Stress Evolution of Thin Film RuO2 Li-ion Battery Electrodes
by
Brian Mills
Submitted to the Department of Physics in partial fulfillment of the requirements for the degree of
Bachelor of Science in Physics
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
May 2020
© Massachusetts Institute of Technology 2020. All rights reserved.
Author ...... Brian Mills Department of Physics May 8, 2020
Certified by...... Carl Thompson Professor of Material Science and Engineering Thesis Supervisor
Certified by...... Joseph Checkelsky Associate Professor of Physics Thesis Supervisor
Accepted by ...... Nergis Mavalvala Associate Department Head, Department of Physics
Stress Evolution of Thin Film RuO2 Li-ion Battery Electrodes
by
Brian Mills
Submitted to the Department of Physics on May 8, 2020 in partial fulfillment of the requirements for the degree of Bachelor of Science in Physics
Abstract Thin Film Li-ion batteries (TFB) are seen as a promising candidate for powering small, low power microelectronic devices as they exhibit high energy density and can operate reliably at low voltages. Currently the biggest obstacle to TFB battery development is high volume expansion and material degradation in electrodes with high theoretical Li ion capacities. Among these materials is RuO2, which exhibits excellent capacity and great potential for use as either cathode or anode in low power electronics. In order to better understand the mechanisms that underlie mechanical failure in RuO2, we perform the first in situ measurement of mechanical stress evolution in thin film RuO2 electrodes. The results of these measurement reveal a very unique stress evolution pattern in RuO2, which has not been observed or modeled in any previous experiment, exhibiting near zero stress delithiation and linear increase in stress during lithiation. These results point to a mode of failure of RuO2 which does not occur in other materials currently being studied.
Thesis Supervisor: Carl Thompson Title: Professor of Material Science and Engineering
Thesis Supervisor: Joseph Checkelsky Title: Associate Professor of Physics
3 Contents
1 Introduction5
1.1 Thin Film Battery Materials...... 6 1.2 Ruthenium Oxide Electrodes...... 9
2 Stress in Thin Films 11
2.1 Introduction...... 11 2.2 Stress-Strain Curve...... 15 2.3 Thin Film Curvature...... 17 2.4 In Situ Stress Measurements...... 20
3 Experimental Setup 22
3.1 Sample Fabrication...... 22 3.2 Cell Construction and Cycling...... 23
4 Results and Discussion 25
4.1 Stress and Electrochemical Measurements...... 25 4.2 Discussion...... 29 4.2.1 Crack Closure...... 30 4.2.2 Conversion Reaction Film Growth...... 32 4.2.3 General Remarks...... 32
5 Summary and Future Work 33
4 1 Introduction
Li-ion batteries (LIB) have seen extensive use in the production of modern day elec- tronics. Their high power density and capacity, as well as reliable cycling has made them the ideal choice for secondary (ie., rechargeable) batteries in many applications. Advances in LIB design have been necessary in accommodating the development of more powerful and smaller electronic devices [1]. The sustained movement towards internet of things (IoT) technology has further increased demand for highly efficient, reliable and compact power supplies, especially for systems with low operating volt- ages and currents. An exceedingly promising candidate for these types of applications is the thin film Li-ion battery (TFB), which has shown considerably larger gravimetric (per unit mass) and volumetric (per unit volume) capacities than other battery designs [2]. TFBs are also entirely built from solid state components, which addresses long- standing safety concerns with the volatility of conventional liquid electrolyte batter- ies. Fabrication of TFBs generally consists of sequentially depositing 100 nm-1 µm thick layers of various materials using film deposition techniques such as DC and RF sputtering. Sometimes conditioning sequences, such as thermal annealing, are con- ducted between these depositions to optimize performance of certain materials [3]. Figure 1.1 shows the layout of a typical TFB.
Figure 1.1: A schematic of a typical layering of materials used to construct a TFB.
5 1.1 Thin Film Battery Materials
Advances in TFBs have largely been driven by breakthroughs in new materials and fabrication methods. There are significant benefits to the development of TFBs in which all components are compatible with complementary metal-oxide-semiconductor (CMOS) fabrication methods that are the standard in manufacture of modern micro- electronics. Such an advancement would allow for microchips with completely inte- grated power supplies, constituting a large increase in the efficiency of manufacturing as a whole.
Many of the currently commercialized TFBs for integrated circuits follow a stan- dard design: lithium metal-oxide cathode, solid electrolyte, and lithium metal anode. The electrode materials are reactants in an oxidation reduction reaction, requiring the exchange of Li ions and electrons between the cathode and anode. The elec- trolyte allows for movement of Li ions between the electrodes while also electrically isolating them. In order to complete the reaction, electrons flow through an exter- nal circuit that connects the electrodes, generating current. Lithium metal-oxides currently in use as cathode materials, such as LiCoO2 and LiMn2O4, have theoretical −1 gravimetric capacities of 150-300 mA h g and theoretical volumetric capacities that −3 −3 exceed 1000 mA h cm for some species, with ~500 mA h cm being successfully harnessed in commercialized designs. The solid electrolyte that is almost exclusively used in TFBs is lithium phosphorus oxynitride — LiPON. LiPON is a glassy, amor- phous solid and is appealing as an electrolyte because of its relatively good ionic conductivity and mechanical stability it provides to battery structure. It is this added stability from LiPON which allows Li metal to be used as the anode in TFBs, as it is found that it suppresses dendrite growth in the cell which lead to electrical shorts
6 between the electrodes – a leading cause of exploding batteries[4]. In bulk batter- ies where Li metal cannot reliably be used, graphite is the preferred material, and as a result has also been extensively studied for use in TFBs. It has theoretical gravi- −1 −3 metric and volumetric capacities of 339 mA h g and 760 mA h cm respectively and is used as the anode material in many TFBs. The main appeal of graphite is the stability of its crystalline structure during charge and discharge, which results in reliable cycling and a longer lifetime. The common thread between the current materials of choice for TFB—and a large portion of Li-ion batteries in general—is that they incorporate Li ions through intercalation. In intercalation materials, Li ions are incorporated into a material’s existing crystallographic structure, for example be- tween layers of LiCoO2, and can be reversibly extracted without drastically altering the surrounding material structure. These intercalation type materials have been the focus of extensive research, which has shown the deep relationships between crystal structure, insertion mechanisms, Li ion capacity and reaction reversibility [5].
Extensive research has also be conducted to assess and increase the viability of alloy- ing materials for use as both positive and negative electrodes in Li ion batteries. As opposed to intercalation materials, these alloying materials have significant structure changes when they react with Li ions, but tend to also have much higher theoreti- cal capacities. Group IV elements such as Si and Ge that form alloys with Li, have −3 low potential vs Li, and can have theoretical capacities as high as 3000 mA h cm for some elements. However, disruption of crystal structure in these materials gen- erally results in extremely high volume change during lithiation (addition of Li) and deliathion (removal of Li)—over 240% for Si, Ge and Sn, compared to only 10% in graphite [1][2][6]. High volume expansion leads to pulverization of films and loss
7 Figure 1.2: Capacity trends for different types of electrode materials for Li-ion batteries. of electrical contact between regions of the electrode and current collector and thus irreversibly reduces battery capacity. For these reasons there has been considerable effort in creating alloying materials that maximize structural integrity while main- taining high specific capacity. This is often done through a combination of making composite materials, such as Mg2Si or Si-C composites, and creating nanostructures within the electrode material to accommodate the large volume changes [7][8].
While experimentation with new materials has yielded some improvement in cycling performance of TFB, the exact mechanisms through which mechanical stress in thin film electrodes evolves during lithiation and delithiation, and how that stress may be alleviated, is not well understood. Establishing the link between electrode volume ex- pansion and internal film stress which drives degradation of the electrode is crucial to the development of TFB electrodes with reliable cycling performance. Study of these
8 phenomena in intercalation and alloying materials has yielded insightful results that deepen our understanding of how they operate within a TFB cell [9][10][11]. This motivates a continuation of these analyses for conversion materials. A conversion material is one in which incorporation of Li ions entails the replacement of the ex- isting crystal matrix at the electrode with one of significantly different structure and chemical properties. Like alloying materials, they often feature significantly higher capacities than intercalation materials, but also suffer from high volume expansion and irreversible capacity loss.
1.2 Ruthenium Oxide Electrodes
RuO2 is a common material used in the research of supercapacitors [12][13] but has recently seen some application as a Li-ion battery electrode [14][15]. With 2.2 V potential vs Li, RuO2 electrodes can feasibly be used as either cathode or anode materials in low power applications in many types of microelectronics [16]. Mea- −1 sured first cycle specific capacities for RuO2 are reported to be >1000 mA h g and −3 ~500 mA h cm [17], with conversion reaction and interfacial storage between RuO2 and converted Ru metal nanocrystals accounting for most of the capacity [18]. RuO2 is also compatible the complementary metal-oxide-semiconductor (CMOS) fabrica- tion techniques which are the standard in microelectronics fabrication industry. This is a hugely beneficial property for an electrode material, as an entirely CMOS com- patible TFB would greatly streamline the manufacturing process by eliminating fab- rication steps that would be necessary for certain electrode materials. However, irre- versible capacity loss due to volume expansion (~100% in RuO2) has been a major obstacle in creating viable RuO2 electrodes for TFBs. This thesis covers how to con-
9 duct in situ stress measurements of RuO2 films during battery cycling, and analyzes the unique stress evolution that RuO2 exhibits.
10 2 Stress in Thin Films
2.1 Introduction
In a broad sense, stress and strain are generalizations of the concepts of force and displacement that allow us to describe how non-rigid bodies deform under the appli- cation of forces. In the simplest case, forces acting along the axis of a rod L0 which stretches the rod by some amount ΔL is described in terms of the strain and stress in the rod. The stretching of the rod behaves similarly to a spring and follows Hooke’s law.
F = kΔL
Figure 2.1: Sketch of simple deformation of a rod along its axis described by Hooke’s law.
By normalizing this equation with respect to the cross sectional area of the rod we arrive at the stress strain relation for 1 dimension
= E" (2.1)
F ΔL ≡ " ≡ where the stress, A , is the force per unit area and strain, L0 is the fractional change in the length of the rod. E is the Young’s modulus, which is analogous to the spring constant. Materials that exhibit this linear relationship between stress and
11 strain are called linear-elastic materials.
In order to describe deformations caused by arbitrary forces acting on an object, the stress and strain are decomposed into components that are perpendicular and parallel to the surface they act on, resulting in the tensor equation:
ij = Cijkl"kl (2.2) which is known as generalized Hooke’s law, where ij is the stress tensor, "ij is the strain tensor and Cijkl is the stiffness tensor. We use Einstein notation to denote summing over indices. Indices i and j represent, respectively, the axis along which the force acts and the direction of the normal vector of the surface on which it is acting. Similarly k and l describe the directions of the deformations with respect to the surface normal vector.
(a) A normal stress. (b) A shear stress.
Normal stresses and strains are given by diagonal components ii and "kk respectively, and shear stresses and strains are described by other components. What was simply the Young’s modulus in Eq 2.1 has become a more complicated rank 4 stiffness tensor
Cijkl. This is because in general a stress in a given direction may cause strain in
12 many different directions. Applying this to our original 1-D example: 11 = E"11,
so C1111 = E. It is also common to see the generalized Hooke’s law written in inverse form as
"ij = Sijklkl (2.3) where the compliance tensor Sijkl is the inverse of Cijkl.
Figure 2.3: Sketch of the transverse response to axial stress on an object.
As mentioned before, a single stress may cause a variety of strains. For example, as shown in Fig 2.3, an axial stress on an object may stretch it along that axis but contract
it along perpendicular axes. The ratio between the axial strain "11 and transverse strains "22 and "33 is known as the Poisson ratio, and is represented by the symbol " = " = −" = − . For an isotropic material, the resulting strains are 22 33 11 E 11. The negative is because in general we expect a material that is being stretched in one direction to contract in others; although there are special materials that violate this. It
13 can also be shown that the shear moduli, which are components of Sijkl where i = k j = l 1+ and , can be expressed as E in isotropic materials. The compliance tensor that describes this system under an arbitrary stress is then given by
⎧ ⎪ 1 , i = j = k = l ⎪E ⎪− , i = j, k = l, i k ⎪ E ≠ Sijkl = ⎨1+ ⎪ , i = k, j = l, i ≠ j ⎪ E ⎪ ⎪0 Otherwise ⎩ which when used in Hooke’s law yields the constitutive equations
1 " = ( (1 + ) − ) ij E ij ij kk (2.4)
The energy stored within the material as a result of elastic deformations can also be described using the stress and strain components. Starting from the usual energy in terms of the force and displacement of the object we write an expression for energy density u for a simple one dimensional case.
dU dU = F dx → = d" dV
u = d" (2.5) Ê
For a linear elastic material this becomes
1 u = E"2 2
To generalize to our strain tensor we simply sum the contribution to the energy of
14 each component 1 1 u = " = C " " 2 ij ij 2 ijkl ij kl (2.6)
2.2 Stress-Strain Curve
So far we have focused on systems that exhibit linear elasticity, ie. systems that fol- low Hooke’s law Eq 2.2. If we were to plot the stress vs strain on an object as it is deformed, this curve would be linear, with the elastic modulus defining the slope. However, if an object is stressed passed a certain point, called the yield stress, the relationship between the stress and strain in the object is no longer linear. Deforma- tions in this regime of non-linear stress and strain are referred to as plastic deforma- tions. Behavior of a material in this plastic regime is more complex than in the elastic regime, but forms the basis from which properties such as brittleness and ductility can be understood. The way in which a material deforms in the plastic regime is called plastic flow.
An important property that is often exhibited by materials that have been plastically deformed is elastic unloading. This means that once the external force causing the strain is released, the object relaxes by following a linear stress-strain curve. Typi- cally this means that plastically deforming a material then allowing it to relax will result in an unstressed length that is different form the original. A typical stress strain curve is shown in Fig 2.4 with labeled linear elastic and plastic regions, as well as linear unloading.
For thin film batteries, the stress vs capacity curve is synonymous with the stress- strain curve of the film. This is because in general the Li content in a film is directly
15 proportional to the amount of volume expansion, or bulk strain, it is subject to. This proportionality has been shown experimentally in the case of Si films [19]. Electrodes are compressed ( < 0) during lithiation and are under tensile stress ( > 0) during deliathion. The stress capacity curve in a battery electrode during electrochemical cycling typically follows a pattern of
Elastic Unloading → Elastic Loading → Plastic Flow (Delithiation)
→ Elastic Unloading → Elastic Loading → Plastic Flow (Lithiation)
creating a hysteresis loop like the sketch shown in Fig 2.5.
Yield Point Plastic Flow
Elastic Loading