Synthesis of high-density olivine LiFePO₄ from paleozoic siderite FeCO₃ and its electrochemical performance in lithium batteries

Lithium-ion batteries are an essential energy storage technology that has undergone widespread market introduction for many applications. Of note is the movement of the technology toward satisfying the need for low-cost grid storage. Toward this goal, efforts are under way worldwide to deliver cathode materials for lithium-ion batteries with higher capacity, excellent stability, and high-abuse tolerance, all in an affordable package. The affordability and safety of the battery materials are the important factors critical for the success of the technology. To date, a few MWh installations have been made operational with varied chemistries, such as lithium transition metal oxides^1,2 and LiFePO₄ olivine.³ Certainly, olivine LiFePO₄ (LFP), in particular, is one such battery chemistry capable of long cycle life that could find increased usage.

Olivine LFP was first shown by Padhi et al.⁴ to be electrochemically active for lithium-ion batteries. It was later nano-sized processed, doped, carbon-coated, and developed into cathode materials with high-rates, good stability, and excellent safety.^5,6 The LFP provides a 50% state-of-charge voltage of ∼3.4 V and a theoretical capacity of 170 mAh g⁻¹_LFP. Many synthesis routes for LFP have been successfully demonstrated, including solid-state reaction, sol-gel, hydrothermal, co-precipitation, microwave synthesis, polyol and solvothermal, micro-emulsion, spray technique, template method, and mechanical activation (see Ref. 7 for a review of these synthesis approaches). For solid-state methods, in particular, the choice of precursors has a significant impact on the properties of LFP. For example, LFP synthesized from hydrated iron(II) oxalates (FeC₂O₄) and a lithium salt (Li₂CO₃) reacted together with ammonium hydrogen phosphate often shows the presence of a γ-Fe₂O₃ impurity.^8,9 Samples free of impurity can be formed from stoichiometric amounts of FePO₄(H₂O)₂ and Li₂CO₃.⁷ Iron (III) precursors can be used in carbothermal reduction reactions as well.¹⁰ LFP prepared by the hydrothermal reaction gives control over the chemical composition and crystallite size, and the electrochemical performance is among the best reported.^11–14 In particular, Chen and Dahn,¹⁵ Yamada et al.,¹⁶ and Julien et al.¹⁷ optimized the synthesis and properties of LFP to yield improved electrochemical performance. Later work by Dominko et al.¹⁸ on optimizing the carbon coating also significantly improved the rate capability.

In this study, a new synthesis route to LiFePO₄ from the mineral siderite FeCO₃ is introduced for the first time. FeCO₃ provides an elegant, simple two-component reaction and can lead to a LFP material with greater tap densities than what is commercially available. Note that FeCO₃ is inherently difficult to be directly synthesized in the lab with co-precipitation methods due to the facile oxidation of the carbonate to iron oxides. In this work, we provide not only a perspective on siderite materials and their origin but also discuss the notion of their use in battery applications. The synthesis and characterization of LFP derived from siderite are first presented; then, the electrochemical activity of the LFP made by this route is verified by galvanostatic cycling.

Here, we take an interdisciplinary approach, drawing from fields of geology, paleontology, chemistry, and battery science, to convert mineral siderite into a functional battery material. Siderite is a naturally occurring mineral composed of FeCO₃ and is commonly found not only in bedded sedimentary formations and hydrothermal metallic veins but also in concretions. Siderite beds can be exceedingly old, dating to formation 1.8 billion years ago when the level of carbon dioxide in the Earth atmosphere was 100 times greater than today.¹⁹ A unique form of siderite concretions from the upper Carboniferous period of the Paleozoic Era can contain plant and animal fossils. Such a source exists in old coal mines of northern Illinois, USA. These fossil concretions are commonly known as Mazon Creek fossils, which were formed 307 million years ago.²⁰ Examples of these splendid fossils and discussion of their formation can be found in Refs. 21 and 22. These concretions can be collected from the Earth’s surface at the Mazonia-Braidwood State Fish & Wildlife Area in northern Illinois, USA (with fossil collecting day permit; not for commercial purposes).

Isolating the interior of the rock and exposing the most pure FeCO₃ are critical. Thus, several freeze–thaw cycles in water cause the rock to break open along its fault, exposing the fossil on a plane of a largely pure FeCO₃ matrix. Over time and with exposure to air, these siderite concretions can decompose into iron oxides; therefore, in this work, only freshly opened concretions were used with a focus on the pure center core. In Fig. 1, the photographs show the siderite FeCO₃ rock specimens used in subsequent LFP synthesis. To explore the homogeneity, specimens were collected from various points from the exterior (S1–S2) and interior (S3–S6) of the rock. These were ground to a fine powder, which is pictured in Fig. 1(d). Specimens S1 and S2 collected from the exterior of the siderite are orange-ish in color, indicative of an oxidized form, such as iron oxide, as the outside of the rock is more exposed to weathering. Siderite specimens from the interior are originally dark in color, but upon grinding, they become more light gray. The thermal properties of the siderite were determined by thermogravimetric analysis (TGA), as shown in Fig. 2. A single endothermic process is observed beginning at ∼420°C with a corresponding 32% weight loss. This is consistent with the literature for siderite decomposition to magnetite (Fe₃O₄).²³ Carbon dioxide is lost from the material, and slight oxidation by oxygen in the air flow occurs [see Eq. (1)]. According to the following equation, the theoretical weight loss is 33%, which is quite close to the experimental 32% observed in Fig. 2(a):

It is extremely likely that the siderite samples contain natural impurities. For distinguishing the kind of elemental impurities, energy-dispersive x-ray spectroscopy (EDX) measurements were carried out on S1–S5 [representative spectrum shown in Fig. 3(a)]. These results suggest that the predominant impurities are Mg, Ca, Si, Al, K, and Mn; the weight percentages are indicated in Fig. 3(b). Si is the largest contaminant at 3.2 wt. %. From these predominant impurities, we calculate that FeCO₃ is about 92 wt. % pure. Additionally, inductively coupled plasma mass spectrometry (ICP-MS) trace chemical analysis was conducted on specimen S5. The results are presented in Table I (major elemental impurities are Mg, Ca, Al, and Mn) and confirm the majority of impurities found in the EDX analysis.

The specimen physical appearance and impurity analysis are consistent with XRD patterns collected on these specimens [Fig. 4(a)]. All specimens from the interior of the rock show reflections consistent with FeCO₃. As expected, reflections from iron oxide are detected in S1 and S2 (marked with *), although samples from the interior of the rock are free of oxidation. A crystalline impurity phase is detected across all samples (reflection at 26.7° 2θ), although its phase fraction appears to diminish at the interior of the rock. We tentatively assign this feature to the intense 101 reflection from quartz silica, SiO₂. There is some variation in the position of the FeCO₃ reflections between the database pattern²⁴ and our specimens and across samples S2–S6, indicative of slight differences in the unit cell. Consequently, the lattice parameters for S5 were determined by Le Bail analysis to be a = 4.7058(2) Å and c = 15.471(1) Å (details in Fig. S1). These values are larger by 0.0142(6) and 0.0914(3) Å, respectively, relative to the reference Inorganic Crystal Structure Database (ICSD) pattern values [a = 4.6916(4) Å and c = 15.3796(16) Å],²⁴ representing an overall unit cell volume expansion of 1.3%. This discrepancy may indicate some substitution chemistry of other divalent atoms, such as Mn, Mg, or Ca (see impurities listed in Fig. 3 and Table I). For example, replacing Fe²⁺ with some Ca²⁺ would expand the unit cell due to the larger ionic radii of Ca compared to Fe. However, divalent substitution in Fe_1−xM_xCO₃ (M = Mn, Ca, or Mg) could alter the chemistry of the phase and ultimately the LFP electrochemical activity. Other contaminants such as Al³⁺ and Si⁴⁺ are likely tied up as alumino-silicates (or clays) that are persistent in the geology of these concretions and would render them electrochemically inactive but also present as “dead” weight.

To understand the reaction taking place during synthesis between siderite FeCO₃ and lithium hydrogen phosphate (LiH₂PO₄), a TGA experiment under flowing argon was performed [Fig. 2(b) and differential plot in Fig. 2(d)]. From the differential plot, two main weight loss processes are observed. The first is an endothermic process beginning around 200°C with an associated ∼8% weight loss. This is consistent with the simultaneous LiH₂PO₄ melting and loss of water to form lithium hydrogen pyrophosphate (Li₂H₂P₂O₇). After 240°C, mass is lost at a steady −0.0396(4)% °C⁻¹ rate until 390°C, after which mass loss accelerates to −0.107(1)% °C⁻¹ (fitting shown in Fig. S2). These processes can be attributed to further loss of water from Li₂H₂P₂O₇ to form lithium phosphite (LiPO₃) and/or formation of LiFePO₄ (evolving carbon dioxide and water). Beyond 550°C, the reaction is apparently complete and no further mass loss occurs. The overall mass loss is 24%, which is consistent with the proposed reaction in Eq. (2) (theoretical mass loss 28%) when ∼8% impurities in FeCO₃ are taken into account,

The section titled “Experimental” discusses in detail how the LFP in this work is prepared from siderite. To note, the Li:Fe atomic ratio changes the color of the powder post synthesis, as shown in Fig. S3. Different shades of gray are evident, with the darker powders formed when the Li/Fe fraction is <1 and lighter for Li/Fe > 1. The XRD pattern of LFP synthesized from siderite with Li/Fe = 0.9 is given in Fig. 4(b). This shows that LFP can be synthesized at 700°C in argon from the reaction of natural siderite FeCO₃ with LiH₂PO₄. Two impurity reflections are noted in the diffraction pattern at 26.7° and 31.3° 2θ. The first corresponds to quartz silica, present in the FeCO₃ precursor [see Fig. 4(a)], which is thermally stable under the conditions employed during synthesis. The 31.3° 2θ reflection may be from unreacted LiH₂PO₄. Lattice parameters for the synthesized LFP were determined to be a = 10.2910(4) Å, b = 5.9910(3) Å, and c = 4.6879(2) Å. These compare favorably with that of LFP in the literature²⁵ and with the commercial LFP in Fig. 4(b) [a = 10.2942(4) Å, b = 5.9899(3) Å, and c = 4.6769(2) Å]. Note that when the Li/Fe fraction was 1 or 1.1, longer heating times were required to remove additional lithium hydrogen iron pyrophosphate and lithium iron monodiphosphate [Li₉Fe₃(P₂O₇)₃(PO₄)₂]²⁶ impurities.

Before testing the electrochemical properties of the LFP, the powder was carbon coated (details are in the section titled “Experimental”). Scanning electron microscope (SEM) images in Fig. 5 show the morphology after coating. Figure 5(a) shows the EDX spectroscopy for P, O, and Fe K edges of the synthesized LFP, from which a homogeneous distribution of elements can be observed. Investigation of the material via SEM in Figs. 5(b) and 5(c) shows secondary particles (roughly 30–60 µm in size) composed of smaller primarily particles of 2–10 µm in size. The primary particles show inhomogeneous geometries but are roughly similar in size.

Galvanostatic cycling was employed to test the electrochemistry of the carbon-coated siderite-derived LFP. Potential profiles collected at various rates at 30°C are shown in Fig. 6 (more details are given in Table S1). First, the 3.5 V plateau on charge and 3.43 V plateau on the discharge characteristic of LFP are observed [see the differential capacity plot in Fig. 6(b)], confirming typical lithium (de)intercalation into the siderite-derived LFP.^4,5 Typically, high performance LFP, as shown by Julien et al., shows a class leading specific capacity of 160 mAh g⁻¹_LFP at C/4 in the voltage range 2.2–4.0 V vs Li⁰/Li⁺, which is stable over 120 cycles.^7,17 In our case, however, at a relatively slow C/10 rate, the delithiation capacity on the second cycle is only 70 mAh g⁻¹_LFP. After 100 cycles at C/10, 17% capacity fade is observed and the material (de)lithiates with 99.6% efficiency. The capacity is only marginally improved by using very slow rates and increasing the cycling temperature, with a C/100 cycle at 30°C, giving rise to 95 and 110 mAh g⁻¹_LFP using C/100 at 55°C.

The analysis of volumetric capacity allows one to compare the packing in a battery pouch cell, for example. It is measured as mAh/cm³; thus, it takes the powder density into account. For practical volumetric capacity, we do not use the theoretical crystal density, but, instead, the tap density. As tap density is measured as g/cm³, then simply one multiplies the tap density together with the gravimetric density in mAh/g. For commercial LFP with 0.76 g/ml tap density, this value is 114 mAh/cm³_LFP, and for siderite derived LFP, it is 157 mAh/cm³_LFP or an improvement of 1.38×. Note this gravimetric capacity is obtained at C/100.

To understand the relatively low capacity of siderite-derived LFP, it is worth considering the historic development of LFP as a battery cathode material. Initially, the low intrinsic electronic conductivity of the material limited the electrochemical performance. This was overcome by coating the particles with conductive carbon, essentially improving the electronic contact between particles.^6,15,27,28 Carbon coating has also been shown to have two additional benefits.^7,16 The first is the conversion of Fe³⁺, found in the surface layers of uncoated particles, to Fe²⁺. Second, the disorder found in the surface layer is reduced, which significantly improves the electrochemical performance. In this work, the siderite-derived LFP is carbon coated; however, this process is not optimized. Another major hurdle in the development of LFP was the low rate capability. This was primarily addressed by the reduction of the particle size to the nano-scale.^29–31 Owing the low electronic and ionic conductivity of LFP,^32,33 reducing the particle size has a significant impact on the performance, particularly at high C-rates. SEM images in Figs. 5(b) and 5(c) show that the particles used in this work are, by comparison, large (primary particles 2–10 µm). Optimizing the processing of siderite FeCO₃ through nano-size processing (e.g., high-energy ball milling) and optimizing the synthesis conditions (e.g., temperature and time) will likely improve the performance of siderite-derived LFP. Finally, it is well documented that impurity phases severely and adversely affect the electrochemical performance of LFP.^8,34 Various impurity phases have been identified in phospho-olivine materials,¹⁷ for example, γ-Fe₂O₃, Fe₃O₄, Li₃PO₄, Li₃Fe₂(PO₄)₃, Fe₂P₂O₇, LiFeP₂O₇, Fe₂P, and Fe₃P. Synthesis from unrefined mineral siderite concretions introduces additional impurities, as noted above in the EDX (Fig. 3), ICP (Table I), and XRD [Fig. 4(a)]. The 1D lithium channels in the olivine structure^35,36 make LFP performance highly sensitive to impurities and structural faults that block the channels. Furthermore, when structural disorder is introduced in the form of impurity phases, the coherence of the lattice is lost. It is well understood that the (de)lithiation of LFP is a cooperative process, operating via an interfacial reaction front between LiFePO₄ and FePO₄, i.e., a two-phase reaction.^37–40 Structural defects would interrupt the displacement of the reaction front, pinning the reaction at defect sites/phase boundaries.⁷ Impurities native to mineral siderite and present in siderite-derived LFP synthesized in this work are therefore playing a significant role in limiting the electrochemical performance.

Consideration of the redox activity and reported electrochemical performance of other battery materials sourced from natural compounds can also provide some perspective for the performance of yet unoptimized siderite-derived LFP. For example, Ravet and co-workers⁶ reported the electrochemical activity of natural triphylite ore sourced from mines in New Hampshire, Maine, and South Dakota. This relatively abundant mineral is on the Fe-rich side of the LiMn_1−xFe_xPO₄ solid solution. After ball milling and carbon coating, a discharge capacity of 85 mAh g⁻¹ was obtained at a low cycling rate (∼C/40) and at 80°C. This compares well with this first report of siderite-derived LFP cycled at C/50 and at 30°C, which yielded a discharge capacity of 92 mAh g⁻¹_LFP.

Another analogous example is the development of mined natural graphite for lithium-ion battery anodes. Sanyo was the first to introduce purified natural graphite as an anode to commercial lithium-ion batteries in patents filed in 1993 and 1996.^41,42 Graphite as an intercalation compound for lithium, however, had been demonstrated almost 40 years prior in the mid-1950s by Hérold⁴³ and by electrochemical intercalation in 1983 by Yazami and Touzain.⁴⁴ The literature from the late 1990s and early 2000s clearly highlights that the raw material processing, and in particular the purity, surface moisture, particle size, and shape, is highly influential on the electrochemical performance of natural graphite.^45–52 For example, Manev et al.⁴⁵ and Kenji et al.⁴⁷ reported that untreated natural graphite sourced from Brazil and Madagascar, respectively, yielded a first cycle reversible capacity of less than 210 mAh g⁻¹, <55% of the theoretical 372 mAh g⁻¹ value. However, with material purification, particle modification, and/or layer foliation, capacities much closer to the theoretical value were obtained. Zaghib et al.⁵¹ demonstrated a clear relationship between the natural graphite degree of purity and the reversible capacity. Materials with 99.94% purity gave a reversible capacity of ∼355 mAh g⁻¹, ∼15 mAh g⁻¹ lower than those with 99.99% purity. The performance of the unrefined material, with a purity of 98.5%, was not given. From the above discussion, it is clear that increasing the purity and optimizing the physicochemical properties of natural graphite have resulted in near theoretical electrochemical utilization.

A similar journey toward full electrochemical utilization is probable for siderite-derived LFP. The siderite specimens used in this work were merely picked out of convenience to take advantage of a local source of siderite, which also happens to be world famous (Mazon Creek Fossils). There are much larger deposits of siderite throughout the world that could be considered. A recent article⁵³ discussed using the major reserves of siderite ore in China and Austria to produce iron metal in methods that would have significantly lower emissions of carbon dioxide per ton of metal as compared to iron production from traditional iron oxides. The scale of readily available siderite ores is massive. Two million tons of iron ores containing high (currently unusable) amounts of siderite are stockpiled at mines near Anshan, Liaoning Province, China.⁵⁴ It is possible that these sources of siderite may have higher purity, which may lessen the degree of purification required.

As a final note, the volumetric energy density of battery electrode materials has become increasingly important, particularly in space-limited applications, such as electric vehicles. Consequently, along with the specific capacity, the material tap density is a critical metric to assess the energy density of a battery electrode material. The tap density of a commercial LFP [XRD pattern shown in Fig. 4(b)] was measured to be 0.76 g cm⁻³. In comparison, the siderite-derived LFP was found to be twice as dense, with a final tap density of 1.65 g cm⁻³. Within a lithium-ion battery, using an electrode material with a higher tap density results in higher volumetric capacity, thinner electrodes, and shorter electron pathways for the same mass loading. If the gravimetric capacity of LFP synthesized from siderite deposits can be improved, the advantages of using siderite FeCO₃ as a starting material for LFP could open up a promising opportunity utilizing high tap density LFP for high energy density lithium-ion batteries.

We have demonstrated the synthesis of electrochemically active LiFePO₄ derived from a natural mineral, siderite or FeCO₃. FeCO₃, which is over 300 million years old, was collected from the ground and directly used in a chemical reaction with the LiH₂PO₄ salt to make the LiFePO₄ product. To our knowledge, FeCO₃ itself, whether synthesized in the lab or not, has never been previously used to synthesize LiFePO₄. Following the reaction, the material was fully characterized using XRD, SEM/EDX, and battery cycling testing in lithium coin cells.

In a lithium battery, siderite-derived LiFePO₄ yields about 70 mAh g⁻¹_LFP in initial cycles at C/10 and fades to about 58 mAh g⁻¹_LFP after 100 cycles. We believe that impurities in the sample detract from the electrochemical performance of the material, possibly by occupying key lattice positions, blocking the 1D motion of lithium ions, and interrupting the progress of the two-phase reaction front. Interestingly, the voltage profile indicates the likeliness that the impurities are not electrochemically active, suggesting, in this regard, that only dead weight is added to the system. Nevertheless, the tap density is higher than commercial LiFePO₄, which would equate to a high volumetric energy density after material optimization. Toward improving the capacity of siderite-derived LFP, we propose that effort to decrease the particle size, optimize the carbon coating, and purify the precursors and/or product will yield the most promising results.

A few siderite (FeCO₃) concretions were collected from old coal mines located in northeastern Illinois, USA, which is described in the section titled “Results and Discussion.” Removal of the outer layer of iron oxide around the concretions via tool filing allowed for isolation of gray-colored FeCO₃ from the interior of the concretion. Aside from a light water rinse and drying, no other purification of FeCO₃ was conducted.

Iron carbonate (FeCO₃) was ground in a high-energy ball mill (stainless steel) and then passed through a 25 µm stainless-steel sieve in order to ensure a small particle size for synthesis. Under an argon atmosphere, specific quantities of ground FeCO₃ and LiH₂PO₄ (Sigma-Aldrich, 99%) were mixed by high energy ball milling. Although various Li:Fe ratios were tested (0.9:1, 0.95:1, 1:1, 1.1:1), this article deals with LiFePO₄ (LFP) formed with a 1:1 or 1.1:1 ratio, unless otherwise stated. When the mixture was deemed homogeneous, it was transferred to an alumina crucible and heated at a rate of 10°C per minute under argon to 700°C and held at this temperature for 16 h. Carbon coating of the synthesized LFP was conducted by adding the LFP to a solution of methanol (Sigma-Aldrich, 99+% anhydrous) and fructose (8 wt. % relative to LFP, Sigma-Aldrich, >99.99%) and heating at 55°C until all methanol was evaporated and a dry brown powder remained. The powder was then placed in an alumina crucible and heated at a rate of 10°C per minute under argon to 450°C and held at this temperature for 12 h. After cooling, the carbon coated LFP was passed through a 25 µm stainless steel sieve and stored under nitrogen.

Scanning electron microscopy (SEM) and energy-dispersive x-ray spectroscopy (EDX) were performed using a Hitachi S-4700-II microscope in the Electron Microscopy Center of Argonne National Laboratory. To reduce charging on some samples, a thin layer (<10 nm) of gold/palladium (Au/Pd) was sputter-coated onto the electrode material prior to analysis.

X-ray diffraction (XRD) patterns were collected on a Bruker D8 Advance powder diffractometer with Cu Kα radiation between 10° and 80° 2θ with a 0.02° step size and 1.0 s step time.

Thermal gravimetric analysis (TGA) of the electrode starting materials and products was conducted with a Netzsch STA 449 F3 Jupiter analyzer. The samples were heated at a rate of either 5°C per minute under flowing air (siderite FeCO₃, 50 ml min⁻¹) or 10°C per minute under flowing argon (siderite FeCO₃ plus LiH₂PO₄, 50 ml min⁻¹) from 30 to 900°C.

Tap density measurements were taken using a Pharma Alliance Group Tap Density Tester. Two samples were measured, one commercial LFP sample (Hanwha Chemical, China) and one siderite-derived LFP sample, with a starting mass and volume of 3.01 g and 4.95 ml and 5.25 g and 3.95 ml, respectively. Taps were done in 1000 tap increments until the volume change was less than 5%.

To make inductively coupled plasma–mass spectroscopy (ICP-MS) measurements, ∼25 mg of each sample was digested in a hydrochloric nitric acid mixture. The samples were heated to near dryness and diluted to 15 ml. Another 1:10 dilution in 2% nitric acid was performed. The digested and diluted samples were measured using a Perkin Elmer DRC II inductively coupled plasma mass spectrometer.

Electrodes were fabricated as follows: The positive electrode contained a coating of 90 wt. % carbon-coated siderite-derived LiFePO₄ (as synthesized above), 5 wt. % C45 (Timcal), and 5 wt. % polyvinylidene fluoride binder (PVDF, Solvay 5130). The laminate was prepared on a 20 µm aluminum current collector and had an active material loading of ∼4.7 mg cm⁻² with a coating thickness of 70 µm. Electrode laminates were not calendered.

Electrochemistry of the LFP was investigated using 2032-type coin cells. Before cell assembly, the LFP electrodes were dried at 100°C and the separator at 70°C in vacuum ovens. All coin cells were assembled in an argon glove box (<0.1 ppm O₂) using a 14 mm diameter LFP positive electrode, 15 mm diameter lithium chip (MTI Corporation), and 16 mm diameter separator (Celgard 2325). The electrolyte was composed of 1.2M LiPF₆ in ethylene carbonate:ethyl methyl carbonate (EC:EMC, 3:7 wt./wt., Tomiyama), and all coin cells used excess electrolyte (“flooded cells”).

Electrochemical cycling of the LFP//Li half cells was conducted on a Series 4000 Test Unit battery cycler (MACCOR) at 30 or 55°C. Two sets of protocols were used: an aging protocol and a rate capability protocol. The aging protocol consisted of repeated C/10 cycling for 100 cycles. The rate capability protocol performed the following rate tests to measure the effect of impedance on the discharge capability: C/100, C/50, C/20, C/10, and C/5.

See the supplementary material for additional x-ray diffraction, thermogravimetry analysis, and electrochemistry data, and a photograph of siderite-derived LFP.

This work was supported as part of the Center for Electrochemical Energy Science (CEES), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences. The submitted manuscript has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory (“Argonne”). Argonne, a U.S. Department of Energy Office of Science laboratory, is operated under Contract No. DE-AC02-06CH11357.

The authors have no conflicts to disclose.

A.N.J. and C.S.J. performed conceptualization. C.P., W.M.D., A.N.J., and C.S.J. gave the methodology. C.P., W.M.D., and J.B. carried out the investigation. C.P. and W.M.D. wrote the original draft. C.P., W.M.D., A.N.J., and C.S.J. reviewed and edited the original draft. A.N.J., C.L., and C.S.J. supervised the work.

W.M.D. and C.P. contributed equally to this work.

The data that support the findings of this study are available from the corresponding author upon reasonable request.