Coherent Evolution of Signal Amplification by Reversible Exchange in Two Alternating Fields (alt-SABRE).

Parahydrogen (pH2 ) is a convenient and cost-efficient source of spin order to enhance the magnetic resonance signal. Previous work showed that transient interaction of pH2 with a metal organic complex in a signal amplification by reversible exchange (SABRE) experiment enabled more than 10 % polarization for some 15 N molecules. Here, we analyzed a variant of SABRE, consisting of a magnetic field alternating between a low field of ∼1 μT, where polarization transfer is expected to take place, and a higher field >50 μT (alt-SABRE). These magnetic fields affected the amplitude and frequency of polarization transfer. Deviation of a lower magnetic field from a "perfect" condition of level anti-crossing increases the frequency of polarization transfer that can be exploited for polarization of short-lived transient SABRE complexes. Moreover, the coherences responsible for polarization transfer at a lower field persisted during magnetic field variation and continued their spin evolution at higher field with a frequency of 2.5 kHz at 54 μT. The latter should be taken into consideration for an efficient alt-SABRE. Theoretical and experimental findings were exemplified with Iridium N-heterocyclic carbene SABRE complex and 15 N-acetonitrole, where a 30 % higher 15 N polarization with alt-SABRE compared to common SABRE was reached.

The amplification of nuclear spin polarization or hyperpolarization has enabled many exciting applications in chemistry,[ , ] catalysis, biology and, in particular, medical diagnostics.[ , ] For the latter, the major challenge is a relatively short lifetime of the enhancement (typically <1 minute) and a one‐time administration of hyperpolarized media (bolus) to the object of interest.[ , ] Although continuous polarization was demonstrated,[ , , , ] it does not have yet a medical application, as e. g. an alternative to computed tomography angiography where rapid and multiple injections of contrast are needed. Therefore, the goal remains to polarize a given sample to the maximum, and the quest for such methods is ongoing.

Parahydrogen (pH2) based techniques provide one of the most cost‐efficient ways to enhance the MR signal with 13C polarization over 50 % and 15N polarization over 20 %. pH2 is a spin‐isomer of molecular hydrogen whose nuclear spins are in the singlet spin state with and being spin states with parallel and anti‐parallel projections of spin‐1/2.

Signal amplification by reversible exchange (SABRE) of pH2 with a substrate, S, allows to convert the para‐order into observable polarization continuously and within seconds. SABRE is a dynamic process and polarization transfer occurs only in the transient Ir‐pH2‐S complex. Still the contact time of pH2 and S, or lifetime of such Ir‐complex, , is sufficiently long to allow polarization transfer by free evolution or RF‐pulses. The most efficient “spontaneous” polarization transfers in the static magnetic field in SABRE happens at or near a level anti‐crossing (LAC) magnetic field B LAC. It was shown that the increase of temperature or the exchange rates (or decrease of ) results in broadening of the magnetic field region of effective polarization transfer and moves its maximum up‐field. At a LAC or near B LAC, the singlet state of pH2 couples with its triplet states such that the polarization is distributed between the crossed states. For 1H‐SABRE, B LAC is on the order of 6 mT (LAC of pH2 and 1H of substrate), while fields on the order of 1 μT are needed to transfer spin order from pH2 to 15N. The latter experiment was dubbed SABRE in SHield Enables Alignment Transfer to Heteronuclei (SABRE‐SHEATH).

Recently it was shown that fast and repetitive alternation (1–50 Hz) between B low∼1 μT and a field in the range of 50 μT

Figure 1

Sequence schematics. (a) Magnetic field pattern for alt‐SABRE‐SHEATH and (b) for SABRE‐SHEATH. (c) A simple free induction decay (FID) readout consisting of a pulse excitation of 1H and 15N spins (B 1) and signal acquisition at B 0≈54 μT with duration t acq. Alt‐SABRE‐SHEATH consists of a hyperpolarization stage of length t hyp, where the magnetic field is alternated between B low and B high with duration t low and t high respectively. This part is repeated n times (total hyperpolarization time t hyp). In our experiments, the hyperpolarization stage is followed by a simple FID readout. During the entire experiment pH2 is continuously supplied with a flow rate of ≈2.5 L/h for convenience. Susceptibility effects on magnetic field homogeneity are negligible that is demonstrated by a spectral linewidth <0.5 Hz at B 0≈54 μT.

Sequence schematics. (a) Magnetic field pattern for alt‐SABRE‐SHEATH and (b) for SABRE‐SHEATH. (c) A simple free induction decay (FID) readout consisting of a pulse excitation of 1H and 15N spins (B 1) and signal acquisition at B 0≈54 μT with duration t acq. Alt‐SABRE‐SHEATH consists of a hyperpolarization stage of length t hyp, where the magnetic field is alternated between B low and B high with duration t low and t high respectively. This part is repeated n times (total hyperpolarization time t hyp). In our experiments, the hyperpolarization stage is followed by a simple FID readout. During the entire experiment pH2 is continuously supplied with a flow rate of ≈2.5 L/h for convenience. Susceptibility effects on magnetic field homogeneity are negligible that is demonstrated by a spectral linewidth <0.5 Hz at B 0≈54 μT.

In SABRE, polarization transfer takes place in the active SABRE complex like [IrH2S2] (Figure 2a, SI section 1), which comprises an Ir‐complex with ligands, H2 and two equatorial substrates S. The chemical kinetics between S and [IrH2S2] can be approximated using a dissociation rate constant k d and pseudo‐first‐order association rate constant k’ a.[ , , ] The theoretical findings illustrated in Figure 2 were validated experimentally using 15N‐acetonitrile as a substrate and [IrIMesCODCl] as a precursor of active SABRE complex with IMes=1,3‐bis(2,4,6‐trimethylphenyl) imidazole‐2‐ylidene and COD=1,5‐cyclooctadiene in methanol‐H4. We expect that the method also works for other SABRE‐active substrates. The SABRE complex was activated as described before.[ , ]

Figure 2

Simulations of (alt‐)SABRE‐SHEATH experiments. (a) AA'XX’ system and SABRE exchange model.[ , ] (b) 15N SABRE‐SHEATH (Figure 1a) polarization as a function of magnetic field B hyp, hyperpolarization time t hyp and polarization build‐up at B hyp=0.7 μT. (c) 15N alt‐SABRE‐SHEATH (Figure 1b) polarization of S as a function of B low and t low for t high=100 ms and 10 ms and an example of polarization oscillations. (d) 15N alt‐SABRE‐SHEATH polarization of S as a function of t low and t high for B low=1 μT and B high=54 μT. When both time periods are short the faster oscillations are visible (oscillation frequencies are given for the example kinetics – right plots). B low=1 μT in (d) is an example, experimentally alt‐SABRE‐SHEATH at several fields was measured (SI, Figure S5–S7). System parameters used for simulations were[ , ] =2 s−1, =20 s−1, t ramp=0, n=100; the complex consisted of two protons and two 15N with a high field T 1 of 1 s and 6 s, chemical shifts are −22 ppm and 0 ppm and =−24 Hz, =0 and =−8 Hz; after dissociation of 15N its high‐field T 1 was 60 s and chemical shift 100 ppm. Two substrates S in the Ir‐complex were considered; each consisted of one 15N. Note the change of the frequency of polarization transfer in (c). Three purple lines on (b), (c), and (d) indicate the cut plane for the graphs on the right. One purple box in (d, left) describes the sector which is expanded in the middle graph.

Simulations of (alt‐)SABRE‐SHEATH experiments. (a) AA'XX’ system and SABRE exchange model.[ , ] (b) 15N SABRE‐SHEATH (Figure 1a) polarization as a function of magnetic field B hyp, hyperpolarization time t hyp and polarization build‐up at B hyp=0.7 μT. (c) 15N alt‐SABRE‐SHEATH (Figure 1b) polarization of S as a function of B low and t low for t high=100 ms and 10 ms and an example of polarization oscillations. (d) 15N alt‐SABRE‐SHEATH polarization of S as a function of t low and t high for B low=1 μT and B high=54 μT. When both time periods are short the faster oscillations are visible (oscillation frequencies are given for the example kinetics – right plots). B low=1 μT in (d) is an example, experimentally alt‐SABRE‐SHEATH at several fields was measured (SI, Figure S5–S7). System parameters used for simulations were[ , ] =2 s−1, =20 s−1, t ramp=0, n=100; the complex consisted of two protons and two 15N with a high field T 1 of 1 s and 6 s, chemical shifts are −22 ppm and 0 ppm and =−24 Hz, =0 and =−8 Hz; after dissociation of 15N its high‐field T 1 was 60 s and chemical shift 100 ppm. Two substrates S in the Ir‐complex were considered; each consisted of one 15N. Note the change of the frequency of polarization transfer in (c). Three purple lines on (b), (c), and (d) indicate the cut plane for the graphs on the right. One purple box in (d, left) describes the sector which is expanded in the middle graph.

For the sake of simplicity, the spin system of [IrH2S2] can be simplified to an AA′XX′ type 4‐spin system: AA′ are the protons from pH2 and XX′ are the 15N nuclei of the equatorial substrates. The AA’ as well as the XX’ system can be described using the singlet‐triplet basis: , , , . Subsequently, the spin state of the complete system can be described with a ket vector where K stands for the AA’ states and M for the XX’ part.

Below 1 μT there is a LAC between two states: and with an energy level splitting on the order of , where describes the J‐coupling constant for opposite 1H and 15N and for neighboring nuclei. This is the one pair of coupled spin states that is responsible for SABRE‐SHEATH polarization transfer from pH2 to 15N. The state is overpopulated when fresh pH2 binds to [Ir], and the interaction between crossing states and results in a maximum SABRE‐SHEATH polarization at a LAC (Figure 2b). In fact, because  Hz< 24 Hz the LAC for SABRE‐SHEATH (SI, Figure S9) is not as defined as in the case of 1H‐SABRE.[ , ] The analytical expressions for LACs in 3 spin systems consisting of two protons and one X‐nucleus were obtained by Eills et al.

Other energy states (and coherences) are also populated when pH2 is added and affect the spin order transfer. It is important that the total spin projection of the states is constant (e. g. 0 for and states) because scalar spin‐spin interaction cannot change the total projection of the state. This means that zero‐quantum coherences (or transitions) transfer spin order in this case.

When the magnetic field is higher or lower than B LAC, the mixing efficiency of the crossing states decreases while the rate of polarization transfer increases (or decreases) since there is no complete match of the Zeeman interaction and J‐coupling constants. alt‐SABRE SHEATH reproduces this LAC analysis (Figure 2c).

The coherently driven polarization transfer from pH2 to S near B LAC with a frequency on the order of the scalar spin‐spin coupling constants (Figure 2c) was already experimentally verified by Lindale et al. for relatively long pH2 refreshing times t high >200 ms at B high=55 μT. When t high (Figure 2c, left), all substrates and pH2 will exchange with the one in solution and [Ir] with refreshed pH2 is ready for the next cycle of polarization transfer at B low. This effect can be well reproduced using a SABRE model based on either Markov chain Monte Carlo simulations of chemical exchange,[ , ] infinite‐order perturbation approach or the modified Liouville von Neumann equation with chemical exchange superoperators used here (Figure 2).[ , ]

Interestingly, if B low B high and the magnetic field variation is short (for simplicity we assume each inerval is fast: t low, t high, t ramp =1/kd=50 ms) then the 1H−15N zero‐quantum coherences (1H15N−ZQ) created at B low are retained and continue to evolve at B high at an elevated frequency of (γ 1H B high−γ 15N B high) 1H+ (Figure 2d). Here, γ x is the magnetogyric ratio and X is the corresponding Larmor frequency with X=1H or 15N. Note that γ 15N is negative, so, the resulting frequency of the polarization oscillations is higher than 1H. Here, 1H15N−ZQs are spin orders that involve two spins: one 1H of hydride protons of [Ir] and one 15N of bound substrate. Their chemical dissociation results in a collapse of the wave function, meaning that the population of this spin order will be lost. This effect is illustrated on Figure 2c (right plot), where the oscillation amplitude decreases for long t low.

Note the difference between the two cases of alt‐SABRE‐SHEATH illustrated in Figure 2c and 2d. When t low∼1/JNH (Figure 2c), the polarization transfer primarily happens at B low. However, if t low 1/JNH, the polarization transfer still can happen, but at B high (Figure 2d). In this case, it will be additionally modulated by Zeeman interactions. In both regimes, alt‐SABRE‐SHEATH provides significantly higher polarization levels of the free substrate than SABRE‐SHEATH (simulations, Figure 2b vs. 2 c,d).

For experimental validation, we used a SQUID‐based ultra‐low field (ULF) NMR spectrometer (SI, Figure S10), allowing the simultaneous observation of all MR visible nuclei (in our case 1H and 15N). Although the chemical shift resolution is missing due to ULF detection at B 0=54 μT, the field homogeneity of the setup is sufficient to obtain J‐resolved NMR spectra with a linewidth <0.5 Hz during constant pH2 bubbling (Figure 1).

The setup was proved to be very robust for long lasting SABRE experiments, which was demonstrated through ULF‐SABRE‐correlation spectroscopy (COSY) lasting more than 8 hours straight at room temperature and pressure with continuous pH2 supply. After mixing [IrIMesCODCl], 15N‐acetonitrile in methanol‐H4 and flushing pH2 for 5 minutes at atmospheric pressure, the SABRE complex was activated, which followed from constant polarization produced in SABRE‐SHEATH experiment (SI, Figure S3).

In a common SABRE‐SHEATH experiment, the system was flushed with pH2 at a given field B hyp for a time period t hyp, and then shuttled to the observation field (Figure 1a). In our case, we simply increased B hyp after polarization to reach B 0∼54 μT for ULF NMR observation. The maximum polarization was achieved at 1.2 μT (SI, Figure S2).

In the next step, the slow alt‐SABRE‐SHEATH oscillations near B low=1 μT were demonstrated for different t high and B low (Figure 3). The amplitude of 15N (Figure 3a) and 1H signals was found to oscillate as function of t high as expected (Figure 3b). Figure 3c shows the integrated real part, ReI, of the 15N spectrum, complemented with a sinusoidal fit of the polarization oscillation frequency ν.

Figure 3

Experimental alt‐SABRE‐SHEATH 15N and 1H spectra (a, b) and integral over the real part of the 15N signal (c) as a function of t low at B low=2.6 μT. Note that both 15N and 1H signals of acetonitrile oscillate as a function of t low (these protons are neglected in the simulations). Other experimental parameters: t hyp=30 s, t high=10 ms, and B high, B 0=54 μT. The kinetics (c) were fitted by (fit, line) with fitting parameters:  Hz and  ms.

Experimental alt‐SABRE‐SHEATH 15N and 1H spectra (a, b) and integral over the real part of the 15N signal (c) as a function of t low at B low=2.6 μT. Note that both 15N and 1H signals of acetonitrile oscillate as a function of t low (these protons are neglected in the simulations). Other experimental parameters: t hyp=30 s, t high=10 ms, and B high, B 0=54 μT. The kinetics (c) were fitted by (fit, line) with fitting parameters:  Hz and  ms.

The lowest ν∼70 Hz was near B LAC∼1.2 μT was (SI, Figure S4c) and increased to ν∼119 Hz at ∼2.6 μT (Figure 3c and SI, Figure S5c) and ν∼185 Hz at ∼3.9 μT (SI, Figure S6c). In the case of short‐lived SABRE complexes, ν is equivalent to the rate of polarization transfer, meaning higher ν can achieve higher signal gain, despite the fact, that B low is not fulfilling the optimum B LAC condition.

Next, we kept both t low and t high shorter than 1/J .These settings resulted in an observation of a much faster oscillation with a frequency close to 1H+ as function of t high (Figure 2d and Figure 4). As discussed above, the frequency is equal to the 1H−15N two‐spin order zero quantum coherence at B 0. And, as expected, both 15N (Figure 4a and 4 c) and 1H signals (Figure 4b) showed these oscillations. This fast oscillation was observed for different parameter settings (SI, Figure S7 and S8). Due to the long hyperpolarization build up times (>10 s) and limited availability of the involved substances (acetonitrile, Ir‐catalyst), only a very narrow parameter space of the performed simulations could be experimentally assessed. However, these results confirm that although the polarization is generated at low fields, the evolution continues at high fields, where only refreshment of the Ir‐complex was assumed. Here, we show that the length of t high is also important for alt‐SABRE‐SHEATH hyperpolarization.

Figure 4

Experimental alt‐SABRE‐SHEATH 15N and 1H spectra (a, b) and integral over the real part of the 15N spectrum (c) as a function of t high at B high=54 μT. Note that both 15N and 1H signals of acetonitrile oscillate as a function of t high (these protons are neglected in the simulations). Other experimental parameters: B low=2.6 μT, t hyp=30 s, t low=1.5 ms, and B 0=54 μT. The kinetics (c) were fitted by (fit, line) with fitting parameters: .

Experimental alt‐SABRE‐SHEATH 15N and 1H spectra (a, b) and integral over the real part of the 15N spectrum (c) as a function of t high at B high=54 μT. Note that both 15N and 1H signals of acetonitrile oscillate as a function of t high (these protons are neglected in the simulations). Other experimental parameters: B low=2.6 μT, t hyp=30 s, t low=1.5 ms, and B 0=54 μT. The kinetics (c) were fitted by (fit, line) with fitting parameters: .

Interestingly, that with alt‐SABRE‐SHEATH spin order oscillates between 1H and 15N of acetonitrile (Figure 3 and 4), thus increasing the signal intensity for this specific nucleus. The importance of multiple quantum coherences for SABRE experiments was already shown in Ref. [31], where the COSY spectrum of a hyperpolarized spin system with multiple orders was predicted and experimentally verified. It was shown that for a conventional SABRE experiment, the spin‐order of pH2 distributes among all coupled spins: their magnetization and various zz‐spin orders were found to be populated.

We were able to probe the fast spin dynamics of the active SABRE complex and achieved a 30 % higher 15N polarization with alt‐SABRE‐SHEATH compared to common SABRE‐SHEATH of equal t hyp (SI, section 8), in accordance with the theory. This was accomplished using the sequence parameters depicted in Figure 4.

The analysis of alt‐SABRE performance showed that instead of non‐adiabatic (fast in respect to LAC frequency splitting) variation of external magnetic fields, slower ramps maybe beneficial, which will prevent such a fast and detrimental oscillation at B high. Moreover, adiabatic passages through the LAC condition were reported to be beneficial for polarization transfer.[ , ]

In parallel with this work, it was suggested to replace the (thigh, Bhigh) block with two blocks: (thigh/2, Bhigh) and (thigh/2, −Bhigh). As a result, the effect of Zeeman interaction is refocused while J‐coupling evolution is not: the polarization transfer at B high continues while fast oscillations will not cause problems for the experimental settings. This refocused type of alt‐SABRE‐SHEATH similar to refocused INEPT (INEPT+) we suggest calling alt‐SABRE‐SHEATH+.

The oscillations of 1H/15N amplitude at B low with frequency ν (close to LAC condition) and at B high with frequency 1H+ (weak 1H−15N coupling) were uncovered experimentally and confirmed by simulations. For t low, t high shorter than 1/J, we showed that a higher polarization than for common SABRE‐SHEATH can be achieved. Since we only assessed a narrow parameter space further investigations may provide yet higher polarization. The presented simulations and experiments were focused on 15N of acetonitrile. If and to what extend the polarization of other nuclei and other substrates can be boosted has to be investigated, too. Still, we expect that this approach will allow to enhance the signal of many other SABRE substrates.

Methods

SABRE‐SHEATH (Figure 1a, SI, Figure S1–S3). B hyp was varied in the range of −0.5 to 3.5 μT to find B LAC. pH2 was flushed continuously. Close to 100 % enrichment pH2 was prepared using in house built liquid helium pH2 generator.

Alt‐SABRE‐SHEATH (Figure 3 and 4, SI, Figure S4–S8). Alt‐SABRE‐SHEATH sequence (Figure 1b) has four experimental parameters, which were varied: t low, t high, B low and B high. Ramp time t ramp was kept constant and equal to 10 ms. The alt‐SABRE cycle was repeated n‐times to reach a constant hyperpolarization time. Then a 90° RF‐pulse flipped the 15N and 1H spins into the transversal plane for NMR signal observation. Since a broad band SQUID based detector was used, the 15N and 1H signals are read out simultaneously. Parahydrogen was bubbled through the reactor during the entire experiment. Since the variation of all parameters is very time consuming due to a long hyperpolarization period t hyp>10 s only some projections of this 4‐dimensional experimental space were measured (Figure 3 and 4, SI, Figure S4–S8). The alternating magnetic fields were created by a homemade current source that was controlled by an arbitrary waveform generator (model Rohde & Schwarz HM8150).

Sample preparation. The samples consisted of 40 μL acetonitrile‐15N (98 % 15N, Merck, CAS: 14149–39‐4) and 3 mg [IrIMesCODCl] (in house synthesis) catalyst dissolved in 14 mL methanol‐H4 (99 % purity). The sample was conducted into a 2 mL reactor. Due to constant pH2 bubbling the sample evaporated at a rate of ≈1 mL/h enabling a maximum measurement time of ≈10 h (signal stability, SI, Figure S3).

ULF MRI setup. The home‐made ULF MRI setup consists of a broad band SQUID based detector, sitting inside a low noise fiberglass dewar. A homogenous magnetic field is achieved by using a tetracoil with a sphere radius of 260 mm. A Helmholtz coil accomplishes RF transmission and spin rotation. The whole system sits inside a cubic shielding chamber consisting of two layers of μ‐metal for shielding low frequency magnetic fields and one aluminum layer for shielding RF fields.

The volume of the hyperpolarization reactor is about 2 mL. It has an inlet for the pH2 and in‐ and outlets to the adjacent reservoir. As soon as the sample level in the reactor decreases, new sample from the reservoir flows into the reaction chamber.

SABRE simulations. SABRE experiments were simulated using the MOIN spin library and the quantitative SABRE theory. The effective association and dissociation exchange constants for the substrate were =2 s−1, =20 s−1. The parameters of [IrH2(S=15N)2]=AA'XX’ : high‐field T 1 of 1H was 1 s for 1H and 6 s for 15N. Their respective chemical shifts were −22 ppm and 0 ppm. J‐coupling constants were =−24 Hz, =0 and =−8 Hz. The parameters of free substrate (15N) were: high field T 1=60 s and chemical shift 100 ppm.

Author Contributions

KB, ANP, NK: conceptualization, methodology, writing – original draft, preparation. KB, ANP, NK, MP: investigation. MP: [Ir] synthesis. KB, NK: experiments. ANP: simulations. KS, JB, JBH: supervision. KS, JBH, ANP: funding acquisition. KS, JB: provided the environment. All authors contributed to discussions, writing and helped interpreting the results and have given approval to the final version of the manuscript.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.