Operando bulk and interfacial characterization for electrochemical energy storage: Case study employing isothermal microcalorimetry and X-ray absorption spectroscopy

The global shift to electricity as the main energy carrier will require innovation in electrochemical energy storage (EES). EES systems are the key to the “electron energy economy,” minimizing losses and increasing reliability between energy supply and demand. However, steep challenges such as cost, cycle/calendar life, energy density, material availability, and safety limit widespread adoption of batteries for large-scale grid and vehicle applications. Battery innovation that meets today’s challenges will require new chemistries, which can originate from understanding charge transport phenomena at multiple time and length scales. The advancement of operando characterization can expedite this progress as changes can be observed during battery function. This article highlights progress in bulk and interfacial operando characterization of batteries. Specifically, a case study involving Fe3O4 is provided demonstrating that combining X-ray absorption spectroscopy and isothermal microcalorimetry can provide real-time characterization of productive faradaic redox processes and parasitic interfacial reactions during (de)lithiation.


Operando characterization of electrochemical energy storage
Electrochemical energy storage is critical to achieve an "electron energy economy, " where electricity generated by renewable resources powers all end-use applications. Rechargeable lithiumion batteries are among the most adopted energy storage systems; however, demands for higher power and energy densities, longer lifetime, and more rigorous safety assurance require significant material chemistry and engineering innovation. Challenges for lithium-ion batteries include sluggish ion transport due to inhomogeneous material phase transitions, irreversible electrode-electrolyte parasitic reactions, and energy losses arising from heat dissipation. Therefore, the ability to characterize the structural evolution, electrolyte decomposition, and thermal behaviors concomitant with the battery's electrochemistry represent an important approach toward solving the complex technical challenges of rechargeable batteries.
Operando characterization is at the forefront of battery research [1]. Unlike ex-situ characterization, which requires removing battery components from the cell body, operando characterization is non-destructive revealing synchronous electrochemical information that reflects working conditions when current is applied. The majority of operando battery characterization relies on electromagnetic (X-rays, optical) or particle radiation including angle and energy dispersive X-ray diffraction (ADXRD, EDXRD), X-ray absorption spectroscopy (XAS), transmission X-ray microscopy (TXM), X-ray microfluorescence (μ-XFM), neutron powder diffraction (NPD), transmission electron microscopy (TEM), nuclear magnetic resonance (NMR), and Raman spectroscopy. These techniques have led to advances in understanding the charge storage mechanisms that govern next generation battery chemistries by detecting structural and chemical changes at the electrodes and electrode/electrolyte interphases. However, the listed optical and X-ray based operando techniques often require distinct cell configurations, access to specialty facilities such as synchrotron light sources, and significant experience in conducting electrochemical measurements and collecting/analyzing large data sets of spectra. Additionally, exposure to light, electrons, neutrons, and X-rays can potentially change battery materials overtime leading to possible ambiguity in the results depending on the system. Therefore, non-radiation operando methods can prove indispensable as the risk of radiation damage is eliminated and instrumentation is more accessible.
Generally, the energy balance of a battery is expressed as: [6] where C p , Q , and q are the heat capacity, the rate of heat generated from an active cell, and the rate of heat dissipation from a battery to ambient conditions, respectively. The heat generated from an electrochemical cell under the passage of current leads to a local temperature change, which dissipates to its surroundings. In calorimetry, the thermal flux measured by the instrument represents the heat dissipation from the cell through the heat transfer media (i.e., air, oil etc.). In an isothermal microcalorimeter (IMC) heat flow is monitored at constant temperature; therefore, the capacitive heat ( C p dT dt ) and the heat transfer coefficients ( h =q A(�T) ) associated with heat transfer media are neglected. Thus, the rate of heat (referred as "heat flow" herein) dissipated from the cell ( q ) represents total heat flow generated by the cell ( Q ) in IMC experiments.
Precedent literature has described theoretically and shown experimentally that the total heat flow measured by the IMC is composed of three predominate sources of heat generation: (1) polarization, (2) entropy, and (3)  where, E represents the potential of the cell under the passage of current (I), E eq is the thermodynamic equilibrium potential, T is temperature of the isothermal bath, and dE eq dT is the equilibrium potential-temperature coefficient. Polarization heat flow arises from any processes that causes the delivered potential to deviate from the cells equilibrium potential at a given lithiation state. The polarization of an electrochemical cell encompasses ohmic, activation, and concentration polarizations [10]. Entropic heat flow arises from entropy change due to the structural disorder in the system, including sources of configurational entropy, phonon entropy, and electronic entropy [11]. The phonon entropy and electronic entropy are associated with lattice vibration and thermal disorder of the near-Fermi electron state, respectively. The configurational entropy originates from the disorder of Li atom placing vacancies, which is regarded as the dominant factor of the cell's entropy change [12]. Parasitic heat flow is attributed to parasitic reactions, which include electrolyte decomposition and SEI formation. By measuring heat flow generated upon

Bulk and interfacial operando characterization of conversion materials
The need for operando characterization is pressing as many battery chemistries with superior energy densities operate via more complex, conversion mechanisms rather than traditional intercalation processes. During a conversion reaction, electron transfer and ion transport cause the changes in the electrode material structure. These structural changes can be severe and irreversible, often resulting in significant volume expansion and particle fracture limiting capacity retention. Additionally, solid electrolyte interphase (SEI) formation and evolution on conversion electrodes is more complex. Controlled formation of SEI is critical and occurs when a non-aqueous liquid electrolyte decomposes on the solid electrode surface [13,14]. An SEI of appropriate thickness can protect the electrode from side reactions (i.e., dissolution), thereby, extending cycle life. However, for conversion electrodes, structural and volumetric changes can lead to SEI instability, where new active sites for electrolyte reduction are continuously exposed in each cycle [15,16]. Continuous decomposition of electrolyte consumes lithium ions and solvent, increases SEI layer thickness, and limits ionic and electronic transport resulting in capacity loss [17,18]. Notably, few techniques can dynamically characterize the formation and evolution of the SEI, as the layer is thin (< 100 nm), amorphous, and sensitive to air/moisture [19]. Therefore, interfacial operando characterization of conversion materials, in addition to operando bulk characterization, is of the upmost importance for developing innovative electrochemical energy storage systems with higher energy densities.

Electrolyte modification for SEI stability on conversion electrodes
The stability of the SEI on conversion anode materials can be improved by using an electrolyte additive or co-solvent. While ethylene carbonate (EC) is commonly used in lithium-ion battery electrolytes, fluoroethylene carbonate (FEC) is a well explored electrolyte additive. The addition of the fluorine group to the cyclic alkyl ester reduces the energy of the lowest unoccupied molecular orbital (LUMO), enabling FEC to reduce at higher potentials (~ 1.1-2 V) [20] relative to the non-fluorinated ethylene carbonate (EC) analog (below 1 V) [20,21]. EC reduction results in the generation of insoluble lithium carbonates and organic species (Fig. 2a), which can increase electrolyte and overall cell resistance [22,23]. The overall reduction of FEC yields a metastable FEC anion that further facilitates radical polymerization ( Fig. 2b) [24,25]. The elastomeric structure in the initial SEI layer formed through the recombination of radical polymerization [26] can adjust to accommodate the significant volume expansion that the electrode undergoes during lithiation, therefore, resulting in improved cycle life. The production of such fluorinated SEI species (LiF) facilitates Li + ion transport and reduces the charge-transfer resistance at the electrode/electrolyte interface [27]. Thus, FEC has been shown to generate a denser, homogenously-distributed, low resistance, and more elastic SEI, which improves cycle life of conversion and alloying anodes [5,28]. The electrochemical benefits of FEC and subsequent SEI modifications have been experimentally observed in lithium-ion batteries with silicon [26][27][28][29][30][31][32][33][34][35] and transition metal anodes [36,37].

Magnetite (Fe 3 O 4 ): a model insertion-conversion electrode material
Magnetite (Fe 3 O 4 ) is a model anode material for understanding insertion and conversion reactions as it undergoes both processes electrochemically. Abundant in nature, Fe 3 O 4 can theoretically deliver eight electrons per formula unit, resulting in a high theoretical specific capacity (924 mAh/g). The redox active centers, consist of Fe 3+ and Fe 2+ and occupy both tetrahedral (8a) and octahedral (16d) sites within a cubic close-packed (ccp) oxygen anion framework to form an inverse spinel structure [38]. During lithiation, Li + ions intercalate into octahedral sites, iron cations then migrate to form a rock-salt-like phase before finally converting to nanosized Fe 0 and Li 2 O, Eqs. 3-7 [39].
(1) The first Li + equivalent intercalates into an interstitial octahedral (16c) site (2) The second Li + equivalent inserts into the Fe 3 O 4 lattice, displacing the Li + located in the 16c site. The two Li + redistribute between 8a, 48f, and 8b interstitial tetrahedral sites

The overall lithiation reaction for Fe 3 O 4 is
The application of Fe 3 O 4 as an anode material is limited by poor capacity retention associated with significant volume expansion resulting from phase changes during cycling and the repetitive formation of an unstable solid electrolyte interphase (SEI) [17,18,39]. FEC has been shown to improve cycle life of Fe 3 O 4 cells by modifying the SEI layer [16,40,41]. For example, Li/Fe 3 O 4 cells with 1.33 M FEC demonstrated a 280% increase in delivered capacity relative to non-FEC cells after 100 cycles [20]. X-ray photoelectron spectroscopy (XPS) characterization of Fe 3 [16]. In addition, the Fe 3 O 4 electrodes lithiated with FEC showed a decrease in inorganic and organic carbonate species relative to EC, suggesting presence of an elastomeric, cross-linked polymer product or in-situ CO 2 formation [16]. Operando atomic force microscopy (AFM) also showed FEC electrolyte produces a denser SEI layer compared to EC electrolyte on Fe 3 O 4 thin film electrodes during cyclic voltammetry testing [41].

Operando bulk and interfacial characterization of Fe 3 O 4 : an IMC and XAS approach
Herein we apply operando IMC and XAS to characterize bulk and interfacial changes of  :7)]. The Fe 3 O 4 crystallite size was controlled via coprecipitation synthesis [42,43], and characterized via XRD and TEM. In electrodes, the size of the active material will significantly govern the area and morphology of the interface. Intuitively, electrodes packed with finer or porous-structured active materials could have a larger contact area with the electrolyte. On one hand, this increased contact facilitates the ion transport; yet on the other hand, enlarged electrolyte-electrode interaction area results in more interfacial parasitic reactions. Therefore, controlling the parameter of size is critical for our study in order to accurately quantify the different interfacial parasitic reaction mechanisms introduced solely by the electrolyte solvents. Mechanisms associated with parasitic reactions that contribute to non-reversible capacity loss were evaluated by IMC. Chemical and structural differences produced through the incorporation of a fluorinated species within the electrolyte were evaluated by analysis of the X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) regions of the operando XAS spectra. Thus when employed in concert, IMC and XAS are shown to be effective and complementary tools for understanding interfacial parasitic reactions, including SEI formation operando providing comprehensive understanding of electrolyte impact on the electrochemical Fe 3 O 4 conversion process.

Results and discussion
Materials characterization X-ray diffraction (XRD) and transmission electron microscopy were employed to characterize the as-synthesized Fe 3 O 4 nanopowder. The inverse spinel structure was confirmed via Rietveld refinement ( Fig. 3a; Table S1). A crystallite size of 11 ± 1 nm calculated from the (220) peak using the Scherrer equation was comparable to the refinement derived crystallite size of ~ 14 nm. TEM images demonstrate that the Fe 3 O 4 material consists of spherical particles (Fig. 3b). Particle size estimates from the TEM images demonstrate a minimum particle size of 3 nm, maximum of 18 nm, and an average particle size of 10 nm (Fig. 3c).

Operando isothermal microcalorimetry (IMC)
The voltage and heat flow profiles for the first lithiation of Li/  Figure S1. From the lithiation mechanism of Fe 3 O 4 (Eq. 5), changes in enthalpy (ΔH rxn ), entropy (ΔS rxn ) and Gibbs free energy (ΔG rxn ) were estimated using standard heats of formation (H f ) and standard entropies (S f ) of Fe 3 O 4(s) , Fe (s) , Li (s) , and Li 2 O (s) (Table S2) [44]. For an eight electron equivalent reduction of Fe 3 O 4 , thermodynamic considerations estimate a release of ~ 5300 J/g of heat. The energy released from the Li/Fe 3 O 4 cells (ΔE) measured in Fig. 4a, b was calculated by summing the electrical work (W) and the total heat measured by the microcalorimeter (Q). Electrical work was determined by integrating the cell potential with respect to specific capacity, whereas the total heat was determined by integrating the heat flow with respect to time (Table S2). For both the EC and FEC-containing Li/Fe 3 O 4 cells, the measured energy change (ΔE) exceeds the theoretical value of thermodynamic Gibbs energy change (ΔG rxn ) (Eq. 5; Table S2), indicating electrochemical kinetics under a loaded current also contributes to the heat output. Even at a moderate rate of discharge, electrochemical lithiation does not coincide with a thermodynamic equilibrium state, especially in conversion-type electrode materials whose kinetic resistance can be significant. As a result, the  Analysis of differential capacity (dQ/dV) plots (Fig. 4c, d) in relation to the heat flow profiles (Fig. 4e, f)  Notably, the initial insertion reactions and beginning of the FeO conversion reaction occur at higher potentials and deliver greater capacities when FEC electrolyte is present. Additionally, the magnitude of the heat flow at these transition points are larger in the FEC than the EC cell. For example, the maximum heat flow for the FEC cell is -56 mW/g, compared to -40 mW/g for the EC cell. The differences in heat flow between EC and FEC cells early in the first lithiation suggest differences in electrolyte reactivity [20].
Moreover, the difference in heat generation between EC and FEC cells arises predominately at the early lithiation state (0-400 mAh/g), shown by Table S3. The total heat is calculated by integrating the heat flow curve measured by the IMC. Before 400 mAh/g, the EC cell releases -312.3 mWh/g of heat, which is ~ 25% less than the FEC cell (-413.6 mWh/g). In contrast, the total heat dissipated during the first lithiation to 0.03 V and after 400 mAh/g of lithiation is comparable between the FEC and EC cells. Similar trends are observed when total heat is normalized to capacity (normalized heat). Early in lithiation, the FEC cell exhibits greater normalized heat, whereas throughout and at the end of the first lithiation, normalized heats among EC and FEC cells are comparable (Table S3).
Moreover, both the EC and FEC cells experience the largest normalized heat during cycle 1 lithiation relative to cycle 1 delithiation and cycle 2 ( Figure S2, Figure S3, Table S4). During cycle 1 delithiation, the EC and FEC cells deliver 575 and 473 mAh/g, corresponding to a total heat generation of 230.3 and 176.2 mWh/g, respectively. In cycle 2, the EC cell produces total heats of 365.1 and 156.1 mWh/g, reaching lithiation and delithiation capacities of 691 mAh/g and 341 mAh/g, respectively. In cycle 2, the FEC cell generates slightly less heat during lithiation (225.4 mWh/g) and delithiation (103.8 mWh/g), consistent with lower capacities of 492 mAh/g and 243 mAh/g for lithiation and delithiation, respectively. The consistency between capacity and heat generation during cycle 1 delithiation and cycle 2 (de) lithiation further indicates that the significant differences among the EC and FEC cells occur during the first lithiation before 400 mAh/g of lithiation.

Polarization and entropy
Polarization and entropic heat flow contributions were determined using Galvanostatic intermittent titration technique (GITT) and potential-temperature measurements (Fig. 5, Figure S4). The specific capacity of the GITT tests were normalized to the IMC cell by equalizing capacities at the lithiation voltage limit of 0.03 V ( Figure S5). In both EC (Fig. 5a) and FEC (Fig. 5b) cells, the polarization heat flow is the major heat flow source compared to the entropic contribution. The polarization and entropic heat flow contributions are most distinct between the EC and FEC cells before two molar electron equivalents of lithiation (< 2 ee), consistent with total heat flow measurements (Fig. 3) and total heat calculations (Table S3). From 0 to 2 ee, the FEC cell exhibits larger polarization heat flow ~ 40 mW/g, whereas the EC cell shows polarization heat flow range between ~ 15 mW/g to ~ 30 mW/g. The higher initial polarization in FEC cells before 2 ee is surprising, as FEC is proposed to reduce resistance at the electrode interface through stable SEI formation [27]. The IMC data indicate that in this lithiation regime, a stable SEI has not yet formed on the electrode surface.   (Fig. 6). The difference between the polarization and entropic heat flow and total heat flow curves represents parasitic heat flow. Before 2 ee, the sum of entropic and polarization heat flow contributions matches the total heat flow (Fig. 6a) for the EC cell consistent with previous reports, suggesting that interfacial parasitic reactions are not a significant factor as polarization and entropic heat flow fully capture and describe all heat generated. The two curves begin to deviate after 2 ee of lithiation (~ 0.9 V), indicating the onset of parasitic reactions. The separation of two curves increases upon the lithiation after 2 ee, suggesting continuous reduction of electrolyte co-solvents and the formation of the SEI. Parasitic heat flow in the EC cell becomes a major contributor (~ 80%) to the observable heat flow at the end of lithiation.
For the FEC cell (Fig. 6b), total heat flow is ~ 20 mW/g greater than the polarization and entropic heat flow contributions before 2 ee of lithiation, indicating earlier onset of the parasitic reactions relative to EC, which is consistent with reports indicating FEC reduction and SEI formation occurs at higher potential [5,33]. After 2 ee, total heat flow remains higher than the polarization and entropic heat flow curve indicating parasitic reactions continue to evolve in the presence of FEC during lithiation.
Heat generated by polarization and entropy was calculated by integrating the polarization and entropic heat flow curve over time (Table S5). The polarization and entropic heat were subtracted from total heat calculated in Table S3 to quantify the parasitic heat contribution. The greatest difference in parasitic heat generation occurs before 2 ee of lithiation, where parasitic heat contributes to 3% and 10% of total heat flow for the EC and FEC cells, respectively. Moreover, before 400 mAh/g of lithiation, 17% of total heat generated by the FEC cell is associated with parasitic contribution, in contrast to 12% for the EC cell. The greater parasitic heat generation observed for the FEC cell early in the lithiation is consistent with greater interfacial reactively. Taking into account the entirety of the first lithiation, parasitic heat accounts for 40% and 36% of total heat generated by the EC and FEC cells, respectively suggesting parasitic heat accounts for more of the total heat in the EC condition at the end of lithiation. The dominance of parasitic heat generation at the end of lithiation for the EC cell is confirmed by quantifying parasitic heat > 400 mAh/g, where 51% of total heat is attributed to parasitic reactions relative to 46% for the EC cell.

Operando X-ray absorption spectroscopy (XAS)
Operando XAS was collected during lithiation of in-situ Li/ Fe 3 O 4 cells at 115 mA/g with either EC or FEC electrolyte to discriminate bulk changes in the Fe oxidation state reduction and local atomic structure from the X-ray absorption near edge fine structure (XANES) and extended X-ray absorption fine structure (EXAFS), respectively. During the first 2 ee lithiation, the Fe K-edge shifts from 7126.5 eV to 7124.0 eV for both the EC (Fig. 7a) and FEC (Fig. 7d) cells, indicating a reduction of Fe 3+ to Fe 2+ . Additionally, a notable change in the pre-edge feature is observed. Fe 3 O 4 has a characteristic pre-edge feature associated with Fe 3+ tetrahedral atoms. Upon reduction, the pre-edge feature decreases in intensity and the peak broadens, signifying a migration of Fe from tetrahedral to octahedral sites ( Figure S6). These changes in edge energy and pre-edge feature are observed from 0 to 2 ee for both the EC (Fig. 7b) and FEC (Fig. 7e) cells. From 2.1 to 4.0 ee, the pre-edge feature broadens and edge energy remains constant for both the EC (Fig. 7c) and FEC (Fig. 7e) cells.
The edge energy (Fig. 7a) and centroid position of the preedge (Fig. 7b) were tracked to evaluate differences in Fe oxidation state during Fe 3 O 4 lithiation. A decrease in edge energy and pre-edge centroid position can signify a decrease in Fe oxidation state [45,46]. At each electron equivalent, small differences in the edge energy and centroid position are observed (< 0.4 eV), suggesting bulk reduction of Fe progresses as a similar rate in both electrolytes, suggesting differences in the voltage profile (Fig. 7c) between the two cells can be attributed to surface phenomena rather than bulk Fe oxidation state changes.
The progression of the EXAFS spectra during the lithiation also suggest bulk changes in the local atomic environment of Fe are similar in the presence of EC (Fig. 8a, b) and FEC (Fig. 8c, d) electrolyte. The EXAFS spectra of Fe 3 O 4 have two characteristic peaks attributed to Fe-O and Fe-Fe interactions. From 0 to 2 ee, the Fe-O peak at 1.4 Å shifts to higher interatomic distances at a similar rate for both the EC and FEC cells ( Figure S7). Additionally, the second shell peak near 2.5 Å narrows due to a loss in intensity of the shoulder contribution at ~ 3.1 Å. These observations are consistent with the conversion of Fe 3 O 4 inverse spinel structure to a FeO-like phase and are supported by EXAFS modeling (Tables S6-S9). From 0 to 1.6 ee, the EXAFS region for both EC and FEC cells fits the Fe 3 O 4 inverse spinel structure with good agreement (R factor < 2%). At 1.9 ee, the EXAFS spectra for both FEC and EC cells are best fit with an FeO rock-salt like structure. From 2.1 to 4.0 ee, the EXAFS fits the FeO-like structure and Fe metal is not detected in either cell. Thus, the progression of the EXAFS spectra during lithiation suggests that bulk changes in the Fe atomic environment are not significantly impacted by electrolyte modifications.

Conclusions
Operando bulk and interfacial characterization provide realtime information on the dynamic processes inside energy storage systems. The acquired information can lead to further electrochemical insights that serve as a catalyst for innovation in energy storage. This article highlights two complimentary operando approaches: isothermal microcalorimetry (IMC) and X-ray absorption spectroscopy (XAS). Isothermal microcalorimetry, which measures total heat flow dissipated from a battery during operation, can be analyzed to track onset and evolution of parasitic reactions. These parasitic reactions include interfacial phenomena, such as solid electrolyte interphase (SEI) formation, therefore, making it one of the few techniques that can probe interfaces operando. X-ray absorption spectroscopy (XAS) is a complimentary bulk operando technique, which can follow reduction and oxidation of a redox active species and changes in local atomic environment of the designated element through the X-ray absorption near edge fine structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectra.
Herein we apply operando IMC and XAS to understand bulk and interfacial changes of  LiPF6 FEC/DMC (3:7)] electrolyte. Similarities in the differential capacity (Fig. 4), XANES (Fig. 7) and EXAFS (Fig. 9) profiles of FEC and EC cells suggest the bulk Fe 3 O 4 lithiation mechanism, Fe oxidation state, and local atomic environment of Fe changes are not impacted by changes in electrolyte. In contrast, dramatic differences in the IMC heat flow profiles before 400 mAh/g of lithiation ( Fig. 4; Table S3) are observed between EC and FEC cells. Analysis of the IMC heat generation and heat flow into polarization, entropic, and parasitic components (Figs. 5, 6; Table S5) indicates the FEC cell generates 7% (< 2 ee) and 5% (< 400 mAh/g) more parasitic heat than the EC cell, consistent with electrolyte reduction at the electrode interface at higher potential. Thus, the IMC and XAS data presented suggests that electrolyte modification influences interfacial phenomena rather than bulk lithiation properties.
This article demonstrates how operando IMC and XAS in tandem can elucidate bulk and interfacial information, leading to a more holistic understanding of the electrode-electrolyte interactions within a conversion electrode. Equipped with knowledge of charge transport phenomena at multiple time and length scales within an EES system, we can expedite solutions for current challenges in energy storage.   Fe 3 O 4 was synthesized via co-precipitation consistent with previous reports, where FeCl 2 ·4H 2 O (aq) and FeCl 3 ·6H 2 O (aq) were added to deoxygenated ammonium hydroxide (NH 4 OH) and H 2 O [47]. The product was washed with H 2 O until neutral and dried under vacuum at room temperature. X-ray powder diffraction was conducted with a Rigaku Miniflex X-ray diffractometer with a Cu Kα source (λ = 1.5406 Å). Crystallite size of Fe 3 O 4 was calculated using the Scherrer equation based on the (220) reflection and corrected for instrumental broadening using a LaB 6 standard. The (220) peak was fit with a Pearson VII Area function using the PeakFit software v4.12. Lattice parameters were acquired by refining the powder diffraction pattern using an inverse spinel Fe 3 O 4 starting structure [38] and the GSAS-II software package [48]. Transmission electron microscopy (TEM) of the as-synthesized Fe 3 O 4 powder was collected using a JEOL JEM-1400 instrument equipped with a field-emission electron gun that operated at 120 kV. Particle size distribution determination used ImageJ image software where 133 particles were chosen for the analysis [49].

Electrochemical measurement
A slurry of 90 wt.% Fe 3 O 4 and 10 wt.% PVDF in NMP was cast onto copper foil. Stainless-steel coin cells were assembled using the

Isothermal microcalorimetry
A TA instrument TAM III microcalorimeter was used for IMC measurement and connected to a BioLogic VSP potentiostat via low thermal conductivity wires. The temperature of the IMC oil bath was 30 °C for all measurements. The cells were lithiated at 40 mA/g to 0.3 V.

Polarization heat contribution determination
Galvanostatic Intermittent Titration Technique (GITT) was employed to estimate the equilibrium potentials of EC-or FEC -containing Li/Fe 3 O 4 systems at different lithiation states. Li/ Fe 3 O 4 cells with EC or FEC electrolytes were discharged to different states of lithiation and then allowed to rest at each state. The experimental GITT curves were further normalized based on lithiation contents of cells cycled in IMC. The open-circuit potential recovery profile was numerically fit to a logarithmic function to extrapolate the open-circuit potential at 500 h. The relaxation potential at 500 h was used to calculate polarization heat flow at each lithiation increment, as Fe 3 O 4 electrodes in lithium cells have been shown to reach equilibrium after 500 h of rest [50]. The polarization heat flow was calculated as the difference between the extrapolated equilibrium potential and loaded voltage multiplied by the discharge current.

Entropic heat flow contribution determination
Methods for elucidation of entropic heat flow were adapted from those described previously [5]. At

Operando X-ray absorption spectroscopy
Two in-situ cells were fabricated with a Fe 3 O 4 /PVDF (90/10) cathode, polyethylene separator, Li metal anode, and either 1 M LiPF 6 in EC/DMC or 1 M LiPF 6 in FEC/DMC electrolyte. In-situ cells were lithiated at 115 mA/g to 4 molar electron equivalents. Operando X-ray absorption spectroscopy (XAS) measurements at the Fe K-edge (7112 eV) were collected in transmission mode and recorded during lithiation of the EC and FEC-containing Li/Fe 3 O 4 pouch cells at beamline 7-ID QAS in the National Synchrotron Light Source II within Brookhaven National Laboratory. An iron metal reference foil was collected simultaneously with each measurement and was used for calibration during analysis.
XAS spectra were aligned, normalized, merged, and background subtracted (R bkg = 1.0) using the Athena software [51,52]. The edge energy was determined using maximum of the first derivative of μ(E) spectrum. The pre-edge peak feature from the μ(E) spectrum was background subtracted using a spline function then fit using two or three Voigt functions analogous to previously reported methods [45,46]. Normalized EXAFS spectra were fit using the Artemis program with theoretical structural models (inverse spinel Fe 3 O 4 [38], rock-salt like FeO, and body-centered cubic Fe metal [53]) created with FEFF6 [52,54]. The rock-salt like FeO structure was generated from a non-lithiated Fe 3 O 4 spinel structure modified so that all Fe centers occupy octahedral 16c sites. The Fourier transform of χ(k) (|χ(R)|) used a k-range of 2-11 Å −1 and Hanning window (dk = 2). EXAFS fitting was conducted using k, k 2 , and k 3 weights simultaneously and an R-range of 1-3.5 Å to fully encompass the first and second shells of |χ(R)|.

Acknowledgments
Materials synthesis, characterization, electrochemistry, and isothermal microcalorimetry were supported by the Center for Mesoscale Transport Properties, an Energy Frontier Research Center supported by the DOE-BES, under Award #DE-SC0012673. The XAS measurements used the Quick X-ray Absorption and Scattering (QAS) 7-BM beamline at the National Synchrotron Light Source II, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Brookhaven National Laboratory under Contract DE-SC0012704. G.D.R acknowledges the support of the Turner Fellowship Program. E.S.T. acknowledges funding from the William and Jane Knapp Chair for Energy and the Environment.

Data availability
Relevant data for analysis, including the detailed results and approaches for IMC and XAS analysis, are provided as part of supplementary information.

Conflict of interest
The authors declare no conflict of interest.

Open Access
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.