Multinuclear MR and MRI study of lithium-ion cells using a variable field magnet and a fixed frequency RF probe

Abstract

An exploratory multinuclear magnetic resonance (MR) and magnetic resonance imaging (MRI) study was performed on lithium-ion battery cells with ⁷Li, ¹⁹F, and ¹H measurements. A variable field superconducting magnet with a fixed frequency parallel-plate radiofrequency (RF) probe was employed in the study. The magnetic field was changed to set the resonance frequency of each nucleus to the fixed RF probe frequency of 33.7 MHz. Two cartridge-like lithium-ion cells, with graphite anodes and LiNi_0.5Mn_0.3Co_0.2O₂ (NMC) cathodes, were interrogated. One cell was pristine, and one was charged to a cell voltage of 4.2 V. The results presented demonstrate the great potential of the variable field magnet approach in multinuclear measurement of lithium-ion batteries. These methods open the door for developing faster and simpler methods for detecting, quantifying, and interpreting MR and MRI data from lithium-ion and other batteries.

Graphical abstract

1. Introduction

Rechargeable lithium-ion batteries (LIBs) are the preferred battery technology for powering electric vehicles [1,2]. They are also the most suitable technology for large-scale energy storage applications [3]. LIBs present high volumetric and gravimetric densities, which make them light and compact [1]. Additionally, they possess a low self-discharge rate, a long life cycle, and a fast charging rate, which are desirable features for such applications. However, improving LIB performance is essential to maximize their benefits and widen their range of applications. Optimizing LIB performance is often precluded by the complexity of the materials employed in their assembly and their complex interactions, making it challenging to analyze LIBs operationally by most analytical techniques [4,5].

Magnetic resonance (MR) and magnetic resonance imaging (MRI) have proven to be excellent tools for interrogating LIBs [[6], [7], [8]]. MR/MRI studies have provided information that helps improve LIB performance and safety [[9], [10], [11], [12]]. MR/MRI studies have also provided fundamental insights on topics such as lithium plating [8,[13], [14], [15]], lithium intercalation [16,17], electrolyte degradation [18,19], and solid electrolyte interface (SEI) formation [20,21]. Most MR and MRI studies of LIBs focus on ⁷Li experiments, leading to incomplete information about battery material interactions during cell operation.

Multinuclear MR studies of LIBs have proven valuable for the characterization and quantification of the different chemical species present in the cell [[22], [23], [24], [25]]. Combined information from different MR active nuclei will be informative; however, multinuclear MR and MRI studies of LIBs have obvious hardware complexity. The most common experimental approaches are 1) a single magnet with different RF probes for different nuclei, 2) complex multi-resonant RF probes with a single magnet, or 3) several magnet systems with single-resonance RF probes. The complexity of these approaches leads to increased experiment difficulty and precludes a better characterization and quantification of the species in the LIBs. Additionally, due to this hardware complexity, dedicated RF probes and realistic Li-ion cells are lacking for MR studies of LIBs.

We present an exploratory multinuclear MR and MRI study of lithium-ion battery cells for the nuclei ⁷Li, ¹⁹F, and ¹H. A variable field superconducting magnet with a fixed frequency parallel-plate RF probe, 33.7 MHz, was employed for the first time in LIB studies. The variable field superconducting magnet reduces the complexity of the multinuclear MR experiment. The resonance frequency of each nucleus is matched to the RF probe frequency by changing the static magnetic field. The parallel-plate RF probe was combined with a cartridge-like Li-ion cell to maximize sample volume and avoid RF attenuation. Combining the parallel-plate RF probe with the cartridge-like cell has proven well-suited for LIB studies [6]. The results presented here open new MR and MRI avenues to interrogate LIBs.

Several multinuclear studies have been performed on LIBs. Meyer et al. [24] employed a multinuclear (⁷Li, ¹⁹F) MR approach to identify and quantify LiF in the SEI. The data was collected using two separate magnet systems. A 16.5 T superconducting magnet was employed for ⁷Li studies at a ⁷Li frequency of 272 MHz, and a 14.0 T superconducting magnet was employed for ¹⁹F studies at a¹⁹F frequency of 565 MHz. Huo et al. [22] performed a multinuclear (¹H, ⁷Li, ¹⁹F, ³¹P) study to determine the composition of the SEI in a Li–FeSn₂ electrode. Their experiments were performed on a wide-bore 2.4 T superconducting magnet for ¹H, ⁷Li, and ³¹P and in a wide-bore 7.0 T superconducting magnet for ¹⁹F and ¹⁹F–³¹P cross-polarization. Wan et al. [23] also employed a multinuclear (⁶Li, ¹⁹F, ¹³C, ¹H) approach to determine the composition of the SEI in Cu/Li cells. Solid-state magic angle spinning (MAS) MR experiments were performed in a 20 T spectrometer with 4 mm double resonance pencil-like RF probes. Krachkovskiy et al. [25] also employed a double resonance pencil-like RF probe for the multinuclear (⁷Li, ¹⁹F) study of LIBs. An 8 mm diameter RF probe was employed to study the ionic concentration in the electrolyte solution during in situ battery operation. The experiments were performed in a 7.0 T superconducting magnet.

Multinuclear studies of a LIB can be complicated and time-consuming due to the complexity of the cell materials and the need for different magnets or multi-resonant RF probes to interrogate the Li-ion cell. Cell complexity involves small sample volume, low sensitivity nuclei, and RF attenuation due to metallic components of the cell. Using different magnet systems is problematic. Multi-resonant RF probes are expensive and difficult to build [26,27]. They also have different sensitivities depending on the working frequency. Another disadvantage of common multinuclear studies is the inability to interrogate an operational LIB due to hardware complexity. This complexity precludes in situ MR studies of operational LIBs, which are essential for optimizing battery performance.

In this exploratory study, two cartridge-like Li-ion cells were studied, one a pristine cell and one a charged cell, with a cell voltage of 4.2 V. Both cells were assembled with a graphite anode and a LiNi_0.5Mn_0.3Co_0.2O₂ (NMC) cathode. Although in situ and in operando studies can be performed, only ex situ (not attached to the charging system) experiments were performed to reduce the experiment time and complexity. These two cells were chosen to represent logical endpoints of battery operation. After a cell is charged, new Li species are formed, and some electrolyte is expected to be consumed or degraded. These changes are expected to be detected with the MR and MRI experiments performed.

The experiments presented here focus on measuring bulk relaxation times T₁ and T₂ and T₁-T₂ relaxation correlation. Additionally, two-dimensional (2D) SPRITE images were acquired for each nucleus in both cells. No spectroscopy experiments were performed due to the low frequency of the magnet employed.

2. Experimental

2.1. Cell assembly and charging protocol

Two identical cells were assembled. The electrodes, the separators, and the cartridge parts were dried before assembly. They were dried at 80 °C for 36 h. All LIB cells were assembled inside an argon-filled glovebox (7 ppm oxygen level, 0.5 ppm moisture level, box pressure 6.9 mbar). The separators of both cells were soaked with 1.2 mL of the electrolyte solution. Lithium hexafluorophosphate (LiPF₆) solution, 1.0 M, in ethylene carbonate (EC), C₃H₄O₃, and dimethyl carbonate (DMC), C₃H₆O₃, was employed as the electrolyte (1.0 M LiPF₆ in EC/DMC = 50/50 (v/v)) (Sigma-Aldrich, Ontario, Canada). The electrodes and the separators were cut square with an area of 9 cm². The cells were assembled with four paper filters as the separator. The final separation between the electrodes was approximately 1 mm. One cell was charged in constant current mode with a 2400 Source Meter (Tektronix, Inc., Beaverton, OR) at a low rate of C/35 (500 mA) until the cell voltage reached 4.2 V. The cell was charged over 33 h.

All cells were assembled with a graphite anode and a Lithium Nickel Manganese Cobalt Oxide cathode, NMC. Both electrodes have a capacity of ∼2 mAh/cm² (NEI Corporation, Somerset, NJ). The graphite electrode was coated on a copper current collector. The graphite had a porosity of 35 % and a thickness of 60–65 mm. The NMC electrode was coated on an aluminum current collector with an active material loading of 11.83 mg/cm² and a 70–75 mm thickness. Whatman glass microfibre filters were used as separators, Grade GF/C, 250 mm thick (Sigma-Aldrich, Ontario, Canada).

2.2. MR experiments

A variable field magnet with a maximum field of 3 T (MR Solutions, Guildford, Surrey, UK) was employed for all MR and MRI measurements. The magnet was run at three different magnetic fields, 0.791 T, 0.841 T, and 2.035 T, for ¹H, ¹⁹F, and ⁷Li studies, corresponding to a resonance frequency of 33.7 MHz for each nucleus. The switching time between selected fields was less than 45 min. The gradient coils were driven by amplifiers (Performance Controls, Inc., PA, US), providing maximum gradient strengths of 66.4 G/cm, 64.9 G/cm, and 87.8 G/cm in the x, y, and z directions.

A homemade parallel-plate resonator (PPR) [6], tuned and matched at 33.7 MHz, was employed for all measurements. The separation between the plates was 0.7 cm. The plates of the PPR are 5 cm long and 3.5 cm wide. The PPR was driven by a 2 kW Tomco RF amplifier (Tomco Technologies, Stepney, Australia). The input power for the PPR was set at 58 W. For this input power, the 90^°-pulse durations were 11 μs, 12 μs, and 27 μs, for ¹H, ¹⁹F, and ⁷Li, respectively. The quality factor Q of the PPR was unchanged at 143 for both the unloaded and loaded probe after cell insertion.

Bulk T₂ measurements were performed on both cells for each nucleus. The echo time, TE, was 400 μs for ¹H and ¹⁹F experiments and 600 μs for ⁷Li experiments. The number of echoes acquired was 9375 for ¹H, 17500 for ¹⁹F, and 11667 for ⁷Li. In all T₂ measurements, the number of averages was 128.

For T₁ measurements, the inversion recovery method was employed with 40 inversion recovery steps. The minimum recovery delay time was 0.25 ms in all measurements. The maximum recovery delay time was 7966 ms for ¹H and 11380 ms for ¹⁹F and ⁷Li. In all T₁ measurements, the number of averages was 128. The measurement time for T₁ measurement was 44 min for ¹H and 60 min for ¹⁹F and ⁷Li.

Bulk T₁-T₂ measurements were performed with the same parameters for T₁ and T₂ bulk measurements. The number of averages was 8 in all T₁-T₂ measurements. The measurement time for the T₁-T₂ relaxation correlation experiment was 66 min for ¹H, 113 min for ¹⁹F, and 102 min for ⁷Li.

2D images were acquired using a spiral SPRITE sequence [28]. The RF pulse duration was 3 μs for ¹H, 6 μs for ¹⁹F, and 9 μs for ⁷Li. A 64 × 64 k-space data set was acquired with 128 averages for all images. The image field of view (FOV) was 7 cm × 7 cm. Four interleaves were acquired in each spiral SPRITE image acquisition. The acquisition time for the 2D ¹H image was 23 min, and the acquisition time for ¹⁹F and ⁷Li images was 32 min. The encoding time was 600 μs for the ¹H and ⁷Li images and 400 μs for the ¹⁹F image. The signal-to-noise ratio (SNR) of all images was calculated in the image domain. The SNR was calculated as the ratio of the average signal intensity of the object in the image divided by the standard deviation of the noise. The average signal intensity was calculated from 576 pixels in the center of the image. The noise standard deviation was calculated from 640 pixels in the image outside the object.

3. Results and discussion

3.1. Magnet system and RF probe

The variable field superconducting magnet provides new opportunities to characterize electrochemical devices using multinuclear studies. The magnet's magnetic field can be smoothly varied from 0 T to 3 T. Switching between fields is relatively fast and straightforward, which permits the study of different nuclei by changing the magnetic field and using the same RF probe.

Employing a single magnet system and an RF probe with fixed frequency allows for better and more quantitative experiments. Different RF probes possess different B₁ magnetic field inhomogeneities and sensitivities, making accurate multinuclear nuclei quantification difficult. Imagine we have two RF probes for a conventional experiment at two different resonant frequencies, and we desire a quantitative MR experiment from each RF probe. If the homogeneous B₁ regions do not correspond in the conventional RF probes, or their deadtimes differ, or the innate sensitivities of the two RF probes are not the same, we will not have an equivalent quantitative result. These difficulties are avoided by using the same RF probe for each MR measurement at the same resonance frequency. The new variable field result is also advantageous due to sample positioning. When changing between nuclei with a variable field magnet, the sample remains in place, avoiding sample perturbation, misplacement, or possible damage. Not disturbing the sample during the MR experiment is essential for in situ and in operando studies of electrochemical devices.

Table 1 shows MR-accessible nuclei in electrochemical devices [7] that can be studied at 33.7 MHz by varying the magnet's magnetic field. The large variety of nuclei that can be interrogated with a single RF probe is the most significant advantage of using a variable field superconducting magnet. This flexibility simplifies hardware complexity and permits in situ and operando study of operational LIBs. It is also important to highlight that the first five isotopes in the table, from ¹H to ²⁷Al, are among the ten most sensitive MR nuclei. Studying these highly sensitive nuclei is advantageous because it permits less time-consuming experiments.

Table 1.

MR accessible nuclei using the variable field superconducting magnet for a resonance frequency of 33.7 MHz. The isotopes are listed in descending order based on their absolute sensitivity.

isotope	spin	n.a/%	absolute sensitivity	B0/T
1H	1/2	99.98	1.00	0.79
19F	1/2	100	0.83	0.84
51V	7/2	99.76	0.38	3.00
7Li	3/2	92.58	0.27	2.03
27Al	5/2	100	0.21	3.03
11B	3/2	80.1	0.13	2.47
23Na	3/2	100	9.25E-2	2.98
31P	1/2	100	6.63E-2	1.95
119Sn	1/2	8.59	4.44E-3	2.12
13C	1/2	1.07	1.76E-4	3.13

Although LIBs are the most reliable rechargeable battery type, current research interests are moving toward safer and cheaper replacements for lithium battery technology. Sodium-ion [29,30] and aluminum-ion [31] batteries are alternative promising battery types. Sodium and aluminum are plentiful and less expensive than lithium. There is also a great demand for large stationary energy storage, where high energy density is not required. For this type of application, vanadium redox flow batteries [32] are suitable candidates due to their stability, long life, and low-capacity degradation. All of these battery candidates may be readily interrogated using multinuclear MR and MRI.

In the current LIB study, the magnetic field was varied for ¹H, ¹⁹F, and ⁷Li studies. These are the most sensitive isotopes in the electrolyte solution. We did not measure ³¹P in the LiPF_6, given the reduced sensitivity compared to ¹⁹F. A single RF probe was employed for all the nuclei studied.

Metallic components in the LIB may attenuate the RF signal if the proper RF probe is not employed. A PPR RF probe was employed in all MR studies. This RF probe has proven to be an excellent choice for interrogating LIBs [6,13,15]. The PPR was tuned and matched at 33.7 MHz. The RF probe was untouched as one nucleus and then another was interrogated. The PPR with a B₁ magnetic field parallel to the cell's electrodes avoids RF attenuation during excitation and reception. Its flat geometry ideally accommodated the cartridge-like battery cells. The PPR works in a probe noise-dominated regime. The quality factor of the probe did not change when the cell was inserted.

Combining a variable field magnet and a single PPR RF probe reduces the complexity of the multinuclear MR experiments. The variable field magnet reduces the need for different magnet systems, multiple RF probes, or complex multi-resonant RF probes. However, this multinuclear approach might preclude multinuclear study of fast dynamic processes, such as lithium intercalation. Since in operando MR battery experiments are usually executed over time scales of hours and days, it is easy to imagine interleaving multinuclear field switching experiments given this time scale. Changing the magnetic field requires approximately 45 min.

3.2. Relaxation measurements

Bulk T₁, T₂, and T₁-T₂ relaxation measurements for ¹H, ¹⁹F, and ⁷Li were performed in this study. The measurements were performed on one pristine uncharged cell and one charged cell. The pristine and fully charged were selected for the experiments because they represent two crucial endpoints in the charging of the LIB. In the pristine cell, minimum chemical changes have occurred, while in the charged cell, significant changes are observed. These two endpoints were chosen for better comparison. T₁ and T₂ relaxation distributions and T₁-T₂ correlation maps are presented for both cells.

Fig. 1 shows the T₂ relaxation distributions for the three MR nuclei studied. Fig. 1A presents the ¹H T₂ relaxation distribution for the pristine and charged cells. Three well-defined peaks are observed in both T₂ distributions. These peaks are associated with three different ¹H populations in the cell. A strong signal from the electrolyte was expected because the two organic solvents, EC and DMC, are rich in ¹H. The peak with the longest T₂ lifetime, 700 ms, is associated with ¹H in the electrolyte solution. The middle peak, in the relaxation distribution of the pristine cell, with a lifetime of 50 ms, is associated with the silicon rubber gasket employed to seal the cell. Control experiments confirmed this. As expected, the location and amplitude of this peak did not change significantly with charging. Using a T₂ cutoff range from 5 ms to 150 ms, it was calculated that the sum of amplitudes for this peak was only 2 % different compared with the pristine cell. No difference was expected. This slight difference is due to the processing algorithm. Thus, it could be used as a reference for calibration and quantification purposes. The third peak in the relaxation distribution, with the shortest T₂ lifetime of 2 ms, is associated with ¹H in the printed circuit board (PCB) employed to build the PPR. This was also confirmed with control measurements. This peak did not change either after the cell was charged. However, for calibration and quantification purposes, the peak associated with the rubber gasket is a better choice. It is important to notice that the peak associated with the electrolyte changed when the cell was charged. The peak has a wider T₂ relaxation distribution and a shorter T₂ lifetime of 500 ms. A change in the relaxation distribution for the electrolyte was expected. During the charging process, some electrolyte degrades, forming new chemical species and the SEI. For a T₂ cutoff range from 150 ms to 1s, for the electrolyte peak, the sum of amplitudes was 17 % smaller than in the pristine cell. This significant change in amplitude was expected for the electrolyte peak. However, the result is sensitive to the selection of the T₂ cutoff range because there is not a clear separation between the rubber gasket and the electrolyte peaks.

Fig. 1.

T₂ distribution for the three different nuclei in the Lithium-ion cells. T₂ distributions are shown for the pristine uncharged cell (solid line) and charged cell (dashed line). A) T₂ distribution for ¹H. The electrolyte peak amplitude for the charged cell decreases, and the T₂ lifetime of the peak shifts to 500 ms. B) T₂ distribution for ¹⁹F. The electrolyte peak amplitude decreases for the charged cell, but the T₂ lifetime remains unchanged at 1 ms. C) T₂ distribution for ⁷Li. The T₂ lifetime of the peak for the charged cell remains unchanged at 500 ms, but the peak amplitude decreases, and the peak broadens. The echo time, TE, was 400 μs for ¹H and ¹⁹F experiments and 600 μs for ⁷Li experiments.

Fig. 1B illustrates the ¹⁹F T₂ relaxation distribution for pristine and charged cells. In both cases, two peaks are observed. The peak with the shortest T₂ lifetime of 4 ms is associated with the ¹⁹F background signal of the RF probe. To hold the RF probe, as well as, the tuning and matching circuit, a Teflon support base is embedded in the housing of the RF probe. The short T₂ lifetime signal likely comes from the Teflon. As expected, this peak remains unchanged in charged cells.

The peak with the longest T₂ lifetime is associated with the electrolyte. The inorganic lithium salt LiPF₆ is dissolved in the organic solvents EC and DMC to form the electrolyte solution. After the cell is charged, the peak associated with the electrolyte broadens to cover a wider T₂ range with a shift to shorter T₂ lifetimes. This T₂ broadening causes a reduction of the peak amplitude. A key component, LiF, of the SEI is formed during cell charging. LiF is formed by the reaction of lithium in the charged electrode with F^- in the electrolyte [21]. Due to the solid-like nature of the SEI, it is expected that ¹⁹F in the SEI will have a shorter T₂ lifetime than ¹⁹F in the electrolyte solution. It is too short to be observed in this experiment.

Fig. 1C illustrates the ⁷Li T₂ relaxation distribution for a pristine and a charged cell. A single peak associated with the lithium in the electrolyte solution is observed in both distributions. The T₂ lifetime of the peak remains unchanged with charging at 500 ms. However, the peak amplitude is significantly reduced after the cell is charged, and the peak is no longer well defined, as in the pristine cell. The relaxation distribution of the charged cell is wider and shifts toward shorter T₂ lifetimes. During cell charging, the SEI is formed. The structure of this complex layer consists of multiple inorganic and organic products from electrolyte decomposition, such as oligomers (or polymers) and lithium alkyl carbonates [21]. Thus, a wide range of T₂ lifetimes was expected.

No signal from the NMC paramagnetic cathode nor the lithium intercalated into the anode was observed. The T₂ lifetime for lithium ions in the cathode and lithium ions intercalated into the graphite anode are too short to be observed with the TE employed, 600 μs. With a shorter TE, less than 150 μs, lithium intercalated into the graphite anode can be observed [6]. No signal from metallic lithium was observed. The cell was charged slowly to avoid lithium plating and dendrite formation. However, metallic lithium, plating the anode, would be formed if a faster charging protocol were employed. A short TE acquisition would allow observation of this plating.

Fig. 2 shows T₁ relaxation distributions for the three MR active nuclei studied for both lithium-ion cells. In all cases, a single peak associated with the electrolyte is observed. T₁ relaxation distributions for ¹H and ¹⁹F are reported in Fig. 2A and B, respectively. The T₁ distribution does not change significantly with the cell charging. This result is expected because ¹H and ¹⁹F are marginally involved in cell changes when charged and discharged. However, a more visible change in the T₁ relaxation distributions for ⁷Li was expected. In Fig. 2C, a significant change in the T₁ relaxation distribution amplitude was observed after charging. A broader distribution was observed while the peak lifetime remained unchanged at around 1 s. This change in the T₁ relaxation distribution may be related to the SEI formation and its interaction with the electrolyte solution.

Fig. 2.

T₁ distribution for the three different nuclei in the Lithium-ion cell. T₁ distribution for the pristine uncharged cell (solid line) and the charged cell (dashed line) are shown. A) T₁ distribution for ¹H, B) T₁ distribution for ¹⁹F, and C) T₂ distribution for ⁷Li. The inversion recovery method was employed with 40 inversion recovery steps. No significant changes in the T₁ distribution are observed for ¹H and ¹⁹F after the cell was charged. The T₁ distribution for ⁷Li was broader, with a reduced peak amplitude after the cell was charged.

It is important to highlight that the T₁ relaxation distribution was unable to resolve the different ¹H populations in the LIB because the T₁ lifetimes were too close to each other. This makes the T₁ relaxation distribution ill-suited for ¹H species quantification in this case.

T₁-T₂ relaxation correlation techniques are well suited to discriminating different populations in complex samples. They often work well when species are challenging to resolve with a one-dimensional relaxation distribution. Fig. 3A and B show ¹H T₁-T₂ relaxation correlation maps for the pristine and charged cell, respectively. In both correlation maps, three well-defined peaks are observed. The peaks are associated with the PCB, the rubber gasket, and the electrolyte, as presented in the T₂ relaxation distribution. As expected, the peaks are better resolved in the T₁-T₂ relaxation correlation map than in the T₁ relaxation distribution. The T₁ lifetime of the electrolyte can be distinguished from the rubber gasket lifetime in the T₁-T₂ relaxation correlation map. Resolving the different ¹H populations is necessary for good quantification.

Fig. 3.

¹H T₁-T₂ relaxation correlation A) pristine, uncharged cell, B) charged cell. The echo time, TE, was 400 μs for ¹H. Three well-defined peaks are observed on both maps. They are associated with the electrolyte, the rubber gasket, and the RF probe background.

Fig. 4A and B depict the ¹⁹F T₁-T₂ relaxation correlation map for the pristine and charged cell, respectively. A well-defined peak associated with the electrolyte solution is observed. The peak with a T₁ lifetime of 1 s agrees with the T₁ and T₂ relaxation distribution results. A reduction in the amplitude of the electrolyte peak was expected after charging. Some fluorine is converted to the inorganic compound LiF, forming the SEI. The peak amplitude associated with the electrolyte decreased by 9 % after the cell was charged. The ¹⁹F T₂ relaxation distribution peak associated with the Teflon background signal is not well-resolved in the ¹⁹F T₁-T₂ relaxation correlation map due to low signal. Increasing the number of averages would help resolve the peak. However, this peak does not carry any valuable information.

Fig. 4.

¹⁹F T₁-T₂ relaxation correlation A) pristine, uncharged cell, B) charged cell. The echo time, TE, was 400 μs for ¹⁹F experiments. One well-defined peak associated with the electrolyte solution was observed in both maps. The peak elongates along the T₂ axis in the charged cell.

Fig. 5A and B illustrate the ⁷Li T₁-T₂ relaxation correlation maps for the pristine and charged cells, respectively. Fig. 5A shows a well-defined peak with a T₁ lifetime of 1000 ms and a T₂ lifetime of 500 ms. This peak is associated with Li ions in the electrolyte solution. The signal from the Li ions in the NMC cathode electrode was not observed due to the short T₂ relaxation time [33]. The NMC cathode is highly paramagnetic because it includes several paramagnetic transition metal species [34]. In Fig. 5B, no signal from the Li intercalated into graphite was observed due to the long TE employed. Previous work [6] showed that the T₂ lifetime of lithium intercalated into graphite was approximately 1 ms with a T₁ lifetime above 1 s.

Fig. 5.

⁷Li T₁-T₂ relaxation correlation A) pristine, uncharged cell, B) charged cell. The echo time, TE, was 600 μs for ⁷Li experiments. A single peak associated with lithium in the electrolyte solution was observed in both maps. The peak elongates along the T₂ axis in the charged cell.

Although no quantification for lithium species in each cell was made, a decrease in the signal intensity of the electrolyte in the charged cell in Fig. 5B was expected compared to the electrolyte signal in Fig. 5A. During cell charging, some electrolyte is irreversibly consumed in the SEI formation [35,36].

Fig. 6A and B show ¹H 2D SPRITE images for the pristine and charged cells, respectively. The images were in the plane of the LIB. The signal from the electrolyte and the rubber gasket was observed in both images. A slight difference in the electrolyte image intensity was observed between the two cells. This result is corroborated by the SNR of the image, which was 11.8 for the pristine cell and 10.6 for the charged cell. The cells were tightly sealed; thus, no electrolyte evaporation was expected during the experiment. However, for the charged cell, it was anticipated that a small quantity of the electrolyte would degrade during the charging process. Thus, a lower SNR was expected for the image of the charged cell. In both images, the signal intensity from the rubber gasket at the top and bottom of the image is reduced. This signal decreases at the top and bottom because the rubber gasket is outside the RF probe.

Fig. 6.

2D SPIRAL SPRITE images of the pristine and charged lithium-ion cells. A) ¹H pristine cell, B) ¹H charged cell, C) ¹⁹F pristine cell, D) ¹⁹F charged cell, E) ⁷Li pristine cell, F) ⁷Li charged cell. The FOV was 7 cm × 7 cm in all images, with the imaging plane coincident with the plane of the battery. A 64 × 64 k-space dataset was acquired with 128 averages for all images. Images can be easily acquired for all three nuclei.

Fig. 6C and D show ¹⁹F 2D images for the pristine and charged cell, respectively. In both images, only the signal from the electrolyte was observed. A reduction in the image intensity is expected during the charging of the cell. During the SEI formation, fluorine is irreversibly consumed, forming an inorganic compound, lithium fluoride (LiF) [21]. The SNR is low in both images. The SNR of the pristine cell was 3.6, while the SNR of the charged cell was 2.3. The results corroborate the expected difference in the SNR of the images. The artifact on the right side of the image is an aliasing artifact. This artifact is due to the signal of the Teflon in the frame of the RF probe. The ¹⁹F background signal is outside the normal FOV. This result is consistent with the short component peak observed in Fig. 1B and the partial peak observed on the T₁-T₂ relaxation correlation map, in Fig. 4. This result was confirmed with a control experiment.

Fig. 6E and F show ⁷Li 2D images for the pristine and charged cell, respectively. The SNR of both images is low. The SNR for the pristine cell was 5.2, while the SNR of the charged cell was 4.4. During cell charging, some lithium-ions are irreversibly consumed in SEI formation. This reduction can be as significant as 15 % compared to the pristine cell [35,36].

As previously discussed, no signal from lithium ions intercalated into the graphite electrode was observed due to the long encoding time. No lithium signal from the paramagnetic cathode electrode is observed due to the short lifetime in this environment. Metallic lithium formation was precluded due to the low-level current employed during charging. Thus, no signal from metallic lithium was observed.

No significant difference was observed between the two images. However, the results presented here can be further improved by increasing the number of averages and reducing the encoding time in the SPRITE method.

4. Conclusion

A variable field superconducting magnet and a single PPR RF probe were employed for a multinuclear MR/MRI study of lithium-ion cells at 33.7 MHz. The use of a variable field magnet reduces experiment complexity. This approach reduces experiment time and allows for better quantification of the results because a single RF probe is employed. A pristine and a charged lithium-ion cell were studied ex situ using bulk relaxation measurements and 2D SPRITE images. T₂ relaxation distributions and T₁-T₂ relaxation correlation measurements are suitable for differentiating cell species. This study can be expanded to other types of batteries and MR-sensitive nuclei essential for developing alternative rechargeable battery technologies. Multinuclear MR studies could also be employed during in situ and in operando studies to detect and quantify species in transient processes.

CRediT authorship contribution statement

Andrés Ramírez Aguilera: Data curation, Investigation, Methodology, Visualization, Writing – original draft, Writing – review & editing. Florin Marica: Data curation, Investigation. Kevin J. Sanders: Conceptualization, Writing – review & editing. Md Al Raihan: Resources. C. Adam Dyker: Resources, Writing – review & editing. Gillian R. Goward: Conceptualization, Funding acquisition, Writing – review & editing. Bruce J. Balcom: Funding acquisition, Supervision, Writing – review & editing, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Biographies

Andrés Ramírez Aguilera is a Ph.D. Candidate at the University of New Brunswick (UNB), Canada. His research, supervised by Professor Bruce J. Balcom at the UNB MRI Centre, explores using novel RF probe designs for MR and MRI studies of lithium-ion batteries. Before arriving at UNB, Aguilera received an M.Sc in Physics from the University of Havana in 2010 and a B.Sc in Physics from the University of Oriente in 2005 Cuba. He worked as a research associate at the Biophysics Medical Centre and Assistant Professor in the Physics Department of the University of Oriente. Email: aramire1@unb.ca.

Florea Marica is a Senior Research Scientist at the UNB MRI Centre. He specializes in designing, developing, and implementing magnetic resonance methodologies for analyzing fluid dynamics in porous media. Marica has a Ph.D. in Physics from the University of Bucharest, Romania. Email: fmarica@unb.ca.

Kevin J. Sanders received BA (2011) and M.Sc (2013) degrees in Chemistry from The Ohio State University under the supervision of Professor Philip J. Grandinetti. He obtained a Ph.D. in 2018 from the École Normale Supérieure de Lyon, France, where he studied paramagnetic inorganic materials by NMR spectroscopy in the group of Dr. Guido Pintacuda. In 2019 he moved to his current position as a post-doctoral fellow in the group of Professor Gillian R. Goward at McMaster University in Hamilton, Ontario. His research focuses on the development of new methods to study batteries by in situ NMR. Email: sandek1@mcmaster.ca.

Adam Dyker was born in Saint John, New Brunswick, and obtained his B.Sc from the University of New Brunswick in 2002. He completed his Ph.D. in the area catena-phosphorus cations under the supervision of Dr. Neil Burford at Dalhousie University. As an NSERC Postdoctoral Fellow, he studied under Dr. Guy Bertrand at the University of California, where he developed stable bent allenes. In January 2010, Adam joined the Department of Chemistry at the University of New Brunswick, where he is a full Professor. His current research interests include the development of organic redox flow batteries and the exploitation of ylidic pi-donor substituents for new advances in organic and inorganic chemistry.

Gillian R. Goward completed her undergraduate degree in Chemistry at McMaster University in 1995, and her Ph.D. at the University of Waterloo in 1999 under the supervision of Profs. Linda Nazar and Bill Power. Her doctoral thesis combined materials chemistry and solid-state NMR of lithium-ion battery materials. She was awarded an NSERC post-doctoral fellowship at the Max Planck Institute in Mainz, in the group of Prof. Hans W. Spiess, studying dynamics in proton-conducting polymers for fuel-cell applications. In 2002 she began her independent career as a faculty member at McMaster University. She leads a research team focusing on the application of advanced magnetic resonance techniques to materials for energy storage and conversion applications. Email: goward@mcmaster.ca.

Bruce J. Balcom received an undergraduate degree in Chemistry from Mount Allison University in 1985. He attended the University of Western Ontario for his Ph.D., graduating in 1990 supported by an NSERC fellowship. He pursued post-doctoral studies at Cambridge University from 1990 to 1992 with an NSERC post-doctoral fellowship. Balcom is a Professor of Physics and Director of the MRI Centre at the University of New Brunswick. He is best known for the Single-Point Ramped Imaging with T₁ Enhancement MRI method. A major focus of Balcom's research is the development of quantitative MRI methods for use in material science. He was an NSERC Steacie fellow (2000–2002) and a Canada Research Chair in Material Science MRI (2002–2023). Email: bjb@unb.ca.