Abstract
To address the need for observing the electrolyte influenced by dissolved metal ions during the charge and discharge process of lithium-ion battery, a fast-switching broadband radio frequency (RF) switch for broadband nuclear magnetic resonance (NMR) systems was designed. This design specifically aims to overcome the challenge where signals exhibiting fast relaxation-often containing critical information-are lost due to the long “dead time” of conventional NMR instruments. The switch utilizes a differential architecture paired with a single-ended drive circuit. Its innovative feature is the inclusion of common-mode chokes in the bias circuit, which suppresses switching spikes and accelerates the transition speed. Electrical testing results confirm its superior performance: within a 3 MHz–100 MHz bandwidth, the switch achieves an insertion loss of less than 0.86 dB and isolation greater than 85 dB, coupled with an input third-order intercept point (IIP3) linearity of 55.50 dBm. Crucially, the measured switching time is only 175.3 ns, and the switching spike is limited to 6.09 mV. Experimental validation with paramagnetic additives (Mn2+, Fe2+) demonstrated that the switch's short dead time preserves the signal-to-noise ratio (SNR) of short transverse relaxation time (T2) signals while a 5 μs delay reduced the SNR by over 45%, underscoring the practical importance of fast switching. In practical NMR experiments using LiPF6 electrolyte, the switch successfully detected the free induction decay (FID) signals of 1H, 7Li, and 19F nuclei. The resultant short switching time minimizes the NMR front-end dead time, which is highly advantageous for detecting samples with short T2 relaxation and broad spectra.
Keywords: Broadband RF switch, Nuclear magnetic resonance (NMR), Lithium-ion batteries, Dead time, Multi-nuclear detection
Graphical abstract
Design a Fast Broadband RF Switch for Multi-Nuclear NMR in Lithium-Ion Batteries.
1. Introduction
The electrochemical stability of lithium-ion batteries is intrinsically linked to the complex decomposition pathways and solvation dynamics of the electrolyte. To fully elucidate these degradation mechanisms and transport properties, a holistic observation of distinct chemical environments is required, specifically involving the solvent matrix (1H), the charge carriers (7Li), and the anion components (19F) [[1], [2], [3], [4]]. While nuclear magnetic resonance (NMR) spectroscopy is a powerful tool for such element-specific characterization, analyzing the dynamic evolution of electrolytes sometimes faces a significant physical barrier: the distinct relaxation behaviors of species due to dissolved metal ions from the cathode (e.g., Mn2+ and Fe2+). This paramagnetic ion accelerates nuclear spin relaxation, reducing the transverse relaxation time (T2) and leading to significant line broadening [5,6]. In conventional NMR measurements, these rapidly decaying signals—which often contain the most critical information regarding interfacial reactions and intermediate states—are frequently lost. They decay during the probe's acoustic ringing and electronic recovery time, known as the “dead time”. Consequently, a “blind spot” is created, where the broad signals of immobilized or rapidly reacting species become invisible, leaving only the sharp signals of the bulk liquid detectable.
Capturing these rapidly decaying multi-nuclear signals presents a dual challenge for conventional NMR instrumentation. First, standard high-resolution probes act as narrowband resonant circuits. They lack the agility to switch rapidly between widely separated Larmor frequencies (e.g., from 1H to 7Li and 19F) without time-consuming hardware adjustments, hindering the real-time monitoring of correlated changes across different nuclei. Second, and critical for quantitative accuracy, the detection limit is strictly governed by the instrument's dead time. To visualize the full picture of the electrolyte system—from mobile bulk components to restricted interfacial species—the utilization of broadband NMR probes featuring ultra-short dead times is indispensable. This methodology enables the seamless, concurrent detection of 1H, 7Li, and 19F nuclei while minimizing the delay between excitation and detection, thus recovering the “lost” short-lived signals essential for a comprehensive understanding of electrolyte behaviors [7,8].
The dead time in NMR systems mainly depends on the dead time of the radio frequency (RF) front-end. Fig. 1 shows the schematic block diagram of the RF front-end for broadband NMR system. A broadband RF front-end typically uses a differential design to improve transmission and reception efficiency [9,10]. In this setup, transmit switches (SWT+/−) and receive switches (SWR+/−) compose transmit-receive (T/R) switch. The power amplifier (PA) sends signals to the coil through SWT+/−, while the coil connects to the low-noise amplifier (LNA) via SWR+/−. During excitation, SWT ± open and SWR ± close to allow high-power RF signals to reach the coil. In reception mode, SWT ± close and SWR ± open to route the NMR signals to the LNA. Faster transitions minimize the RF front-end's dead time. Increased switching speed improves the SNR of captured NMR signals [11].
Fig. 1.
Schematic block diagram of differential RF front-end for broadband NMR system.
The dead time of the RF front-end is primarily governed by the switching speed of SWR+/−. Additionally, these switches must exhibit high isolation and low insertion loss. During the excitation state, the PA transmits high-power RF signals (40–60 dBm) to the coil to excite the NMR sample. In this scenario, SWR ± must exhibit sufficiently high power handling capability and isolation (>60 dB) to protect the LNA. In the receive state, the NMR signal detected by the coil (typically on the order of microvolts) needs to be delivered to the LNA. Given the extremely weak nature of NMR signals, SWR ± must exhibit low insertion loss (<1 dB) to ensure the SNR of the NMR signal is preserved.
Positive-intrinsic-negative (PIN) diode RF switches feature low insertion loss, high isolation, and excellent linearity, making them commonly employed as switching elements in NMR RF front-ends [[12], [13], [14]]. PIN diode switches require dedicated driver circuits to provide bias current, making the design of their biasing networks the primary challenge for achieving rapid switching transitions. These bias circuits deliver direct current (DC) bias while incorporating inductive components (inductors or ferrite beads) in series to prevent RF signal interference with DC biasing paths [15]. These components limit the rate of bias current changes, thereby increasing switching time. Additionally, the energy storage properties of inductors induce voltage spikes during rapid switching transitions [16], which tend to saturate the receiver and extend the dead time [17]. Although symmetric positive/negative drive configurations can mitigate spikes and enhance switching speed, this approach introduces increased circuit complexity [13]. Moreover, single-ended RF switch architectures are fundamentally incompatible with differential NMR RF front-end designs.
This work presents a broadband RF switch aiming for the detection of the electrolytes of the batteries, based on a differential architecture with single-ended drive and integrated common-mode chokes. The key innovation lies in the choke configuration that suppresses switching spikes and accelerates switching speed while maintaining excellent broadband performance (insertion loss <0.86 dB, isolation >85 dB over 3–100 MHz) [18]. To underscore its practical significance for battery detection, we experimentally quantified the impact of paramagnetic ions (Mn2+, Fe2+) on 7Li T2 relaxation and demonstrated that deliberate switch delays as short as 5 μs can degrade the signal-to-noise ratio by over 45%. Thus, the switch's 175.3 ns dead time is crucial for capturing the fast-relaxing signals that carry information about interfacial processes in operating batteries.
2. Differential receive switch design
As shown in the schematic diagram of the RF switch section (Fig. 2), it contains a pair of differential input ports (IN+/−) and differential output ports (OUT+/−), with the DRVIN port serving as the input for driving current.
Fig. 2.
Schematic of broadband RF switch.
The switch conducts when the DRVIN port is forward-biased and disconnects when reverse-biased. In RF switches, the PIN diodes are divided into four groups, with diodes in each group connected in series in the same direction. Two differential loops each connect two groups of PIN diodes in a back-to-back series configuration. When the RF switch is turned off or powered down, this back-to-back series connection prevents high-power RF signals from simultaneously activating both PIN diode groups, thereby enhancing power blocking capability. For a single PIN diode group, the maximum blocking power (PMAX) is given by Equation (1) [15].
| (1) |
where N is the number of PIN diodes in a single group, 0 VR is the reverse voltage of the PIN diode, and Z0 (50 Ω) is the input/output impedance. This shows that increasing the number of PIN diodes and their reverse voltage enhances the maximum power blocking capability. Adding more PIN diodes also improves isolation (ISO), but the insertion loss (IL) will be increased, as described by Equation (2) and Equation (3) [15].
| (2) |
| (3) |
where Rs is the forward resistance and Ct is the capacitance of the PIN diode. Therefore, selecting the appropriate PIN diode model and quantity requires a trade-off to maintain low insertion loss while enhancing the switch's isolation and power blocking capability. In this work, the BAP65-03 PIN diode from NXP Semiconductors was selected, with VR = 30 V, Rs = 0.5 Ω, and Ct = 0.5 pF. Each group contains three series-connected diodes, resulting in a maximum blocking power of 46 dBm per group. The differential RF switch, utilizing four PIN diode groups, theoretically achieves a total blocking power of 52 dBm. Using Equation (2) with N = 3, the calculated insertion loss is 0.25 dB. The isolation calculated from Equation (3) uses f = 3 MHz, yielding approximately 70 dB.
The drive current enters through the opposite dotted terminals of L2 and exits from the remaining two ports to the center points of the four PIN diode groups. After passing through the four PIN diode groups, the drive current flows to ground via the opposite dotted terminals of L1 and L3. This configuration enables the magnetic fields generated by the drive current in the common-mode chokes to cancel each other out, thereby accelerating drive current switching speed and reducing switch spikes. Conversely, the magnetic fields induced by RF signals in the common-mode chokes reinforce each other, increasing the RF signal's impedance to ground. The common-mode chokes must maintain stable inductance across a broad bandwidth to ensure the RF switch's wide-frequency performance. This work employs Coilcraft's PFD2015-822MEC common-mode chokes, which deliver a stable 8.2 μH inductance below 100 MHz. Simulation results indicate that the three inductors contribute only 0.1–0.5 dB of additional insertion loss across the 3 MHz–100 MHz range. C1–C4 are DC-blocking/AC-passing capacitors, requiring sufficiently large capacitance values to ensure low insertion loss.
To drive the RF switch, a switch driver is designed as shown in Fig. 3. The switch driver schematic features a GATE input port and a DRVOUT port for driving current. When the GATE port receives a high-level signal, the DRVOUT port outputs forward bias, whereas a low-level signal at the GATE port triggers reverse bias output from the DRVOUT port.
Fig. 3.
Schematic of switch driver.
In the switch driver, operational amplifier U1 (LT1220, Analog Devices) detects the logic level at the GATE terminal to provide appropriate operating voltage for MOSFETs (Metal Oxide Semiconductor Field Effect Transistors) Q1 and Q2 (SQJ570EP, Vishay). The output of U1 connects to the gates of Q1/Q2 through Zener diodes D1 and D2 (BZX585-5V1, Nexperia) with a working voltage of 5.1 V, preventing the shoot-through conduction in both MOSFETs simultaneously. When U1 output exceeds 0 V, D1 clamps Q1's gate voltage to 5 V, turning Q1 off. When U1 output is below 0 V, Q2 remains off. Since the MOSFETs' gate threshold voltage is 2 V, U1's output voltage (VOUT) must exceed ±2 V to ensure proper turn-on of Q1/Q2. Concurrently, VOUT should remain below ±5.7 V to prevent reverse current flow into power supply. The relationship between VOUT and resistors R1, R4, VREF (voltage of R3) is defined by Equation (4).
| (4) |
with R4 being twice the value of R1 and VREF set to 5/3 V, a 0–5 V input is converted by U1 into a ±5 V output.
Q1 and Q2 feature a VDS rating of 100 V, protecting the MOSFETs from breakdown due to high-power RF signals originating from the RF switch. Additionally, D3 and D4 (1N4148UR-1) form back-to-back configurations with the body diodes of Q1 and Q2, respectively, further preventing high-power RF signal leakage into the power supply. The output current of the switch driver is controlled by resistor R7. When the RF switch is activated, the forward current of each PIN diode group, IF, is given by Equation (5).
| (5) |
where VBIAS = 5 V, N is the number of PIN diodes in a single group, VF (0.7 V) is the forward conduction voltage of the PIN diode, and RL (1.5 Ω) represents the DC resistance of the common-mode chokes. In this work, R7 is set to 51 Ω, resulting in an operating current of 10 mA per PIN diode.
The switching time of the RF switch is influenced by three primary factors: the switching time of the PIN diodes, the turn-on time of Q1 and Q2, and the slew rate of U1. The PIN diode switching time is given by Equation (6) [15].
| (6) |
where τ denotes the charge carrier lifetime. For the BAP65-03 diodes with τ = 0.17 μs, the calculated switching time is 17 ns. The average response and turn-on time for Q1/Q2 is 50 ns. With U1's slew rate of 250 V/μs, the transition time for ±5 V signal is 40 ns. Consequently, the theoretical switching time of the RF switch totals 107 ns. Generally, MOSFETs with a higher VDS exhibit longer turn-on times and require higher gate-drive voltages, which are the primary factors limiting the driver's switching speed. However, low-voltage-rated MOSFETs cannot withstand the high-power RF transmission encountered in NMR systems. Hence, the present design parameters represent an optimal compromise after thorough consideration of these trade-offs. To suppress switching spikes during driver current transitions, capacitor C1 (470 pF) is connected in parallel between the switch driver output and ground. This capacitance value was specifically chosen such that the 2RC time constant (with resistor R7) remains shorter than the MOSFET turn-on time, thus preventing C1's charging duration from compromising switching speed.
3. Experimental results
This paper demonstrates the broadband performance, linearity, and switching speed of the RF switch through electrical testing. The actual performance of the RF switch in broadband NMR was validated by acquiring multi-nucleus signals from LiPF6 electrolyte.
Fig. 4 shows the physical implementation of the RF switch. The 54.5 mm × 83.5 mm four-layer printed circuit board (PCB) integrates the RF switch, switch driver, and gate delay. The gate delay is composed of Analog Devices' LTC6994-1, which can achieve delays ranging from 1 μs to 33.6 s to accommodate various conditions in the RF front-end. Input/output RF ports and gate port feature U. FL connectors. ±20 V DC input power is regulated to ±12 V and ±5 V through low-dropout regulators (LDOs).
Fig. 4.
Picture of differential RF switch.
Fig. 5 presents the S-parameter measurement setup and results for the RF switch. The measurements were performed using a Keysight Technologies E5080B vector network analyzer (VNA), with the interconnection scheme between the RF switch and VNA detailed in Fig. 5 (a). Measurements were performed across 3–100 MHz with 0 dBm input power. Fig. 5 (b) illustrates the results, indicating an insertion loss ranging from 0.33 dB to 0.86 dB. Fig. 5 (c) presents the return loss, which ranges from 11.2 dB to 28.8 dB. Fig. 5 (d) demonstrates small-signal isolation exceeding 85 dB for the RF switch.
Fig. 5.
S-parameter measurement from 3 MHz to 100 MHz. (a) Testing photography. (b) Insertion loss and (c) return loss test results when switched on. (d) Isolation test results when switched off. The values in (b), (c), and (d) are shown as absolute for clarity.
The nonlinear behavior of the RF switch is characterized through input third-order intercept point (IIP3) testing, with Fig. 6 presenting the measurement setup and corresponding results. As shown in Fig. 6 (a), a signal generator (33600A, Keysight Technologies) supplied dual-tone signals at 19.2 MHz and 23.6 MHz with −5 to 15 dBm power. These signals passed through a 19.2–23.6 MHz bandpass filter and balun before entering the RF switch. The output was connected to a spectrum analyzer (N9020A, Keysight Technologies) via another balun to measure fundamental (23.6 MHz) and third-order intermodulation (28 MHz) powers. After calibrating nonlinear contributions from the filter and balun, Fig. 6 (b) shows the extrapolated IIP3 of 55.50 dBm.
Fig. 6.
IIP3 measurement. (a) Simplified schematic block diagram. (b) Measured IIP3 results obtained from fitting.
The test setup and results for RF switch switching performance are presented in Fig. 7. As shown in Fig. 7 (a), by using a differential function generator (81150A, Keysight Technologies) as the signal input of RF switch, the turn-on spike and time of each OUT port were measured using RF+ and RF− ports of the oscilloscope (MSO9104A, Keysight Technologies). DIF channel displayed the difference signal between RF+ and RF− ports. DRV port measured the output of switch driver. A gate signal controlled the turn-on and turn-off of the RF switch and triggered the oscilloscope. To ensure a clearer view during turn-on spike measurement, the outputs of the differential function generator were both closed. As shown in Fig. 7 (b), the measured results demonstrated that the peak voltage of the turn-on spike for DIF port was 6.09 mV. In Fig. 7 (c), a 20 MHz, 0 dBm differential signal was applied using the differential function generator. The turn-on time was measured at 175.3 ns, while the turn-off time was recorded at 97.0 ns in Fig. 7 (d). The turn-on time is the dead time, which determines the duration for the NMR RF front-end to switch from the RF transmission state to the reception state.
Fig. 7.
Transient characteristics test. (a) Simplified schematic block diagram. (b) Result of turn-on spike. (c) Result of turn-on time. (d) Result of turn-off time.
Based on the measurements, the RF switch's performance is summarized in Table 1, and compared to other broadband switches. The presented switch achieves 0.33 dB–0.86 dB insertion loss (3–100 MHz), >85 dB isolation, 55.50 dBm IIP3, 52 dBm max power handling, 175.3 ns switching time, and 6.09 mV switching spike. M. A. Abou-Khousa [12] reported a low-loss T/R switch, but its inductor increased driver current change time, lengthening dead time. D. N. Elluru [16] shortened dead time by increasing PIN driver voltage, yet transient spikes (∼10 V) could saturate the RX preamp. J. Xu [13] reduced the switch's insertion loss to below 0.2 dB and used a symmetrical structure to suppress switching spikes. However, the parallel symmetrical structure decreased the switch's isolation, and the more complex drive circuit made it difficult to further improve the switching speed. The PE42850 (pSemi) reaches 45.5 dBm maximum power with superior wideband insertion loss and linearity. However, its 15 μs switching delay and lower isolation make it unsuitable for direct deployment in NMR RF front-ends.
Table 1.
Comparison with the state-of-the-art RF switches.
| Reference | BW /(MHz) |
IL /(dB) |
ISO /(dB) |
Max Power /(dBm) |
Dead Time /(ns) |
IIP3 /(dBm) |
Spike /(mV) |
|---|---|---|---|---|---|---|---|
| This Work | 3–100 | 0.33–0.86 | >85 | 52 | 175.3 | 55.5 | 6.09 |
| 2024 [13] | 10–330 | 0.09–0.18 | 54.1–83 | 48.9 | 313.3 | – | 3.3 |
| 2022 [12] | 30–220 | 0.11–0.37 | 63–80 | 51 | 1900 | – | – |
| 2022 [16] | 170–470 | <0.8 | >80 | 53 | <350 | – | 10,000 |
| pSemi PE42850 | 30–1000 | 0.25–0.35 | 30–36 | 45.5 | 15,000 | 42 | – |
In this work, the NMR detection capability of the RF switch was tested under a 1.5 T superconducting magnet. Fig. 8 illustrates the actual experimental setup and its schematic diagram. Fig. 8 (a) shows the schematic block diagram of the component connections. The NMR spectrometer console and PA are located outside the shielded room, while inside the shielded room are the 1.5 T superconducting magnet, NMR probe, sample, receive unit, and two commercial RF amplifiers (MAX2611) with a gain of 18 dB each. The NMR spectrometer console controls the PA and the receive unit to achieve correct RF transmission and NMR signal reception. The NMR spectrometer console is a self-developed device from our research group, and the PA is a commercial broadband amplifier, BT00500-AlphaSA (frequency range 0.5 MHz–150 MHz, maximum power 500 W). Fig. 8 (b) is a physical photograph of the receive unit, where the RF switch and a broadband LNA are co-mounted inside an aluminum shielding box. The broadband LNA, also self-developed by our group, provides a gain of 20 dB and a noise figure of 0.6 dB within 100 MHz. Since the broadband LNA has a single-ended input, a balun (ADT2-1T+) is used to connect the differential output of the RF switch to the LNA. Two MAX2611 are cascaded after the receive unit, resulting in an overall gain of 54 dB. Fig. 8 (c) shows a physical photograph and the schematic of the probe. The probe employs a transmit-receive separate design: the transmit coil is a tuned saddle-shaped coil, while the receive coil is an untuned solenoidal coil. The saddle-shaped transmit coil has a diameter of 12 mm and a height of 20 mm; the receive solenoidal coil has an inner diameter of 5 mm, a height of 10 mm, and consists of 15 turns. Both coils are wound with 0.5 mm enameled wire and fixed inside 3D-printed frames. The sample tube has an outer diameter of 5 mm. A tuned transmit coil was used to match the commercial amplifier for improved transmission efficiency. Since the purpose of the NMR test was to verify the broadband performance of the receive switch, having only the receive path broadband is sufficient for validation. Fig. 8 (d) depicts the experimental layout inside the 1.5 T superconducting magnet shielded room. The NMR probe is placed at the center of the magnet, while the receive unit is placed on the floor away from the magnet to avoid magnetic saturation of inductive components in the circuit.
Fig. 8.
NMR experimental setup and schematic diagram. (a) Schematic diagram of the experimental setup. (b) Physical photograph of the receive unit. (c) Physical photograph and schematic of the probe. (d) Positions of the probe and receive unit inside the shielded room.
To demonstrate the multi-nuclear detection capability of the switch, a standard electrolyte for lithium-ion battery — 1 M LiPF6 EC/EMC (1 M LiPF6 dissolved in EC/EMC with the volume ratio of 3:7, EC: ethylene carbonate, EMC: ethyl methyl carbonate, purchased from Suzhou Duoduo Chemical Technology Co., Ltd.) — was employed as the test sample. This formulation allows simultaneous observation of 1H, 7Li and 19F signals. A sample tube was filled with 0.5 mL of the electrolyte. When detecting different nuclei, the transmit coil was tuned to the corresponding frequency while the rest of the setup remained unchanged. A free induction decay (FID) sequence was used to detect each nucleus, with 100 repeated acquisitions per nucleus and a repetition interval of 2 s to ensure complete recovery of the nuclear magnetization to equilibrium. The results are shown in Fig. 9. FID signals from all three nuclei in the electrolyte were correctly acquired, demonstrating the broadband receiving capability of the RF switch.
Fig. 9.
FID signals of 1H (63.87 MHz), 7Li (24.83 MHz) and 19F (60.08 MHz) in 1 M LiPF6 EC/EMC.
At last, to evaluate the effect of switching speed on electrolyte signals affected by dissolved metal ions, paramagnetic additives were introduced into the electrolyte to shorten the T2 relaxation time, thereby testing how different switch delays affect the SNR. Mn(CH3COO)2 (purchased from Shanghai Yuanye Bio-Technology Co., Ltd.), a common additive in aqueous batteries, and LiFePO4 (purchased from Shanghai Macklin Biochemical Co., Ltd.), a typical electrode material, provide paramagnetic Mn2+ and Fe2+ ions, respectively. Their effect on T2 is summarized in Fig. 10. Because Mn(CH3COO)2 is often used in aqueous electrolytes, the T2 of 7Li was measured in a 10 M LiCl (purchased from Jiangsu Aikon Biopharmaceutical R&D Co., Ltd.) aqueous solution with and without 1 M Mn(CH3COO)2. For these FID measurements, 512 accumulations were used to achieve sufficient SNR. Fig. 10 (a) shows the normalized FID of 7Li in 10 M LiCl aqueous solution, yielding a T2 of 48 ms. Fig. 10 (b) shows that after adding 1 M Mn(CH3COO)2, the T2 dropped to 406 μs. LiFePO4, on the other hand, is present as a solid electrode in real cells, with the electrolyte infiltrating the electrode pores. To mimic this situation, a sufficient amount of LiFePO4 powder was mixed into the 1 M LiPF6 EC/EMC. Due to the signal attenuation caused by the LiFePO4 powder, this comparative experiment employed 16,384 accumulations to enhance the SNR. Fig. 10 (c) gives the T2 of 7Li in the pure organic electrolyte as 51 ms, while Fig. 10 (d) shows that after adding LiFePO4 powder, the T2 shortened to 10.9 μs.
Fig. 10.
T2 relaxation times of 7Li FID signals for different lithium battery electrolytes. (a) 10 M LiCl aqueous electrolyte; (b) 10 M LiCl with 1 M Mn(CH3COO)2 aqueous electrolyte; (c) 1 M LiPF6 EC/EMC; (d) 1 M LiPF6 EC/EMC with LiFePO4 powder added.
The LiFePO4 powder produces a more pronounced T2 reduction, bringing it to about 10 μs — a regime that places stringent demands on switch performance. To evaluate how different switching times affect the NMR signal, we introduced controlled delays to the switch-control signal, thereby emulating various switching speeds. Direct comparison with other commercial switches was avoided because differences in their insertion loss, linearity, and other performance parameters would confound the isolation of switching-time effects. Since the dead time of conventional high-end NMR consoles is typically around 10 μs, three conditions were tested in the LiFePO4-doped 1 M LiPF6 EC/EMC: no added delay (intrinsic switching time 175.3 ns), a 5 μs delay, and a 10 μs delay. The complete transmit-receive sequence and the acquired FID signals (both time-domain and frequency-domain) for these three cases are presented in Fig. 11. Fig. 11 (a) corresponds to the no-delay case: the receive switch turns on immediately after the RF transmission ends. The self-developed LNA used in this work has a startup time of approximately 5 μs after the receive switch is enabled; however, this startup period does not affect the switching of the receive switch itself and therefore does not compromise the validity of the SNR comparison across different delay settings. After the LNA startup is complete, the FID is acquired, and its Fourier transform yields an SNR of 693.0. Fig. 11 (b) shows the result when the receive switch is turned on 5 μs after the transmission ends. Compared with the no-delay case, the time-domain signal amplitude drops to 58.7%, and the FFT-derived SNR is 379.1. Fig. 11 (c) shows the case with a 10 μs delay, where the time-domain signal amplitude further drops to 35.7% and the SNR decreases to 247.6. Compared to the no-delay baseline, the 5 μs and 10 μs delays reduce the SNR by 45.3% and 64.3%, respectively. These results clearly demonstrate the critical role of fast switching in preserving the SNR of short-T2 signals.
Fig. 11.
7Li FID signals received at different gating delays for 1 M LiPF6 EC/EMC with LiFePO4 powder added. (a) No delay, (b) 5 μs delay, and (c) 10 μs delay.
4. Conclusion
To address the challenges in lithium-ion batteries detection using NMR, this work designed a fast-switching broadband RF switch. Within a bandwidth of 3 MHz–100 MHz, it achieves an insertion loss of less than 0.86 dB, isolation greater than 85 dB, an IIP3 of 55.5 dBm, and a switching time of only 175.3 ns. Experiments with paramagnetic additives confirmed that the switch's short dead time is essential for preserving the SNR of short-T2 signals while a 5 μs delay reduced the SNR by 45.3% and a 10 μs delay reduced it by 64.3%. In NMR experiments on LiPF6 electrolyte, FID signals from nuclei at different frequencies 1H, 7Li and 19F were successfully detected. The short switching time enables the NMR front-end to achieve a very short dead time, which is particularly advantageous for detecting samples with short T2 relaxation and broad spectra. The RF switch in this work uses commercial common-mode chokes. Their magnetic cores are prone to saturation under strong magnetic fields, which makes the switch's insertion loss higher. As a result, the switch cannot be placed near the NMR coil to get better reception sensitivity. This drawback can be addressed in the future by designing custom air-core or ceramic-core common-mode chokes that match the specific NMR operating bandwidth and meet the inductive requirements.
CRediT authorship contribution statement
Shouquan Yao: Writing – original draft, Visualization, Formal analysis, Data curation, Conceptualization. Juncheng Xu: Writing – review & editing, Formal analysis, Data curation. Yiqiao Song: Writing – review & editing, Formal analysis, Data curation. Bingwen Hu: Writing – review & editing, Supervision, Formal analysis, Conceptualization. Yu Jiang: Writing – review & editing, Resources, Investigation, Conceptualization.
Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Yiqiao Song and Bingwen Hu are editorial board members of Magnetic Resonance Letters but were not involved in the editorial review or the decision to publish this article.
Acknowledgment
This work was supported by National Key Research and Development Program of China (2024YFE0115300) and the National Natural Science Foundation of China (No. 22272055, 22402178). Thanks to Keysight Technologies (China) Co., Ltd. Shanghai Branch Open Lab for offering 4-channel vector network analyzer (E5080B, Keysight Technologies) and support on differential S-parameter measurement.
Biographies

Shouquan Yao received his B.S. in Physics in 2018 and his M.S. in Radio Physics in 2021, both from East China Normal University. He is currently pursuing his Ph.D. at East China Normal University. His research focuses on NMR hardware, specifically on broadband NMR RF front-end hardware design and the development of lithium battery NMR detection devices.

Yu Jiang earned his master's and doctoral degrees in Radio Physics from East China Normal University in 1993 and 2002, respectively. With over 35 years of experience in magnetic resonance instrument hardware technology, he has conducted in-depth research on magnetic resonance spectrometer consoles. He has led the development of various systems, including medical magnetic resonance imaging spectrometers, single-board magnetic resonance spectrometer consoles, and handheld magnetic resonance spectrometer consoles.

Bingwen Hu, received the B.S. degree from Fudan University in 2003, the master degree from Fudan University in 2006 and Ph.D. degree from University of Lille-1 (France) in 2009. He is holding Zijiang Professorship in East China Normal University. He is currently deputy director of Magnetic Resonance Sub Association of China Association for Instrumental Analysis and senior editor of the Journal of Physical Chemistry Letters. His research interests are nuclear magnetic resonance (NMR), electron paramagnetic resonance (EPR) and their applications for lithium-ion batteries and other batteries-related materials.
Footnotes
Peer review under the responsibility of Innovation Academy for Precision Measurement Science and Technology (APM), CAS.
Contributor Information
Bingwen Hu, Email: bwhu@phy.ecnu.edu.cn.
Yu Jiang, Email: yjiang@phy.ecnu.edu.cn.
References
- 1.Pecher O., Carretero-González J., Griffith K.J., Grey C.P. Materials' methods: NMR in battery research. Chemistry of Materials. 2017;29(1):213–242. doi: 10.1021/acs.chemmater.6b03183. [DOI] [Google Scholar]
- 2.Allen J.P., Grey C.P. Solution NMR of battery electrolytes: assessing and mitigating spectral broadening caused by transition metal dissolution. The Journal of Physical Chemistry C. 2023;127(9):4425–4438. doi: 10.1021/acs.jpcc.2c08274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Xu K. Electrolytes and interphases in Li-ion batteries and beyond. Chemical Reviews. 2014;114(23):11503–11618. doi: 10.1021/cr500003w. [DOI] [PubMed] [Google Scholar]
- 4.Hayamizu K., Aihara Y., Arai S., Martinez C.G. Pulse-gradient spin-echo 1H, 7Li, and 19F NMR diffusion and ionic conductivity measurements of 14 organic electrolytes containing LiN(SO2CF3)2. The Journal of Physical Chemistry B. 1999;103(3):519–524. doi: 10.1021/jp9825664. [DOI] [PubMed] [Google Scholar]
- 5.Geng F., Yang Q., Li C., Hu B., Zhao C., Shen M., Hu B. Operando EPR and EPR imaging study on a NaCrO2 cathode: electronic property and structural degradation with Cr dissolution. The Journal of Physical Chemistry Letters. 2021;12(2):781–786. doi: 10.1021/acs.jpclett.0c03327. [DOI] [PubMed] [Google Scholar]
- 6.Liu Y., Wu X., Feng H., Lu G., Jiang Y., Lou X., Li C., Shen M., Geng F., Hu B. Interfacial deterioration in highly fluorinated cation-disordered rock-salt cathode: carbonate-based electrolyte vs. ether-based electrolyte. Energy Storage Materials. 2024;73 doi: 10.1016/j.ensm.2024.103789. [DOI] [Google Scholar]
- 7.Trease N.M., Zhou L., Chang H.J., Zhu B.Y., Grey C.P. In situ NMR of lithium ion batteries: bulk susceptibility effects and practical considerations. Solid State Nuclear Magnetic Resonance. 2012;42:62–70. doi: 10.1016/j.ssnmr.2012.01.004. [DOI] [PubMed] [Google Scholar]
- 8.Britton M.M. Magnetic resonance imaging of electrochemical cells containing bulk metal. ChemPhysChem. 2014;15(9):1731–1736. doi: 10.1002/cphc.201400083. [DOI] [PubMed] [Google Scholar]
- 9.Han V., Reeder C.P., Hernández-Morales M., Liu C. Any-nucleus distributed active programmable transmit coil. Magnetic Resonance in Medicine. 2024;92(1):389–405. doi: 10.1002/mrm.30044. [DOI] [PubMed] [Google Scholar]
- 10.Fan S., Zhou Q., Lei K.M., Martins R.P., Mak P.I. A miniature multinuclei NMR/MRI platform with a high-voltage SOI ASIC. IEEE Journal of Solid-State Circuits. 2024;60(6):2013–2024. doi: 10.1109/JSSC.2024.3485123. [DOI] [Google Scholar]
- 11.Watzlaw J., Glöggler S., Blümich B., Mokwa W., Schnakenberg U. Stacked planar micro coils for single-sided NMR applications. Journal of Magnetic Resonance. 2013;230:176–185. doi: 10.1016/j.jmr.2013.02.013. [DOI] [PubMed] [Google Scholar]
- 12.Abou-Khousa M.A., Mustapha A.A. Wideband RF transmit-receive switch for multi-nuclei NMR spectrometers. IEEE Transactions on Circuits and Systems II: Express Briefs. 2022;69(3):904–908. doi: 10.1109/TCSII.2021.3121210. [DOI] [Google Scholar]
- 13.Xu J., Yao S., Song Y., Li J., Jiang Y. A balanced T/R switch for wideband NMR front-end. IEEE Transactions on Circuits and Systems II: Express Briefs. 2024;71(5):2834–2838. doi: 10.1109/TCSII.2023.3348798. [DOI] [Google Scholar]
- 14.Ghorbani F., Dayan A., Zhou J., Huang Y., Schuchinsky A.G., Gustafsson M. Linearity enhancement in PIN-Diode-Based switches. IEEE Transactions on Microwave Theory and Techniques. 2025;73(2):789–800. doi: 10.1109/TMTT.2024.3441948. [DOI] [Google Scholar]
- 15.Microsemi-Watertown . 1998. The Pin Diode Circuit Designers' Handbook. [Google Scholar]
- 16.Elluru D.N., Awasthi A.K., Gogineni S.P., Taylor D., Shahabi A., Lemmon A.N., Chung C., Lee J. Design of an absorptive high-power PIN diode switch for an ultra-wideband radar. IEEE Journal of Microwaves. 2022;2(2):286–296. doi: 10.1109/JMW.2021.3138889. [DOI] [Google Scholar]
- 17.Lei K.M., Mak P.I., Law M.K., Martins R.P. A μ NMR CMOS transceiver using a butterfly-coil input for integration with a digital microfluidic device inside a portable magnet. IEEE Journal of Solid-State Circuits. 2016;51(10):2274–2286. doi: 10.1109/JSSC.2016.2579158. [DOI] [Google Scholar]
- 18.Reddivari R., Jena D. A negative embedded differential mode Γ-source inverter with reduced switching spikes. IEEE Transactions on Circuits and Systems II: Express Briefs. 2020;67(10):2009–2013. doi: 10.1109/TCSII.2019.2941597. [DOI] [Google Scholar]












