In situ and ex situ NMR for battery research

The efficient storage of energy has seen a remarkable escalation in importance as an increasingly mobile and energy-demanding global population arises. Since the first voltaic piles were produced over 200 years ago, progressive improvements to this technology have been made to decrease the size, increase the capacity, lengthen the service life, and even allow for the reuse of the cell. Due to the convenience of recharging, batteries, in particular Li–ion batteries, have become the dominant power source for portable electronic devices and mobile electrically-powered vehicles [1, 2]. Consumer demand for devices with enhanced power cycle lifetimes has driven the development of rechargeable batteries with continually increasing energy density. As these improvements are realized, a given battery class may reach the theoretical maximum energy density limit, requiring the use of different internal materials for continued progress. As such, extensive research efforts have focused on improving the performance of these energy storage devices by utilizing alternative materials within the battery. Judicious electrochemical improvements require a thorough understanding of the chemical interactions taking place within the battery, prompting a need for thorough scientific research into these technologies.

The chemistry of batteries occurs with four basic components: two electrodes (cathode and anode), a separator, and an electrolyte. The anode (negative terminal) serves as the source of electrons in the electric circuit. The cathode (positive terminal) accepts these electrons. The electrons are transferred through the electrolyte as ions, allowing for the flow of current. Each of the electrolyte constituents (i.e. solvent, salts and other additives) has a dramatic impact on the performance of the battery and can impact the chemistry of the other materials. The operation of a battery is reliant upon electron and ion transfer across solid–solid and solid–liquid interfaces. Within a given phase, changes in one component can impact the response of electron and ion transfer as well as the reversibility. For rechargeable battery systems, the repeated cycling induces the formation of microstructures at the solid-electrolyte interface (SEI) as well as in the electrode itself due to the migration of ionic species, which is often responsible for battery failure. Thus, it is necessary to have a deep understanding of each component and how they will impact the battery’s performance.

The physical and chemical properties of electrolytes are directly derived from the composition of the electrolyte. Maximizing the conductivity, ionic fluidity, non-reactivity with electrodes, and stable operating temperature range are important attributes of an electrolyte. Anodes should be efficient reducing agents that exhibit good conductivity, high coulombic output, and stability. Cathodes, on the other hand, should be efficient oxidizing agents over the desired voltage and stable when in contact with the electrolyte. The separator is a porous membrane that prevents the direct contact of the two electrodes. It must be stable under the reactive conditions and uniformly permeable to the ions or electrolytes carrying the ions.

There exist a number of important considerations regarding the electrode design beyond electrochemical potential. On one hand, the formation of soluble intermediate species is critical to harnessing the power of the cell, but dissolved species can also be an indicator of failure, requiring the use of carefully designed materials to limit the diffusion of soluble species, as in the case of utilizing sulfur-impregnated carbon nanotubes as a cathode to mitigate diffusion and improve cycle stability [3]. Refinements to the design of electrodes can also be made to prevent significant volumetric expansion of the cell, which is highly undesirable in commercial applications. In one such example, a sulfur-titania yolk-shell architecture was used to preemptively provide void space for volumetric expansion [4]. Other architectures can be envisioned to improve the conductivity of the electrodes [5].

Despite these efforts, there is still much to learn about the species generated on electrodes and in the SEI during operation. Their nature and the mechanisms by which these microstructures arise are difficult to ascertain. Further, the specific reasons for a given electrolyte showing promising performance over another are not always intuitive, requiring careful molecular level experimental observations to understand the fundamental principles at work. The SEI is often poorly understood as well, despite being a key component in advanced electrochemical devices. This interface forms as solvent and electrolyte salts are reduced to oligomers and crystals on the electrode surface. These electrolytes are stabilized at potentials beyond their thermodynamic limits, enabling the reversibility of the reaction. However, the chemistry, formation mechanism, and cycle-induced evolution of these are largely unknown, promoting a strong desire for detailed investigations [6].

A variety of techniques have been employed to evaluate the performance and chemistry of batteries to better understand their operation and failure mechanisms for continued enhancements to performance. Techniques unrelated to nuclear magnetic resonance (NMR) have provided excellent insight for battery technology. Many of these methods will be briefly discussed to provide an overview of the insight they provide. Cyclic voltammetry, for instance, is extensively used to characterize the general electrochemical performance of a battery, such as the stability and reversibility of the reaction at a given potential [7]. More detailed analysis allows for a measurement of the electrode reaction kinetics [8]. This technique, however, fails to provide a clear and detailed picture of the molecular changes that occur throughout the repetitive cycling process. In particular, this technique lacks a decisive description of the chemical evolution of the various battery components and how this dynamic process impacts the properties of the cell.

Beyond the standard electrochemical performance evaluations, characterization of the materials encourages a deeper understanding of system limitations and potential enhancement routes. One such avenue for characterization is the direct imaging of the battery with microscopy techniques [9, 10]. The advent of highly-developed aberration-corrected scanning and transmission electron microscopy (TEM) has propelled the capability to atomic-scale imaging [11–16]. Ex situ beam imaging has been widely applied to analyze the structural features of battery constituents, but failed to capture information related to the dynamics of the charge/discharge cycles [17–19]. In situ investigation of batteries with TEM is challenging owning to the high-vacuum operation of the equipment and subsequent incompatibility with liquid electrolytes. To mitigate this, three strategies have been explored to provide an in situ analysis of these materials. The first, an open-cell configuration utilizing ionic liquid-based electrolytes, takes advantage of nanowire anodes that are wetted by the ionic liquid to mimic an electrolyte system [20, 21]. It has been used to study a number of anode materials, including Si, Ge, Al₂O₃, SnO₂, ZnO, graphene, Sn, and carbon nanotubes [22–28]. This concept, however, suffers from ionic liquid polymerization in the presence of the electron beam, significantly shortening the number of charge/discharge cycles allowable by the system. The next, another open-cell configuration but using metal and metal oxide electrolytes, has been used for Li metal and silicon electrodes with a lithium oxide electrolyte [29]. Variations exist, but this design can be used with Na-, Mg-, and Ca–ion battery systems as well [30, 31]. Again, this configuration is not representative of a truly rechargeable battery due to limitations on the charging cycles and beam-induced effects, but these open cell configurations offer significant advantages in spatial resolution imaging and chemical analysis by electron energy loss spectroscopy [31]. Finally, a closed liquid cell battery design has been employed to better mimic the electrolyte diffusion and electrode interaction characteristics exhibited in a typical cell. Initiated by studies on copper nanoparticle growth from an electrolyte solution, electrochemical cells for in situ TEM were developed to push the technology forward [32, 33]. Working battery cells soon followed that allowed for the direct observation of the lithiation, ion transport, and SEI layer formation [34–37]. However, replicating a realistic number of charge/discharge cycles is still a concern with this technology. In addition to these, liquid observation is also possible to monitor changes in electrolytes, but further development in the field is required to match the performance of analyzing the solid components [10].

Raman spectroscopy can also be used to characterize changes in battery components. Ex situ Raman spectroscopy and ellipsometry have been employed to characterize the carbon films used as electrodes in Li–ion cells [38]. For example, they have shown that the structure of different carbon samples is degraded slightly through progressive discharge cycles. In addition to observing changes to the carbon anodes, cathodic LiCoO₂ rock-salts have been shown to deintercalate lithium species and result in seemingly random Li⁺ occupations of the available sites on the lattice [39]. In addition to the information ex situ Raman can offer, in situ Raman sampling cells have also been employed to observe the specific cycling stages in which intercalation of electrolyte ions into graphitic electrodes occurs [40, 41]. For instance, similar technologies have been used to show the specific redox steps proceeding at given voltage transitions for Li–ion and Li–S batteries, for example [42, 43]. Surface reaction mechanisms at the electrolyte-electrode interface can be probed as well, as was performed for Li–ion and Li–V₂O₅ systems [44–46]. It has even been used to show the complex and oscillatory distribution of ions in various regions of the separator and other fragments of the battery [47, 48]. UV–vis spectroscopy has also been shown to provide quantitative evidence on the polysulfide chain growth during Li–S battery operation [49]. Similarly, infrared methods have played a role in better understanding the nature of the SEI [50, 51]. These real-time capabilities have notably contributed to a deeper understanding of the changes that occur within the battery.

A number of x-ray methods have contributed to the scientific understanding of battery systems by analyzing samples both in situ and ex situ [52]. X-ray diffraction (XRD), for instance, can be used to understand the phase and layer spacing of electrodes [53]. An early demonstration of in situ XRD measurements for electrochemical applications focused on understanding the phases generated during the intercalation of lithium into graphite during electrochemical cycling [54]. This work studied various Li:C₆ ratios and found a random distribution of Li at low Li content with more distinct phases arising with the intensification of metal content, accompanied by an increasing carbon layer spacing as the lithium content increased. Utilizing synchrotron radiation and a specially designed cell, Walus et al were able to separate the diffraction pattern contributions from individual electrodes in a Li–S battery from the signal of the entire cell [55]. This demonstrated the formation of Li₂S on the anode at the early stages of the discharge cycle and complete consumption during the subsequent charge. Additionally, XRD was used to demonstrate the formation of a different sulphur allotrope after recrystallization. Numerous additional in situ XRD studies have focused on the crystalline phases present for a variety of battery systems; Li–Si, Li–MO_x, Li–air, Li–ion, Na–ion, etc [56–63]. An in situ XANES cell has also been used, though less commonly due to the necessity of a synchrotron, to investigate electrodes and electrolytes for sulfur speciation on the sulfur K-edge, showing the impact of sulfur species on capacity [64, 65].

X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES) have been used both in situ and ex situ to understand the bonding and oxidation state of the electrodes throughout the cycling process, but these techniques suffer from drawbacks related to the need to operate under ultra-high vacuum (UHV) conditions and difficulties with mitigating the relevance of closed cell designs with open cell detection sensibility [66, 67]. To address the concern about operating under UHV conditions, recent XPS efforts on Li–air battery research take advantage of a special in situ ‘ambient pressure’ cell that allows for reaction monitoring up to 500 mtorr (0.0007 bar), which demonstrated notable differences in the Li species present upon discharge as compared to UHV studies [68]. Though an impressive improvement, this pressure falls significantly short of ambient (1 bar). Despite these shortfalls in realistic condition representation, key insights have been gained from employing the technique with complementary experiments [69–71]. Additionally, XPS benefits from the ability to detect the presence of organic impurities on the electrode surface that arise from power cycling [72].

In situ and ex situ NMR spectroscopies have contributed significantly to the field. NMR allows detailed molecular level information related to electrolyte, separator, and electrodes all to be observed if contained within the detection RF (radio frequency) coil. Although not without its own limitations in sensitivity, it holds many unique advantages and has been thoughtfully developed over the last two decades. For one, NMR is (generally) a quantitative technique that allows for tracking the concentration of species regardless if they are liquid, solid, gaseous phases or a mixture of these phases from sample to sample or over time as the electrochemical reaction proceeds. For solid phases, both the amorphous and the crystalline phase can be captured at the same time. It should be noted that quantitative measurements are limited to skin depth when applied to metals and thus the thickness of the microstructure and bulk surface area must be considered [73]. Another significant advantage to NMR is that it is non-destructive, allowing non-invasive detection of the process under true in situ conditions where unique species may only exist. The true in situ conditions include, but are not limited to, operating at a wide temperature range, e.g. from well below 0 °C to well over 100 °C, and operating at pressures ranging from well below ambient to 100 bars and above [74–77]. Furthermore, the layout of the battery components and the size of the battery can be made the same as commercially available mini-sized batteries. The only exception here is that the outmost stainless container is replaced by a plastic holder to allow the RF field to penetrate into the battery electrodes, separator and electrolytes. In combining the in situ NMR with the detailed molecular information of stable species obtained by ex situ methods, NMR can be an invaluable addition to the experimental arsenal of a battery investigation [78–80].

Two additional differences between characterization by NMR and other techniques are the intrinsic length and time scales of the measurements performed. Herein, we seek to provide an overview of these scales for NMR, but would refer the reader to other literature on the spatial and temporal limits of other methods [81]. The time scale associated with NMR is not a straight-forward, singular concept. In fact, it should be regarded as a collection of scales due to the abundance of phenomena that this spectroscopy technique may probe. While changes to specific chemical signals can be monitored during a process, the mobility of these species and their interactions with surrounding media reveal unique transient information. These phenomena will be discussed to give the reader a sense of the capabilities NMR processes that can be employed to better understand battery systems.

The local chemical environment of the nucleus being analyzed directly impacts observed chemical shifts. Such a measure for the local environment relinquishes insight on the bonding structure and lengths, torsion angles, hydrogen bonding networks, secondary structures, and electric charges [82]. Due to the multiple factors determining the observed frequency, theoretical calculations that simulate bonding structure and secondary structures are often necessary for a deeper interpretation of the observed signals. Such calculations have shown the important impact of a second solvation shell on the observed solute chemical shifts, suggesting a length scale on the order of nm [83]. Nucleus-electron coupling has also been shown to impact NMR signals, suggesting a range of a few angstroms [84]. Further lengths scales could arise from the molecular motion of the nucleus itself as it is exposed to different environments. Environment-specific information is commonly used to provide dynamic speciation of observed signals. Though it has relatively low temporal resolution, chemical species transformation during a time-lapsed in situ experiment is one direct measure of the time scale for NMR spectroscopy. This type of experiment monitors changes in the chemical structure over a period of time. For example, in section 5.3 (titled In situ NMR on Li–ion battery research) we recount the evolution of ⁷Li species on the electrode during electrochemical cycling and highlight the changes in species identity over time. The time scale to analyze such transformations is heavily dependent upon the system. Factors such as the nucleus, relaxation environment, spin abundance, and experiment type all play an important role in determining the minimum time for analysis. This type of experiment can be conducted on a wide range of time scales, typically on the order of seconds (or fractions of a second) for an abundant ¹H system to hours or even days on a less favorable nucleus. Similar time resolutions (though often extended) can be applied to observe the change (or derivative) of other transient behaviors described, such as a change in the chemical exchange rate during a reaction.

In addition to the chemical shift value describing length-based information, the linewidth of a given signal provides the dynamic properties of a molecule, such as the rotational and translational behaviors. Though related to the relaxation mechanisms (see below), this motion is also related to the interchanging of species environment. This phenomenon is termed chemical exchange and is readily elucidated by NMR [85]. In this, a given nucleus may sample a range of magnetic environments, such as an interchanging ligand state, during the exchange process and the resulting NMR spectrum will provide kinetic data on the exchange rate. The concept is analogous to the uncertainty principle whereby the uncertainty in the resonance frequency is inversely proportional to the lifetime in a given state. For instance, when the exchange rate is relatively high, the lifetime in a given state is low and the resonant frequency is apparent. The temporal limitations of chemical exchange are determined by the frequency difference between two species, and thus the nucleus and magnetic field, according to equation (1).

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle {{\tau }_{{\rm coalescense}}}={{\left(\sqrt{2}\pi \Delta v \right)}^{-1}}={{k}^{-1}}.\nonumber \end{align} \tag{ 1 }$$

Here, Δν represents the frequency difference between the magnetic environments. For a slowly exchanging process, each resonance is well-resolved. Exchange spectroscopy (EXSY) has been used in this regime to quantify dynamics on the ms time scale [86]. As the exchange rate increases, the lines broaden and migrate towards the weighted average signal location. When the exchange rate reaches about an order of magnitude higher than the frequency difference, the resonances are averaged to a single broad line. On this time scale, fractions of ms, the Carr–Purcell Meiboom–Gill relaxation dispersion (CPMG RD) approach can be applied to refocus the exchange broadening with the spin-echo pulse sequence to elucidate the relationship between this observed broadening and the exchange rate [87]. This single line narrows to reveal a sharper resonance up until about three orders of magnitude faster exchange than the frequency difference [88]. To visualize the range of time scales for the exchange process, the reader can imagine the case of two resonances at the extremes of a given nucleus’ chemical shift range. The time scale involved in the averaging of these two species can be approximated by equation (1), bearing in mind the effect of magnetic field on the frequency difference. As shown in table 1, the NMR time scale for ¹H exchange at 7 T varies from 0.2 s (given 1 Hz resolution) down to 58 µs for the full range of 3.9 Hz. Since the frequency range (columns 3 and 5) vary based on the field strength, so too does the NMR time scale for exchange, where a proportional reduction is observed between columns 4 and 6.

The observation of dipole-dipole coupling by NMR can also indicate the time scale of molecular motion. The scalar coupling constant can be applied to equation (1) to provide the threshold for the exchange rate. If the lifetime of the species in a given environment is short compared to the inverse of the coupling constant, no coupling will be observed in favor of a single resonance. Typical coupling constants for ¹H–¹H, ¹³C–¹H, and ¹⁵N–¹H may be around 10, 150, and 50 Hz. These correspond to about 22, 1.5, and 4.5 ms, respectively, though this can be reduced to the ps time scale in specialized applications [89]. Since J-coupling is not impacted by field, these time scales are field independent and directly impacted by the chemical bonding.

In addition to gaining insight on the time scales involved with dynamic chemical environments, NMR is widely used to study the relaxation time scales of chemical species. T₁, for instance, is a measure of the longitudinal NMR relaxation that describes the rate in which excited nuclear states relax to equilibrium magnetization in the direction of the applied field. This relaxation is governed by modulation of the nuclear spin energy levels from fluctuations in magnetic interactions with components that match the Larmor frequency of the nucleus. This could arise from such physical processes as vibrations, torsion angle rotations, and looping motions [88]. The spin-lattice dipolar relaxation rate (reciprocal time) is described by equation (2) and shows a maximum rate at ωτ_C  =  0.6158, where ω is the resonant frequency and τ_C is the correlation time [90]. Though an indirect measure of the time scale, it is important to note that it is dependent upon the Larmor frequency, and thus the magnetic field, where the time scale of molecular motion associated with this relaxation is on the order of (2πϒ)⁻¹. Though spin-lattice relaxation times are typically on the order of seconds, the corresponding time scale for molecular motion (translation, rotation, or vibration) is on the order of 0.1–10 ns. In considering the relaxation initiated by paramagnetic centers, such as a rotating metal complex, the time scale for this motion can extend down to tens of picoseconds [91]. On a similar time order, but often smaller in magnitude to the spin-lattice relaxation time, transverse relaxation via spin-spin interactions can similarly be detected. The characteristic dipolar equation describing T₂ can be found in equation (3). Further system insight can be obtained through measurements of the two relaxation rates. Through this, time scale uncertainties can be drastically reduced to reveal the correlation time between species, τ_C.

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \frac{1}{{{T}_{1}}}=\frac{3}{160}\frac{{{\mu }_{0}}{{\gamma }^{4}}{{\hslash }^{2}}}{{{r}^{6}}}\left[ \frac{{{\tau }_{{\rm C}}}}{1+{{\omega }^{2}}\tau _{{\rm C}}^{2}}+\frac{4{{\tau }_{{\rm C}}}}{1+4{{\omega }^{2}}\tau _{{\rm C}}^{2}} \right]\nonumber \end{align} \tag{ 2 }$$

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \frac{1}{{{T}_{2}}}=\frac{3}{320}\frac{{{\mu }_{0}}{{\gamma }^{4}}{{\hslash }^{2}}}{{{r}^{6}}}\left[ 3{{\tau }_{{\rm C}}}+\frac{5{{\tau }_{{\rm C}}}}{1+{{\omega }^{2}}\tau _{{\rm C}}^{2}}+\frac{2{{\tau }_{C}}}{1+4{{\omega }^{2}}\tau _{{\rm C}}^{2}} \right] \nonumber \end{align} \tag{ 3 }$$

where γ is the gyromagnetic ratio (rad/s/T), ħ is the reduced Plank’s constant (1.054 56  ×  10⁻³⁴ m² kg s⁻¹), ω is the Larmor frequeny (s⁻¹), and µ₀ is the magnetic permeability of free space (1.2566  ×  10⁻⁶ H m⁻¹). Transverse relaxation in the rotating frame can also be employed to elucidate the time scale for fast exchange rates that are challenging to determine by other methods (such as CPMG) due to the maximum pulsing rate [92]. For these high exchange rates, a technique exists that enables the determination of T_1ρ. This approach is further advantaged by the enhanced sensitivity of minor species in an exchange process and by the requirement of only one static field strength. This method provides the potential to identify exchange rates on the order of ms to µs [93]. The interested reader is directed to other sources for details on the pulsing schemes for T_1ρ determination [94].

Given the variety of mechanisms by which NMR can provide insight into the time scales of the chemical interactions, it can certainly be considered an attractive option to probe chemical systems. Though this review and many published works primarily analyze the time scales associated with changes to the species identity (chemical shift), the framework for additional insight exists, demonstrating the power for NMR to supply a wealth of information about chemical systems in situ. Figure 1 provides a summary of the applicable time scales for NMR.

Zoom In Zoom Out Reset image size

A recent review by the Grey group has described their own thorough endeavors related to NMR spectroscopy applied to batteries [95]. The article outlines the purpose for utilizing NMR spectroscopy to understand battery systems as well as specific instructions for sample and experimental preparation and design. Other battery-class specific reviews are described in the relevant sections. Herein, we describe the development of in situ and ex situ solid-state NMR for battery research and discuss key examples that highlight the power of NMR in battery research. The included examples are specifically centered on chemical shift-structure relationships to identify the types of species present in batteries and electrolytes. These methods are complimented with computational chemistry to make accurate assignments of the observed spectral featured. An understanding of the temporal resolution on the time scale of several seconds and higher is described to monitor the chemical transformations during cycling.

Distinct advantages present themselves in conducting in situ observation. Apart from the advantage of observing the state of the material during the entire process instead of only discrete stages of a transformation, in situ investigation does not require numerous samples to be prepared for a full understanding of an entire process, saving time and promoting sample consistency. Instead, one representative sample can be scanned throughout the entire process. Importantly, some transient species can only exist when the electric wires are attached to the battery systems with current constantly applied [96, 97]. There are, however, technical challenges in successfully devising an apparatus suitable for in situ investigation. A number of iterations have taken place to promote in situ capabilities that have better reliability, sensitivity, and longevity.

The first demonstration of in situ NMR for batteries came from technology developed by Rathke et al at Argonne National Laboratory [98]. Their design employed a toroid cavity in their custom built probe similar to that developed for magnetic resonance imaging [99, 100]. The first battery experiments used Li–ion coin and cylindrical cell containers that integrated operation with the RF coil [101, 102]. This was an exciting advancement that allowed for ⁷Li NMR observations on a live battery system where charge-discharge cycling took place directly in the magnet. The drawback on employing the toroid coil was that the same copper disc (or central wire) was simultaneously used as both the current collector and radio frequency conductor for NMR signal excitation and detection. This means that for in situ measurement to begin, the battery operation had to stop periodically or non-periodically. Still, the battery was charged and discharged within the probe and magnet and measurements were taken shortly after the cycling was stopped, but this was a severe limitation in terms of truly in situ measurement.

Though the toroid design for in situ experiments provided a new frontier for the investigation of a functioning battery, it suffered from challenges in conformity with standard (lithium ion) battery designs in addition to severely limited signal-to-noise ratios. In the mid-nineties, small, flexible, plastic Li–ion batteries with the capability of being recharged were demonstrated by Bellcore, offering improved application flexibility [103]. To address the aforementioned disadvantages of the toroid design, this plastic cell design was sized for utilization as an in situ NMR cell by Chevallier et al in Orléans, France in 2003 [104]. This allowed for direct use inside an NMR probe for truly in situ electrochemical cycling since recording in open-circuit mode was no longer a limitation [105]. A schematic of the plastic bag cell and the accompanying detection and cycling system is shown in figure 2. In this design, each electrode was supported on a copper current collector before being treated in a 1% solution of polyvinylidene fluoride-hexaflouropropylene (PVDF-HFP) in acetone and laminated (130 °C, 20 psi) to opposite sides of a separator membrane composite composed of PVDF-HFP and dibutyl phthalate. With the stack activated with electrolyte prior to the second electrode being laminated to the separator, the two electrodes, separated by a membrane and resting between current collectors, were then packaged into an ultra-thin aluminum blue bag measuring 4 cm by 8 mm. These plastic bag cells provided a low cost and flexible option for in situ measurement, but experienced relatively short lifetimes (five days has been reported) due to cell breakage and permeability. The cells can also experience large electrical resistance [106]. Al-coated bags have been used to extend the cell lifetime, but these decrease the sensitivity of the experiment [105]. Early demonstrations of ⁷Li in situ NMR for battery technology with these plastic cells allowed for up to 2 mg of Li and 16.7 mg of graphite to be involved in the intercalation process [107]. This design allowed for the successful observation of reversible lithium intercalation into the carbon layer, where a direct charge transfer from lithium is realized. After the formation of these intercalated species, lithium is capable of penetrating the graphitic nanopores to form quasi-metallic species that registered downfield [104]. It has also allowed for the firm identification of the dense interstitial LiC₆ and LiC₁₂ as well as clarified the true identity of the dilute LiC₉ stage, demonstrating the importance of careful sample preparation (washing stage) for the ex situ NMR data collected in the decades prior [108–110].

Zoom In Zoom Out Reset image size

While the flat plastic cell propelled in situ NMR for battery applications and reduced the metallic components contained within, it suffered from key drawbacks such as the restriction to plastic electrodes and fragility that leads to solvent leakage and a loss of electrical contact. To address this, an improved design made especially for Li–ion batteries was constructed from a cylindrical plastic housing with a threaded seal, reducing the sealing area and metallic components while providing good cycling characteristics, as demonstrated with ³¹P NMR for on Cu₃P electrode in the Li–ion battery [111]. Another design is based on a cylindrical micro-battery developed for a telemetry system for salmon tracking [112]. This jelly roll structure was subsequently used to study Li–S batteries where the Kel-F holder containing the battery cell sat directly inside the RF coil and a charge cycler could be connected to the apparatus outside of the magnet [96]. Similar versions of the cell have been used for lithium alloy and graphene oxide battery studies [97, 113]. Briefly, the battery was comprised of cells symmetrically fabricated by two flat Li metal foils (7 mm  ×  7 mm  ×  0.75 mm) separated by a flat glass fiber and inserted into the Kel-F holder. An electrolyte, such as propylene carbonate, was incorporated and the entire system was glued/epoxy-sealed with a Kel-F cap containing a 1 mm hole in the center for the copper wires carrying the current [114]. The cell and glass spacers fit tightly inside the plastic holder, which is capped and glued shut prior to placement inside the probe and magnet. The design can be subsequently improved by removing the need to rely on glue to form the seal by utilizing O-rings. Further improvements come from stacking, but separating, electrodes and connecting them electrically in parallel to enhance the sensitivity of the spectroscopy. The use of O-rings will improve operating conditions and quality of the seal.

The most recent design of an in situ NMR battery cell was developed in Germany by Kayser et al [116]. The design incorporated carefully 3D printed components that enabled a rigid seal, reproducible cell compression, and long-run capabilities (tested up to 2400 h). The battery cell is placed on the bottom of a hallow plastic cell body. The body has a smaller hole which the battery covers wherein a contacting stick will consistently apply the same pressure to the bottom of the cell due to a change in diameter of the shaft and hole. A top piece slides into the bore of the cell body to contact the top of the cell. This also has a hole for a second contacting stick to similarly apply a consistent contact to the battery cell. The entire apparatus is sealed tight with nuts that compress bellowed cutting rings. This method has shown good sealing with a loss of just 1 bar of 15 after 1 d. Based on the material components used, the cell should be able to tolerate up to 120 °C and exhibit good chemical resistance over many years. The entire cell fits into a homemade saddle coil positioned with careful attention to the alignment within the magnetic field to ensure consistent and reproducible measurements [106, 117, 118]. Though still a new development, this cell paves the way in understanding the degradation pathways for batteries across hundreds of cycles due to its reliability and robustness.

Nowak et al have developed an in situ battery cell for liquid electrolyte measurements [119]. This device employs a cylindrical battery that rests atop the electrolyte measurement area, i.e. an extended compartment containing only liquid electrolytes intentionally placed inside the NMR detection coil. The authors report ¹H and ¹⁹F measurements to monitor signals associated with fluoroethylene carbonate and fluoromethyl methyl carbonate during electrochemical cycling. Through these initial demonstrations, a detection limit of 200 µmol l⁻¹ was determined for ¹H and ¹⁹F in this cell.

While in situ techniques are certainly the pinnacle of representing a battery’s operational condition, difficulties can arise in capturing the species of nuclei that are relatively more challenging to probe in terms of both sensitivity and linewidth. While ¹H, ⁷Li, and in some cases ¹³C are great candidates for time-resolved NMR spectroscopy, low sensitivity, low abundance, and quadrupolar nuclei strongly rely on magic angle spinning (MAS) and longer spectral acquisitions for a detailed and highly resolved picture of the sample’s chemical state, especially when in the solid state. Battery applications employ an array of atoms that can be probed with NMR, including ¹H, ²H, ⁶Li, ⁷Li, ¹³C, ¹⁵N, ¹⁷O, ¹⁹F, ²⁵Mg, ²⁹Si, ³¹P, ⁵¹V, ¹³³Cs, and others [39, 114, 120–126]. Many of these are either quadrupolar or low abundance nuclei.

While informative, the highly abundant and spin ½ nuclei do not always provide the clearest understanding of the chemical system. To compensate, less abundant or quadrupolar nuclei are often probed to investigate the detailed molecular environments in battery materials. Low abundance nuclei are a particular challenge for this relatively insensitive spectroscopic technique ¹⁷O, for example, is a nucleus with natural abundance of only 0.038%. Moving to a high field can help promote a stronger oxygen signal through a more pronounced Zeeman interaction. However, more distinct advantages can come from employing a probe that allows for a larger sample volume since the detected signal is directly proportional to the number of spins in the coil. If isotopic enrichment is not practical, an increase in sample volume is generally the simplest way of boosting a signal to acquire detailed information about a chemical system. ¹⁷O is also a quadrupolar nucleus, resulting in relatively broader lines. A number of interactions contribute to the resulting spectrum, as indicated by this summation of Hamiltonians:

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle {{H}_{{\rm Int}}}={{H}_{{\rm Z}}}+{{H}_{{\rm D}}}+{{H}_{{\rm CS}}}+{{H}_{{\rm K}}}+{{H}_{{\rm J}}}+{{H}_{{\rm P}}}+H_{{\rm Q}}^{1}+H_{{\rm Q}}^{2}+H_{{\rm Q}}^{n}+\ldots \nonumber \end{align}$$

where the Zeeman (Z), dipolar (D), chemical shielding (CS), Knight shielding (K), spin-spin coupling (J), paramagnetic (P), and nth-order quadrupolar (Q) interactions can all play a role in the resulting NMR signal. Quadrupolar nuclei generally suffer from low symmetry and wider lines due to the quadrupolar interactions that arise from the anisotropyof the nucleus charge distribution with the electron field gradient. This quadrupolar effect can be partially reduced by employing more powerful magnetic fields. In cases where the quadrupolar nucleus also resides in a solid, which may be in the case of extracted SEI, more advanced techniques are required for high resolution spectroscopy. The advent of MAS has also had a dramatic impact on the interpretation of NMR spectra [127]. This technique allowed for solid samples to approach resolutions obtained with liquid samples by averaging out many of the interactions that impact line-width. Odd-ordered Quadrupolar Hamiltonians converge to zero when the sample is rotated at 54.74° with respect to the external magnetic field, leaving behind only the even-ordered quadrupolar terms and dramatically improving the signal to noise ratio of a spectrum. As outlined later, the strategies of employing larger sample volumes, higher magnetic field, and MAS have all been used to make experimental observations on natural abundance, quadrupolar and solid samples.

While highly abundant nuclei are relatively more studied due to the ease of acquisition, in some cases they are not the best suited for an application. Lithium-based batteries, in particular, have a large body of research where ⁷Li and ⁶Li NMR have been both utilized to gain insight on the system [113, 114, 128–131]. ⁷Li is a spin 3/2 nucleus of 92.5% natural abundance, making its acquisition easy relative to ⁶Li (I  =  1, 7.5% nat. abundance). Further, ⁶Li also suffers from low sensitivity due to its generally longer relaxation times; however, the quadrupole moment is smaller than that of ⁷Li (~2.4 · 10⁻³⁰ versus ~4 · 10⁻³⁰ e m²) and dipole-dipole interactions are also much weaker, allowing for higher spectral resolution and significantly enhanced structural information [132–134]. This can provide a direct benefit in situations in which two Li are near each other, a closer approximation of the isotropic chemical shift is desired, or the source of relaxation is not well-understood. For batteries, the relatively more challenging ⁶Li might be a preferred technique if a variety of Li species are present and high spectral resolution is required to distinguish the signals. An additional advanced method for ex situ analysis of Li-based batteries includes pulsed field gradient (PFG) NMR, which has been employed to secure a firmer understanding of the role of temperature in the solvation structure of Li, Cs, and H species [135].

While significant strides have been made to push traditional battery system investigations towards in situ conditions, redox flow technology is primarily probed ex situ due to the difficulty of pumping steady streams of electrolytes along an interface while simultaneously detecting NMR-active species. These studies typically involve detailing the solubility and solvation structures of electrolytes in a given system to better understand solvent and ion interactions at different solute concentration levels to show the concomitant fluctuations of contact-ion pair presence and solvent-solvent or solvent-ion interactions [136]. Ex situ NMR has been shown to reveal the key insights on the impact of functionalizing electrolyte components, such as ferrocene, towards enhancing the solubility to overcome performance barriers of non-aqueous electrolytes [137, 138]. Similarly, the rational design of electrolytes has been demonstrated for the vanadium redox system where the effects of chlorine have been discerned [139–142]. Indeed, NMR has become an important technique for characterizing the chemistry of electrolytes in redox flow battery systems [143–148]. This same strategy can be broadly applied to electrolyte systems of an array of battery classes. An example for understanding the solvation structure of electrolytes for Mg-based batteries will be detailed in section 5.5. It is important to mention that the NMR interpretation for many of these electrolyte systems is aided by computational modeling, which will be discussed in the subsequent section.

While direct experimental observations are essential to a deep scientific understanding of a given battery process, computational approaches are gaining both popularity and importance for their predictive capabilities in chemical systems. Computational approaches become further important for relatively complicated spectroscopic studies where a given chemical system does not have a well-defined library of reference compounds that represent the chemical environments being probed. This might arise from unique environments only observed during a reactive process, or an entirely new material that has never before been probed. In these cases, it is often beneficial to complement NMR studies with computational simulations. As it applies to interpreting the NMR spectra related to batteries, two techniques serve as the primary aides in elucidating the signals observed; density functional theory (DFT) and molecular dynamics (MD).

DFT modeling is a computer-based method that uses quantum mechanics to describe the electronic structure of a many-body system. Functionals of the electron density are used to compute system properties. This method relies on exchange-correlation functional approximations that account for expectation differences when wave functions overlap and the influence of other electrons on a translating particle. This method allows for one to calculate the geometry and electron density of a system that can be used to compare to experimental observations. As it applies to battery systems, DFT can be used to better understand which structures show preferential formation during lithiation or model the effects of defects on cathodes [149]. It can also be used to calculate NMR parameters for comparison to collected spectra. A variety of methods can be used to accomplish this task [150]. A common method includes using the gauge-including atomic orbital (GIAO) approach [151–153]. In this, the shielding of a nucleus can be expressed as a second derivative of the total electronic energy referenced to an external magnetic field and magnetic moment of the nucleus. This method overcomes the unphysical ‘gauge’ complication of coordinate origin selection in the calculation and has been shown to deliver accurate results in comparing calculations to experimental results. Though many codes lack the ability to accommodate periodic NMR chemical shift calculations, the gauge-including projector augmented wave (GIPAW) method has enabled such computations [154]. This has reduced barriers to computations on systems that would typically require very large clusters to accurately model. Along with Quantum Espresso, PARATEC, and others, the CASTEP package, in particular, has been extensively used to apply this periodic method to battery research [155]. Much like experimental methods, the computational shielding calculations require the use of a reference compound, a library of which are available across the literature. All electron Slater basis sets tend to be used for these types of calculations since they describe core electrons more accurately than pseudo-potentials or Gaussians. Core electrons are generally free but the importance of accounting for these is not absolute given that sufficient shell electrons are accounted for [156]. This approach has been widely used to compare computational and experimental spectra ¹⁷O or ²⁵Mg NMR in Li and Mg batteries [124, 157]. The DFT approach has also been used to understand electrolyte systems for which solvent effect codes, such as the conductor-like screening model (COSMO), or the inclusion of sufficient solvent shells are required [158]. This has been applied to better understand the solvent effects in a system [159, 160].

An additional challenge can present itself in applying DFT to heavier elements in batteries, such as for vanadium- or cesium-containing battery components, where the electrons of the 4d orbitals begin to exhibit strong relativistic effects, requiring the use of the more computationally expensive (relative to the Schrödinger equation) Dirac equation. DFT-based total shielding calculations can be regarded as a sum of paramagnetic, diamagnetic, and spin–orbit coupling contributions. While relativistic effects may be accounted for without considering spin–orbit coupling, their applicability to accurate NMR calculations can be limited [161]. As such, NMR calculations considering relativistic effects often take advantage of the zeroth-order regular approximation (ZORA) [162, 163]. This method accounts for special relativity in heavier elements as electron speeds become more significant. ZORA is an excellent approximation to the fully relativistic Dirac equation, especially near the valance region. This improves the identification of species continuing heavier elements and secures greater confidence in the computational results. This computational method can be used to predict the NMR chemical shifts of structures that are likely to be present in the system. It should be noted that pure DFT calculations will not natively account for thermal effects on chemical shift since the calculations are based on 0 K. Effects such as chemical exchange between two species will require more robust methods to elucidate the structures present in the sample.

Still today, it can be difficult for DFT to accurately describe intermolecular interactions, especially dispersion forces where further perturbation terms may be necessary [164]. It is also challenging to model very large systems with hundreds of atoms to account for extended-order solvation schemes. To help determine the structure of a many-molecule system, MD simulations can be employed as a useful tool in understanding the nature of a battery system. Molecular Dynamics is a computational tool for understanding the motion of a many-bodied system based on Newton’s equations of motions. Ab initio molecular dynamics simulations take this concept a step further and compute forces using Shrödinger’s equations [165]. This powerful technique has been widely applied to study the SEI formation on Li–ion batteries [166–169]. It has also been used to understand the properties of electrolytes in batteries [170, 171]. Diffusion coefficients, for instance, can be approximated and compared to MRI values as well to better predict the state of the electrolyte during battery operation. MD can also be used for NMR calculations by taking the structures most likely to be present based on MD simulations and calculating the chemical shielding with DFT.

What may be a more robust approach would be to directly combine the strengths of the two methods to generate an aiMD-DFT, time-averaged picture of the chemical system. In this method, the MD simulation generates a series of snapshots from the trajectory for the molecules in the system. These configurations are used in DFT calculations to determine the NMR shielding tensors, which can be used to generate a weighted-average chemical shift of the configurations present for a given system. This combinatorial approach has been used extensively in solvated ionic platinum species, but can be useful in understanding complex electrolyte systems for battery applications [172–174].

The extraordinary importance of understanding the fundamental interactions in a given electrochemical cell has driven extensive research in the area with powerful in situ NMR investigations. Though numerous studies have been conducted that take advantage of this tool, the following important works demonstrate well the capabilities of these methods across a number of battery systems.

5.1. In situ NMR on Li-composite ion battery research

Composite anodes consisting of lithium and other metals (such as group IV Sn, Ge, Si) are attractive alternatives for traditional lithium anodes in Li–ion batteries. For instance, the use of Li–Sn composites have been investigated with NMR to demonstrate a 50% enhancement of lithium adsorption over carbon [175]. A brief review of in situ⁷Li NMR efforts in this area is available elsewhere that highlights the use of silicon, of interest due to its relatively high theoretical capacity [79]. Despite the interest in such materials, the precise structure of the lithiated species can be elusive to diffraction efforts due to the amorphous nature of the species. Lithium silicides, for example, have four reported stable crystalline phases, but electrochemical lithiation at room temperature favors an amorphous silicide. NMR has been applied to bridge the gap for these species by monitoring the transformation of lithium species during electrochemical cycling [115].

Silicon presents a popular material for composite anodes that has been investigated to better understand delithiation [131]. One prominent example of Li–Si composites is reproduced in figure 3. This illustrates the evolution of ⁷Li species with time, whose signals strongly depend on their coordination environment with Si. Initially, four peaks are present between 0 and 8 ppm which relate to Li in the electrolyte, SEI on the positive and negative electrodes, and Li_XC. A broad peak ascribed to small silicon clusters begins to form around 18 ppm during the discharge process, which migrates to 14 ppm (A2). Additionally a resonance at 4.5 ppm (magenta) related to silicon clusters forms with discharging, indicating further breakage of the silicon linkages. The peak that forms in the later stages of discharge at  −10 ppm arises from overlithiated, crystalline Li₁₅Si₄, which was not observed ex situ. The use of NMR helped solidify the understanding of Li–Si systems, explaining the nature of species present under reaction conditions and showing the reactivity of amorphous lithium silicides. Additionally, the authors discovered that excess-Li could facilitate self-discharge at low voltages, resulting in a loss of capacity for Si batteries.

Zoom In Zoom Out Reset image size

Balancing the anode’s high energy densities (992, 1623, and 4200 mAh g⁻¹ for Sn, Ge and Si respectively) with excellent substrate lithium diffusivity and electrical conductivity places Li–Ge as an attractive option [176–178]. NMR has also been used to better understand strategies for stabilizing the structure of this composite. For instance, conventional methods of incorporating the group IV metals result in poor capacity retention across cycling. This effect stems from the drastic volume changes during the lithiation and delithiation process coupled with unstable formation of the SEI film. However, embedding Ge on a conductive scaffold can assist in the minimization of the detrimental volume fluctuations; however, inhomogeneity of metal deposition still negatively impacts the retention of energy storage capacity [179, 180]. To mitigate the effect, core–shell structures have been employed that show enhanced resistivity towards storage degradation [181, 182]. Despite these enhancements, the mechanisms by which Ge core-shells impact Li cycling were unclear, complicating performance optimization. Tang et al employed Ge nanorodes encapsulated by bamboo-type multiwall carbon nanotubes (Ge@CNT) for a Ge–Li planar half-cell battery and in situ NMR to better describe the roles of Ge in the Li–Ge battery [97].

Figure 4 shows the in situ⁷Li NMR results obtained during the first two charging cycles of a Ge@CNT Li–Ge battery half-cell, plotted as a function of cell capacity. They observed four ⁷Li resonance lines centered at 24, 13, 10, and  −24 ppm which were monitored together with the ⁷Li signals of the electrolyte and SEI (0 ppm). From a detailed analysis of the collected data, supplemented with XRD, TEM, and STEM, it was concluded that the significantly shielded signal at  −24 ppm could be attributed to an overlithiated Li_15+ΘGe₄ phase that was only present during the electrical contact. This phase was uniquely identifiable by NMR (not XRD or TEM) under reaction conditions. The absence of this overlithiated phase promoted the formation of a broad signal downfield that was attributed to a more amorphous Ge phase that may enhance the lithiation/delithiation rate. The remaining peaks could be assigned likewise by the properties exhibited during the various experiments (see table 2). They showed much greater stability towards a lack of electrical contact.

Table 2. Summary of in situ⁷Li NMR Observations. Reprinted with permission from [97]. Copyright 2015 American Chemical Society.

Resonance index	Assigned phase	Chemical shift/ppm	Potential/capacity*	Crystal property	Observed by
R1	Li2.25Ge	24	0.17/839	Amorphous (in situ XRD [183] and TEM)	NMR (in situ7Li)
R2	Li3.5Ge	13	0.14/1086		NMR (in situ7Li)
R3	Li15Ge4	10	0.11/1325	Crystalline phase (in situ XRD [183] and TEM)	NMR (in situ7Li), in situ XRD [183]
R4	Li15+ƟGe4	−24	0.005/1653	Over-lithiated	NMR (in situ7Li)

Zoom In Zoom Out Reset image size

This study enabled key understandings surrounding the Li–Ge battery system, in particular, the reaction pathway during electrochemical cycling. The in situ⁷Li NMR studies revealed that the reversibility of lithiation in Ge@CNT is mediated by coexisting amorphous and crystalline phases present in the battery. It also enabled a deep understanding of the system’s electrically-driven metastable Li_15+ΘGe₄ phase that was only possible during electrical contact on the carbon walls that encapsulated the Ge. Using a combination of the techniques, the authors were able to propose a reaction pathway that accounts for all spectroscopic and microscopic observations as well as phase transitions available during cycling:

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle & {\rm Ge}(c)\to {\rm L}{{{\rm i}}_{2.26}}{\rm Ge}(a)\to {\rm L}{{{\rm i}}_{15}}{\rm G}{{{\rm e}}_{4}}(c)+{\rm L}{{{\rm i}}_{3.5}}{\rm Ge}(a)\nonumber \\ & \to {\rm L}{{{\rm i}}_{15+\theta }}{\rm G}{{{\rm e}}_{4,{\rm metastable}}}\left(c \right)+{\rm L}{{{\rm i}}_{3.5}}{\rm G}{{{\rm e}}_{{\rm metastable}}}(a)\nonumber \end{align}$$

where (c) indicates a crystalline phase and (a) is the designation for amorphous. This model well-reflects the progression of phases identified with the in situ NMR to describe the lithiation process during electrochemical cycling. It was speculated that the enhanced capacity from such carbon-encapsulated Ge was related to the formation of these metastable states during electrical contact. Such observations are unique to in situ NMR and promote a deeper understanding of composite battery systems. A number of additional in situ NMR studies have focused on composite anode and cathode materials using ⁶Li and ⁷Li, to provide a detailed understanding of the microstructure formation of Li dendrites, ion hopping, and phase transitions with a variety of electrode materials [115, 184–186].

5.2. In situ NMR on Li–S battery research

Lithium-sulfur electrochemical cells are battery systems that hold great potential advantages over conventional Li–ion batteries due to improvements in specific energy (~2500 Wh kg⁻¹) and low-cost production options [187–189]. In these systems, soluble lithium polysulfide species are formed as intermediates important to the energy transfer chemistry; however, dissolved species may also lead to poor electrochemical performance and failure. Though a number of in situ methods have highlighted the importance of the reaction in system understanding, they failed to provide a quantitative look at the evolution of the present species, instead focusing on identification of discrete sulfur or lithium polysulfide compounds [49, 53]. In situ NMR has the distinct advantage of simultaneously identifying and quantifying all NMR-active species during the charge, discharge, and standby periods of battery cycling. This concept was applied to the Li–S system using the cylindrical microbattery in situ NMR cell pictured in figure 5(A) [96]. A lithium metal anode (100 µm thick in blue) was separated from the cathode with Celgard 2500 of thickness 25 µm (red). The sulfur-based cathode film (around 175 µm thick) was laminated onto an aluminum mesh (green) where binding utilized polytetrafluoroethylene. In this system, the transient nature of Li species was monitored over several charge/discharge cycles to capture the microstructural evolution of the species in the various components of the battery.

Zoom In Zoom Out Reset image size

The resulting ⁷Li spectra were plotted as a function of time, where each spectrum was collected in just 4 minutes (see figure 5(B)). Representative spectra were selected and deconvoluted (figure 5(C)) to identify eight unique lithium species that could be identified with high confidence due to a more solid understanding of the evolution of these during the charge/discharge process (figure 5(D)) not available during ex situ measurement [64]. Though ex situ measurement of each the separator, cathode, and anode could be compared, a more clear assignment of these peaks arose based on the observed chemical shift, known species previously identified at a given voltage, and the line intensity change as the experiment progressed. This first in situ study using the cylindrical micro Li–S battery demonstrated that the reaction pathways were composed of mixed species in a complex environment instead of discrete step-by-step reactions.

For example, peak 1 at  −128 ppm was identified as a soluble long-chain polysulfide since S₈ is expected to reduce to this species at around 2.3 V. This was confirmed by the maximum intensity arising during the 2.1 V charge plateau, a minimum during the end of the discharge when insoluble Li₂S₂/Li₂S dominate, and on the basis of chemical shift value. This species was shown to directly and rapidly convert to long-chain polysulfides instead of relying on the accumulation of additional LiS₄ species. Further, this peak and its analog at  −100 ppm were not visible except under in situ conditions and is attributed to Li⁺ interacting with long-chain polysulfides as well as sulfide radicals. The second peak (−71 ppm) was similarly associated with soluble long-chain polysulfides such as Li₂S₈ or Li₂S₆. The remaining peaks, high field peaks 3 (−34 ppm) and 4 (−7 ppm), were described as insoluble Li₂S₂ or Li₂S on the lithium electrode.

Peaks 5, 6, and 7 were challenging to clearly identify, but on the basis of Li metal shifts near 250 ppm (Knight Shift from electrons in the conduction band) these are tentatively assigned as quasi-metallic, porous lithium species. These species also follow the opposite trend as the polysulfide signals. Peak 5 reaches a maximum intensity as the cell is fully charged whereas peaks 6–8 reached their maximum at 50% charge and declined to a local minimum at full charge. This suggests the growth of the porous lithium species in the deposition process and the concomitant reduction of the signals for peaks 6–8. These peaks were further found to relate to Li involved in creating the SEI. In particular, signal 8 (254 ppm) is assigned as a dendritic or thin mossy Li environments on the anode material.

In situ NMR also supports the hypothesis that the reaction pathways during charge and discharge differ due to the evolution of species observed and the differing lengths of the two plateaus of the charge-discharge curve.

This important in situ work was able to clarify the evolution of species involved in SEI layer formation as well as anode surface roughening and dendritic nucleation. Unique polysulfide environments became apparent during the charge and discharge process that cannot be captured by ex situ investigation. Confirmation of transient free radical presence offers key insights into the degradation of battery performance. Observations like those collected for this Li–S battery study offer molecular-level insights that are crucial to the accelerated development of the technology. A number of ex situ NMR studies employing ⁷Li and ¹H NMR have continued to investigate the key uncertainties surrounding this technology to compliment in situ studies such as See et al’s attempts to better describe the Li–S phase diagram [64, 190–192].

5.3. In situ NMR on Li–ion battery research

Numerous highly respectable efforts have employed in situ NMR to detail the chemical transformations occurring, such as extraction, insertion, and dendritic growth [81, 111, 118, 193]. For instance, the effect of carbon coating graphite for use as an anode was shown to have two sites for lithium; the graphite core and soft-carbon shell for intercalation and storage, respectively. The ⁷Li NMR showed the the relative abundance of the two sites was well correlated to the extent of coating, explaining the decrease in irreversible capacity and increase in coulombic efficiency [194]. Recently, lithium-ion battery technology utilizing reduced graphene oxide (rGO) has attracted great interest as a high capacity alternative to graphite anodes due to its high surface area and conductivity [195–200]. However, the lithiation mechanism for rGO was poorly understood relative to that of graphite-based cells, where differences in the mechanisms are evidenced by its increased capacity over graphite (600–1000 mA h g⁻¹ versus 372 mA h g⁻¹). To better understand the evolution of the SEI layer on rGO, in situ studies employing Raman spectroscopy and an NMR battery cell were conducted to monitor the temporal evolution of the G-band and ⁷Li signal during the charge and discharge cycle [113].

For both the graphite and rGO anodes, the in situ Raman shows a decrease in the G-band at ~1582 cm⁻¹ as the battery was discharged. This was ascribed to a resonance loss by the intercalation of Li, increasing the homogeneity of the sample. In the case of graphite, a 9 cm⁻¹ upshift is observed during cathodic polarization due to the donated electrons to the carbon. Contrasting the graphite’s loss of G-band below 0.2 V, the rGO sees this disappearance around 0.3 V. To provide additional insight, Tang et al used in situ⁷Li NMR to monitor the specific chemical environments of lithium species. A number of distinct lithium species were detected in the in situ experiments. Upfield, a broad peak at  −3 ppm and a few small, sharp peaks around 0 ppm were present. The sharp peaks at 0 are indicative of solvated lithium species while the broad signal is attributed to a combination of lithium intercalated into the rGO lattice and adsorbed on the surface functional groups present on rGO [201, 202]. Downfield, three peaks at 244, 250, and 265 ppm are present that relate to the Li–metal electrode. The peaks at 244 and 250 were attributed to a 10 µm skin layer on the Li–metal surface while the 265 ppm peak was assigned to the Li–metal fibers perpendicular to the metal surface [73]. Notable changes were observed for the broad peak at  −3 ppm and the peak at 265 ppm that provided insight on the electrochemical process. As seen in figure 6, the evolution of the various chemical species present could be tracked along the charge/discharge cycles.

Zoom In Zoom Out Reset image size

The  −3 ppm peak intensity evolution shows the clear enhancement of signal over the course of the experiment, approximately quadrupling in intensity by the end of the third charge cycle. The intensity of the intercalated/adsorbed lithium species is shown to increase over the discharge periods and reduce in abundance as the cell is charged. Conversely, the Li–metal fibers (256 ppm) remain nearly absent until charging initiates, reaching a maximum at the end of the first charging cycle. As discharging is initiated, the peak intensity falls as lithium ions are peeled from the surface of the metal. The concurrent and opposite behaviors during cycling demonstrate well the fate of Li species in electrochemical cycling. Tang et al made the interesting observation that only a small portion of the  −3 ppm species is participating in the cycling, suggesting that a majority of the lithium that had moved to rGO were inactive and participating in the formation of the SEI with the surface functional groups (–H, –OH,–COOH, –C=O, etc). The second charge/discharge cycle demonstrated an increased Li–metal species compared to the first cycle, showing a relative stabilization of the SEI after the first cycle. This observation was confirmed by the third cycle, which may serve to explain the high capacity of rGO during the first cycle and reliability during subsequent cycles, quantified by the area of the Li–metal peak.

Combined with the CV and additional NMR experiments, it was concluded that the primary difference between rGO and graphite was the preference of rGO to intercalate Li via surface adsorption instead of the conventional graphite mechanism. The passive SEI in rGO was formed during the first electrochemical cycle and was stabilized afterward. Approximately 36% of the intercalated Li⁺ was recycled during the first cycle. While the detailed structure of Li in the SEI could not be resolved, the differentiation between Li⁺ in the SEI and the counterpart on the rGO is directly visible as a function of time and electrochemical cycling. This study highlights the power of NMR to observe quantitative, real-time descriptions of battery constituents under unique chemical environments. Additional NMR studies related to this topic have made a variety of contributions to our understanding of Li–ion batteries [79, 111, 118]. Important and decisive observations, such as the mass transport limitations present in high-power Li–ion powered vehicle applications, are greatly assisted by these in situ NMR methods [203].

5.4. In situ and ex situ high-field NMR studies of Li–metal batteries

Li–metal batteries offer a high energy density option relative to Li–ion battery technology, but also pose a number of challenges in the safe recharging of the cell. The origin of the rechargeable Li–metal battery research finds itself in the 1970s [204]. Of interest as an anode due to its low electrochemical potential (−3.04 V with respect to hydrogen electrode), low material density (0.534 g cm⁻³), and high theoretical capacity (3860 mA h g⁻¹), limited success was realized due to poor columbic efficiency and dendritic growth of Li during charging cycles [205, 206]. The cycling stability of Li strongly depends on the interactions with the electrolyte, in particular where organic solvents render Li unstable and promote the formation of these microstructures. It has been identified that addition of certain species to the electrolyte forms a uniform, solid SEI or alloy with Li during deposition that can suppress dendrite formation and thus increase the cycling life [207–211]. For example, one approach to suppress dendritic growth is based on a self-healing electrostatic shield where low concentrations of Cs⁺ or Rb⁻ in the base electrolyte form a positively charged electrostatic shield around the nucleation points (Li tips). This forces deposition to adjacent regions to reduce dendrite formation [212]. Microscopy even identified a dendrite-free Li film, suggesting the presence of synergistic effects between the Cs⁺ and SEI [213]. Previous in situ NMR investigations on Li batteries have quantitatively tracked the real time dynamics under working conditions [80, 101, 105]. One study even highlighted the deposition of dendritic Li on the Li electrode during cycling [73]. This study, however, found a wide range of possible shifts (242–272 ppm) for a given species depending on its orientation with respect to the magnetic field. To amplify potential important signals in this region, the bulk magnetic susceptibility effect was clarified by employing planar symmetric Li metal cells measured as a function of the normal direction relative to the magnetic field for the study of the Li–metal battery system with and without the Cs⁺ additive [114].

The study highlighted the importance of Cs⁺ in the formation of well-aligned Li nanorods and reversibility of the Li electrode. Though a number of studies readily utilize ⁷Li NMR, this one also incorporated ¹³³Cs NMR technics to probe the dynamics of Cs⁺ in the battery. The in situ¹³³Cs NMR clearly showed the migration of Cs⁺ to the Li electrode to form a positively charged electrostatic shield during the charging process, evidenced by a significantly broadened signal compared to the narrow signal of a solvated Cs species. It was found that the Cs intensity was retained and reversibly solvated at the other end of the electrochemical cycle. The results also indicated a much larger fraction of the Li taking part in the cycling process when the additive was present. The rods generated in the absence of Cs tended to be much thicker and lead to a large fraction of inactive lithium species.

In addition to in situ Li–metal investigation, NMR has provided unique insights with ex situ techniques, particularly for low-sensitivity nuclei. The formation of new SEI layers as the battery was cycled that trap metallic lithium in the passivation film poses a potential safety hazard, making the SEI a target for detailed investigation [214, 215]. One notable example of ex situ NMR took advantage of ¹H, ⁶Li, ¹³C, and ¹⁹F NMR to study the SEI formed on the copper electrode. Wan et al employed the unique advantages of ⁶Li NMR over ⁷Li NMR to obtain a higher spectral resolution in the recovered SEI of a number of Cu|Li cells [216].

This work analyzed the SEI extracted from Cu|Li cells using two different concentrations of lithium bis(fluorosulfonyl)imide (LiFSI) in 1,2-dimethoxyethane (DME) as electrolytes (LiFSI-DME samples), including one lithium bis(trifluoromethanesulfonyl)imide (LiTFSI)-DME sample. The cells were cycled such that a 1 h deposition time was utilized to a capacity of 1.0 mA h cm⁻² and the stripping was controlled to an upper cutoff voltage of 0.5 V versus Li/Li⁺. These cells were cycled 200 times prior to SEI extraction, preparation, and loading into a sealed 3.2 mm MAS rotor for ex situ⁶Li MAS NMR measurement. The authors were able to identify a number of lithium species using high-field ⁶Li NMR spectroscopy. Across all samples investigated, metallic Li (264 ppm) was present that can stem from dead Li and broken dendrites that failed to cycle during the final delithiation stage. Relatively more metallic lithium is present in the least concentrated electrolyte (i.e. the 1 M LiFSI-DME), but fewer lithiated species can be observed compared to the other two concentrations, showing the clear impact of solution concentration on lithium species. By employing reference compounds and taking advantage of the higher spectral resolution afforded by ⁶Li NMR over ⁷Li NMR, six additional Li species were identified in the SEI of these cells: LiFSI, LiF, Li₂O₂, Li₂S, and Li₂O. Fluorine-containing compounds were verified by ¹⁹F MAS NMR due to the highly sensitive nature of these signals to the chemical environment of F. ¹H and ¹³C MAS NMR were also used to rule out potential Li species.

The spectra from the 15 to  −15 ppm region can be viewed in figure 7. The 4M LiFSI-DME sample (figure 7(a)) contained Li₂O (2.8 ppm), Li₂S (2.4 ppm), LiOH (1.3 ppm), Li₂O₂ (−0.1 ppm), LiF (−0.9 ppm), and LiFSI (−1.8 ppm). The 1M LiFSI-DME sample (figure 7(b)) could be deconvoluted with peak that aligned well to those of the 4M LiFSI-DME sample, except that no ¹⁹F peak was present, suggesting that the  −1.1 ppm peak is a different species from LiF; a number of options were proposed in the full text. Figure 7(c) shows the spectrum for the 3M LiTFSI-DME sample, which showed similar features as the 4M LiFSI-DME sample, but without Li₂O and the presence of LiTFSI instead of LiFSI. Higher concentrations of LiFSI/DME resulted in an overall thicker SEI with fewer dead Li–metal species. This trend was observable across the three samples and explained the enhanced electrochemical performance at high LiFSI concentrations in DME. While this is one example of ex situ NMR assisting in the characterization of how Li–metal battery formulation impacts the SEI, numerous others are available in the literature [132, 217].

Zoom In Zoom Out Reset image size

5.5. Ex situ NMR studies of solvation structures

A key component to battery systems are the electrolytes responsible for energy transfer. Their observation by NMR can lead to a detailed understanding of the electronic environment and thus their molecular arrangements. While a number of electrolyte components have been mentioned previously, less abundant and recorded nuclei can provide valuable insight. The challenge associated with this, however, is the low concentration and abundance of species, such as ¹⁷O in nonaqueous carbonate electrolytes [123]. In this, Bogle et al were the first to employ natural abundance ¹⁷O NMR measurements to such electrolyte systems, demonstrating a maximum of six ethylene carbonate molecules in the Li⁺ solvation sheath accompanied by dimethyl carbonate.

Often, a confident assignment of signals for such electrolyte systems is hindered by the structural interpretation of the lines observed. Without supporting experimental data for similar structures and trends, it can be very challenging to describe the exact coordination of molecules in the electrolyte, necessitating the application of theoretical calculations. Our group’s recent work highlights the power of combining NMR spectroscopy with high-level theoretical calculations for a deeper understanding of battery-related electrolyte systems [218]. For such applications, our group often utilizes a specialized large sample volume probe established for ultra-high field (850–900 MHz) was employed [74, 219, 220]. The probe permitted up to 2 ml in sample volume, allowing for natural abundance detection of Mg at concentrations as low as 10 mM within just a couple hours of sampling (S/N ~ 60). The first reported application of the probe with respect to battery technology employed the technology for analyzing the natural abundance ¹⁷O signals in lithium-based electrolytes [74]. In this work, Deng et al collected natural abundance ¹⁷O spectra of Li-bis(trifluoromethane sulfonyl) imide (LiTFSI) at concentrations around 20 mM to elucidate the solvation structures of LiTFSI in various solvents and mixtures of the solvents. Accompanying the high-sensitivity NMR studies, DFT-based models of the solvation structures helped determine the coordination and exchange structures of Li⁺ in the solvents at various concentrations in an attempt to understand the barriers a given electrolyte might encounter during battery operation. A more recent study highlights this principle by combining electrochemical evaluation, high-field natural abundance ²⁵Mg NMR, classical molecular dynamics simulations, and DFT-based chemical shift calculations to understand the solvation structure for Mg-based battery technology [83].

Due to a rise in evidence suggesting that the solvation structure of a liquid electrolyte has a dramatic impact on the electrochemical performance of batteries, a detailed investigation of Mg-battery electrolytes was conducted by dissolving magnesium borohydride [Mg(BH₄)₂] and Mg(TFSI)₂ in diglyme (DGM) [160, 221, 222]. The rational development of improved electrolytes requires an understanding of how the components interact with each other, and as such, how the solvation structure plays an important role in the resulting properties. Mg batteries offer an attractive alternative to Li–ion designs due to the potential to triple the volumetric energy density [223]. However, a number of challenges exist in utilizing Mg as an energy carrier. One prominent drawback is the employment of an electrolyte capable of operating across a wide electrochemical window while simultaneously allowing reversible plating/stripping of Mg [30]. Since Mg forms an ionically blocking layer when exposed to oxygen, inhibition of deposition is a difficult challenge for Mg-based battery systems to overcome, driving the need for a designer electrolyte system to either prevent or reduce this interaction of the SEI [224, 225]. Halo, organo, and organo-halo salts are known to allow reversible Mg deposition, but they have limited anodic stability and conductivity [226]. A more conductive salt such as Mg(TFSI)₂ suffers from difficulties in the reversible stripping process. To better understand how continued improvements to these electrolytes can be realized, a detailed assessment of the structural aspects is required to understand the solvation environments.

Our team has investigated the solvation structures of Mg²⁺ with varying concentrations and combinations of TFSI⁻ and ${\rm BH}_{4}^{-}$ mixtures using a combined experimental and theoretical approach, chiefly employing the large sample volume probe for natural abundance ²⁵Mg NMR measurements combined with quantum chemistry calculations and molecular dynamics simulations to relate the electrolyte structures to their electrochemical performance [83]. The experimental results as well as the predicted shift accounting for molecular exchange are presented in figure 8. Figure 8(a) shows the experimental results of the natural abundance ²⁵Mg experiments from pure 0.01 M Mg(BH₄)₂ (a), mixed Mg(BH₄)₂ and Mg(TFSI)₂ electrolytes (b)–(f) to pure 0.4 M Mg(TFSI)₂ (g). The observed experimental shifts of the various counter-ion concentration ratios agree well with those predicted based on the molecular exchange principle:

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle {{\delta }_{m}}=\frac{{{c}_{1}}}{{{c}_{1}}+{{c}_{2}}}{{\delta }_{1}}+\frac{{{c}_{2}}}{{{c}_{1}}+{{c}_{2}}}{{\delta }_{2}}\nonumber \end{align}$$

where c_i is the concentration of a given species and δ_i is the chemical shift of the pure, dilute Mg salt.

Zoom In Zoom Out Reset image size

To interpret the results, molecular dynamics simulations were employed to predict likely structures of the salts at various concentrations. Analyzing the residence time and radial distribution of the molecules in the simulation provided a clearer picture of which structures were most likely and to which electrolyte components Mg shared coordination. The resulting structures were then used as models for DFT-based chemical shift calculations. This approach assisted in interpreting the NMR data to provide clear evidence on the solvation structures of the electrolytes due to the agreement between experimental and computational data.

For the pure Mg(BH₄)₂ salt dissolved in DGM, the first solvation shell contains two ${\rm BH}_{4}^{-}$ anions and one DGM as a tridentate chelating the Mg²⁺, the second shell has 5–6 DGMs. The Mg(TFSI)₂ dissolved in DGM appears to take a first solvation shell containing one DGM as a tridentate ligand and a second as a monodentate ligand. A single TFSI⁻ donates an oxygen and nitrogen atom to the Mg²⁺ cation. The second solvation shell contains ~4 DGM molecules. When the two salts are mixed, a fast exchange between the two pure component structures was discovered that proceeds through a third, intermediate structure containing one of each BH4⁻, TFSI⁻, and DGM in the first solvation shell and 5 DGMs in the second shell. This intermediate structure was classified by molecular dynamics and accurately explains the observed shifts at high salt concentrations. The efficiency of the electrochemical performance was maximized at a BH4⁻:TFSI⁻ ratio of 1:4 and was attributed to the enhanced molecular dynamics and improved stability of the TFSI identified by the combined NMR and computational approaches. This work provides a clear demonstration of the power of NMR in characterizing the effect of the electrolyte solvation structure on electrochemical performance.

Combined computational and NMR efforts have also improved scientific understanding of electrolyte mixtures in redox-flow batteries. Investigations of electrolyte stability led to the development of a mixed-acid system with a dramatic increase in energy density and stability over conventional vanadium redox-flow batteries, for which NMR and theoretical evidence were combined to suggest the effect could be attributed to improved anion selection for the ligand exchange process [139, 140]. Similar concepts were applied to lithium-based redox-flow systems. Comprehensive NMR and DFT analysis was used to design and characterize an ionic-derivatized ferrocene to be the active redox species in the electrolyte. The enhanced dynamic interactions between solvent and electrolyte led to increased solubility that improved battery performance. Characterization identified that the Li-graphite anode prevented dendritic growth and SEI instability to promote stability over many charge-discharge cycles [227]. Indeed, the combination of NMR with DFT calculations of the chemical shift has deepened the pool of knowledge relating to electrolyte structures, but the dynamics of these processes are important too. Briefly, the role of NMR and MD calculations in elucidating battery science will be mentioned.

Since solvation structures are in constant motion and chemical species in electrolytes are diffusing, pulse-field gradient (PFG) approaches have also been employed to elucidate the solvation structure and dynamics [228, 229]. This technique has been used to show the strong relationship between cationic transference number and diffusivity [230]. Robust efforts to predict specific diffusion constants have also been proposed based on extensive NMR efforts [231]. Due to the empirical nature of these models, a more theoretical approach could provide a deeper understanding of diffusion in electrolyte systems. To support measurements of structure and diffusion, computational MD predictions of the diffusion constants can be well-matched with experimental values [232]. In this way, computational efforts that are not DFT-based have also played an important role in characterizing electrolyte systems.

5.6. Magnetic resonance imaging

Given the importance of batteries in modern society, a deep understanding of these chemical systems at the molecular level is of paramount importance for continued technological advancement. Understanding the failure mechanisms and properties of different batteries will direct future research efforts to solving the key challenges in capacity, longevity, and efficiency. Herein, we have outlined the technological advances and techniques that make such discoveries possible. In situ NMR is truly a powerful tool for battery research and has been used as such for about two decades to better understand these electrochemical systems. We have shown the unique advantages NMR processes over other techniques and the importance of in situ measurement where some species can only be observed when current is applied to the real, functioning battery. Additionally, we have shown the unique advantages of conducing ⁶Li, ¹⁷O, and ²⁵Mg NMR at the ultra-high field of 850 MHz for obtaining the detailed chemical compositions in the cycled electrodes and the solvation structures of ions in electrolytes by coupling with molecular dynamics studies and DFT NMR calculations. Continued progress will be made in the area where further research will be focused on applying the newest cell designs with O-ring sealing for handling elevated temperature and pressure and 3D printed bodies for long-run studies of battery cycling. Future efforts to enhance the signal for detection of less abundant nuclei in these batteries is also an area of interest and may be realized by applying additional cells in parallel. Other sensitivity enhancement techniques, notably, the technique of dynamic nuclear polarization (DNP) that can potentially increase an NMR signal by up to 2–4 orders of magnitude via transferring paramagnetic electron polarization from a polarization agent by taking advantage of the much larger electron Zeeman interaction [238–240], may further help to advance the application of NMR in energy storage research. One final area that lacks extensive literature is the detailed time scale information on various processes that NMR can probe [232]. Such information will provide unprecedented insight into the interaction mechanisms and molecular motion at play within these systems. Regardless of the focus of these next developments, NMR will clearly maintain importance in understanding the electrochemical process as new battery systems are developed that progressively improve our current technology as the iterative approach dictates.

This work was supported by the Joint Center for Energy Storage Research (JCESR FWP # 63076), an Energy Innovation Hub funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences (BES). Pacific Northwest National Laboratory is operated by Battelle for the U.S. Department of Energy under Contract DE-AC05-76RL01830. The authors are very grateful to Ms Pamela M Warren for her help with the polishing of grammar and usage throughout the text.