Influence of charge status on the stress safety properties of Li(Ni_1/3Co_1/3Mn_1/3)O₂ cells

Abstract

In order to improve safety management, the stress changes of Li(Ni_1/3Co_1/3Mn_1/3)O₂ (NMC) cells are real-time monitored using non-destructive strain gauges, and the effects of gauge substrate, temperature and state-of-charge (SOC) have been investigated. The shell exhibits elastic deformation behaviour, and the strain–stress relationship is established. As the temperature increases from 25 to 80 °C, the stress of the NMC cells increases from 0 to 275 MPa, especially greatly at 70 °C and sharply at 80 °C after 18 h. The stress increases from 0 to 9.2 MPa when the potential increases from 2.8 to 4.3 V. However, the value rises from 10 to 55 MPa when the voltage increases from 4.6 to 5.0 V during the over-charge process. An obvious increase of stress appears when the cut-off voltage is below 0.6 V during over-discharge tests. The facile method is significant for non-destructive inspection and emergency management of batteries.

---

1. Introduction

Continuous environmental deterioration and traditional fossil fuel resource depletion make (hybrid) electric vehicles highly attractive for city transport. As one of their key power sources, lithium ion batteries have gained tremendous attention because of their unique advantages such as high energy density, low weight and high environmental compatibility. However, the safety issues, especially swelling, leakage, burning and explosion, restrict their sustainable and fast development as well as large-scale applications. Such typical hazards usually occur when the maximum inner stress is higher than the strength of the outer shell and could cause severe damage to the whole battery. Therefore, it is vital and essential to monitor the stress changes and take precautions for the potential dangers for safety management of batteries. A few methods have been reported to investigate the stress changes in batteries, with a focus on the stress of the microscale particles in active materials^1,2 and the safety simulation of the shell of batteries.^3,4 However, the typical safety issues are directly related to the macroscopic stress in a battery shell, which has not yet been carefully considered. Such macroscopic stress, including the stress of particles and other stresses, is highly complex as a result of the interaction between particles, thermal expansion, interaction between the electrode and electrolyte, the electrochemical reaction, and so on. Actually, a battery is a sealed can with a thin shell, of which the inner pressure can be tested via the outer surface. A mechanical strain gauge is a useful device to accurately measure the pressure/stress via small changes in the electrical resistance of the wire grids under strain. If the battery shell is under linear elastic deformation, the strains measured by the strain gauge on the shell surface without destruction can be used to determine the macroscopic stress of batteries according to Hooke’s law.

As one of the popular cathode materials with a high mass density, NMC can meet the requirements of pulse power characteristics for transport applications in a small volume, but the safety problems (especially large deformations and/or leakage due to abuse or elevated temperature) hamper its extensive applications.^5,6 Various studies of NMC have been reported, attempting to understand and solve these problems. For example, the thermal stability of charged NMC⁷ and the gas generation of Li₄Ti₅O₁₂/NMC cells at 80 °C⁸ have been investigated. The production of heat and gas can cause runaway reactions and trigger explosion of batteries,⁹ which will be very hazardous under abnormal operating conditions.¹⁰ According to the production processes, such problems would result in large deformations and then leakage of batteries. If the stresses that are closely related to the deformations of the battery surface can be monitored and corresponding measures can be taken in advance, the hazards like leakage, burning or explosion could be efficiently prevented. Therefore, it is necessary to monitor the macroscopic stress changes of NMC batteries in order to develop further measures for safe applications.

In this paper, strain gauges are adopted to real-time monitor the strains and stresses of Li(Ni_1/3Co_1/3Mn_1/3)O₂ lithium ion cells under various conditions. The reliability of the monitoring tests, along with the relationship between the macroscopic mechanical stress and electrochemical performance, has been analysed. The reasons for such stress changes have also been discussed.

2. Experimental

2.1. Synthesis of NMC

Powders of (Ni_1/3Co_1/3Mn_1/3)(OH)₂ and LiOH·H₂O (molar ratio of 1 : 1.05) were fully mixed to form a rheological precursor using alcohol as the dispersing agent by planetary ball milling at 180 rpm for 4 h. The milled precursor was pre-heated at 550 °C in air for 4 h in a muffle oven and further calcined at 750–850 °C in air for 12 h. The as-synthesized black product was ground for physicochemical characterization and stress tests.

2.2. Physicochemical characterization

In order to eliminate the unexpected influence of anode materials, common lithium cells composed of NMC and lithium metal are chosen for stress and strain monitoring tests under different conditions. The as-synthesized powder sample for electrochemical characterization was well mixed with 10 wt% of conductive additive of super P and 7 wt% of commercial PVDF binder to form a homogeneous viscous slurry to coat a cleaned aluminium foil using a doctor blade. After being dried at 100 °C, the aluminium foils loaded with active materials were cut into 1.2 cm² wafers which were further dried at 100 °C under vacuum for 12 h before being used as the working electrodes. Using pure lithium metal as the counter and reference electrodes, CR2032 coin cells were assembled in an argon-filled glove box by sandwiching a Celgard 2300 microporous separator between the working electrode and lithium disc. The electrolyte was 1 M LiPF₆ in a mixture of ethyl carbonate (EC), dimethyl carbonate (DMC) and ethyl methyl carbonate (EMC) (1 : 1 : 1 in vol, Shenzhen Capchem Chemicals Co. Ltd., China).

Galvanostatic charge/discharge cycling tests of the above devices were carried out with different potential windows, current rates and working temperatures on a Neware battery testing instrument (Shenzhen Neware Technology Ltd., China). Electrochemical impedance spectroscopy (EIS) measurements were conducted before and after stress tests under different conditions in the frequency range from 100 kHz to 0.01 Hz with a sinusoidal excitation voltage of 10 mV, and the impedance curves were fitted using Zsimpwin and Zview softwares. The cyclic voltammetry (CV) technique was used in the voltage window of 2.8–4.3 V to investigate the electrode reaction process after different cycles. Both EIS and CV tests were carried out on an electrochemical workstation consisting of a PAR 273A potentiostat/galvanostat and a signal recovery model 5210 lock-in-amplifier controlled by a Powersuit software (Princeton Applied Research, USA).

In order to compare the changes of phase composition and structure of NMC electrodes under different states, X-ray diffraction (XRD) tests were carried out using Cu Kα radiation in the range from 10 to 70° at a scanning rate of 0.04° s⁻¹ on a Philips X’Pert pro MPD machine.

2.3. Measurement of cell strains

The testing cell is a sealed container composed of 304 stainless steel (SS), so resistance strain gauges for metals (Zhonghang Electronic Measuring Instruments Co. Ltd., China) were chosen for measuring the strains and consequently evaluating the stresses in lithium cells under different conditions. The gauge has a size of 0.3 × 1.8 mm and a maximum strain range of 2%. The accuracy and the resistance are about 1 με and 120 ohm, respectively. The strain gauge consists of a grid of wire filament on an insulated rear side which supports a metallic foil pattern. To allow the strains to be transferred from the test specimen to the foil through the adhesive and strain gauge, the strain gauge needs to be properly mounted onto the shell surface. The cathode shell was firstly polished with sandpaper, and cleaned by ethanol. Glue was pasted on the dried clean surface for adhering the strain gauge. After being well adhered to the cathode shell surface, the gauge was connected to a stress–strain testing system (Donghua testing technology Co. Ltd, China). At the same time, the cell was connected to the electrochemical testing system to evaluate the relationship between strains and the electrochemical performance of NMC cells under different operating conditions, and the principle diagram and images of the testing systems are shown in Fig. 1.

Fig. 1 Images of the NMC coin cell adhered with strain gauges and the corresponding testing platform.

3. Results and discussion

3.1. Influence of strain gauge installation on the electrochemical performance

In order to evaluate the influence of strain gauge installation on the electrochemical performance of the lithium cells, galvanostatic charge/discharge tests before and after installing the strain gauge were carried out at 25 °C, and the typical curves are compared in Fig. 2. After installing the strain gauge, the galvanostatic charge/discharge curves from 2.8 to 4.3 V at a 1.0C current rate are coincident under the same testing conditions, indicating that the installation of the strain gauge has a negligible effect on the electrochemical performance of the NMC cells. Therefore, such a non-destructive method can reflect the stress changes of lithium ion cells without altering their performance.

Fig. 2 Comparisons of the charge/discharge curves of NMC before/after installing gauges.

3.2. Temperature correction and deformation analysis of the cells

Because the most desirable strain gauge materials are sensitive to temperature variations, resistance wires of the strain gauge will deform when the surrounding temperature is changed, which will result in a resistance change and produce test inaccuracies and even errors. Due to the low thermal expansion coefficient,¹¹ quartz glass was chosen to characterize the test accuracy, and the measured strain data will be calibrated to authentically reflect the safety properties of the battery materials and combined electrolyte system. The strain changes could be caused by electrode reaction, the decomposition of active materials or volume changes of electrolyte, and/or the expansion of cell case and related accessories. In order to obtain the strain changes from the NMC electrode material, two types of lithium ion cells were assembled under the same conditions, i.e. one is a normal cell (abbreviated as C) with active material and electrolyte, and the other is a reference cell (abbreviated as R) which is the same as the normal cell but with no active material. These test specimens were heated to the evaluated temperatures and kept for a period before being naturally cooled down by air. As the temperature increases from 25 to 50 °C, the strain changes of the normal cell, reference cell, and quartz glass are compared in Fig. 3(a). The strain values of the quartz glass decrease quickly to −580 με in less than 1 h, while that of the reference cell decreases slowly to −120 με in about 2 h. However, the strain value reaches more slowly +100 με in 5 h when the NMC electrode is added. Noticeably, the strain values of the quartz glass and the reference cell decrease very quickly in a short time before reaching a stable state. However, the strain values of the normal cell increase and reach a constant state much more slowly. The same strain gauges were used and the tests were carried out under the same environmental conditions, so the strain difference is from the substrates. The substrates with larger deformation than that of the resistance wires will cause the negative strain (compression), while deformation of the substrates lower than that of the resistance wires will cause the positive strain (tension). If the substrates change little, the compression and tension of the resistance wires will produce a negative strain and positive strain, respectively. The normal cell has the highest deformation, and the deformation of the reference cell is much higher than that of the quartz glass. The quartz glass has a negligible deformation at 50 °C, so the maximum negative strain is from the compression of resistance wires, and the as-obtained strain values reflect the changes of the strain gauge itself. As the substrate changes from quartz glass to the SS shell of the reference cell, the strain values increase greatly under the same conditions, indicating that the deformation of the SS substrate increases to make the compression of resistance wires weak. The strain values change from negative to positive and reach a maximum when the NMC electrode is added to the reference cell, indicating that electrode causes the increase of the substrate deformation to change the resistance wires from compression to tension. Therefore, the test strain values are the combination from both the strain gauge and test specimens, and the substrate deformation from the swelling of the SS shell causes an increase of strain values. As for the cells at open-circuit states, the normal cell has much higher strain values than that of the reference cell under the same conditions, indicating that the higher deformation of the normal cell is due to the NMC electrode. The quartz glass reflects the temperature effect of the strain gauge, so that all the strain values used for calculating stresses will be corrected by eliminating the temperature effect of the strain gauge.

Fig. 3 (a) Strain curves of the NMC cell, reference cell and quartz glass at 50 °C, and (b) strain curves of the cell shell and quartz glass between 25 and 80 °C.

In order to determine the stresses from the measured strains, the stress–strain relationship of the cathode shell needs to be determined. The deformations of the SS cathode shell and quartz glass were further tested within a wide temperature range of 25–80 °C. The strain values obtained from the tests during the increasing/decreasing processes of temperature are summarized in Fig. 3(b). Apparently the strain values can be recovered to 0 when the temperature decreases from 80 to 25 °C, indicating that only elastic deformation occurs, thus elastic theory can be used for calculating the stresses using the measured strains, as discussed in the next section.

3.3. Strain–stress relationship of lithium cells

The radius (R) and thickness (t) of the SS cathode shell of the cells are 10 mm and 0.15 mm, respectively. Thus the ratio of t/R is 0.15/10 which is much smaller than 0.1, and the cathode shell can be regarded as a thin-walled plate. As mentioned before, the cathode shell of the cells exhibits small elastic deformation during the testing processes. If the distribution of the stresses P_z along the longitudinal direction of the cathode shell is assumed to be uniform, a rigorous relationship between P_z and the stresses can be found using the elastic theory for shells as follows.

The construction of the cell is illustrated in Fig. 4(a), and the cathode shell laminate is cut off by two cylinders and two radial transverse sections to produce a small element, as shown in Fig. 4(a-1) and (a-2). The radius of the two cylinders is r and r + dr, respectively. The angle between the two transverse sections is dθ. From Fig. 4(a-3) and (a-4), the axial bending moment and transverse shear stress on the two cylinders are M_r and M_r + (dM_rr/dr)dr, and Q_r and , respectively. The circumferential bending moment on the two cylindrical surfaces is M_θ. The external stress on the two cylindrical surfaces is P_z.

Fig. 4 (a) Schematic diagrams of the stress distribution (a-1) and (a-2) and internal forces (a-3) and (a-4) of the cell, and (b) deformation analysis when the cathode laminate bends. (c) Shearing force of the laminate under uniformly distributed loading, and (d) normalized stress distribution along the radius.

According to the moment equilibrium of the element, the algebraic sum of all internal moments and external moments on the tangent line of the cylinders is 0:

with when θ is small. Ignoring the second-order terms, this becomes:

The deformation of the mid-plane caused by the uniform pressure is axially symmetrical. Therefore, the deformation, w, only depends on the radius r. As shown in Fig. 4(b), AB is a line on the radial section whose vertical distance to the mid-plane is z. The radius of A and B is r and r + dr, respectively, so AB = dr. The lines of mn and m₁n₁ pass through the m₁n₁ points of A and B, respectively. Both lines are vertical to the mid-plane. When the cathode laminate deforms, A and B move to A₁ and B₁, respectively. Therefore, the strains are:

Under small deformation, , and eqn (3a) and (3b) are rewritten as:

By the Kirchhoff–Love assumption, every point in the laminate is under the two-direction stress state when the cathode laminate deforms. According to the generalized Hooke’s law, the physical equations of the cathode laminate are:

Combining the above equations and substituting (4) into (5), the stresses can be written as:

After integration, the bending moment can be obtained:

where

By substituting (7a) and (7b) into (1) and with some arrangements, the following equation can be obtained:

The distribution of the transverse load is assumed to be uniform. As shown in Fig. 4(c), the shearing force on the cross-section of the cylinder whose radius is r is as follows:

Substituting (9) into (8) gives us:

After integration, the bending deformation of the mid-plane is:

where C₁, C₂, and C₃ are the constants to be determined. Since the deformation and slope of the laminate are limited values, C₂ = 0. The rim of the cathode laminate is assumed to be clamped, so

Finally the deformation is calculated as:

Substituting (12) into (11) gives:

So the stresses are:

The stresses at the centre of the cathode laminate (i.e. r = 0) are found to be:

The material of the cell shell is 304 stainless steel which has a Young’s modulus E of 193 GPa and a Poisson’s ratio μ of 0.28 at 25 °C.¹² Therefore, the stress at the centre of the cell shell can be calculated using the measured strain according to eqn (15). From eqn (14a) and (14b), the maximum stress is at the centre of the cell plate, and the calculated stress distribution along the radius across the whole surface of the cell is supplied in Fig. 4(d), which is of most interest for evaluating the safety issues of the cell. The normalized stress and normalized radius are obtained by dividing the maximum stress and maximum radius, respectively. The stresses along the axial direction and radial direction decrease differently with the increase of radius, which is significant to analyse the stress distribution and the possible failure position of the cell surface.

3.4. Influence of ambient temperature on the stresses in cells

Since both the strain gauge and the specimen are sensitive to ambient temperature, the strain gauges are designed to minimize the sensitivity to temperature by compensating the thermal expansion of the specimen materials. In order to eliminate the influence of electrochemical activation on the stresses, the cells were galvanostatically charged/discharged for 10 cycles and discharged to 2.8 V to reach a stable state. The cycled and stable cells were kept for over 20 h in an oven with various temperatures to evaluate the effects of ambient temperature on the stresses of the electrode. Under a series of ambient temperatures (i.e. 25, 50, 60, 70, 80 °C) the stress changes of the normal cells and the reference cells at the same charge states were monitored, and the testing results are shown in Fig. 5. The stable stress values of the normal cell C and the reference cell R are almost zero and have no change at 25 °C. However, the values of C and R respectively increase gradually to 125 MPa and 95 MPa when the ambient temperature reaches 50 °C. The values of C and R respectively increase quickly to 126 MPa and 113 MPa when the temperature further increases to 60 °C. The values of C and R respectively increase to 175 MPa and 162 MPa when the temperature further increases to 70 °C. As the temperature further increases to 80 °C, the stress values of the normal cell C increase quickly and continuously above 200 MPa, and have a jump above 240 MPa when the heating period is 18 h, while that of the reference cell R has an abrupt increase and reaches a stable value of 170 MPa gradually. Under the same conditions, the stresses of the normal cells are always higher than those of the reference cells, and a higher temperature results in higher stress values. Comparing the stress values of the cells under the same conditions, the higher stress values of the normal cells are from the NMC electrode.

Fig. 5 Stress changes of the NMC cell and reference cell at different ambient temperatures.

No obvious signal for Mn, Ni, and Co is detected in the electrolyte when the charged Li_1.1(Mn_1/3Ni_1/3Co_1/3)_0.9O₂ cell is stored in a 55 °C oven for 3 weeks,⁶ and the amounts of Ni, Co, and Mn dissolved in the electrolyte are respectively 48.0, 50.5, and 42.4 ppm when the Li[Mn_1/3Ni_1/3Co_1/3]O₂ electrode is immersed in the electrolyte at 55 °C for 10 days.¹³ Correlating this information with the stress changes of the normal cells and reference cells, the stress is mainly caused by the thermal expansion of the cell case in the temperature range of 50–60 °C, and the extra stress values of the normal cells are attributed to the volume expansion of the NMC electrode, including the active material, conductive additive and binder. Although the stable stress value of the normal cell at 60 °C is close to that at 50 °C, the stress value at 60 °C increases much faster than that at 50 °C, indicating that the volume expansion of the NMC electrode reaches a maximum state at different rates within the temperature range and a higher temperature enhances the increasing rate of the stress. The stress value of the reference cells at 70 °C is similar to that at 80 °C, indicating that the thermal stress of the SS cell case reaches a maximum state when the temperature is 70 °C, and quickly becomes stable when the temperature is 80 °C. LiPF₆ can be resolved into PF₅ at 70 °C, and PF₅ will react with the solvents of EC and DMC at 85 °C.¹⁴ Therefore, the extra increment of the stress values of the normal cells is mainly from the decomposition of LiPF₆ when the temperature increases to 70 °C, and the much greater increment is mainly from the reaction of the decomposition product of PF₅ with the solvents in the electrolyte at 80 °C. The produced gases in the normal cells will continuously increase the stress with the extension of the heating time at 80 °C.

3.5. Stress changes during cycles

In order to investigate the stress changes during galvanostatic charge/discharge cycles at a low current density, the stresses of the fresh cells were real-time monitored at 25 °C. The stress curves corresponding to the charge/discharge curves at a 0.5C rate between 2.8 V and 4.3 V are shown in Fig. 6. During the charge process of the NMC material, lithium ions are continually extracted from the structure, and the decrease of lithium ions results in the increase of the reduced ions of Ni, Co and Mn in an oxidation state to keep the charge balance, so the lattice constants of a and c change in the α-NaFeO₂ layered crystal structure. a decreases whilst c increases due to an increasing electrostatic repulsion.¹⁵ These changes of the parameters of a and c produce microscale stress between particles, which further develops to macroscale stress monitored by the strain gauge. From the curves, such stress increases at different rates during a charge/discharge cycle. At the beginning of the charge process, the potential increases quickly from 2.8 V to 3.6 V, while the corresponding stress increases gradually from 0.0 to 0.3 MPa. At the potential platform from 3.6 to 3.8 V, the stress value increases from 0.3 MPa to 4.6 MPa, indicating that lithium ion extraction causes a change in the lattice volume of the crystal in the single phase region of the electrode material,¹⁶ which results in the increase of stress. When the potential is increased from 3.8 V to 4.3 V, the stress increases quickly to the maximum value of 9.2 MPa at the highest potential, along with the large volume changes from the high electrostatic repulsion. Therefore, the stress value of the cell is closely related to the charge state of the cells, which has also been reported in ref. 17. The stress increases with the potential within the testing window, and both the stress and potential reach maximum values simultaneously at the end of the charge process, indicating that the increased stress is caused by the volume changes of the NMC electrode into which lithium ions are inserted. During the discharge process, the stress decreases differently with the decrease of potential. During initial discharge, from 4.3 V to 3.8 V, the stress decreases rapidly to 3.9 MPa, followed by a gradual relaxation as more lithium ions are inserted. At the discharge platform from 3.8 V to 3.6 V, the stress value decreases from 3.9 MPa to 1.1 MPa. However, the stress value is 0.5 MPa, not the initial 0, when the potential restores to 2.8 V, indicating that there is a residual stress during the charge/discharge cycle, which is similar to the irreversible increase in stack stress to the permanent volumetric expansion of the graphite anode.¹⁸ Such residual stress will accumulate and lead to the unrecoverable deformation of the cell during cycles, which could evolve into safety problems like swelling or bulging.

Fig. 6 (a) Typical galvanostatic charge/discharge curves, and (b) the corresponding stress changes at a 0.5C current rate. (c) The residual stress during cycles at a 1.0C current rate, and (d) different CV curves of the NMC cell at a scanning rate of 0.2 mV s⁻¹.

In order to investigate the relationship between the capacity fade and residual stress at high current density, the normalized capacities and corresponding residual stress values of the cell during 100 cycles between 2.8 V and 4.3 V at a 1C current rate are summarized in Fig. 6(c). After 100 cycles, the residual stress accumulates from 0 to 37.8 MPa, and the capacity declines to 88.7% of the initial capacity. During the continuous charge/discharge cycles, the capacity fades gradually but the residual stress increases quickly, as a result of the deformation from the loss of cycled lithium ions, similar to that in graphite.¹

CV tests are carried out on the cells which are discharged to 2.8 V after different cycles to better understand the increased residual stress and decreased capacity, and the results are shown in Fig. 6(d). The oxidization potential (E_O), reduction potential (E_R), potential difference between oxidization peak and reduction peak (ΔE), oxidation peak current (I_O), and reduction peak current (I_R) are listed in Table 1. The 2^nd and 3^rd CV curves are almost coincident except for the higher oxidization peak intensity of the 2^nd curve and the lower potential difference between oxidization peak and reduction peak of the 3^rd curve, indicating the higher reaction activity of the 2^nd cycle and the lower polarization degree of the 3^rd cycle. The oxidization peak at around 3.87 V and the reduction peak at around 3.68 V are associated to the de-intercalation and intercalation of lithium ions inside the host matrix, respectively. Ni²⁺/Ni⁴⁺ accounts for the oxidation peak at 3.87 V,¹⁹ and the reduction peak at 3.67 V.²⁰ However, the redox peak intensities decrease greatly after 100 cycles and the potential difference between the oxidization peak and reduction peak increases overwhelmingly, suggesting that the reaction activity decreases and the reaction resistance increases during the cycling process, which results in the decrease of capacity and the increase of residual stress.

Table 1 Values of E_O, E_R, ΔE, I_O and I_R of Li(Ni_1/3Co_1/3Mn_1/3)O₂ after different cycles

Cycle	EO (V)	ER (V)	ΔE (V)	IO × 10^−4 (A)	IR × 10^−4 (A)
2^nd	3.87	3.67	0.20	12.83	−6.16
3^rd	3.84	3.67	0.17	11.93	−6.28
100^th	3.87	3.53	0.34	3.23	−1.80

---

3.6. Stress and structure changes during over-charge and over-discharge processes

From the above analysis results, the charge state is found to greatly affect the stress values. In order to investigate the stress changes of the cells under some extreme conditions like over-charge (abbreviated as OC) and over-discharge (abbreviated as OD), the stress changes were monitored when they are respectively charged to 5 V and discharged to 0 V at a 1.0C current rate at 25 °C, and the results are summarized in Fig. 7. As shown in Fig. 7(a), the stress increases slowly and almost linearly from 0 to 9.9 MPa with a low slope of 7.6 MPa h⁻¹ when the potential of the cell increases from 2.8 V to 4.3 V, close to that in the normal potential window shown in Fig. 6. So, the battery is under a normal state in this charge process. However, the stress increases abruptly from 10.0 MPa to 55.5 MPa when the potential increases from 4.7 V to 5.0 V, and the slope jumps to 225.5 MPa h⁻¹. The cell above 4.7 V is regarded as in an over-charge state,²¹ so the stress increases rapidly in the over-charge state, which is attributed to the decomposed gases from the electrolyte.¹⁵

Fig. 7 The stress changes of the NMC cell during (a) the over-charge process, and (b) the over-discharge process. (c) XRD patterns of the NMC materials, and (d) EIS curves of the NMC cell at different charge states. (e) Part amplified, and (f) the corresponding fitting model of the EIS curves.

During the over-discharge process from 4.3 V to 2.8 V, the stress drops slowly from 10.0 MPa to 0.2 MPa in an almost linear way, as shown in Fig. 7(b), similar to the above results. However, the stress decreases linearly to −1.5 MPa when the potential decreases to 1.2 V. Furthermore, the decreasing slope of the stress keeps constant when the potential drops from 4.3 V to 1.2 V. The cyclability of the cell is affected little by the over-discharge to 2.0 V or 1.5 V, and the capacity loss is close to that of the similar cell cycled between 3.0 and 4.2 V.²² Below 1.2 V, the potential decreases slowly and exhibits a quasi-plateau until 0.6 V, where the corresponding stress decreases in a stair step until −5.0 MPa. Below 0.6 V, the potential decreases quickly to 0.0 V, but the stress increases abnormally in a linear way to −1.8 MPa. The stress values change from positive to negative during the over-discharge process, indicating the stress changes from tension to compression, which is from the negative deformation of the surface shell. With the decrease of the cut-off potential, lithium ions continuously intercalate the electrode material to destroy the crystals,^23,24 which results in the decrease of volume and the increase of the negative stress values when the potential decreases from 1.2 V to 0.6 V. Over-discharge to 1.0 and 0.5 V leads to the cell’s capacity loss of 29 and 38%, respectively, which is significantly higher than that of normally cycled cells.²² A large amount of CO₂ and hydrocarbons as well as CO have been found in the over-discharged cell,²⁵ and electrolyte decomposition is accelerated in the over-discharged state.²⁶ The decomposition of the electrolyte is dependent on the changes of surface conditions at the cathode. Therefore, with the continuous increase of inserted lithium ions, the crystal structure of the NMC material starts to breakdown to reduce deformation when the potential drops to 1.2 V, so the discharge process becomes difficult and the corresponding stress exhibits negative values. The completed destruction results in the lowest negative stress when the potential is 0.6 V. The produced gases or hydrocarbons will increase the anomaly in the stress below 0.6 V, as shown in the curve. Although the potential decreases slowly between 1.2 V and 0.6 V, the stress has several plateaus during the decreasing process. The plateaus may correspond to the damage process of the crystals. The possible mechanisms are further confirmed by the following XRD and EIS tests.

In order to detect the composition and the microstructure of the active materials under different charge states, the NMC electrodes were analysed by the XRD technique, and the patterns are supplied in Fig. 7(c). Considering the low loading density of the active material, the aluminium current collector of the NMC electrode was not removed for XRD characterization. In these patterns, all the index peaks except the obvious Al peak at around 65° are attributed to the samples at different charge states. Noticeably, the XRD patterns of the pristine and the over-charged samples are almost identical and the average crystallite size of the pristine one and the over-charged one is 46.265 and 52.765 nm, respectively, indicating that the crystal structure changes during the over-charge process. All peaks of both samples match well with those of the hexagonal α-NaFeO₂ structure with the R3m space group, indicating the existence of a layered structure.^27–29 After over-charge, lithium ions are extracted from NMC, the peak densities of (101), (102) and (104) decline, and the lattice constants of a and c of NMC change from 2.860 and 14.225 Å to 2.831 and 14.322 Å, respectively. Furthermore, the high potential will lead to decomposition of the electrolyte. These changes cause the increase of stress during the over-charge process, consistent with the above stress analysis results. However, the main characteristic peaks of NMC disappear in the over-discharged pattern, and more amorphous characteristics appear except for the peak at around 38°, suggesting that the crystal structure of the over-discharged sample is destroyed, which also confirms the above stress changes. The amorphous products are probably the X-ray undetectable nano-sized mixtures, similar to those in the over-discharged LiFePO₄, LiNiO₂, and LiMn₂O₄.²⁴ Such products result in the decrease of volume and electrostatic repulsion to produce negative deformation of the shell to produce negative stress, agreeing well with the above results.

In order to compare the interface behaviour of the electrodes under different charge states, the cells were further analysed with the EIS technique, and the curves are supplied in Fig. 7(d). The fresh cells were galvanostatically charged or discharged to the referred potential for the EIS tests. Except for that of the over-discharged curve, all the curves at other charge states have two semicircles in the high-to-medium frequency region, indicating a similar electrode reaction process. The zoomed-in section in Fig. 7(e) clearly shows the constitutions of the semicircles. In the curves of 2.8 V and 4.3 V, the two semicircles grow with the increase of charge voltage upper limit, similar to Zheng’s report.²⁷ The semicircle in the medium-to-low frequency region becomes abnormally big when the cell is over-charged. The intercept of the curves on the Z′ real axis in the high frequency region is attributed to the ohmic resistance of the electrolyte solution (R_s). The semicircle in the high-to-medium frequency region corresponds to the impedance of the solution film resistance (R_f) on the electrode surface. The semicircle in the medium-to-low frequency region is related to the solid electrolyte interface (R_SEI) and charge transfer resistance (R_ct). The oblique line in the low frequency region concerns the Warburg impedance W which concerns the semi-infinite diffusion of lithium ions in the bulk electrode. Using ZsimpWin and Zview softwares with the R(QR)(QR)(QR)W model supplied in Fig. 7(f), the EIS curves are fitted well, indicating that lithium ions diffuse in the crystal structure after reacting with electrode materials. Due to the non-homogeneity such as porosity, roughness, and geometry in the system, a constant phase element (CPE) Q is substituted for traditional capacitance C in the model. In the equivalent circuit, Q_f, Q_SEI and Q_d correspond to the constant phase elements of the electrolyte film, SEI film and film/electrode interface, respectively.³⁰ During the charge process, lithium ions are extracted from the NMC material to form Ni_1/3Co_1/3Mn_1/3O₂ with high metal valence, so the charge transfer reaction becomes difficult and the high metal valence accelerates the oxidization of the electrolyte to produce gases to form a thick film on the electrode surface. All these result in the big semicircles, agreeing with the testing results. According to the simulation results supplied in Table 2, the ohmic resistances of the liquid charge transfer resistance of the 2.8 V and 4.3 V are close, but that of the over-discharged and over-charged ones increases greatly, indicating that the conductivity of the electrolyte near the electrode surface changes, which is from the changed composition after over-discharge or over-charge. The large differences of film resistance and charge transfer reaction resistance show that the electrode surface changes greatly under different charge states, which may be from the different compositions or structure of the NMC electrode material. All these changes cause the stress change in the cell shell to produce possible safety issues.

Table 2 Values of R_s, R_f, R_SEI, R_ct and W

State-of-charge	Rs (ohm cm^−2)	Rf (ohm cm^−2)	RSEI (ohm cm^−2)	Rct (ohm cm^−2)	W × 10^−3 (ohm cm^−2)
0.0 V	9.82	108.07	4.75	1145.61	1.60 × 10^9
2.8 V	3.83	13.33	15.55	224.56	11.52
4.3 V	3.89	12.33	63.60	128.95	6.53
5.0 V	4.50	222.89	2310.53	988.60	1.53

---

In order to detect the stress from the lithium anode, the strain gauge was adhered to an anode shell, and the stresses were real-time monitored under the same testing conditions with the NMC cathode. From the testing results in Fig. S1,† the stress values are close to the minimum detection limit of the strain gauge, indicating that the stress change of the lithium anode shell is too low to be detected by the strain gauge with a detection limit of 1 με. The low strain is from the big buffer of lithium anode and the wave spring between lithium spacer and anode case. Lithium ions are extracted from the NMC cathode to deposit on the lithium surface which is in the electrolyte during the charge process, while lithium ions are extracted from the lithium anode to insert the NMC cathode during the discharge process. As for the lithium anode, lithium ions are excessive in the test cells, and lithium ion extraction/deposition reactions take place at the lithium surface wetted by the electrolyte, instead of the crystal lattice like that in the NMC cathode to cause volume changes. Furthermore, lithium is a soft metal which can resist the strain during the electrode reaction processes, and the flexible polymer separator along with electrolyte can buffer the large volume change of the NMC cathode. Therefore, the stress change from the lithium anode is low enough to be neglected, and the strain of the NMC shell is mainly from the lithium ion processes of the NMC electrode in the test cells.

3.7. Relationship between charge states and stress changes

According to the above analysis results, the relationship between charge states and stress changes at 25 °C is constructed in Fig. 8. The stress continuously increases from −5.0 MPa to 55.5 MPa when the state-of-charge (SOC) changes from over-discharge (OD) to over-charge (OC), but the cell exhibits a good electrochemical performance in the normal region with the SOC increasing from 0 to 100%. Beyond the normal region, the electrochemical performance becomes poor, and the stress increases differently. As the potential discharges to 2.8 V, the residual stress value reaches 0.2–0.5 MPa when the current rate is 1C. As mentioned before, the residual stress accumulates to 37.8 MPa after 100 cycles, and the polarization becomes high and the capacity fades to 88.7%, so the increased residual stress undermines the electrochemical performance. The stress value ranges from 9.9 MPa to 55.5 MPa during the over-charge process from 4.3 V to 5.0 V, while that from 0.2 MPa to −5.0 MPa during the over-discharge process from 2.8 V to 0.0 V. Under the over-discharge state, the stress of the cell first decreases and then increases because of the different reaction processes. Therefore, the emergence of high pressure on the surface of the cell case means degradation of the electrochemical performance, which is significant for the safety management of batteries in applications.

Fig. 8 The relationships between stress changes, state-of-charge and electrochemical performance at room temperature.

4. Conclusions

A non-destructive method has been developed to real-time monitor the surface stress changes of a SS cell case for safety management. The temperature sensitivity and stress properties of the strain gauges have been investigated, and temperature correction has been conducted. The stress–strain relationship of the cell case has been established according to the derived stress calculation formula. The stress values of the NMC cells under different states have been discussed. As the cells are open-circuit, the surface stress mainly originates from the thermal stress of the case at temperatures from 50 to 70 °C, but the continuously increasing stress at 80 °C is due to electrolyte decomposition in addition to the volume expansion of the electrode and thermal stress of the case. The surface stress is greatly affected by the charge state, and the stress value increases from 0.0 to the maximum value of 9.2 MPa when the potential increases from 2.8 to 4.3 V at a 0.5C current rate, but the stress does not restore to 0.0 MPa when the potential decreases to 2.8 V. The accumulated residue stress is 37.8 MPa, while the capacity degradation is 11.3% compared to the initial capacity after 100 cycles at a 1.0C current rate. During the over-charge to 5.0 V, the stress increases slowly to 10.0 MPa at 4.7 V, and then increases quickly to 55.5 MPa. During the over-discharge to 0.0 V, the stress value decreases linearly from 10 to −2.0 MPa when the potential decreases from 4.3 to 1.2 V, and then decreases to −5.0 MPa from 1.2 to 0.6 V, while it increases abnormally to −1.8 MPa from 0.6 to 0.0 V. Various techniques have been used to analyse the electrode reaction processes. The increased surface stress worsens the electrochemical performance, and their relationship has been analysed. The relationship between stress and potential can potentially serve as a useful tool to monitor the electrochemical performance and failure of batteries.

The surface stress changes can be monitored successfully by strain gauges in lithium cells. A sealed battery can be regarded as a pressure vessel where the electrode reaction takes place, and the inner pressure would cause the deformation of the shell to produce stress, so the strain gauge could be used to monitor the strain-stress of a full battery, and some pictures of the stress testing system for a practical full battery are supplied in Fig. S2.† The stress measurements can be further combined with computer simulation calculations on the surface for cross-validation and to predict the stress changes in the complex processes involving chemical reactions and thermal effects. Comparing the surface stress to the stress threshold value of the battery package, special techniques (including ending, blocking, cutting, or injecting chemicals, etc.) could be utilized to eliminate safety issues. Therefore, this initial work demonstrates the possibility to monitor the surface stress changes of cells using strain gauges as an effective and economical way which may greatly improve the safety management of batteries.

Acknowledgements

We gratefully acknowledge the financial support from the National Science Foundation of China (Grant No. 21206099 and 21576170), and the Experimental Technical Project (2015-0141) of Sichuan University.

References

1. J. Cannarella and C. B. Arnold, Stress evolution and capacity fade in constrained lithium-ion pouch cells, J. Power Sources, 2014, 245(1), 745–751.
2. Ö. Ö. Çapraz, P. Shrotriya, P. Skeldon, G. E. Thompson and K. R. Hebert, Role of oxide stress in the initial growth of self-organized porous aluminum oxide, Electrochim. Acta, 2015, 167, 404–411.
3. S. Fadda, A. Cincotti and G. Cao, The effect of cell size distribution during the cooling stage of cryopreservation without CPA, AIChE J., 2009, 56(8), 2173–2185.
4. Y. J. He, J. N. Shen, J. F. Shen and Z. F. Ma, State of health estimation of lithium-ion batteries: a multiscale gaussian process regression modeling approach, AIChE J., 2015, 61(5), 1589–1600.
5. R. Spotnitz and J. Franklin, Abuse behavior of high-power lithium-ion cells, J. Power Sources, 2003, 113(1), 81–100.
6. K. Amine, Mechanism of capacity fade of MCMB/Li_1.1(Ni_1/3Mn_1/3Co_1/3)_0.9O₂ cell at elevated temperature and additives to improve its cycle life, J. Mater. Chem., 2011, 21(44), 17754–17759.
7. K. W. Nam, W. S. Yoon and X. Q. Yang, Structural changes and thermal stability of charged Li(Ni_1/3Co_1/3Mn_1/3)O₂ cathode material for Li-ion batteries studied by time-resolved XRD, J. Power Sources, 2009, 189(1), 515–518.
8. K. Wu, J. Yang and Y. Liu, et al., Investigation on gas generation of Li₄Ti₅O₁₂/Li(Ni_1/3Co_1/3Mn_1/3)O₂ cells at elevated temperature, J. Power Sources, 2013, 237(3), 285–290.
9. C. Y. Hong, T. Z. Yuan, H. Y. Bing and L. Qiang, Research of explosion mechanism of Lithium-ion battery, Electrochemistry., 2006, 18(6), 823–831.
10. F. Larsson and B. E. Mellander, Abuse by external heating, overcharge and short circuiting of commercial Lithium-ion battery cells, J. Electrochem. Soc., 2014, 161(10), A1611–A1617.
11. C. A. Sorrell, H. U. Anderson and R. J. Ackermann, Thermal expansion and the high–low transformation in quartz. II. Dilatometric studies, J. Appl. Crystallogr., 1974, 7(5), 468–473.
12. H. Kim, J. Y. Kang, D. Son, et al., Microstructures and mechanical properties of cold-work tool steels: a comparison of 8%Cr Steel with STD11, CRC Press, 2014, vol. 27(5), pp. 242–252.
13. S. B. Jang, S. H. Kang, K. Amine, Y. C. Bae and Y. K. Sun, Synthesis and improved electrochemical performance of Al (OH)₃-coated Li(Ni_1/3Co_1/3Mn_1/3)O₂ cathode materials at elevated temperature, Electrochim. Acta, 2005, 50(20), 4168–4173.
14. S. E. Sloop, J. B. Kerr and K. Kinoshita, The role of Li-ion battery electrolyte reactivity in performance decline and self-discharge, J. Power Sources, 2003, 119–121(03), 330–337.
15. C. K. Lin, Y. Ren, K. Amine, Y. Qin and Z. Chen, In situ high-energy X-ray diffraction to study overcharge abuse of 18650-size lithium-ion battery, J. Power Sources, 2013, 230(15), 32–37.
16. S. Renganathan, G. Sikha, S. Santhanagopalan and R. E. White, Theoretical analysis of stresses in a lithium ion cell, J. Electrochem. Soc., 2010, 2(2), 156.
17. V. A. Sethuraman, N. V. Winkle, D. P. Abraham, A. F. Bower and P. R. Guduru, Real-time stress measurements in lithium-ion battery negative-electrodes, J. Power Sources, 2012, 206(1), 334–342.
18. J. Cannarella and C. B. Arnold, State of health and charge measurements in lithium-ion batteries using mechanical stress, J. Power Sources, 2014, 269(2), 7–14.
19. Y. Idemoto and T. Matsui, Thermodynamic stability, crystal structure, and cathodic performance of Li_x(Ni_1/3Co_1/3Mn_1/3)O₂ depend on the synthetic process and Li content, SolidState Ionics, 2008, 179(17), 625–635.
20. S. Yang, X. Wang, Q. Chen, X. Yang, J. Li and Q. Wei, Effects of complexants on (Ni_1/3Co_1/3Mn_1/3)CO₃ morphology and electrochemical performance of Li(Ni_1/3Co_1/3Mn_1/3)O₂, J. Solid State Electrochem., 2012, 16(2), 481–490.
21. D. Belov and M. H. Yang, Failure mechanism of Li-ion battery at overcharge conditions, J. Solid State Electrochem., 2008, 12(7), 885–894.
22. H. Maleki and J. N. Howard, Effects of overdischarge on performance and thermal stability of a Li-ion cell, J. Power Sources, 2006, 160(2), 1395–1402.
23. J. Shu, M. Shui and F. Huang, et al., A new look at lithium cobalt oxide in a broad voltage range for Lithium-ion Batteries, J. Phys. Chem. C, 2010, 114(7), 3323–3328.
24. J. Shu, M. Shui, D. Xu, D. Wang, Y. Ren and S. Gao, A comparative study of overdischarge behaviors of cathode materials for lithium-ion batteries, J. Solid State Electrochem., 2012, 16(2), 819–824.
25. H. Yoshida, T. Fukunaga, T. Hazama, M. Terasaki, M. Mizutani and M. Yamachi, Degradation mechanism of alkyl carbonate solvents used in lithium-ion cells during initial charging, J. Power Sources, 1997, 68(2), 311–315.
26. K. Kumai, H. Miyashiro, Y. Kobayashi, K. Takei and R. Ishikawa, Gas generation mechanism due to electrolyte decomposition in commercial lithium-ion cell, J. Power Sources, 1999, 81–82(9), 715–719.
27. H. Zheng, Q. Sun, L. Gao, X. Song and V. S. Battaglia, Correlation between dissolution behavior and electrochemical cycling performance for Li(Ni_1/3Co_1/3Mn_1/3)O₂-based cells, J. Power Sources, 2012, 207(6), 134–140.
28. K. M. Shaju, G. V. S. Rao and B. V. R. Chowdari, X-ray photoelectron spectroscopy and electrochemical behaviour of 4 V cathode, Li(Ni_1/2Mn_1/2)O₂, Electrochim. Acta, 2003, 48(11), 1505–1514.
29. C. S. Kang and J. T. Son, Synthesis and electrochemical properties of Li(Ni_1/3Co_1/3Mn_1/3)O₂ cathode materials by electrospinning process, J. Electroceram., 2012, 29(4), 235–239.
30. H. Kang, G. Wang, H. Guo, M. Chen, C. Luo and K. Yan, Facile synthesis and electrochemical performance of LiFePO4/C composites using Fe–P waste slag, Ind. Eng. Chem. Res., 2012, 51(23), 7923–7931.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra08774d

---