of elements. For example, calculating the interfacial chemical stability of LSPS and LCO would require a 6-dimensional hull corresponding to the set of elements {Li, Si, P, S, Co, O}. The electrochemical stability of this interface is calculated with the system open to lithium, so that lithium is removed from the set and the required hull becomes 5-dimensional ({Si, P, S, Co, O}).
Here we introduce new computational schemata to more efficiently perform interfacial analysis and hence enable effective high-throughput search for appropriate coating materials given both a solid-electrolyte and an operation voltage range. We demonstrate these schema by applying them to search through over 67,000 material entries from the Materials Project (MP) in order to find suitable coating materials for LSPS, which has shown the highest lithium conductivity of around 25 mS cm−1 , in the cases of both anode and cathode operations. Coating material candidates that are both intrinsically stable at the material level and form stable interfaces with LSPS within the prescribed voltage range are termed “functionally stable.”
To establish standards, we focus on finding anode coating materials which are functionally stable in a window of 0-1.5 volts versus lithium metal and cathode coating materials which are functionally stable in a window of 2-4 volts versus lithium metal. These voltage ranges are based on cycling ranges commonly found in today's lithium ion batteries. Within the anode range, we are particularly interested in finding materials that are stable at 0 volts versus lithium metal, as it could enable the use of lithium as a commercial anode material.
Due to remaining computational limitations, this work focuses only on those materials that require an LSPS interfacial hull-dimensionality of less than or equal to 8. In other words, materials were only considered if the elements present in that material consisted of {Li, Si, P, S} plus up to four additional elements. A total of 69,640 crystal structures in the MP database were evaluated for material-level voltage windows. Of those, 67,062 materials satisfied the less than 8-dimensional requirement and were accordingly evaluated for functional stability with LSPS. In total, over 1,000 MP entries were found to be functionally stable in the anode range and over 2,000 were functionally stable in the cathode range for LSPS. Experimental probing of interfacial stability is used for select materials to confirm these predictions.
Results and Discussion
Data Acquisition and Computational Efficiency
To efficiently evaluate the stability of the interface between each of these 67,062 potential coating materials and LSPS, two new computational schemata were developed. To minimize the number of hulls that must be calculated, the coating materials were binned based on elemental composition. Each unique set of elements requires a different hull, but elemental subsets can be simultaneously solved. For example, the calculation of interfacial stability between LSPS and iron-sulfate (Fe2(SO4)3) requires solving for the convex hull of the 6-dimensional element set {Li, Si, P, S, Fe, O}. This hull is the same hull that must be calculated for the interface with LFPO and includes, as a subset, the 5-dimensional hull needed for the evaluation of iron-sulfide (FeS). To capitalize on this, rather than iterate through each of the 67,062 materials and calculate the hull needed for that material, the minimum number of elemental sets that spans the entirety of the materials were determined (
The second schema used to minimize computational cost was a binary search algorithm for determining the pseudo-binary once a hull was calculated. The pseudo-binary approach is illustrated in
(1−x)LSPS+xA→diDi (1)
The pseudo-binary is a computational approach that determines for which value of x the decomposition described by equation 1 is the most kinetically driven (e.g. when is the decomposition energy the most severe). The RHS of equation 1 represents the fraction ({di}) of each of the thermodynamically favored decay products and defines the convex hull for a given x in terms of the products' Gibbs energies (Hull(x)=Σdi(x)Gi). The total decomposition energy accompanying equation 1 is:
G
hull(x)=Σdi(x)Gi−(1−x)GLGPS−xGA (2)
The most kinetically driven reaction between LSPS and the coating material is the one that maximizes the magnitude (i.e. most negative) of equation 2, which defines the parameter xm.
max|Ghull(x)|≡|Ghull(xm) (3)
This maximum decomposition energy is the result of two factors. The first, denoted Ghull0, is the portion of the decomposition energy that is due to the intrinsic instability of the two materials. In terms of the decomposed products of LSPS (DLSPS) and the coating material (DA), Ghull0(x) is the decomposition energy corresponding to the reaction (1−x)LSPS+xA→(1−x)DLSPS+xDA. By subtracting this materials-level instability from the total hull energy, the effects of the interface (G′hull) can be isolated as defined in equation 4.
G′
hull(x)=Ghull(x)−Ghull0(x) (4)
Physically, Ghull0(x) represents the instability of the materials when separated and G′hull(x) represents the increase in instability caused by the interface once the materials are brought into contact.
In this work, to determine the added instability of each interface at the most kinetically driven fraction (G′(xm)), we implement a binary search algorithm (see Methods) that uses the concavity of the hull to find xm to within 0.01% error. This binary search approach finds the xm value in 14 steps of hull evaluations. A more traditional linear evaluation of the hull to 0.01% accuracy would require 10,000 equally spaced evaluations from x=0 to x=1. This increase of speed is leveraged to efficiently search the 67,062 material entries for functional stability.
Functional Stability
Functional stability at a given voltage was determined for each of the 67,062 materials by requiring that (i) the material's intrinsic electrochemical stability per atom at that voltage was below thermal energy (|Ghull(x=1)|≤kBT) and (ii) that the added interfacial instability at the given voltage was below thermal energy (|G′hull(xm)|≤kBT). Under these conditions, the only instability in the system is that of the LSPS intrinsic material-level instability, which can be stabilized via strain induced methods22. Of the 67 k materials, 1,053 were found to be functionally stable in the anode range (0-1.5 V vs. lithium metal) and 2,669 were found to be functionally stable in cathode range (2-4 V vs. lithium metal). Additionally, 152 materials in the anode range and 142 materials in the cathode range were determined to violate condition (i) but only decompose by lithiation/delithation. The practical use of such materials as an LSPS coating material depends on the reversibility of this lithiation/delithiation process, as such these materials are referred to as potentially functionally stable. All functionally stable and potentially functionally stable materials are cataloged in the supplementary information and indexed by the corresponding Materials Project (MP) id.
The correlation between each element's atomic fraction and the interfacial stability is depicted in
Anionic Species Impact on Material-Level Stability
Given the high correlation contrast for anionic species with respect to interfacial stability, analysis of the dataset in terms of anionic composition was performed. To eliminate overlap between the datapoints, the only compounds that were considered were those that are either monoanionic with only one of {N, P, O, S, Se, F, 1} or oxy-anionic with oxygen plus one of {N, S, P}. 45,580 MP entries met one of these criteria as is outlined in Table 3. The percentage of each anionic class that was found to be electrochemically stable at the material-level is also provided.
The average hull energy of each anionic class is given in 0.5V steps from 0-5V in
Anionic Species Impact on Interface-Level Stability
The average values of total decomposition energy (Ghull(xm)) and the fraction that is a result of the interface instability (G′hull(xm)) are depicted in
The average interface-level contribution for electrochemical decomposition is shown in
Anionic Species Impact of Functional Stability
The total number of each anionic class that were determined to be functionally stable or potentially functionally stable are given in
Experimental Comparison
The chemical compatibility between various coating materials and LSPS were tested experimentally by hand-milling the mixture powder of LSPS and coating materials with/without high-temperature annealing, followed by X-ray diffraction (XRD) measurements at room temperature. Any chemical reaction between the powder will cause compositional and structural changes in the original phases, which can be detected by the change of peak positions and intensities in XRD patterns. It is worth noting that even interfacial reactions are predicted to happen based on thermodynamic calculations, a certain amount of energy may be needed to overcome the kinetic energy barrier for these reactions to happen4. Therefore, the mixed powders were annealed at high temperatures (300° C., 400° C., 500° C.) to determine the onset temperature of interfacial reactions as well as the reaction products, and to further assess the role of kinetics by comparing these results with the DFT computed thermodynamic reaction products.
The electrochemical stability of typical coating materials is characterized by Cyclic Voltammetry (CV) technique, in which the decomposition of the tested coating material can be manifested by current peaks at certain voltages relevant to Lithium. Two typical coating materials were used as a demonstration to show good correspondence between our theoretical prediction and experimental observation. The CV test of Li2S (
Methods
Data Acquisition
The data used in this work was the result of prior Density Functional Theory calculations that were performed as part of the Materials Project (MP) and was interfaced with using the Materials Application Programming Interface (API). The Python Materials Genomics (pymatgen) library was used to calculate convex hulls. Of the initial 69,640 structures that were evaluated, 2,578 structures were not considered due to requiring hulls of dimension equal to or greater than 9.
Elemental Set Iterations
To minimize the computational cost of analyzing all 67,062 structures, the smallest number of elemental sets that spanned all the materials were determined. To do this, the set of elements in each structure were combined with the elements of LSPS, resulting in a list of element sets with each set's length equal to the dimensionality of the required hull for that material. This list was ordered based on decreasing length of the set (e.g. ordered in decreasing dimensionality of the required hull). This set was then iterated through and any set that equals to or is a subset of a previous set was removed. The result was the minimum number of elemental sets, in which every material could be described.
Chemical decomposition hulls were calculated using the energies and compositions from the MP. Changes in the volume and entropy were neglected (ΔG≈ΔE). Similarly, electrochemical decomposition hulls were founded by using the lithium grand canonical free energy and subtracting a term μLiNLi from the energies (ΔΦ≈ΔE−μLiΔNLi), where μLi is the chemical potential of interest and NLi is the number of lithium ions in the structure. After a hull was calculated, it was used to evaluate every material that exists within the span of its elemental set.
The Pseudo-Binary
The pseudo-binary, as described in section 2, seeks to find the ratio of LSPS to coating material such that the decomposition energy is the most severe and, hence, is the most kinetically driven. This problem is simplified by using a vector notation to represent a given composition by mapping atomic occupation to a vector element. For example, LiCoO2→(1 1 2) in the basis of (Li Co O), meaning that there are 1 lithium, 1 cobalt, and 2 oxygen in the unit formula. Using this notation, the decomposition in equation 1 can be written in vector form.
Using ū to represent a vector and Ū to represent a matrix, equation 5 becomes:
The relative composition derivatives for each decay product can be found by inverting
∂x
Equation 7 allows for the calculation of the derivative of the hull energy with respect to the fraction parameter x.
By using equation 7, and the fact that the hull is a convex function of x, a binary search can be performed to find the maximum value of Ghull and the value at which it occurs xm. This process consists of first defining a two-element vector that defines the range in which xm is known to exist xrange=(0,1) and an initial guess xD=0.5. Evaluating the convex hull at the initial guess yields the decomposition products {Di} and the corresponding energies {GD
Equations 5-8 are defined for chemical stability. In the case of electrochemical (lithium open) stability, the free energy is replaced with Φi=Gi−μNi where μ is the chemical potential and Ni is the number of lithium in structure i. Additionally, lithium composition is not included in the composition vectors of equation 6 to allow for the number of lithium atoms to change.
X-Ray Diffraction
The compatibility of the candidate materials and solid electrolyte was investigated at room temperature (RT) by XRD. The XRD sample was prepared by hand-milling the candidate materials (LCO, SnO2, SiO2, LTO) with LSPS powder (weight ratio=55:30) in an Ar-filled glovebox. To test the onset temperature of reactions for candidate materials and LSPS solid electrolyte, the powder mixtures were well spread on a hotplate to heat to different nominal temperatures (300, 400 and 500 degree Celsius) and then characterized by XRD.
XRD tests were performed on Rigaku Miniflex 600 diffractometer, equipped with Cu Kα radiation in the 2-theta range of 10-80°. All XRD sample holders were sealed with Kapton film in Ar-filled glovebox to avoid air exposure during the test.
Cyclic Voltammetry
Candidate coating materials (Li2S and SiO2), carbon black, and poly(tetra-fluoroethylene) (PTFE) were mixed together in a weight ratio of 90:5:5 and hand-milled in an Ar-filled glovebox. The powder mixtures were sequentially hand-rolled into a thin film, out of which circular disks ( 5/16-inch in diameter, ˜1-2 mg loading) were punched out to form the working electrode for Cyclic Voltammetry (CV) test. These electrodes were assembled into Swagelok cells with Li metal as the counter electrode, two glass fiber separators and commercial electrolyte (1 M LiPF6 in 1:1 (volumetric ratio) ethylene carbonate/dimethyl carbonate (EC/DMC) solvent).
CV tests were conducted by Solartron 1455A with a voltage sweeping rate of 0.1 mV/s in the range of 0-5V at room temperature, to investigate the electrochemical stability window of the candidate coating materials (Li2S and SiO2).
Conclusion
Our high-throughput pseudo-binary analysis of Material Project DFT data has revealed that interfaces with LSPS decay via dominantly chemical means within the range of 1.5 to 3.5 V and electrochemical reduction [oxidation] at lower [higher] voltages. The fraction of decomposition energy attributed to interfacial effects disappears as the voltage approaches 0V. This result suggests that all material classes tend to decay to maximally lithiated Li binary and elemental compounds at low voltage, in which case the presence of the interface has no impact.
In terms of anionic content, we see that appropriately matching operational conditions to the coating material is paramount. Sulfur and selenium containing compounds, for example, demonstrate a very high chance to be functionally stable (>25% among all sulfides and selenides) in the 2-4V cathode range. However, less than 1% of these same materials form a functionally stable coating material in the 0-1.5V anode range, where iodine, phosphorous and nitrogen have the highest performance. Oxygen containing compounds have a high number of phases that are functionally stable in both voltage regions, but the percentage is low due to the even higher number of oxygen containing datapoints.
We show that an advanced mechanical constriction method can improve the stability of lithium metal anode in solid state batteries with LGPS as the electrolyte. More importantly, we demonstrate that there is no Li dendrite formation and penetration even after a high rate test at 10 mA cm−2 in a symmetric battery. The mechanical constriction method is technically realized through applying an external pressure of 100 MPa to 250 MPa on the battery cell, where the Li metal anode is covered by a graphite film (G) that separates the LGPS electrolyte layer in the battery assembly. At the optimal Li/G capacity ratio, it exhibits excellent cyclic performances in both Li/G-LGPS-G/Li symmetric batteries and Li/G-LGPS-LiCoO2 (LiNbO3 coated) batteries. Upon cycling, Li/G anode transforms from two layers into one integrated composite layer. Comparison between Density Functional Theory (DFT) data and X-ray Photoelectron Spectroscopy (XPS) analysis yields the first ever direct observation of mechanical constriction controlling the decomposition reaction of LGPS. Moreover, the degree of decomposition is seen to become significantly suppressed under optimum constriction conditions.
Design of Li/Graphite Anode
We first investigated the chemical stability between LGPS and (lithiated) graphite through the high temperature treatment of their mixtures at 500° C. for 36 hours inside the argon filled glovebox for an accelerated reaction. XRD measurements were performed on different mixtures before and after heat treatment, as shown in
The Li/graphite anode was designed as shown in
Cyclic and Rate Performance of Li/Graphite Anode
The electrochemical stability and rate capability of Li/graphite (Li/G) anode was tested with anode-LGPS-anode symmetric battery design under 100 MPa external pressure. The comparison of cyclic performance between Li/G-LGPS-G/Li and Li-LGPS-Li batteries is shown in
We also compared the rate performance of Li/G symmetric battery under different external pressures of 100 MPa or 3 MPa as shown in
To further understand the influence of the Li/G composite formed by battery cycling on its high rate performance, a battery test was designed like
Based on the above understanding, we further lowered the current density for the initial cycles to 0.125 mA cm−2 and cycled with the same capacity of 0.25 mAh cm−2 for a more homogeneous Li distribution and storage in the Li/G composite for improved lithium transfer kinetics. As shown in
Li/Graphite Anode in all-Solid-State Battery
We first performed DFT simulations of LGPS decomposition pathways in the low voltage range of 0.0-2.2V versus lithium metal. Mechanical constriction on the materials level was parameterized by an effective bulk modulus (Keff) of the system. Based on the value of this modulus, the system could range from isobaric (Keff=0) to isovolumetric (Keff=∞). Expected values of Keff in real battery systems were on the order of 15 GPa. In the following, these simulation results were used to interpret XPS results of the valence changes of Ge and P from LGPS in the solid state batteries after CV, rate and cycling tests.
As shown in
It is worth noting that while the applied pressure and the effective modulus (Keff) were both measured in units of pressure, they are independent. The effective modulus represents the intrinsic bulk modulus of the electrolyte added in parallel with the finite rigidity of the battery system. Accordingly, Keff measures the mechanical constriction that can be realized on the materials level in any single particle, while the external pressure applied on the operation of solid state battery enforced the effectiveness of such constriction on the interface between particles or between electrode and electrolyte layers. This is because exposed surface was the most vulnerable to chemical and electrochemical decompositions, while a close interface contact enforced by external pressure will minimize such surface. Thus, even though the applied pressure was only on the order of 100 MPa, the effective bulk modulus was expected to be much larger. In-fact, close packed LGPS particles should experience a Keff of approximately 15 GPa. The applied pressure of 100-250 MPa was an effective tool for obtaining this close packed structure. In short, the applied pressure minimizes gaps in the bulk electrolyte, allowing for the effective modulus that represents the mechanical constriction on the materials level to approach its ideal value of circa 15 GPa.
The XPS results of LGPS that was either in direct contact with a lithium or lithium-graphite anode, as well as bulk LGPS during battery cycling are provided in
We first investigate the function of Li/G composite in comparison with pure lithium metal at a slow rate of 0.25 mA/cm2 under 100 MPa external pressure (
When the cycle rate was increased to 2 mA/cm2 and 10 mA/cm2, the observed decompositions on the L/G-LGPS interface under external pressures in
These two competing reactions with thermodynamic and kinetic preferences, respectively, can be understood by considering a current dependent overpotential (η′(i)) for each of these two competing reactions (η→η+η′(i)). This η′ term would arise from kinetic effects such as ohmic losses, etc. When current is small (i≈0), η′ disappears, thus the thermodynamic overpotential (7) dominates and favors the ground state decomposition products of
The impedance profiles before and after CV test (
Conclusion
A lithium-graphite composite allows the application of a high external pressure during the test of solid-state batteries with LGPS as electrolyte. This creates a high mechanical constriction on the materials level that contributes to an excellent rate performance of Li/G-LGPS-G/Li symmetric battery. After cycling at high current densities up to 10 mA cm−2 for such solid-state batteries, cycling can still be performed normally at low rates, suggesting that there is no lithium dendrite penetration or short circuit. The reduction pathway of LGPS decomposition under different mechanical constrictions are analyzed by using both experimental XPS measurements and DFT computational simulations. It shows, for the first time, that under proper mechanical constraint, the LGPS reduction follows a different pathway. This pathway, however, can be influenced kinetically by the high current density induced overpotential. Therefore, the decomposition of LGPS is a function of both mechanical constriction and current density. From battery cycling performance and impedance test, it is shown that high mechanical constriction along with the kinetically limited decomposition pathway reduces the total impedance and realizes a LGPS-lithium metal battery with excellent rate capability.
Methods
Electrochemistry
Graphite thin film is made by mixing active materials with PTFE. The weight ratio of graphite film is graphite:PTFE=95:5. All the batteries are assembled using a homemade pressurized cell in an argon-filled glovebox with oxygen and water <0.1 ppm. The symmetric battery (Li/G-LGPS-G/Li or Li-LGPS-Li) was made by cold pressing three layers of Li(/graphite)-LGPS powder-(graphite/)Li together and keep at different pressures during battery tests. The batteries were charged and discharged at different current densities with the total capacity of 0.25 mAh cm−2 for each cycle. A LiCoO2 half battery was made by cold pressing Li/graphite composite-LGPS powder-Cathode film using a hydraulic press and keep the pressure at 100-250 MPa. The LiCoO2 were coated with LiNbO3 using sol-gel method. The weight ratio of all the cathode films was active materials:LGPS:PTFE=68:29:3. Battery cycling data were obtained on a LAND battery testing system. The cyclic performance was tested at 0.1 C at 25° C. The CV test (Li/G-LGPS-LGPS/C) was conducted on a Solartron 1400 cell test system between OCV to 0.1V with the scan rate of 0.1 mV/s. The LGPS cathode film for CV test is made with LGPS:super P:PTFE=87:10:3.
Material Characterization
XRD: The XRD sample was prepared by hand milling LGPS powder with lithium metal and/or graphite with weight ratio=1:1 in a glovebox. The powder mixtures were put on a hotplate and heated to the nominal temperature (500° C.) for 36 hours and then characterized by XRD. XRD data were obtained using a Rigaku Miniflex 6G. The mixtures of LGPS and graphite before and after high temperature treatment were sealed with Kapton film in an argon-filled glovebox to prevent air contamination.
SEM and XPS: Cross-section imaging of the pellet of Li/graphite-LGPS-graphite-Li was obtained by a Supra 55 SEM. The pellet was broken into small pieces and attached onto the side of screw nut with carbon tape to make it perpendicular to the beam. The screw nuts with samples were mounted onto a standard SEM stub and sealed into two plastic bags inside an argon-filled glove box. FIB-SEM imaging was conducted on an FEIHelios 660 dual-beam system. The XPS was obtained from a Thermo Scientific K-Alpha+. The samples were mounted onto a standard XPS sample holder and sealed with plastic bags as well. All samples were transferred into vacuum environment in about 10 seconds. All XPS results are fitted through peak-differentiating and imitating via Avantage.
Computational Methods All DFT calculations were performed using the Vienna Ab-initio Simulation Package (VASP) following the Material Project calculation parameters.32 A K-point density of 1000 kppa, a cutoff of 520 eV, and the VASP recommended pseudopotentials were used. Mechanically constrained phase diagrams were calculated using Lagrange minimization schemes as outlined in Ref. 13 for effective moduli of 0, 5, 10 and 15 GPa. All Li—Ge—P—S phases in the Material Project database were considered. Bader charge analysis and spin polarized calculations were used to determine charge valence.
In this work, we focused on how the external application of either high-pressure or isovolumetric conditions can be used to stabilize LGPS at the materials level through the control at the cell-level. This advances beyond the microstructural level mechanical constraints present in previous works, where particle coatings were used to induce metastability. Under proper mechanical conditions, we show that the stability window of LGPS can be widened up to the tool testing upper limit of 9.8 V. Synchrotron X-ray diffraction (XRD) and x-ray absorption spectroscopy (XAS) that measure the structure changes of LGPS before and after high-voltage holding show, for the first time, direct evidence of LGPS straining during these electrochemical processes. Both thermodynamic and kinetic factors are further considered by comparing density functional theory (DFT) simulations and x-ray photoelectron spectroscopy (XPS) measurements for decomposition analysis beyond the voltage stability window. These results suggest that mechanically-induced metastability stabilizes the LGPS up to approximately 4V. Additionally, from 4-10V, the local stresses experienced by decomposition amid rigid mechanical constraints leads to kinetic stability. Combined, mechanically-induced metastability and kinetic stability allow expansion of the voltage window from 2.1V to nearly 10V. To demonstrate the utility of this approach for practical battery systems, we construct fully solid-state cells using this method with various cathodes materials. Li4Ti5O12 (LTO) anodes are paired with LiCo0.5Mn1.5O4 (LCMO), LiNi0.5Mn1.5O4 (LNMO) and LiCoO2 (LCO) cathodes to demonstrate the high-voltage stability of constrained LGPS. To further probe the electrochemical window of LGPS, we report the first all-solid-state battery based on lithium metal and LiCo0.5Mn1.5O4, which can be charged to 6-9 V and cycled up to 5.5 V.
Results
To illustrate how mechanical constraint influences the electrochemical stability of LGPS, cyclic voltammetry (CV) tests of LGPS+C/LGPS/Li cells were performed (
As shown in
The synchrotron XRD of LGPS from the isovolumetric cell, as shown in
This strain effect was further elucidated from XAS measurement and analysis.
In theory, given an unconstrained reaction in which LGPS decomposes with a Gibbs energy change of ΔGchem<0, the reaction can be inhibited by the application of a mechanical constraint with effective bulk modulus (Keff) if:
ΔGchem+KeffϵRXN>0 (1)
Where V is the reference state volume and ERXN is the stress-free reaction dilation—in other words ϵRXN is the fractional volume change of LGPS following decomposition in the absence of any applied stress. The effective bulk modulus of equation one is the bulk modulus of the ceramic sulfide (Kmaterial) added in parallel with the mechanical constraint as given in equation 28:
K
eff
−1
=K
material
−1
+K
constraint
−1 (2)
Minimization of free energy in the mechanically constrained ensemble allows for calculating the expanded voltage window and the ground state decomposition products. Using ab-initio data,
The exact decomposition products predicted by DFT without considering the thermal tolerance are shown in
In contrast, the calculated thermodynamic stability limit of LGPS reaches nearly 4V at Keff=15 GPa. Accordingly, there was no oxidization of S and a very small amount of oxidized P was observed in the condition of strongly constrained LGPS at 3.2V in FIGS. 47C3 and D3. This small amount of oxidized P could be attributed to the ineffective constraint from the device or the voltage is close to the thermodynamic voltage. Furthermore, beyond the voltage stability limit for the case of 9.8 V, the solid-state battery showed less oxidized S or P than it was expected. Note that from
The application of the mechanical constraint can greatly reduce the speed at which ceramic sulfides decay as depicted in
The proposed mechanism for mechanically-induced kinetic stability is depicted in
However, even when Equation 1 is violated, the speed with which the front propagates into the pristine LGPS will still be influenced by the application of mechanical constraint. This is illustrated in
Given that ρi(p) can quickly grow with constriction, it is to be expected that this overpotential becomes significant at high pressures. This effect can be seen by comparing the expected constriction with prior molecular dynamics results of constricted cells. The pressure on the decomposition front is given by p=KeffϵRXN and the elastic volume strain of the material at that pressure is p=KmaterialϵV. Since the strain of a single lattice vector is approximately ϵ=⅓ϵy, the strain of the ab-plane of LGPS near the front is expected to be on the order of
For well constrained systems where Keff≈Kmaterial, this strain can easily reach 4%, as ϵRXN exceeds 30% at high voltages. Given that the activation energy for Li migration in LGPS is predicted to increase from 230 meV to 590 meV upon constriction by 4%, the rate at which lithium reordering can occur decreases by a factor of:
This many order of magnitude reduction in the possible reordering rate can explain why, for any voltage below 10V, the isovolumetric cell showed virtually no decomposition current.
XAS measurement shows a pre-edge on the intensity of S element while no pre-edge is found from P (
Interfacial reactions between two materials (i.e. LGPS and a cathode material) present computational challenges as ab-initio simulations of the interface present unique burdens. Instead, the preferred method to simulate both chemical and electrochemical stabilities of interfaces are the so-called pseudo-phase (also known as pseudo-binary) methods. In these methods, a linear combination of the materials of interest are taken and represented as a single phase with both composition and energy given by the linear combination. This phase is the pseudo-phase. Conventional stability calculations can then be applied to the pseudo-phase to estimate the reaction energy of the interface.
Usually, lithium metal is soft and which leads to the difficulty of applying pressure due to the immediate short of lithium through the bulk solid electrolyte. In order to probe the high voltage capability of pressurized LGPS in the system of lithium metal solid-state battery, lithium metal was used as anode with a graphite layer as a protection layer, which allows high pressure applied during battery test. Firstly, lithium metal-LCO batteries were made at different mechanical conditions using Swagelok, aluminum pressurized cell and stainless-steel pressurized cell, as shown in
To contrast this performance with conventional electrolytes,
In summary, we demonstrate how mechanical constraint widens the stability of ceramic solid electrolyte, pushing up its electrochemical window to levels beyond organic liquid electrolytes. A CV test shows that properly designed solid-state electrolytes working under isovolumetric conditions can operate up to nearly 10 V, without clear evidence of decomposition. A mechanism for this mechanically induced kinetic stability of sulfides solid-electrolytes is proposed. Moreover, based on this understanding, it has been shown how several high-voltage solid-state battery cells, using some of the most commonly used and promising cathode materials, can operate up to 9 V under isovolumetric conditions. Therefore, the development of high-voltage solid-state cells is not compromised by the stability of the electrolyte anymore. We anticipate that this work is an import breakthrough for the development of new energy storage systems and cathode materials focused on very-high voltage (>6V) electrochemistry.
Method
Sample Characterization
Structural Analysis
Routine XRD data were collected in a Rigaku Miniflex 6G diffractometer working at 45 kV and 40 mA, using CuKα radiation (wavelength of 1.54056 Å). The working conditions were 26 scanning between 10-80°, with a 0.02° step and a scan speed of 0.24 seconds per step.
Electrochemical Characterization
The LGPS+C/LGPS part of the cells were pellets which were made by pressing the powder at 1T, 3T, 6T, respectively, and put into Swagelok or the homemade pressurized cell. In the CV test, voltage starting from the open circuit voltage to 10 V was ramped, during which the decomposition currents at each voltage were measured. The CV test was conducted on a Solartron 1400 electrochemical test system between OCV to 3.2V, 7.5V, and 9.8V, respectively, with the scan rate of 0.1 mV/s. The CV scan was followed by a voltage hold for 10 hours to make sure the decomposition is fully developed, and it was scanned back to 2.5V before any other characterizations. The electrochemical impedance spectroscopy (EIS) was conducted on the same machine in the range of 3 MHz to 0.1 Hz.
For all-solid-state batteries, the electrode and electrolyte layers were made by a dry method which employs Polytetrafluoroethylene (PTFE) as a binder and allows to obtain films with a typical thickness of 100-200 μm. Additionally, two different kinds of all-solid-state batteries were assembled, using Li4Ti5O12 (LTO) or lithium (Li) metal as anode. In any case, the composite cathode was prepared by mixing the active materials (LiCo0.5Mn1.5O4, LiNi0.5Mn1.5O4 or LiCoO2) and Li10GeP2S12 (LGPS) powder in a weight ratio of 70:30 and 3% extra of PTFE. This mixture was then rolled into a thin film. On the one hand, for those all-solid-state batteries which use LTO as anode, a separator of LGPS and PTFE film was employed with a weight ratio of 95:5. The anode composition consists in a mixture of LGPS, LTO and carbon black in weight ratio 60:30:10 and 3% extra of PTFE. Finally, the Swagelok battery cell of cathode film (using LiCo0.5Mn1.5O4, LiNi0.5Mn1.5O4 or LiCoO2 as active material)/LGPS film/LTO film was then assembled in an argon-filled glove box. The specific capacity was calculated based on the amount of LTO (30 wt %) in the anode film. The galvanostatic battery cycling test was performed on an ArbinBT2000 work station at room temperature. On the other hand, when lithium metal was used as anode, a Li metal foil with a diameter and thickness of ½″ and 40 μm, respectively, was connected to the current collector. In order to prevent interface side reactions, the Li foil was covered by a 5/32″ diameter carbon black film with a weight ratio of carbon black and PTFE of 96:4. After loading the negative electrode into a Swagelok battery cell, 70 mg of pure LGPS powder, which acts as a separator, was added and slightly pressed. Finally, −1 mg film of the cathode composite LCMO was inserted and pressed up to 6 Tn (0.46 GPa) to form the battery, which final configuration was LCMO/LGPS pellet/graphite film+Li metal. For high voltage test in
Computational Simulation
All ab-initio calculations and phase data were obtained following the Material Project calculation guidelines in the Vienna Ab-initio Software Package (VASP). The mechanically-induced metastability calculations were performed following the LaGrangian optimization methods outlined in Small 1901470, 1-14 (2019) and J. Mater. Chem. A (2019). doi:10.1039/C9TA05248H). Pseudo-phase calculations were performed following the methods of J. Mater. Chem. A 4, 3253-3266 (2016), Chem. Mater. 28, 266-273 (2016), and Chem. Mater. 29, 7475-7482 (2017).
Other embodiments are in the claims.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US2019/063354 | 11/26/2019 | WO | 00 |
| Number | Date | Country | |
|---|---|---|---|
| 62771319 | Nov 2018 | US |
