Pyrolysis kinetic study of cathode material derived from spent lithium ion batteries (LIBs): Comparison of different models

ABSTRACT

Separating cathode material and Al foil from spent lithium-ion batteries (LIBs) is a critical step for LIBs recycling. As compared to chemical dissolving and decomposition, the pyrolysis pretreatment is an alternative and simple method. In this work, the pyrolysis kinetics of cathode material were comparatively studied using various isoconversional methods, including Flynn-Wall-Ozawa (FWO), Friedman, Kissinger-Akahira-Sunose, Starink, Tang, and Boswell. The thermal degradation mechanism was investigated by the Coats-Redfern (CR) and master-plot methods as well. The thermogravimetric analysis revealed that cathode material decomposition could be divided into three stages with mass losses of 1.51%, 0.787%, and 0.449%, respectively. Activation energy (Eα) calculated using the six model-free methods showed a similar trend, gradually increasing as the degree of conversion (α) increased from 0.001 to 0.009, and then significantly elevating. The FWO method gave the best fitting and Eα values first increased from 12.032 to 24.433 kJ·mol⁻¹ with α elevating from 0.001 to 0.009, then increased further to 43.187 kJ·mol⁻¹. Both CR and Criado methods indicated that the degradation of cathode material can be explained by the diffusion models.

Implications: The rapid growth in the production and consumption of lithium-ion batteries (LIBs) for portable electronic devices and electric vehicles has resulted in an increasing number of spent LIBs. Thermal treatment offers advantages of high-efficiency and simple operation. Understanding the thermal process of spent LIBs and probing its kinetic are significant for the large-scale treatment. Through this study, it will be significant for the reactor designing and optimizing in practice.

Introduction

Among numerous energy storage devices, the lithium-ion batteries (LIBs) have attracted more attentions due to their higher power density, environmental benignity, and long cycle life. However, concomitant with the fast-growing service consumption of LIBs, the generation of spent LIBs has been increasing immensely (Gao et al. 2019, Li et al. 2018; Wang, Tan, and Li et al. 2018; Yao et al. 2020a). In view of this waste’s negative effects on the environment and valuable metals present within it, recycling the waste is highly desirable. Recently, many sophisticated technologies (Liu et al. 2019; Liu et al. 2019; Lv et al. 2018) – such as mechanical treatment, pyro-metallurgy, hydrometallurgy, bio-metallurgy, and a combination of the mentioned – have been developed. However, separating cathode materials and Al foil are difficult during these recycling processes, because they are firmly adhered by polyvinylidene fluoride (PVDF) binder. The commonly used approaches of detaching cathode materials involve dissolving Al foil using chemicals, dissolving PVDF with organic solvents, and decomposing the binder with oxidants. But applying chemicals will generate wastewater and increase the recycling cost. In addition, it will pose threats to workers’ health.

As a comparison, thermal treatment has the advantages of simple operation and high efficiency (Cheng et al. 2019; Qi et al. 2019; Reis et al. 2019; Yu et al. 2019). Zhang et al. (2019, 2018, 2019) removed the PVDF binder using pyrolysis treatment. The optimum pyrolysis temperature of organic binders in electrode materials was determined as 500°C with a heating rate of 10°C/min and pyrolysis time of 15 min. Wang et al. (2018) removed the binder by roasting at 450°C for 15 min. Wang et al. (2019) developed a novel molten salt technique to degrade the binder. The AlCl₃-NaCl system could melt PVDF efficiently at a temperature of 160°C with a holding time of 20 min. However, there are sparse researches on the thermodynamic kinetics of these treatments, which will be critical for the reactor design, optimization, and scaleup during industrial-scale treatments of spent LIBs. Therefore, in this work, the pyrolysis kinetics of cathode material was investigated by different isoconversional methods. The possible degradation mechanism of cathode material was studied as well.

Material and methods

Materials

The spent LIBs were supplied by local cellular phone repair shops. Prior to usage, they were discharged using 3.5% NaCl solution to mitigate the potential risk of short circuiting or LIB blast. After drying, they were dismantled manually, and the cathode and anode were separated. The cathode was crushed and used as raw material.

Experimental procedure

The cathode powder with weight of approximately 2 mg was put in crucibles and heated from 30°C under a N₂ flow. The experiments were repeated at different heating rates of 5, 10, and 20 K·min⁻¹. Each test was repeated at least three times.

Basic theories

The principle theories of model-free methods and model-fitting methods can be found in literatures (Da Silva et al. 2018; Fernandez et al. 2020, 2019; Qi et al. 2020; Yao et al. 2020; Yao et al. 2020b). In this work, the pyrolysis kinetics of cathode material was studied using six model-free methods, including Flynn-Wall-Ozawa (FWO), Friedman, Kissinger-Akahira-Sunose (KAS), Starink, Tang, and Boswell. The thermal degradation mechanism was investigated by the Coats-Redfern (CR) and master-plot methods.

Mathematical resolution by FWO integral isoconversional method is based on the Doyle approximation equation and can be expressed as

(1) $ln\beta =ln\frac{A{E}_{\alpha }}{Rg(\alpha )}-5.331-1.052\frac{E_{\alpha }}{RT}$(1)

where β and A refer to the heating rate (K·min⁻¹) and pre-exponential factor (min⁻¹). Eα is the apparent activation energy (kJ·mol⁻¹). T and R represent the absolute temperature (K) and universal gas constant (8.314 J·(mol·K)⁻¹), respectively. g(α) shows the integral form of reaction model, f(α).

The Friedman method is based on the assumption that solids decomposition depends only on the rate of mass loss and is independent from temperature. The equations for Friedman method are given below.

(2) $ln(\beta \frac{d\alpha }{dT})=ln\left[Af(\alpha )\right]-\frac{E_{\alpha }}{RT}$(2)

The integral isoconversional method proposed by KAS involves an approximation using the Murray and White equation that can be described as:

(3) $ln\frac{\beta }{T_{\hspace{0.17em}}^2}=ln\frac{AR}{E\alpha g(\alpha )}-\frac{E_{\alpha }}{RT}$(3)

Another integral isoconversional method originates from the Starink approximation equation can be expressed as follows:

(4) $ln\frac{\beta }{T^{1.92}}=-1.0008\frac{E_{\alpha }}{RT}+C$(4)

Tang method was developed based on a different estimation of Arrhenius temperature integral and is given as follows:

(5) $ln\frac{\beta }{T^{1.894661}}=-1.00145033\frac{E_{\alpha }}{RT}+C$(5)

The Boswell method is also an integral isoconversional method and can be expressed as

(6) $ln\frac{\beta }{T}=-\frac{E_{\alpha }}{RT}+C$(6)

For each α, corresponding T and β from thermogravimetric data are used to plot lnβ versus $\frac{1}{T}$, $ln(\beta \frac{d\alpha }{dT})$ versus $\frac{1}{T}$, $ln\frac{\beta }{T_{\hspace{0.17em}}^2}$versus $\frac{1}{T}$, $ln\frac{\beta }{T^{1.92}}$versus $\frac{1}{T}$, $ln\frac{\beta }{T^{1.894661}}$versus $\frac{1}{T}$ and $ln\frac{\beta }{T_{\hspace{0.17em}}^{\hspace{0.17em}}}$versus $\frac{1}{T}$. The Eα values can be determined from the regression slope for these model-free methods.

Results and discussion

Thermogravimetric analysis

The TG-DTG profiles of the cathode material are illustrated in Figure 1. The weight loss profiles showed a similar trend for these three heating rates. The whole decomposition process could be divided into three stages. At the low heating rate of 5 K·min⁻¹, three weight losses of 1.51%, 0.787%, and 0.449% were detected throughout temperatures of 30–500°C, 500–700°C, and 700–1000°C, respectively. The first weight loss was ascribed to the degradation of PVDF, consistent with the reported temperature of below 500°C (Rathore, Madhav, and Jaiswar et al. 2019; Xu et al. 2010, Zhao et al. 2017). The following stage was attributed to the oxidization of acetylene black. Cho et al. (2013) observed an exothermic peak temperature for the decomposition of acetylene black at 604°C, and Nie et al. (2015) revealed the acetylene black oxidization at 620°C. The last weight loss was due to the decomposition of lithium cobalt oxide and cobalt oxide (Antolini and Ferretti 1995; Zhang et al. 2014). In this study, the first two stages were the major decomposition stages for the cathode material, and were thus selected for the following kinetic study.

Figure 1. TG and DTG profiles of cathode at different heating rates

Activation energy calculation

The linear fitting from six model-free methods is displayed in Figure 2. For these methods, a large gap between the straight lines in the α range of 0.001–0.002 was observed, indicating a slower conversion of cathode material at initial stage, i.e., the low temperature cannot offer energy for rupturing the C-F bonds present in PVDF. Subsequently, the temperatures elevated enough to break down the molecules (Yao et al. 2020a).

Figure 2. Linear fitting curves under different conversions for model-free methods

The calculated Eα and R from different model-free methods are shown in Table 1 and Figure 3. The obtained Eα values showed an increasing trend, which slightly increased in the α range of 0.001–0.009, and then significantly increased. The later drastic increasement was representative of the great variety of chemical bonds and the multiphasic character of the conversion. Comparing these different model-free methods, the Eα values were comparable among KAS, Starink and Tang methods. The slight differences could be ascribed to the different approximations of temperature integral (Da Silva et al. 2018) and systematic error. The Eα values displayed a descending order: FWO > Boswell > Friedman > Tang, Starink and KAS. Since the FWO method gave the best fitting, the Eα values it yielded were applied for further decomposition mechanism study. This situation was consistent with that found in other literatures (Khiari, Moussaoui, and Jeguirim et al. 2019, Ma et al. 2018; Sokoto et al. 2016). For FWO method (Figure 3), the obtained Eα values slightly increased from 12.032 to 24.433 kJ·mol⁻¹ with α elevating from 0.001 to 0.009, and increased further to 43.187 kJ·mol⁻¹.

Table 1. Kinetic parameters derived from the model-free methods

	FWO		Friedman		KAS		Starink		Tang		Bosewell
α	Eα (kJ/mol)	R	Eα (kJ/mol)	R	Eα (kJ/mol)	R	Eα (kJ/mol)	R	Eα (kJ/mol)	R	Eα (kJ/mol)	R
0.001	12.032	0.9930	3.326	0.9051	4.215	0.9619	4.549	0.9665	4.653	0.9677	8.436	0.9874
0.002	15.582	0.9957	7.214	0.9791	7.255	0.9867	7.615	0.9876	7.725	0.9879	11.824	0.9935
0.003	18.280	0.9983	10.052	0.9943	9.558	0.9968	9.936	0.9969	10.053	0.9969	14.394	0.9978
0.004	19.554	0.9992	11.344	0.9977	10.475	0.9991	10.869	0.9991	10.990	0.9991	15.522	0.9991
0.005	20.772	0.9999	12.546	0.9999	11.297	0.9999	11.710	0.9999	11.836	0.9999	16.574	1.0000
0.006	21.630	0.9997	13.336	0.9985	11.705	0.9955	12.137	0.9960	12.269	0.9961	17.230	0.9988
0.007	22.411	0.9999	14.052	0.9996	12.120	0.9980	12.568	0.9982	12.704	0.9983	17.848	0.9996
0.008	23.111	0.9983	14.702	0.9962	12.544	0.9969	13.004	0.9970	13.145	0.9970	18.429	0.9979
0.009	24.433	0.9960	16.004	0.9909	13.637	0.9901	14.108	0.9906	14.251	0.9908	19.670	0.9943
0.01	26.640	0.9946	18.233	0.9883	15.658	0.9864	16.140	0.9871	16.286	0.9873	21.842	0.9921
0.011	29.365	0.9958	20.994	0.9917	18.200	0.9906	18.693	0.9910	18.841	0.9912	24.545	0.9941
0.012	31.592	0.9998	23.179	0.9993	20.090	0.9983	20.600	0.9985	20.752	0.9985	26.662	0.9994
0.013	36.702	0.9988	28.442	0.9974	25.146	0.9957	25.664	0.9959	25.817	0.9960	31.878	0.9978
0.014	43.187	0.9960	35.145	0.9928	31.641	0.9899	32.167	0.9903	32.320	0.9904	38.537	0.9938
Average	24.664	—	16.326	—	14.539	—	14.983	—	15.117	—	20.242	—

Figure 3. Comparison of kinetic parameters derived from model-free methods

Estimation of reaction mechanisms

The CR method (Coats and Redfern 1964) is an integral model-fitting method. For a given β and proposed reaction mechanisms g(α), plotting $ln\frac{g(\alpha )}{T^2}$ versus $\frac{1}{T}$ gives a straight line. The linear fittings were plotted and illustrated in Figure 4. The activation energy for all g(α) functions in Table 2 can be obtained from the slope. Comparing the coefficients, they were distinct for these models. The diffusion models showed better linear fitting, with R values of 0.9786 for D1, 0.9787 for D2, 0.9788 for D3, and 0.9788 for D4. The corresponding Eα values were 31.658, 31.694, 31.730, and 31.705 kJ·mol⁻¹, respectively. This situation has also been observed in the pyrolysis of biomass (Tonbul 2008) and plastics (Sinfronio et al. 2005). The degradation mechanism was studied by master-plots method. Figure 4 illustrates the comparison between the theoretical master curves and the experimental curve confirming that the fitting of experimental curve to the diffusion models.

Table 2. Expressions of f(α) and g(α) for reaction mechanisms

Reaction models
Nucleation models	Symbols	f(α)	g(α)
Power law	P1	4α^3/4	α^1/4
Power law	P2	3α^2/3	α^1/3
Power law	P3	2α^1/2	α^1/2
Power law	P4	2/3α^−1/2	α^3/2
Avrami-Erofeev	A2	2(1-α)[-ln(1-α)]^1/2	[-ln(1-α)]^1/2
Avrami-Erofeev	A3	3(1-a)[-ln(1-α)]^2/3	[-ln(1-α)]^1/3
Avrami-Erofeev	A4	4(1-α)[-ln(1-α)]^3/4	[-ln(1-α)]^1/4
Diffusion models
One-dimensional diffusion	D1	1/2α	α^2
Two-dimensional diffusion	D2	[-ln(1-α)]^−1	(1-α)ln(1-α)+ α
Three-dimensional diffusion (Jander equation)	D3	3(1-α)^2/3/2[1-(1-α)^1/3]	[1-(1-α)^1/3]^2
Four-dimensional diffusion (Ginstling-Brounshetein)	D4	3/2[(1-α)^−1/3-1]	1–2α/3-(1-α)^2/3
Reaction order models
First-order	F1	1-α	-ln(1-α)
Second-order	F2	(1-α)^2	1/(1-α)-1
Third-order	F3	(1-α)^3	[1/(1-α)^2–1]/2
Geometrical contraction models
Prout-Tompkins	R1	1	α
Contracting area	R2	2(1-α)^1/2	1-(1-α)^1/2
Contracting volume	R3	3(1-α)^2/3	1-(1-α)^1/3

Figure 4. Reaction mechanism determined using CR and master-plots method

Conclusion

As compared to chemical dissolving and decomposition, the pyrolysis pretreatment is an alternative and simple method. In this work, the pyrolysis kinetics of cathode material were comparatively studied. The thermal degradation mechanism was investigated by the CR and master-plot methods as well. The thermogravimetric analysis revealed that the degradation of cathode material from spent LIBs could be divided into three stages. Weight losses of 1.51%, 0.787%, and 0.449% were detected for the temperature ranges of 30–500°C, 500–700°C, and 700–1000°C, respectively. The Eα values calculated displayed a similar trend and were comparable among KAS, Starink and Tang methods. They slightly increased in the α range of 0.001–0.009, and then significantly increased. The FWO method offered the highest coefficients and the Eα values gradually increased from 12.032 to 24.433 kJ·mol⁻¹ as α increased from 0.001 to 0.009, then increased further to 43.187 kJ·mol⁻¹. Both CR and Criado methods indicated that the degradation of cathode material could be better described by using the diffusion models. This study will give significant references for the reactor design, optimization, and scaleup during industrial-scale treatments of spent LIBs.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was financially supported by the Zhejiang Provincial Natural Science Foundation of China (Grant no.LY19B070008 and LQ19E060008) and National Natural Science Foundation of China (Grant no. 51908171).

Notes on contributors

Shaoqi Yu

Shaoqi Yu is postgraduate in Hangzhou Dianzi University.

Baogui Zhang

Baogui Zhang is an engineer in Beijing Institute of Space Mechanics & Electricity. 

Jingjing Xiong

Jingjing Xiong is postgraduate in Hangzhou Dianzi University. 

Zhitong Yao

Zhitong Yao is a researchers in Hangzhou Dianzi University.

Daidai Wu

Daidai Wu is a researcher in Chinese Academy of Sciences.

Jie Liu

Jie Liu is a researchers in Hangzhou Dianzi University.

Shaodan Xu

Shaodan Xu is a researchers in Hangzhou Dianzi University.

Junhong Tang

Junhong Tang is a researchers in Hangzhou Dianzi University.