Microfluidic Overhauser DNP chip for signal-enhanced compact NMR

Nuclear magnetic resonance at low field strength is an insensitive spectroscopic technique, precluding portable applications with small sample volumes, such as needed for biomarker detection in body fluids. Here we report a compact double resonant chip stack system that implements in situ dynamic nuclear polarisation of a 130 nL sample volume, achieving signal enhancements of up to − 60 w.r.t. the thermal equilibrium level at a microwave power level of 0.5 W. This work overcomes instrumental barriers to the use of NMR detection for point-of-care applications.


Microwave resonator design considerations and characterisation
The unloaded resonance frequency f 0 of a microstripline relates to its length as for length L, wavelength λ , speed of light c, and integer mode number m ∈ {1, 2, · · · }. The unloaded quality factor Q 0 for a microstripline resonator can be determined from with propagation constant β = 2π/λ , and total attenuation constant α accounting for the predominant loss mechanisms involved, here the dielectric loss α d and conductor loss α c . The attenuation due to conductor loss can be estimated by and surface resistivity with µ 0 = 4π × 10 −7 H m −1 the permeability of free space. For a microstrip resonator made from gold with an electric conductivity σ Au and an operating frequency of f 0 = 14 GHz, the attenuation due to conductor loss is determined to be α c ≈ 1.7 Np m −1 . The attenuation due to dielectric loss can be estimated by which results, for the substrate properties given in Table S2, in α d ≈ 2.44 Np m −1 . As determined by Equation (2), the unloaded qualty factor is expected to be Q 0 ≈ 75. The feed line capacitively excites the λ /2-resonator via a narrow air gap of nominal width C gap = 25 µm. The gap is folded and forms a distributed interdigital capacitance C κ . By adjusting the geometry of the coupling gap, the value of C κ and therefore key parameters of the resonator can be tailored. From transmission line theory 1 the value for C κ , required for critical coupling (κ ≈ 1), can be determined as follows For the above mentioned operating conditions and Q 0 ≈ 75, the desired coupling capacitance is found to be C κ ≈ 0.03 pF. As apparent from Equation (6), C κ has a strong impact on the resonator's coupling coefficient κ, resonance frequency, as well as the impedance matching. For example, for decreasing quality factors, the resonator needs to be coupled tighter, i.e., increasingly higher values of C κ are necessary in order to meet the condition of κ ≈ 1 (assuming f 0 and Z 0 remaining constant). However, the practically more relevant effect is a significant change of the resonator's unloaded resonance frequency, as determined via Equation (1), once the resonator is capacitively coupled to the feed line. The required capacitance of C κ ≈ 0.03 pF can be translated into values for C gap , C w , C l for a corresponding gap geometry by the following equations 1 in which C 0 denotes the coupling capacitance of a plain, non-interdigitated, end-coupled resonator. For an interdigital coupling capacitor composed of N = 3 fingers with an approximated finger width of C w = 137 µm and a minimal coupling gap width of C gap = 25 µm, the required finger length can be estimated by Equations (7) to (9) to be C l ≈ 80 µm. Table S2 summarises the estimates of the most important design variables for the resonator, in particular the gap geometry.
To investigate the EM field properties of the resonators, driven numerical EM simulations were performed. The simulation model included the microstrip feed line and the coupling gap structure on a 0.3 mm thick glass substrate, as well as the MW resonator made from 3 µm thick gold. For accurate results, the SMA connector and the MW fixture had to be included into the model, as well as loss effects for all involved materials (tan δ glass = 0.01, tan δ teflon = 0.001). While sweeping model parameters, the distance between the centre of the resonator and the signal launch point at the substrate's edge was kept constant at 35 mm. For continuous MW excitation, a 50 Ω wave port, defined on the front facet of the coaxial connector, was used. The whole structure was placed inside a box-shaped domain to model the surrounding air, whereas absorbing boundary conditions suppressed any back scattering. A free tetrahedral mesh was employed and iteratively refined to resolve the structure accurately, e.g., for small geometrical features such as the interdigitated finger of the distributed coupling capacitor. Figure S2 compares the EM field distributions, at resonance, of the two resonator designs. Shown are the electric field magnitude as well as the H x -field, i.e, the x-component of the MW magnetic field, being orthogonal to the DC B 0 -field of the permanent magnet.

Design parameters
For the NMR resonator, the material and geometrical design parameters employed are listed in Table S1, and for the EPR resonator in Table S2.  Table S2. Estimated values for the design variables of a type 1 λ /2-resonator featuring a three finger interdigital coupling capacitor. The following assumptions were made: critical coupling κ = 1, characteristic impedance of Z 0 = 50 Ω, a resonance frequency of f 0 = 14 GHz, a transmission line width of w = 411 µm.

5/7 3 Experimental setup and procedure
The following additional components are permanently used for the characterisation experiments: Isolator (PE8305, Pasternack, USA), the circulator (K36-1FFF, Aerotek, Thailand), with a 50 Ω termination at port 3, a wide band (5.9 GHz to 18 GHz) MW amplifier (ZVE-3W-183+, 33.5 dB gain at 14 GHz, Mini-Circuits, USA), the shim current driver, as well as a lock-in amplifier. For MW irradiation a solid-state MW sweep oscillator (HP8350B, RF plug-in 0.01 GHz to 26.5 GHz) is employed. To ensure correct MW power level-settings, the nominal output power range (−5 dBm to 20 dBm) of the MW generator, with and without the MW amplifier, was referenced by a MW power meter (ML2495A + MA2442D sensor, Anritsu, Japan). In combination with a set of MW attenuators (6 dB Macom 2082-6041-6, 10 dB Suhner 661019AA, 20 dB Narda microPad 4749-20) output powers within a range of 0.2 mW to 3.45 W can be achieved. A one-port reflectometer setup (see gray lines in the manuscript Figure 5 (c)) allows to determine the MW resonator's frequency by using a digital oscilloscope in xy-mode. The start and stop timings for switching/gating the MW irradiation on/off are controlled by the NMR spectrometer console (Avance III, Bruker, Germany), by connecting its TTL trigger output signal to the "pulse-in" input port of the MW source. Typical MW pulse lengths for steady state ODNP experiments were set to values between 300 ms to 1000 ms, with rise/fall switching times as specified for the MW source of 15 ns. The NMR RF coil is impedance matched and frequency tuned, by a remote capacitive tuning/matching PCB, using fixed as well as variable trimmer capacitors. Impedance matching is achieved by symmetric, serial capacitors of approximately 82 pF on each side. To achieve resonance at around 20 MHz, a total capacitance, parallel to the RF coil, of approximately 320 pF is necessary.
The setup is EPR-ready and contains an optional reference arm, including a MW phase-shifter, a variable attenuator, a zero-bias Schottky detector diode (R451533000, Radiall), as well as an EPR modulation coil. All parts of the MW bridge are connected by semi-rigid, handformable 50 Ω MW coaxial cable (Sucoform 141, Huber+Suhner, Switzerland).

ODNP measurement procedure
In the following the procedure and prearrangements for a typical ODNP experiment are described.
1. For reproducible experiments, particularly for very long measurements, the permanent magnet should be preheated above ambient temperature as described in the Permanent magnet Section of the manuscript (time duration until thermal equilibrium is approximately 70 min). This ensures similar B 0 -field values before each measurement and reduces temperature induced drift over time.
2. As the thermal NMR signal intensities are very low -due to the low field strength of the magnet and low sample volumes (typically 130 nL) -it is very important to shield out external noise from being picked up by the RF coil, in order to prevent the low NMR signals from being obscured by the presence of an excessive noise floor in the recorded spectra. Therefore, the magnet as well as the environmental box (encased by copper) including all exposed RF feed structures, such as the tuning and matching circuit, must be shielded and properly connected (no ground loops) to the ground potential of the NMR RF preamplifier.
3. In order to check the NMR detection of the system, an NMR test measurement can be performed. The xy-linear stages are employed to accurately position the sample at the sweet spot of the magnet. The NMR pilot measurements are ideally performed on microfludic chips featuring a large sample reservoir providing sufficient spin concentration.
4. The permanent magnet's field coils need to be connected to the amplified auxiliary port of the lock-in amplifier, to provide the DC B-field sweep, as known from CW EPR experiments. For phase sensitive EPR detection the modulation coils are connected to one of the signal output ports of the lock-in amplifier. For details, see the main manuscript, Section ODNP setup and signal processing.
5. The MW source is timed and triggered by the NMR console. Therefore, it is necessary to connect the trigger lines from the NMR console to the MW source's blank input port and to include the corresponding trigger command line inside the NMR pulse program (see the main manuscript, Section ODNP setup and signal processing, for details).
6. A MW resonator chip with known unloaded resonance frequency f 0 is selected and loaded with sample solution. Due to the change in the dielectric environment of the resonator upon sample loading, the resonator's resonance frequency shifts (the change in frequency ∆ f L depends on the sample properties and the type of fluidic chip (e.g., relative permittivity, radical concentration, diameter and shape of the sample reservoir); typical ∆ f L for aqueous solution is about −300 MHz) and needs to be re-measured (temporarily unplug the trigger lines from the "pulse-in" port at the MW source, see reflectometer setup (gray lines) in Section Material and Methods). For a centered quasi EPR spectrum (MR intensity vs. ∆B 0 ) the resonance frequency of the chip is ideally fine-tuned (see tuning/matching strip, Fig S3) as close as possible to the frequency-equivalent of the magnet's static B 0 -field value (typically around 13.84 GHz).
7. Carefully, insert the MW resonator chip into the RF coil and attach the MW fixture via two screws to the mount module. Connect the semi-ridged MW coaxial cable to the MW SMA input of the resonator and position the probe head at the sweet spot of the magnet using the linear stages.
8. The equivalent 1 H NMR Larmor frequency is calculated from the set MW resonance frequency f L , in order to set the NMR transmitter frequency at the NMR console by typing the command SFO1 followed by the frequency value.
9. Enter the "wobble-mode" in TopSpin™ by issuing the command wobb to tune and match the NMR coil at set f n , by manually trimming the variable capacitors of the RF coil's tuning and matching board.
10. Set up and check the connections of the MW bridge as shown in the main manuscript (Section ODNP setup and signal processing). Set the MW output power level at the source to 20 dBm. Switch on the MW signal output of the MW source by pushing the "RF button" (note that the MW sweeper is not yet transmitting a signal, as the blank input port of the device is active, see the main manuscript, Section ODNP setup and signal processing for details).
11. Setting up the NMR acquisition parameters: In TopSpin™ the command edpa is entered to set the pulse acquisition parameters. The spectral width parameter of initial experiments is set relatively large, in order to cover a sufficiently broad spectral region. A single π/2-pulse experiment is issued by typing the command zg, followed by the command sequence epf, apk, abs, dpl1 for post-processing of the spectrum. Set time delays and duration for DNP.
12. The command gs is issued to enter the gradient shim module, in order to interactively observe each acquired spectrum of a single shot NMR π/2-pulse experiment. As the MW experiments are acquired and spectra are continuously updated, the electrical current and hence the magnetic field strength of the magnet's auxiliary coil is iteratively changed by small increments until the resonance condition for EPR, manifesting itself in a large (negative) hyperpolarised NMR signal peak. The electrical current is fine-tuned for a maximal NMR signal peak intensity.
13. Once an NMR signal becomes prominent above noise, the single-shot signal is observed via the interactive gradient shim window in TopSpin™ and shimmed using the implemented five channel shim coils. The electrical current through each shim coil is fine tuned via the precision potentiometers at the shim current driver for maximal NMR signal intensity.
14. For high-power ODNP experiments, the MW amplifier is added to the MW bridge (see the main manuscript, Section ODNP setup and signal processing), amplifying the MW signal from the source to a maximum power level of around 3.45 W at 13.8 GHz.