Professional Documents
Culture Documents
Machine Learning Assisted Prediction in The Discharge Capacities of Novel MXene Cathodes For Aluminum Ion Batteries
Machine Learning Assisted Prediction in The Discharge Capacities of Novel MXene Cathodes For Aluminum Ion Batteries
Research papers
A R T I C L E I N F O A B S T R A C T
Keywords: Conventional experimental methods are widely utilized to discover novel cathode materials for aluminum ion
Machine learning batteries (AIB), but their high price and low-efficiency present extra challenges when dealing with the vast family
Aluminum ion batteries of MXene materials. In order to vigorously promote the application and development of MXene cathodes in
MXene cathodes
aluminum ion batteries, a new accelerable method to predict the discharge capacities based on machine learning
Auxiliary means
is proposed. Firstly, we establish and train the machine learning models which extract from experimental results
of several MXene cathodes (Nb2C, Ti3C2, and V2C) to predict the battery status and initial discharge capacity of
mentioned cathodes. Secondly, we refine the models using few Ti2C experimental results to successfully predict
the battery status and initial discharge capacity for new Ti2C cathode, which exhibit higher capacity and longer
cycle life than Nb2C, Ti3C2 and V2C cathodes. This work provides a new method for predicting electrochemical
properties of advanced energy materials beyond MXene materials for aluminum-ion batteries, including but not
limited to carbon materials or transition metal oxides.
* Corresponding author.
** Correspondence to: X. Liu, School of Materials Science and Engineering, Institute of Materials Genome and Big Data, Harbin Institute of Technology, Shenzhen
518055, China.
E-mail addresses: shaobo.tu@ncu.edu.cn (S. Tu), xjliu@hit.edu.cn (X. Liu).
1
Contributed equally to this paper.
https://doi.org/10.1016/j.est.2023.110196
Received 31 October 2023; Received in revised form 1 December 2023; Accepted 18 December 2023
Available online 18 January 2024
2352-152X/© 2023 Published by Elsevier Ltd.
J. Li et al. Journal of Energy Storage 82 (2024) 110196
2
J. Li et al. Journal of Energy Storage 82 (2024) 110196
Fig. 1. Schematic workflow for performance prediction of MXene cathodes for aluminum ion batteries (AIB).
1 M TMAOH. The intercalated powder was washed with deionized water 2.2.3. Characterizations and measurements
for several centrifugation rounds to fabricate few-layered Ti2C MXenes, The structure and morphology of Ti2AlC MAX and Ti2C MXene were
denoted as Ti2C. studied through scanning electron microscopy (SEM: Zeiss Merlin) and
X-ray diffraction (XRD: Bruker D8 Advance). The cycling performance of
2.2.2. Preparation of Al-ion batteries Ti2C cathode was evaluated using a Neware battery testing system in a
The Ti2C cathode for the aluminum-ion battery were fabricated by 2.3 V voltage window (0.1–2.4 V) at a current density of 100 mA g− 1.
uniformly mixing Ti2C with carbon black powders and polyvinylidene
fluoride (PVDF) binder at a 7:2:1 mass ratio in N-Methylpyrrolidone 3. Results and discussion
(NMP) solvent. Then the resulting slurry was directly cast on a piece of
molybdenum foil collector to form the final Ti2C cathode. After drying at Fig. 1 illustrates the workflow of the proposed machine learning
110 ◦ C under vacuum for 12 h, the Ti2C cathode was assembled with Al strategy, which is applied on the speculation of the electrochemical
anode, AlCl3/1-Ethyl-3-methyl imidazolium chloride ([EMIm]Cl) elec performance of MXene cathodes for AIB. Generally, a classification
trolyte with a 1.3:1 M ratio, and Whatman GF/C microporous mem model for predicting battery status was established based on the
branes separator to fabricate a Swagelok-type Al ion battery in the experimental results of several MXene cathodes with over 70 % reten
glovebox. tion rate after 100 cycles at a current density of 100 mA g− 1. Typically, if
Fig. 2. (a) Prediction accuracies of battery status by using different algorithms. (b) The PCA classification results of battery status based on RFC algorithm on the
training set.
3
J. Li et al. Journal of Energy Storage 82 (2024) 110196
Fig. 3. The performance of each ML algorithm in terms of R, R2, MAE and RMSE on the training set by 10-fold cross-validation: (a) R and R2 and (b) MAE and RMSE;
(c) Prediction results of initial discharge capacity based on GBR algorithm.
the capacity retention rate of a AIB after 100 cycles at a current density %. All the relevant parameters of ML algorithm illustrated in Table S1.
of 100 mA g− 1 is <70 %, it is deemed damaged. Therefore, there are two Fig. 2 (b) depicts a PCA classification result about battery status pre
conditions: good or bad battery status. After removing the data of diction based on RFC algorithm with 97 % interpretation degree, which
damaged batteries, a regression model was used to predict the initial indicated that primary information will be largely retained and dimen
discharge capacity of MXene cathodes for AIB. sionality reduction method is feasible.
Fig. 4. The relative importance of different features on (a) battery status based on Random Forest Classification model and (b) initial discharge capacity based on
Gradient Boosting Regression model.
4
J. Li et al. Journal of Energy Storage 82 (2024) 110196
Fig. 5. The PCA classification results of battery status for Ti2C cathode based on RFC algorithm. (a) without optimization (accuracy is 44.44 %); (b) one-time
optimization (accuracy is 88.89 %).
5
J. Li et al. Journal of Energy Storage 82 (2024) 110196
Fig. 6. The predicted precision of regression model for predicting initial discharge capacity of Ti2C cathode before (red points or bars) and one-time optimization
(blue points or bars). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 7. The performance (a) and overall prediction results (b) of optimized ML model in predicting initial discharge capacity of Ti2C cathode.
learning algorithms. This method combines classification and regression capacity of new Ti2C cathodes and only a few Ti2C experimental results
models of supervised learning to predict the battery status and initial are iteratived to refine and improve the model performance. Finally, the
discharge capacity, which can accelerate to discover new MXene cath predictive results of machine learning are verified through experiments
odes for aluminum-ion batteries to achieve higher capacity and longer to determine their accuracy. Above all, the method provides a promising
cycle life. Specifically, we established and trained the machine learning avenue for predicting the battery status and initial discharge capacity of
models extracted from experimental results of several MXene cathodes advanced energy materials beyond Mxene materials for aluminum-ion
(Nb2C, Ti3C2, and V2C) to predict the battery status and initial discharge batteries, such as carbon materials or transition metal oxides materials.
6
J. Li et al. Journal of Energy Storage 82 (2024) 110196
Fig. 8. SEM images of (a) Ti2AlC MAX and (b) Ti2C MXene. XRD patterns (c) of the Ti2AlC MAX and Ti2C MXene. The first (dis)-charge voltage profiles (d) and cyclic
performance (e) of Ti2C cathode at a current density of 100 mA g− 1.
CRediT authorship contribution statement [3] Y. Gao, H. Yang, X. Wang, Y. Bai, N. Zhu, S. Guo, L. Suo, H. Li, H. Xu, C. Wu, The
compensation effect mechanism of Fe-Ni mixed Prussian blue analogues in aqueous
rechargeable aluminum-ion batteries, ChemSusChem 13 (2020) 732–740, https://
Jiahui Li: Conceptualization, Data curation, Formal analysis, doi.org/10.1002/cssc.201902884.
Investigation, Methodology, Writing – original draft, Writing – review & [4] A. Vahidmohammadi, A. Hadjikhani, S. Shahbazmohamadi, M.A. Kiani,
editing. Shengkun Xi: Data curation, Methodology, Conceptualization, H. Naghibi, M. Salavati-Niasari, Two-dimensional vanadium carbide (MXene) as a
high capacity cathode material for rechargeable aluminum batteries, ACS Nano 11
Formal analysis, Investigation, Writing – original draft, Writing – review (2017) 11135–11144, https://doi.org/10.1021/acsnano.7b05350.
& editing. Tongxing Lei: Investigation. Ruijun Yao: Methodology. [5] W. Lv, F. Zhang, J. Wang, Y. Du, J. Li, H. Ma, L. Zheng, Y. Zhao, Two-dimensional
Fanshuai Zeng: Methodology. Junwei Wu: Methodology. Shaobo Tu: V2C@Se (MXene) composite cathode material for high-performance rechargeable
aluminum batteries, Energy Storage Mater. 46 (2022) 138–146, https://doi.org/
Data curation, Methodology. Xingjun Liu: Conceptualization, Investi 10.1016/j.ensm.2021.11.019.
gation, Methodology, Project administration, Resources. [6] T. Bashir, M.Z. Iqbal, R. Javaid, M. Waqar, Progress in 3D-MXene electrodes for
lithium/sodium/potassium/magnesium/zinc/aluminum-ion batteries,
Electrochem. Energy Rev. 6 (2023), https://doi.org/10.1007/s41918-022-00285-
4.
Declaration of competing interest [7] M. Okubo, A. Sugahara, S. Kajiyama, A. Yamada, Hybrid batteries: state of the art
and future directions, Acc. Chem. Res. 51 (2018) 591–599, https://doi.org/
The authors declare that they have no known competing financial 10.1021/acs.accounts.7b00657.
[8] B. Anasori, M.R. Lukatskaya, Y. Gogotsi, 2D metal carbides and nitrides (MXenes)
interests or personal relationships that could have appeared to influence for energy storage, Nat. Rev. Mater. 2 (2017) 1–17, https://doi.org/10.1038/
the work reported in this paper. natrevmats.2016.98.
[9] Y. Wang, T. Guo, Z. Tian, K. Bibi, Y.Z. Zhang, H.N. Alshareef, MXenes for Energy
Harvesting, Adv. Mater. 2108560 (2022), https://doi.org/10.1002/
Data availability adma.202108560.
[10] F. Shahzad, M. Alhabeb, C.B. Hatter, B. Anasori, S. Man Hong, C.M. Koo,
The authors do not have permission to share data. Y. Gogotsi, Electromagnetic interference shielding with 2D transition metal
carbides (MXenes), Science 353 (2016) 1137–1140, https://doi.org/10.1126/
science.aag2421.
Acknowledgements [11] M.R. Lukatskaya, O. Mashtalir, C.E. Ren, Y. Dall’Agnese, P. Rozier, P.L. Taberna,
M. Naguib, P. Simon, M.W. Barsoum, Y. Gogotsi, Cation intercalation and high
volumetric capacitance of two-dimensional titanium carbide, Science 341 (2013)
This work was supported by the Shenzhen Science and Technology 1502–1505, https://doi.org/10.1126/science.1241488.
Program (Grant No. JCYJ20220531095212027) and the National Nat [12] S. Tu, L. Xu, J.K. El-Demellawi, H. Liang, X. Xu, S. Lopatin, S. De Wolf, X. Zhang, H.
ural Science Foundation of China (Grant No. 52203092). N. Alshareef, Autonomous MXene-PVDF actuator for flexible solar trackers, Nano
Energy 77 (2020) 105277, https://doi.org/10.1016/j.nanoen.2020.105277.
[13] Y. Pei, X. Zhang, Z. Hui, J. Zhou, X. Huang, G. Sun, W. Huang, Ti3C2Tx MXene for
Appendix A. Supplementary data sensing applications: recent progress, design principles, and future perspectives,
ACS Nano 15 (2021) 3996–4017, https://doi.org/10.1021/acsnano.0c10525.
[14] F. Ming, H. Liang, G. Huang, Z. Bayhan, H.N. Alshareef, MXenes for rechargeable
Supplementary data to this article can be found online at https://doi. batteries beyond the lithium-ion, Adv. Mater. 33 (2021) 2004039, https://doi.org/
org/10.1016/j.est.2023.110196. 10.1002/adma.202004039.
[15] Y.Z. Zhang, Y. Wang, Q. Jiang, J.K. El-Demellawi, H. Kim, H.N. Alshareef, MXene
printing and patterned coating for device applications, Adv. Mater. 32 (2020)
References 1908486, https://doi.org/10.1002/adma.201908486.
[16] M. Han, K. Maleski, C. Shuck, Y. Yang, J.T. Glazar, A.C. Foucher,
K. Hantanasirisakul, A. Sarycheva, N. Frey, S.J. May, V.B. Shenoy, E.A. Stach,
[1] K. Chayambuka, G. Mulder, D.L. Danilov, P.H.L. Sodium-ion battery materials and
Y. Gogotsi, Tailoring electronic and optical properties of MXenes through forming
electrochemical properties reviewed, Advanced Energy Materials 8 (8) (2018),
solid solutions, J. Am. Chem. Soc. 142 (2020) 19110–19118, https://doi.org/
https://doi.org/10.1002/aenm.202001310 (1800079.1-.49).
10.1021/jacs.0c05652.
[2] H. Yang, H. Li, J. Sun, Z. He, K. Cheng, H. Li, Feng, Angew. Chem. 58 (2019),
https://doi.org/10.1002/anie.201902987.
7
J. Li et al. Journal of Energy Storage 82 (2024) 110196
[17] Y.Z. Zhang, J.K. El-Demellawi, Q. Jiang, G. Ge, H. Liang, K. Lee, X. Dong, H. [27] Y. Ho, D. Pepyne, Simple explanation of the no-free-lunch theorem and its
N. Alshareef, MXene hydrogels: fundamentals and applications, Chem. Soc. Rev. 49 implications, J. Optim. Theory Appl. 115 (2002) 549–570, https://doi.org/
(2020) 7229–7251, https://doi.org/10.1039/D0CS00446K. 10.1023/A:1021251113462.
[18] J. Li, F. Zeng, K. Jehad El-Demellawi, Q. Lin, S. Xi, J. Wu, J. Tang, X. Zhang, X. Liu, [28] J. Yu, S. Guo, Y. Chen, J. Han, Y. Lu, Q. Jiang, C. Wang, X. Liu, A two-stage
S. Tu, Nb2CTx MXene cathode for high-capacity rechargeable aluminum batteries predicting model for γ′ solvus temperature of L12-strengthened Co-base superalloys
with prolonged cycle lifetime, ACS Appl. Mater. Interfaces 14 (40) (2022) based on machine learning, Intermetallics 110 (2019) 106466, https://doi.org/
45254–45262, https://doi.org/10.1021/acsami.1c17318. 10.1016/j.intermet.2019.106466.
[19] Xiaogeng Huo, et al., Two-dimensional composite of D-Ti3C2Tx@S@TiO2 [29] P. Belhumeur, J. Hespanha, D. Kriegman, Eigenfaces vs. fisherfaces: recognition
(MXene) as the cathode material for aluminum-ion batteries, Nanoscale 12 (5) using class specific linear projection, IEEE Trans. Pattern Anal. Mach. Intell. 19
(2020) 3387–3399, https://doi.org/10.1039/C9NR10663K. (1997) 711–720, https://doi.org/10.1109/34.598228.
[20] Z. Li, X. Wang, W. Zhang, S. Yang, Two-dimensional Ti3C2@CTAB-Se (MXene) [30] J. Yu, C. Wang, Y. Chen, C. Wang, X. Liu, Accelerated design of L12-strengthened
composite cathode material for high-performance rechargeable aluminum Co-base superalloys based on machine learning of experimental data, Mater. Des.
batteries, Chem. Eng. J. 398 (2020) 125679, https://doi.org/10.1016/j. 195 (2020) 108996, https://doi.org/10.1016/j.matdes.2020.108996.
cej.2020.125679. [31] P. Arora, R.E. White, M. Doyle, Capacity fade mechanisms and side reactions in
[21] A. Byeon, M.Q. Zhao, C.E. Ren, J. Halim, S. Kota, P. Urbankowski, B. Anasori, lithium-ion batteries, J. Electrochem. Soc. 140 (1993) 1–10, https://doi.org/
M. Barsoum, Y. Gogotsi, Two-dimensional titanium carbide MXene as a cathode 10.1002/chin.199847297.
material for hybrid magnesium/lithium-ion batteries, ACS Appl. Mater. Interfaces [32] X. Zhang, X. Zhang, X. Sun, Y. An, S. Song, C. Li, K. Wang, F. Su, C.M. Chen, F. Liu,
9 (2017) 4296–4300, https://doi.org/10.1021/acsami.6b15530. Z.S. Wu, Y. Ma, Electrochemical impedance spectroscopy study of lithium-ion
[22] N. Nosengo, Can artificial intelligence create the next wonder material? Nature 533 capacitors: modeling and capacity fading mechanism, J. Power Sources 488 (2021)
(2016) 22–25, https://doi.org/10.1038/533022a. 229454.
[23] P. Raccuglia, K.C. Elbert, P.D.F. Adler, C. Falk, A.J. Norquist, Machine-learning- [33] Y. Yang, B. Huang, V. Srinivasan, Machine learning approaches for predicting
assisted materials discovery using failed experiments, Nature 533 (2016) 73–76, capacity fade of lithium-ion batteries, J. Mater. Chem. A 5 (2017) 23117–23124.
https://doi.org/10.1038/nature17439. [34] Q. Ma, Y. Hu, P. Zuo, The role of interlayer distance in improving lithium storage
[24] S. Xi, J. Yu, L. Bao, L. Chen, Z. Li, R. Shi, C. Wang, X. Liu, Machine learning- performance of two-dimensional Ti3C2 MXene, J. Power Sources 489 (2021)
accelerated first-principles predictions of the stability and mechanical properties of 229442, https://doi.org/10.1016/j.jpowsour.2020.229442.
L12-strengthened cobalt-based superalloys, J. Mater. Inf. 2 (2022) 15, https://doi. [35] T. Lv, Y. Huang, F. Wang, W. Zhang, S. Yu, Crystal water enlarging the interlayer
org/10.1007/s43681-022-00048-y. spacing of ultrathin V2O5⋅4VO2⋅2.72 H2O nanobelts for high-performance
[25] R. Anand, S. Kaur, A. Goswami, A.K. Singh, V. Chaudhary, A. Verma, Doped MXene aqueous zinc-ion battery, Chem. Eng. J. 411 (2021) 128533, https://doi.org/
combinations as highly efficient bifunctional and multifunctional catalysts for 10.1016/j.cej.2020.128533.
water splitting and metal-air batteries, J. Mater. Chem. A 10 (2022) 22500–22511, [36] P. Wang, D. Xie, L. Wu, An interlayer spacing design approach for efficient sodium
https://doi.org/10.1039/d1ta10021h. ion storage in N-doped MoS2, Nanoscale Horiz. 8 (2023) 473–482, https://doi.org/
[26] M.Z.L. He, Machine learning and symbolic regression investigation on stability of 10.1039/D2NH01914F.
MXene materials, Comput. Mater. Sci. 196.1 (2021), https://doi.org/10.1016/j.
commatsci.2021.110578.