•A low-SOC resistance signal measured after formation can accurately predict cycle life•The signal is an indicator for lithium consumed during formation•The signal improves the detectability of lithium consumption over standard measures•The signal is measurable immediately after manufacturing using ordinary equipment Despite recent progress in battery development, electric vehicles remain unaffordable for many. A key enabler for less expensive electric vehicles is lowered battery manufacturing costs, a significant portion of which is due to the formation and aging process. Although some fast formation protocols have been proposed, a one-size-fits-all solution is unlikely to succeed in practice since an optimized formation protocol for one battery design will, in general, not be optimal for another. New formation protocols need to be vetted for their impacts on long-term battery lifetime—a slow process that hinders the discovery of optimal formation protocols. Here, we identify a scalable method for predicting the effect of new formation protocols on cycle life. The method is obtained directly at the end of the manufacturing line and can be deployed immediately in mass production settings to improve diagnostics of new formation protocols. Increasing the speed of battery formation can significantly lower lithium-ion battery manufacturing costs. However, adopting faster formation protocols in practical manufacturing settings is challenging due to a lack of inexpensive, rapid diagnostic signals that can inform possible impacts to long-term battery lifetime. This work identifies the cell resistance measured at low states of charge as an early-life diagnostic feature for screening new formation protocols. We show that this signal correlates to cycle life and improves the accuracy of data-driven battery lifetime prediction models. The signal is obtainable at the end of the manufacturing line, takes seconds to acquire, and does not require specialized test equipment. We explore a physical connection between this resistance signal and the quantity of lithium consumed during formation, suggesting that the signal may be broadly applicable for evaluating any manufacturing process change that could impact the total lithium consumed during formation. Increasing the speed of battery formation can significantly lower lithium-ion battery manufacturing costs. However, adopting faster formation protocols in practical manufacturing settings is challenging due to a lack of inexpensive, rapid diagnostic signals that can inform possible impacts to long-term battery lifetime. This work identifies the cell resistance measured at low states of charge as an early-life diagnostic feature for screening new formation protocols. We show that this signal correlates to cycle life and improves the accuracy of data-driven battery lifetime prediction models. The signal is obtainable at the end of the manufacturing line, takes seconds to acquire, and does not require specialized test equipment. We explore a physical connection between this resistance signal and the quantity of lithium consumed during formation, suggesting that the signal may be broadly applicable for evaluating any manufacturing process change that could impact the total lithium consumed during formation. IntroductionWith the increasing demand for electric vehicles, global lithium-ion battery manufacturing capacity is quickly approaching the terawatt-hour scale.1Australian TradeInvestment CommissionThe lithium-ion battery value chain: new economy opportunities for Australia. (Australian Trade and Investment Commission), 2018https://apo.org.au/node/210341Google Scholar, 2Benchmark minerals intelligence EV battery arms race enters new gear with 115 megafactories, Europe sees most rapid growth.2019https://www.benchmarkminerals.com/ev-battery-arms-race-enters-new-gear-with-115-megafactories-europe-sees-most-rapid-growth/Google Scholar, 3Wood Mackenzie Global lithium-ion cell manufacturing capacity to quadruple to 1.3 TWh by 2030.2020https://www.woodmac.com/press-releases/global-lithium-ion-cell-manufacturing-capacity-to-quadruple-to-1.3-twh-by-2030/Google Scholar A key step in battery manufacturing is formation/aging, which has been estimated to account for up to 30% of total manufacturing costs.4Liu Y. Zhang R. Wang J. Wang Y. Current and future lithium-ion battery manufacturing.iScience. 2021; 24: 102332https://doi.org/10.1016/j.isci.2021.102332Google Scholar, 5Nelson P.A. Gallagher K.G. Bloom I. Dees D.W. Modeling the performance and cost of lithium-ion batteries for electric-drive vehicles. Argonne National Laboratory, 2011https://publications.anl.gov/anlpubs/2011/10/71302.pdfGoogle Scholar, 6Duffner F. Mauler L. Wentker M. Leker J. Winter M. Large-scale automotive battery cell manufacturing: analyzing strategic and operational effects on manufacturing costs.Int. J. Prod. Econ. 2021; 232: 107982https://doi.org/10.1016/j.ijpe.2020.107982Google Scholar, 7Küpper D. Kuhlmann K. Wolf S. Pieper C. Xu G. Ahmad J. The future of battery production for electric vehicles.2018https://www.bcg.com/publications/2018/future-battery-production-electric-vehiclesGoogle Scholar, 8Wood D.L. Li J. Daniel C. Prospects for reducing the processing cost of lithium ion batteries.J. Power Sources. 2015; 275: 234-242https://doi.org/10.1016/j.jpowsour.2014.11.019Google Scholar The formation/aging process involves charging and discharging hundreds of thousands of cells in environmentally controlled chambers, an expensive process that takes days to weeks to complete but is necessary to improve battery performance and lifetime.9Winter M. The solid electrolyte interphase - the most important and the least understood solid electrolyte in rechargeable Li batteries.Z. Phys. chem. 2009; 223: 1395-1406https://doi.org/10.1524/zpch.2009.6086Google Scholar, 10An S.J. Li J. Daniel C. Mohanty D. Nagpure S. Wood D.L. The state of understanding of the lithium-ion-battery graphite solid electrolyte interphase (SEI) and its relationship to formation cycling.Carbon. 2016; 105: 52-76https://doi.org/10.1016/j.carbon.2016.04.008Google Scholar, 11Wang A. Kadam S. Li H. Shi S. Qi Y. Review on modeling of the anode solid electrolyte interphase (SEI) for lithium-ion batteries.NPJ Comput. Mater. 2018; 4https://doi.org/10.1038/s41524-018-0064-0Google Scholar, 12Peled E. Menkin S. Review—SEI: past, present and future.J. Electrochem. Soc. 2017; 164: A1703-A1719https://doi.org/10.1149/2.1441707jesGoogle Scholar, 13Goers D. Spahr M.E. Leone A. Märkle W. Novák P. The influence of the local current density on the electrochemical exfoliation of graphite in lithium-ion battery negative electrodes.Electrochim. Acta. 2011; 56: 3799-3808https://doi.org/10.1016/j.electacta.2011.02.046Google Scholar, 14Lu P. Li C. Schneider E.W. Harris S.J. Chemistry, impedance, and morphology evolution in solid electrolyte interphase films during formation in lithium ion batteries.J. Phys. Chem. C. 2014; 118: 896-903https://doi.org/10.1021/jp4111019Google Scholar Given the high cost burden, manufacturers are incentivized to develop new formation processes that decrease the total time consumed by formation/aging. A variety of fast formation strategies have been studied in academic literature, which employ some combination of rapid charge-discharge cycles, restricted voltage windows, and optimized temperature.10An S.J. Li J. Daniel C. Mohanty D. Nagpure S. Wood D.L. The state of understanding of the lithium-ion-battery graphite solid electrolyte interphase (SEI) and its relationship to formation cycling.Carbon. 2016; 105: 52-76https://doi.org/10.1016/j.carbon.2016.04.008Google Scholar,15An S.J. Li J. Du Z. Daniel C. Wood D.L. Fast formation cycling for lithium ion batteries.J. Power Sources. 2017; 342: 846-852https://doi.org/10.1016/j.jpowsour.2017.01.011Google Scholar, 16Wood D.L. Li J. An S.J. Formation challenges of Lithium-Ion Battery Manufacturing.Joule. 2019; 3: 2884-2888https://doi.org/10.1016/j.joule.2019.11.002Google Scholar, 17Mao C. An S.J. Meyer H.M. Li J. Wood M. Ruther R.E. Wood D.L. Balancing formation time and electrochemical performance of high energy lithium-ion batteries.J. Power Sources. 2018; 402: 107-115https://doi.org/10.1016/j.jpowsour.2018.09.019Google Scholar, 18Müller V. Kaiser R. Poller S. Sauerteig D. Schwarz R. Wenger M. Lorentz V.R.H. März M. Introduction and application of formation methods based on serial-connected lithium-ion battery cells.J. Energy Storage. 2017; 14: 56-61https://doi.org/10.1016/j.est.2017.09.013Google Scholar, 19Antonopoulos B.K. Stock C. Maglia F. Hoster H.E. Solid electrolyte interphase: can faster formation at lower potentials yield better performance? Electrochim.Acta. 2018; 269: 331-339https://doi.org/10.1016/j.electacta.2018.03.007Google Scholar, 20Zhang S.S. Xu K. Jow T.R. Optimization of the forming conditions of the solid-state interface in the Li-ion batteries.J. Power Sources. 2004; 130: 281-285https://doi.org/10.1016/j.jpowsour.2003.12.012Google Scholar, 21Heimes H.H. Offermanns C. Mohsseni A. Laufen H. Westerhoff U. Hoffmann L. Niehoff P. Kurrat M. Winter M. Kampker A. The effects of mechanical and thermal loads during lithium-ion pouch cell formation and their impacts on process time.Energy Technol. 2020; 8: 1900118https://doi.org/10.1002/ente.201900118Google Scholar, 22Münster P. Diehl M. Frerichs J.E. Börner M. Hansen M.R. Winter M. Niehoff P. Effect of Li plating during formation of lithium ion batteries on their cycling performance and thermal safety.J. Power Sources. 2021; 484: 229306https://doi.org/10.1016/j.jpowsour.2020.229306Google Scholar, 23Pathan T.S. Rashid M. Walker M. Widanage W.D. Kendrick E. Active formation of Li-ion batteries and its effect on cycle life.J. Phys. 2019; 1: 044003https://doi.org/10.1088/2515-7655/ab2e92Google Scholar, 24Müller V. Kaiser R. Poller S. Sauerteig D. Importance of the constant voltage charging step during lithium-ion cell formation.J. Energy Storage. 2018; 15: 256-265https://doi.org/10.1016/j.est.2017.11.020Google Scholar, 25Rago N.D. Basco J.K. Vu A. Li J. Hays K. Sheng Y. Wood D.L. Bloom I. Effect of formation protocol: cells containing Si-graphite composite electrodes.J. Power Sources. 2019; 435: 126548https://doi.org/10.1016/j.jpowsour.2019.04.076Google Scholar, 26Lee H.-H. Wang Y.-Y. Wan C.-C. Yang M.-H. Wu H.-C. Shieh D.-T. A fast formation process for lithium batteries.J. Power Sources. 2004; 134: 118-123https://doi.org/10.1016/j.jpowsour.2004.03.020Google Scholar Recent studies have shown that formation time can be decreased while preserving battery lifetime,16Wood D.L. Li J. An S.J. Formation challenges of Lithium-Ion Battery Manufacturing.Joule. 2019; 3: 2884-2888https://doi.org/10.1016/j.joule.2019.11.002Google Scholar,22Münster P. Diehl M. Frerichs J.E. Börner M. Hansen M.R. Winter M. Niehoff P. Effect of Li plating during formation of lithium ion batteries on their cycling performance and thermal safety.J. Power Sources. 2021; 484: 229306https://doi.org/10.1016/j.jpowsour.2020.229306Google Scholar,25Rago N.D. Basco J.K. Vu A. Li J. Hays K. Sheng Y. Wood D.L. Bloom I. Effect of formation protocol: cells containing Si-graphite composite electrodes.J. Power Sources. 2019; 435: 126548https://doi.org/10.1016/j.jpowsour.2019.04.076Google Scholar although conclusions remain tenuous due to the limited sample sizes typically used.In real manufacturing settings, a “one-size-fits-all” formation protocol is unlikely to exist since cell designs with different electrolytes, electrodes, and active materials influence important formation factors such as charging capability, electrode wettability, and solid electrolyte interphase (SEI) reaction pathways. However, cycle life testing often takes months or years to complete, posing a significant barrier to the adoption of new, potentially cost-saving formation protocols. While characterization techniques, such as volume change detection,27Wang X. Sone Y. Segami G. Naito H. Yamada C. Kibe K. Understanding volume change in lithium-ion cells during charging and discharging using in situ measurements.J. Electrochem. Soc. 2007; 154: A14https://doi.org/10.1149/1.2386933Google Scholar, 28Bauer M. Wachtler M. Stöwe H. Persson J.V. Danzer M.A. Understanding the dilation and dilation relaxation behavior of graphite-based lithium-ion cells.J. Power Sources. 2016; 317: 93-102https://doi.org/10.1016/j.jpowsour.2016.03.078Google Scholar, 29Mohtat P. Lee S. Sulzer V. Siegel J.B. Stefanopoulou A.G. Differential expansion and voltage model for Li-ion batteries at practical charging rates.J. Electrochem. Soc. 2020; 167: 110561https://doi.org/10.1149/1945-7111/aba5d1Google Scholar impedance spectroscopy,15An S.J. Li J. Du Z. Daniel C. Wood D.L. Fast formation cycling for lithium ion batteries.J. Power Sources. 2017; 342: 846-852https://doi.org/10.1016/j.jpowsour.2017.01.011Google Scholar,30Zhang Y.Y. Tang Q. Zhang Y.Y. Wang J. Stimming U. Lee A.A. Identifying degradation patterns of lithium ion batteries from impedance spectroscopy using machine learning.Nat. Commun. 2020; 11: 1706https://doi.org/10.1038/s41467-020-15235-7Google Scholar acoustic spectroscopy,31Bommier C. Chang W. Lu Y. Yeung J. Davies G. Mohr R. Williams M. Steingart D. In Operando acoustic detection of lithium metal plating in commercial LiCoO2/Graphite pouch cells.Cell Rep. Phys. Sci. 2020; 1: 100035https://doi.org/10.1016/j.xcrp.2020.100035Google Scholar, 32Davies G. Knehr K.W. Van Tassell B. Hodson T. Biswas S. Hsieh A.G. Steingart D.A. State of charge and state of health estimation using electrochemical acoustic time of flight analysis.J. Electrochem. Soc. 2017; 164: A2746-A2755https://doi.org/10.1149/2.1411712jesGoogle Scholar, 33Knehr K.W. Hodson T. Bommier C. Davies G. Kim A. Steingart D.A. Understanding full-cell evolution and non-chemical electrode crosstalk of Li-ion batteries.Joule. 2018; 2: 1146-1159https://doi.org/10.1016/j.joule.2018.03.016Google Scholar, 34Deng Z. Hu X. Lin X. Xu L. Che Y. Hu L. General discharge voltage information enabled health evaluation for lithium-ion batteries.IEEE/ASME Trans. Mechatron. 2020; 26: 1295-1306https://doi.org/10.1109/TMECH.2020.3040010Google Scholar and X-ray tomography,35Pietsch P. Wood V. X-ray tomography for lithium ion battery research: a practical guide.Annu. Rev. Mater. Res. 2017; 47: 451-479https://doi.org/10.1146/annurev-matsci-070616-123957Google Scholar,36Wood V. X-ray tomography for battery research and development.Nat. Rev. Mater. 2018; 3: 293-295https://doi.org/10.1038/s41578-018-0053-4Google Scholar have been proposed for use in manufacturing settings, these methods can be costly to implement since the metrology will need to be deployed at scale in the battery factory. Diagnostic features obtainable from already existing cycling equipment and especially those using only current-voltage signals37An S.J. Park J.Y. Song J. Lee J. Kim G.H. Yoon J. Oh B. A fast method for evaluating stability of lithium ion batteries at high C-rates.J. Power Sources. 2020; 480: 228856https://doi.org/10.1016/j.jpowsour.2020.228856Google Scholar are thus highly attractive.In this work, we show that the cell resistance at low states of charge (SOC) can be used to screen new formation protocols and predict battery lifetime. Our work shows that this signal, measured at the beginning of life, is a stronger predictor of battery lifetime than conventional signals such as Coulombic efficiency (CE). This metric can be measured within seconds and integrated directly into the battery manufacturing process with no additional capital costs. This low-SOC resistance metric can, thus, be deployed in practical manufacturing settings to accelerate the evaluation of new formation protocols. We further demonstrate that the low-SOC resistance (RLS) decreases as the quantity of lithium lost to the SEI during formation increases. With our physical insight, we propose that RLS, in principle, can also be used to diagnose the impact of any manufacturing process that alters the total lithium consumed during formation.Results and discussionFast formation experimental designTwo formation protocols have been implemented in this work: a fast formation protocol previously reported by Wood et al.15An S.J. Li J. Du Z. Daniel C. Wood D.L. Fast formation cycling for lithium ion batteries.J. Power Sources. 2017; 342: 846-852https://doi.org/10.1016/j.jpowsour.2017.01.011Google Scholar,16Wood D.L. Li J. An S.J. Formation challenges of Lithium-Ion Battery Manufacturing.Joule. 2019; 3: 2884-2888https://doi.org/10.1016/j.joule.2019.11.002Google Scholar that completes within 14 h (Figure S1B) and a baseline formation protocol (Figure S1C) that completes in 56 h. The fast formation protocol maximizes the time spent at low negative electrode potentials to promote the creation of a more passivating SEI.15An S.J. Li J. Du Z. Daniel C. Wood D.L. Fast formation cycling for lithium ion batteries.J. Power Sources. 2017; 342: 846-852https://doi.org/10.1016/j.jpowsour.2017.01.011Google Scholar,38Attia P.M. Harris S.J. Chueh W.C. Benefits of fast battery formation in a model system.J. Electrochem. Soc. 2021; 168: 050543https://doi.org/10.1149/1945-7111/abff35Google Scholar, 39Kim S.-P. van Duin A.C.T. Shenoy V.B. Effect of electrolytes on the structure and evolution of the solid electrolyte interphase (SEI) in Li-ion batteries: a molecular dynamics study.J. Power Sources. 2011; 196: 8590-8597https://doi.org/10.1016/j.jpowsour.2011.05.061Google Scholar, 40Zhang S. Ding M.S. Xu K. Allen J. Jow T.R. Understanding solid electrolyte interface film formation on graphite electrodes.Electrochem. Solid-State Lett. 2001; 4: 206-209https://doi.org/10.1149/1.1414946Google ScholarForty nickel manganese cobalt (NMC)/graphite pouch cells with a nominal capacity of 2.36 Ah were built for this study (Table S1). Half of the cells underwent fast formation, and the remaining cells underwent baseline formation. Cells were further subdivided into “room temperature” and “45°C” aging groups for cycle life testing. The cycling profile was identical for all cells: 1 C charge to 4.2 V with a constant voltage (CV) hold to 10 mA and 1 C discharge to 3.0 V. Reference performance tests (RPTs)41Dubarry M. Baure G. Perspective on commercial Li-ion battery testing, best practices for simple and effective protocols.Electronics. 2020; 9: 206https://doi.org/10.3390/electronics9010152Google Scholar were inserted throughout the cycle life test, which includes slow (C/20) charge and discharge curves as well as a hybrid pulse power characterization (HPPC) sequence42Christopherson J.P. Battery test manual for plug-in hybrid electric vehicles. Idaho National Laboratory, 2015https://www.osti.gov/biblio/1186745-battery-test-manual-electric-vehicles-revisionGoogle Scholar used to extract the cell internal resistance as a function of SOC.Our experimental design (Figure S1A) uses larger samples sizes (n=10 per group) compared with those typically reported in the literature, which often use three cells or fewer per group. The increased sample size enables a more statistically rigorous analysis of the impact of different formation protocols on cell characteristics at the beginning and the end of life.Fast formation cells had longer cycle lifeFast formation cells had higher average lifetimes than the baseline formation cells under the cycle life test, as shown in Figure 1. The degradation rate of fast formation cells initially track the baseline formation cells closely under both temperatures tested (Figures 1A and 1C). However, after 250 cycles, all cells begin to lose capacity rapidly. The fast formation cells sustained over 100 cycles longer before reaching the end of life, defined as when cells reach 70% of their initial measured capacity (Figures 1B and 1D). This result is highly statistically significant (p-value < 0.001). The general result that fast formation improved lifetime performance holds across multiple performance metrics, including CE (Figure S4), voltage efficiency (Figure S5), as well as when plotted against equivalent cycles (Figure S7). Together, these results support the growing body of evidence that well-designed fast formation protocols can improve cycle life.15An S.J. Li J. Du Z. Daniel C. Wood D.L. Fast formation cycling for lithium ion batteries.J. Power Sources. 2017; 342: 846-852https://doi.org/10.1016/j.jpowsour.2017.01.011Google Scholar,22Münster P. Diehl M. Frerichs J.E. Börner M. Hansen M.R. Winter M. Niehoff P. Effect of Li plating during formation of lithium ion batteries on their cycling performance and thermal safety.J. Power Sources. 2021; 484: 229306https://doi.org/10.1016/j.jpowsour.2020.229306Google Scholar,38Attia P.M. Harris S.J. Chueh W.C. Benefits of fast battery formation in a model system.J. Electrochem. Soc. 2021; 168: 050543https://doi.org/10.1149/1945-7111/abff35Google ScholarFinding diagnostic signals at the beginning of lifeGiven the demonstrated impact of formation protocol on battery cycle life, we next investigate methods to quantify the impact of fast formation on the initial cell state. Differences in the initial cell state (e.g., the amount of lithium consumed during formation) may offer clues as to how fast formation could have improved cycle life. We focused on studying signals directly obtainable from full cell current-voltage data, which offer the lowest barrier-to-entry for deployment in real manufacturing settings.Conventional metrics of formation efficiencyFigures 2A–2C show standard measures of formation efficiency extracted from the formation cycling data. The discharge capacity, Qd, was measured at the end of each formation protocol during a C/10 discharge step from 4.2 to 3.0 V. Qd corresponds to the capacity of cyclable lithium excluding the contribution from lithium irreversibly lost to the SEI during formation. Fast formation decreased Qd by 0.3%, a small but statistically significant difference (p=0.01). The charge capacity, Qc, was taken during the initial charge cycle and includes both the capacity of cyclable lithium as well as the capacity of lithium lost irreversibly to the SEI. The quantity of lithium inventory lost to the SEI can be calculated as QLLI=Qc−Qd (Figure 2B). Note that while the two formation protocols differed in the initial charging rate, Qc remains a fair comparison metric since both charge protocols ended on a potentiostatic hold at 4.2 V until the current dropped below C/100. Fast formation increased QLLI by 23 mAh (p=0.03). Finally, we also included another common evaluation metric, the formation CE, defined as CEf=Qd/Qc (Figure 2C), which shows that fast formation decreased CEf by 0.8% (p=0.02). Measured values are summarized in Table 1. Together, the results show that fast formation marginally increased the amount of lithium consumed during formation. A p-value of less than 0.05 in all cases indicate that the measured differences, while small, are statistically significant to a least a 95% confidence level.Figure 2Diagnostic signals for differences in the initial cell stateShow full caption(A) Final discharge capacity.(B) Capacity of lithium inventory lost during formation.(C) Formation Coulombic efficiency, measured from the formation protocol.(D) 10-s resistance obtained from the hybrid pulse power characterization test prior to the start of the cycle life test.(E) Magnification of the 10-s resistance at low SOCs.(F) Distribution of 10-s resistance at 5% SOC comparing between the two formation protocols.(A–C) Are extracted directly from the formation test protocol run on each cell. (D–F) Are extracted from the initial reference performance test from the 45°C cycle life test (see Figure S9 for the results from the room temperature cycle life test). “∗”—statistically significant with p-value < 0.05. “∗∗∗”—statistically significant with p-value <0.001.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Table 1Comparison of initial cell state metricsMetricUnitTemperatureBaseline formationFast formationDelta (abs)Delta (%)p-valueQdmAhroom temp2370 (11)2362 (7)−8−0.3%0.01QLLI (Qc−Qd)mAhroom temp346 (27)369 (35)+23+6.6%0.03CEf%room temp87.3 (0.9)86.5 (1.1)−0.8−0.9%0.02R10s,5%SOC (RLS)mΩroom temp139.7 (2.9)130.0 (2.3)−9.7−6.9%< 0.001R10s,5%SOC (RLS)mΩ45°C48.7 (1.6)43.8 (1.1)−4.9−10.0%< 0.001R10s, 90%SOCmΩroom temp23.6 (0.1)23.9 (1.0)+0.3+1.3%0.28R10s, 90%SOCmΩ45°C14.5 (0.4)14.9 (0.5)+0.4+2.8%0.10Values are reported as mean (standard deviation). Qd, QLLI, and CEf are extracted directly from the formation test protocol. Resistance metrics are extracted from the initial reference performance test at the beginning of the cycle life test. Open table in a new tab Low-SOC resistanceFollowing formation, the cell internal resistance was measured using the hybrid pulse power characterization (HPPC) technique42Christopherson J.P. Battery test manual for plug-in hybrid electric vehicles. Idaho National Laboratory, 2015https://www.osti.gov/biblio/1186745-battery-test-manual-electric-vehicles-revisionGoogle Scholar prior to the start of the cycle life test. During this test, a series of 10-s, 1 C discharge pulses were applied to the cell at varying SOCs, and the resistance is calculated using Ohm’s law (Figure S8). The 10-s resistance, R10s, was plotted against SOC for all cells cycled at 45°C (Figure 2D). R10s generally remained flat at mid-to high SOCs. The peak at 55% SOC corresponds to the stage 2 solid-solution regime of the graphite negative electrode.43Dahn J.R. Phase diagram of LixC6.Phys. Rev. B. 1991; 44: 9170-9177Google Scholar R10s rose sharply below 10% SOC. Focusing on the low-SOC region (Figure 2E), we observed that R10s measured at 4% and 8% SOC were lower for fast formation cells compared with that of baseline formation cells. This result was highly statistically significant, with a p-value less than 0.001 (Figure 2F). A similar result held when R10s was measured at room temperature (Figure S9). At mid to high SOCs, differences in R10s between fast formation and baseline formation cells were generally not statistically significant (Figure S9). Thus, differences in resistance between the two formation protocols appeared uniquely at low SOCs. All initial cell state metrics are summarized as part of Table 1.To study the robustness of the low-SOC resistance signal, we varied the SOC set-point between 4% and 10% and also computed the resistance under 1 and 5-s pulse durations. In all cases, the resistance metric provided a high degree of contrast between the two different formation protocols (Figures S10 and S11). The lowest SOC measured in our dataset was 4% SOC.The remainder of the paper will focus on the resistance measured at 5% SOC and with a 10-s pulse duration. From hereon, this metric will be referred to as the “low-SOC resistance,” RLS.Low-SOC resistance as a diagnostic signal: A data-driven perspectiveLow-SOC resistance correlates to cycle lifeTo evaluate the merit of RLS as a diagnostic feature, we explored the correlations between the initial cell metrics (Figure 2) and cycle life, defined as cycles to 70% of the initial capacity. The results are shown in Figure 3. Out of all metrics studied, RLS was the only signal with a meaningful correlation to cycle life, with a correlation coefficient of ρ=−0.84. Other metrics such as Qd and CEf were poorly correlated to cycle life (|ρ|<0.5). We attribute the weakness of these correlations to the poor signal-to-noise inherent in cell capacity measurements in the absence of high-precision cycling,44Smith A.J. Burns J.C. Xiong D. Dahn J.R. Interpreting high precision coulometry results on Li-ion cells.J. Electrochem. Soc. 2011; 158: A1136-A1142https://doi.org/10.1149/1.3625232Google Scholar,45Fathi R. Burns J.C. Stevens D.A. Ye H. Hu C. Jain G. Scott E. Schmidt C. Dahn J.R. Ultra high-precision studies of degradation mechanisms in aged LiCoO 2 /graphite Li-ion cells.J. Electrochem. Soc. 2014; 161: A1572-A1579https://doi.org/10.1149/2.0321410jesGoogle Scholar a topic we explore in detail later. The resistance measured at high SOCs also did not correlate to cycle life. From these results, we observe that the low-SOC signal uniquely holds information related to cycle life. These results have been reproduced for different end-of-life definitions ranging between 50% and 80% (Figures S13 and S14), as well as for charge pulses (Figure S15).Figure 3Correlation between early-life diagnostic signals and cycle lifeShow full caption(A–D) Correlations under room temperature cycling.(E–H) Correlations under 45°C cycling.Cycle life is defined as cycles to 70% of initial capacity. QLLI and CEf are taken directly from the formation test. R10s,5%SOC (RLS) and R10s,90%SOC are measured at the beginning of the cycle life test and, thus, share the same temperature as the cycle life test.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Low-SOC resistance predicts cycle lifeTo understand if RLS can be used to improve battery lifetime prediction, we trained univariate prediction models with regularized linear regression models inspired by Severson et al.46Severson K.A. Attia P.M. Jin N. Perkins N. Jiang B. Yang Z. Chen M.H. Aykol M. Herring P.K. Fraggedakis D. et al.Data-driven prediction of battery cycle life before capacity degradation.Nat. Energy. 2019; 4: 383-391https://doi.org/10.1038/s41560-019-0356-8Google Scholar The performance of the predictive models are summarized in Table 2. A dummy regressor, which predicts the mean of the training