Advancing homogeneous catalysis for parahydrogen-derived hyperpolarisation and its NMR applications

Parahydrogen-induced polarisation (PHIP) is a nuclear spin hyperpolarisation technique employed to enhance NMR signals for a wide range of molecules. This is achieved by exploiting the chemical reactions of parahydrogen (para-H2), the spin-0 isomer of H2. These reactions break the molecular symmetry of para-H2 in a way that can produce dramatically enhanced NMR signals for reaction products, and are usually catalysed by a transition metal complex. In this review, we discuss recent advances in novel homogeneous catalysts that can produce hyperpolarised products upon reaction with para-H2. We also discuss hyperpolarisation attained in reversible reactions (termed signal amplification by reversible exchange, SABRE) and focus on catalyst developments in recent years that have allowed hyperpolarisation of a wider range of target molecules. In particular, recent examples of novel ruthenium catalysts for trans and geminal hydrogenation, metal-free catalysts, iridium sulfoxide-containing SABRE systems, and cobalt complexes for PHIP and SABRE are reviewed. Advances in this catalysis have expanded the types of molecules amenable to hyperpolarisation using PHIP and SABRE, and their applications in NMR reaction monitoring, mechanistic elucidation, biomedical imaging, and many other areas, are increasing.


Introduction
Nuclear Magnetic Resonance (NMR) spectroscopy and imaging are widely known methods that examine nuclear spins within molecules and materials. Consequently, they provide important structural, dynamic, and spatial information about matter without Ben. J. Tickner did his PhD research (2017-2020) in the group of Simon Duckett at the Centre for Hyperpolarisation in Magnetic Resonance, University of York. His research was focused on developing new iridium SABRE catalysts to hyperpolarise target molecules such as pyruvate for applications in biomedical imaging, chemosensing, and reaction monitoring. Since then he has worked with Ville-Veikko Telkki and Vladimir Zhivonitko on hyperpolarised ultrafast Laplace NMR as a postdoctoral researcher at the University of Oulu, Finland (2020-2021). He is currently continuing work on hyperpolarisation with Lucio Frydman at the Weizmann Institute of Science, Israel.
Associate Professor Vladimir V. Zhivonitko studied coordination and organic chemistry at the Novosibirsk State University (2000)(2001)(2002)(2003)(2004)(2005). He received his PhD degree in 2008 in the group of Professor Igor V. Koptyug at the International Tomography Center, Novosibirsk, for studies of nonlinear catalytic processes using NMR imaging. He continued at ITC as a postdoc, researcher, and senior researcher in the eld of nuclear spin hyperpolarization and catalysis. Since 2017 he has worked at the NMR Research Unit of the University of Oulu. His research interests include signal enhancement in NMR, applications of NMR and MRI in catalysis and materials science, Laplace and ultrafast NMR methods.
using ionising radiation that can oen destroy the samples under examination. Moreover, they can be applied to study any state of matter: gas, liquid, or solid. For these reasons, NMR-based techniques have become a method-of-choice for many scientists to interrogate the structure and dynamic properties of small molecules, proteins, catalytic and functional materials, tissues and many other systems. 1 Despite these advantages, they are considered relatively insensitive as concentrated samples ($mM concentrations) are oen required to get strong enough signals to draw any conclusions. This insensitivity stems from the small population differences across closely spaced nuclear spin energy levels which is dictated by the Boltzmann distribution under standard thermal equilibrium conditions. This is evident from the tiny fraction of nuclear spins (e.g., 1 in every 32 000 1 H nuclei at 9.4 T) that effectively contribute to a detected NMR signal. The sensitivity issue becomes even more pronounced for nuclei that possess a smaller gyromagnetic ratio (such as 13 C, 15 N, 19 F, 31 P and many others) and at lower magnetic elds, as nuclear spin states become closer in energy. This is usually mitigated by using concentrated samples and/or time-consuming signal averaging to generate spectra of sufficient signal-to-noise ratio (SNR).
Great effort is focussed towards instrumental and methodological developments that improve the sensitivity of NMR. 2,3 As a result, approaches collectively known as hyperpolarisation have arisen from these efforts. Hyperpolarisation refers to creation of a non-equilibrium nuclear spin state with dramatically larger population differences between the energy levels relative to those for thermally polarised samples (i.e., those coming from a Boltzmann distribution). NMR signals can be orders of magnitude more intense as a consequence. 3 Hyperpolarisation is an active topic of research as it can signicantly reduce experiment times since large numbers of repetitions are no longer required to generate discernible NMR signals. The enhanced NMR signals provided by hyperpolarisation have advantages for many applications such as detecting reaction intermediates and reaction monitoring, mechanistic elucidations, analysis of low concentration mixtures, and metabolic imaging. [4][5][6][7] Several hyperpolarisation techniques are known to derive strong non-equilibrium nuclear spin polarisation via various sources. 4 For instance, Spin Exchange Optical Pumping (SEOP) 8 and dissolution Dynamic Nuclear Polarisation (DNP) 9 utilize circularly polarised light and polarisation of electron spins, respectively, to hyperpolarise nuclear spins. To a great extent, these methods are underpinned by advanced physics and have both been known for at least the last 65 years. 4 In this review, we focus on hyperpolarisation techniques that are much more chemical by nature since they utilise catalytic transformations of parahydrogen (para-H 2 ), the spin-0 isomer of H 2 . Para-H 2 serves as both a chemical reagent and the source of hyperpolarisation at the same time. Methods that exploit para-H 2 spin order as a source of hyperpolarisation are collectively termed as ParaHydrogen-Induced Polarisation (PHIP) and have been developed since the 1980s. 6,10-12 They provide a low-cost route to hyperpolarise target molecules because expensive equipment is not required and the para-H 2 feedstock is cheap and easy to produce. 13 The para-H 2 molecule is NMR silent (i.e. it does not give any NMR signal). However, breaking the symmetry of para-H 2 in a chemical reaction can lead to a dramatic enhancement of NMR signals of the reaction products. 14,15 This chemistry underpins the PHIP effect and can be hugely benecial for the detection of low concentration ((1 mM) or short-lived molecules (<1 minute) that might be challenging to observe using conventional NMR with thermal polarisation. Herein, we discuss various types of chemical reactions that can break para-H 2 symmetry and give rise to PHIP. These reactions are oen catalysed by transition metal complexes, and we discuss the role these species play in achieving PHIP. In general, many researchers are working to increase the substrate scope, efficiency, conversion rates, lifetime of hyperpolarised states, solvent tolerability, and many other properties of these catalysts which are directly linked to the NMR signal enhancements that can be produced. Therefore, we focus on recent advances in homogeneous catalyst design which have allowed the hyperpolarisation of a wider range of molecules with greater efficiency using PHIP and related SABRE 16,17 effects. We also give some key examples of the tremendous value these novel catalysts have in the areas of mixture analysis, mechanistic elucidation, reaction monitoring, and biomedical imaging. In addition to metal complexes, recent emerging metal-free catalysts are also discussed.

Para-H 2 : properties, production and PHIP
In this review we describe the properties and production of para-H 2 only briey and refer the interested reader to other works 6,[17][18][19] for specic details. Generally, at room temperature dihydrogen gas (H 2 ) exists as a mixture of ortho and para nuclear spin isomers quite accurately in a 3 : 1 ratio, and conversion between these isomers is slow (on the order of days and even longer) due to the symmetry disallowed nature of this transition. 18 H 2 can be enriched in the lower energy para isomer (para-H 2 ) via interactions with a metal centre or a paramagnetic material (catalyst) at low temperatures. [20][21][22] Para-H 2 enrichments of up to 50% can occur at liquid nitrogen temperatures (77 K) and even higher values of 98% can be achieved at 28 K. 5 This enrichment survives removal of the spin interconversion catalyst and return to ambient temperature, 13,20,22 and it is important since analogous PHIP reactions with hydrogen gas thermally equilibrated at or near room temperature does not typically produce NMR signal enhancements.
Para-H 2 has only one nuclear spin manifold and its wavefunction is denoted as (jabi À jbai)/ ffiffiffi 2 p or jS 0 i (nuclear spin singlet). Ortho-H 2 has a nuclear spin degeneracy of three in the absence of a magnetic eld, and these states can be described as jaai, (jabi + jbai)/ ffiffiffi 2 p , and jbbi, or by using the notation jT +1 i, jT 0 i, and jT À1 i (nuclear spin triplet), respectively. In this notation, a and b are spin-up and spin-down states of individual nuclei in H 2 molecules. As para-H 2 has a nuclear spin quantum number of 0, it does not respond with a signal to the radiofrequency excitation applied during NMR pulse sequences and is oen referred to as being 'NMR silent'. The symmetry of para-H 2 must be broken to allow the pair of 1 H nuclei to become observable using 1 H NMR. This can be achieved in a catalytic process in which, e.g., the 1 H nuclei of the para-H 2 molecule are incorporated into a target molecule (Fig. 1a). 9,24,25 These newly introduced protons can exhibit dramatically enhanced 1 H NMR signals. The nuclear hyperpolarisation, however, is typically short-lived and decays reaching the thermal equilibrium level due to nuclear spin relaxation which is typically characterised by the time taken to establish thermal longitudinal magnetisation, T 1 .
A typical requirement for such reactions to yield strong PHIP effects is that catalysed incorporation of para-H 2 into the target substrate must occur in a pairwise manner. In other words, both introduced 1 H nuclei should originate from the same para-H 2 molecule for them to transfer the two-spin correlation (nuclear spin order) inherited from para-H 2 to the substrate. In cases of non-pairwise addition, the spin correlation is not transferred to the substrate and, as a consequence, PHIP effects will not be observed. At the same time, PHIP can be observed for catalytic intermediates if initial para-H 2 activation by the catalyst active centre is pairwise, although the full cycle can be non-pairwise. Based on this possibility, effects such as oneH-PHIP 23-28 can be observed wherein only one of the para-H 2 originating protons ends up in the hyperpolarised product. There are also reactions in which molecules can become hyperpolarised without chemical incorporation of para-H 2 into the target structures, as takes place in SABRE. 16,17,29 These processes still contain a pairwise symmetry-breaking step in the catalytic cycle and rely on other means (e.g., level anticrossings, rf pulsing) to transfer the polarisation or the spin order to the target sites. Therefore, an understanding of the catalysis underpinning these reactions is clearly vital to rationalise and account for these differing effects.
The appearance of PHIP NMR spectra can differ if the catalytic reaction takes place directly at the high magnetic eld of the spectrometer (while the sample is inside the magnet) or outside the magnet at the Earth's magnetic eld (sometimes referred to as 'low' eld) before the sample is transferred to the spectrometer for 'high' eld NMR detection. 14,18,30,31 For instance, Fig. 1b and c shows schematically the inuence of these two different (spectrometer and Earth's) elds on the populations of the nuclear spin states of para-H 2 and the resultant reaction product(s) in the case when the para-H 2 originating protons form a weakly coupled AX spin system at high eld. These regimes have been termed Parahydrogen And Synthesis Allow Dramatically Enhanced Nuclear Alignment (PASADENA) 14 for symmetry-breaking at high eld and Adiabatic Longitudinal Transport Aer Dissociation Engenders Nuclear Alignment (ALTADENA) 32 for symmetry-breaking at low eld. Hyperpolarisation created under ALTADENA and PASA-DENA conditions may display different 1 H NMR signal shapes ( Fig. 1b and c), although both approaches can be described more generally as PHIP.

Hyperpolarised metal dihydrides: ligation and oxidative addition of para-H 2
There are different types of chemical reactions that can break the symmetry of para-H 2 in a pairwise fashion. Perhaps the simplest of these is ligation of para-H 2 to a metal centre to form an h 2 -H 2 complex. The implications of this for catalysis are that the improvement in NMR sensitivity provided by PHIP can Fig. 1 (a) Parahydrogen is 'NMR silent' and can only be observed by NMR if its symmetry can be broken in a pairwise reaction, often catalysed by a metal centre. Products containing 1 H nuclei that were originally located in para-H 2 can exhibit NMR signals orders-ofmagnitude more intense than would be recorded under 'thermal' (i.e. Boltzmann) controlled conditions. (b) Under PASADENA conditions, the para-H 2 symmetry-breaking process in a chemical reaction occurs at the spectrometer (high) field and results in population of both the jabi and jbai nuclear spin energy levels of the product. (c) Under ALTADENA conditions the reaction occurs at Earth's (low) field outside the spectrometer before insertion into the spectrometer for NMR detection. As a result, only one of the jabi or jbai levels of the product are populated. In this depiction the jabi state is shown lower in energy and populated. Note that the two protons in the product are assumed to form a weakly coupled AX spin system and the energies are not shown to scale. NMR spectra may show a different appearance when recorded using different flip angles. In these depictions, the population of each state is indicated by the thickness of the line. make detection of h 2 -H 2 complexes feasible. An excellent example is the indirect detection of a [Ni(h 2 -H 2 )(L-k 2 P,P') 2 ] 2+ L ¼ P Ph 2 N Ph2 2 complex that is impossible to discern using standard thermally-polarised 1 H NMR measurements. 33 The shortlived nature of h 2 -H 2 species usually requires advanced NMR methods (such as Chemical Exchange Saturation Transfer (CEST) and partially negative line effect) to detect them indirectly. [33][34][35] Further reaction of h 2 -H 2 complexes to form metal dihydrides can result in PHIP effects as oxidative addition of para-H 2 to a metal centre also breaks para-H 2 symmetry. 11,36-39 Activation of dihydrogen by transition metal centres has been known since the 1960s. 40,41 This typically occurs via a concerted mechanism where the newly introduced groups are located in a cis relationship. 9,10 These examples provide a fertile playground for PHIP if this activation is pairwise and faster than nuclear spin relaxation. Early examples in the late 1980s were dihydride complexes of Rh, Ir and Ru. 14,15,30,42 Many of these are short-lived intermediates in catalytic processes 27,[43][44][45] (such as hydrogenation or hydroformylation), whereas others are stable hydride species whose ligand exchange processes may be reversible or even light-activated. 36 )] and is the rst step in the hydrogenation of alkynes to alkenes using this catalyst (Fig. 2b). 31 In these contexts, the enhanced 1 H NMR signals for the hydride sites of transition metal complexes are extremely useful for enabling their detection, particularly in cases where the metal complex may be at low concentration. Namely, hydride NMR signals for catalytic intermediates such as [Rh(H) 2 Cl(PPh 3 ) 2 (styrene)], which is formed following phosphine loss from [RhCl(H) 2 (PPh 3 ) 3 ] and subsequent alkene coordination, can be made visible using PHIP. 45 This intermediate is involved in the hydrogenation of alkenes and the detection of such species can be important in determining reaction mechanisms. 7,10,11 In many of these mechanisms, activation of para-H 2 is irreversible and, as a consequence, PHIP effects can only be observed over a narrow time window in which reactivity is occurring and nuclear spin relaxation is yet to dominate. Many catalysts, particularly those based on iridium complexes, have been shown to reversibly activate para-H 2 and therefore enhanced hydride signals can be observed at will upon reaction with fresh para-H 2 . 29,35,48 This is particularly useful in mechanistic studies involving iridium dihydride complexes 38 and the detection of low concentrations of metal complexes for which enhanced hydride resonances may be diagnostic of particular ligands in complex mixtures. 38,[49][50][51] Reactivity of this type is exploited in SABRE as enhanced hydride polarisation can be transferred to other sites in the catalyst (this is discussed in more detail in Section 7). 29 Other catalysts utilise photochemistry to promote ligand loss pathways and produce enhanced hydride NMR signals. 52 (Fig. 2c). In this example, PHIP enables the formation of metal hydrides over millisecond timescales to be monitored and kinetic rate constants for this process to be determined: feats that are impossible to achieve using thermal NMR polarisation due to the need to perform time consuming signal averaging. 52 As the systems that undergo homolytic dihydrogen splitting have increased rapidly in recent years, an analogous increase in PHIP catalysts has accompanied this. Nowadays, metal catalysts based on Pd, 47,53-56 Co 26,27,57-59 and Pt 23 have all been reported to break para-H 2 symmetry via oxidative addition. We also note that metal clusters 11 and heterogeneous supported catalysts 60 have emerged as valuable catalysts that can activate para-H 2 giving rise to PHIP and SABRE effects. The heterogeneous systems have been described in detail elsewhere. 5,61,62 4. Metal-free PHIP: heterolytic para-H 2 splitting and beyond Over recent years, many metal-free catalysts have been developed to activate para-H 2 . Most of these systems are based on the use of frustrated Lewis pairs (FLPs) 63 and can split para-H 2 heterolytically in a way that also produces enhanced NMR signals for sites within the FLP. [64][65][66][67][68][69] The heterolytic activation occurs in a pairwise manner for hyperpolarised reaction products to be observed and can provide valuable insight into the mechanism of H 2 activation by FLPs. These effects were rst demonstrated using ansa-aminoborane FLPs such as 1-{2-[bis(pentauorophenyl)boryl]benzyl}-2,2,4,7-tetramethyl-1,2,3,4-tetrahydroquinoline (QCAT) (Fig. 3a). 66 Activation of para-H 2 by QCAT leads to formation of a QCAT-H 2 adduct in which the 1 H NMR signals for those nuclei originating from para-H 2 are enhanced by up to 30-fold. Polarisation transfer from these sites to 11 B nuclei was achieved using an INEPT-type pulse sequence and yielded a 10-fold enhancement in 11 B NMR signal intensity. Since then, a wider range of ansa-aminoboranes have been used for metal-free PHIP, including those activating para-H 2 under ambient temperatures in a reversible fashion. 65 Spontaneous polarisation transfer from para-H 2 derived 1 H nuclei to 11 B and 15 N sites in a 7 T spectrometer is also possible with signicant 350-fold signal enhancements recorded for 15 N sites of the FLP. 64 Activation of para-H 2 using FLPs based on P-C 68 and P-Sn 70 have also been documented.
In addition to FLPs, P-P 67,69 and As-P 69 biradicaloids, an alternative class of metal-free para-H 2 activators, have also been used in PHIP with up to 2000-fold signal enhancements for 31 P sites at 9.4 T. In this case, the activation is homolytic in nature leading to the formation of hyperpolarised biradicaloid-H 2 adducts. In addition to enhanced para-H 2 -derived 1 H sites, the hyperpolarisation of 31 P sites could be observed either spontaneously or by using rf pulsing to transfer polarisation from 1 H to 31 P, depending on the structure of the biradicaloid (Fig. 3b).
Generally, P-P, As-P and As-As biradicaloids containing fourand ve-membered rings were shown to activate para-H 2 , in many cases reversibly. 69 This reversibility allowed the observation of hyperpolarised starting para-H 2 metal-free activators (FLPs 64 or biradicaloids 69 ) without actual modication of their structures and demonstrates features of a metal-free SABRE effect. In these cases, PHIP has been used as a great mechanistic tool to study para-H 2 activation by such systems.
While current examples of metal-free systems for para-H 2 activation have been reported in organic solvents such as toluene-d 8 , 66,69 dichloromethane-d 2 , 64,65 and bromobenzene- d 5 , 68 they may in the future provide a promising route for PHIP in aqueous solvents without the use of toxic metal catalysts. Generally, metal-free PHIP using FLPs offers an alternative route to break para-H 2 symmetry that does not rely on a transition metal complex for traditional oxidative addition. Homolytic H 2 activation typically involves complex orbital overlap effects. 11 Repulsive interactions between lled orbitals on the metal centre and the H 2 ligand can be a barrier to oxidative addition that must be overcome. For some metal complexes, this barrier can be so large that heterolytic splitting of H 2 becomes feasible. 71 Therefore, examples of metal-catalysed heterolytic para-H 2 addition 34,71,72 to generate enhanced NMR signals for reaction products may emerge in the future.

Transition metal catalysed alkyne and alkene hydrogenation
Many catalytic processes involve H 2 activation as an initial step, including metal-catalysed hydrogenation or hydroformylation. Early PHIP studies that involved para-H 2 activation by square planar group 9 precatalysts of rhodium 30,42 and iridium 31,73,74 were accompanied by transfer of these newly introduced hydrides to an alkyne or alkene (Fig. 4a). If this reaction occurs rapidly relative to the relaxation rate of the hyperpolarised metal dihydride then enhanced polarisation may survive transfer into the product. 14,30,31,45 Early PHIP studies involved hydrogenation of simple alkenes and alkynes such as styrene and phenylacetylene, 15,30,75 although in more recent years PHIP has been extended to signicantly more complex molecules including amino acid derivatives 76 and even complex synthetic oligopeptides which have been of use to probe protein-ligand interactions. [77][78][79][80][81][82] Heteronuclear polarisation within the transition metal hydrogenation catalysts, or hydrogenated reaction products, can also be achieved following either dedicated pulse sequences, 74,76,83-85 spontaneous low eld transfer, 86 or eld cycling. [87][88][89] The most commonly used catalysts for these PHIP hydrogenation reactions remain based on rhodium, and to a lesser extent iridium (Fig. 4b), although there were early examples of less well-dened systems using ruthenium 15 to achieve hyperpolarised signals for hydrogenated products. Rapid hydrogenation is required to observe PHIP effects before nuclear spin relaxation destroys the hyperpolarisation. Therefore, optimisation of these catalysts is essential as the level of hyperpolarisation achieved strongly depends on factors such as hydrogenation rate. 90 Signicant efforts have been directed at the optimisation of catalysts for standard hydrogenation reactions, which usually focus on improving the catalyst turnover, reactivity and selectivity. Such efforts are not typically applied to the optimisation of hydrogenation catalysts for PHIP reactions, which oen use commercially available Rh catalysts (such as [Rh(diene)(dppb)][BF 4 ], diene ¼ norbornadiene or 1,5-cyclooctadiene) that are sometimes incapable of completely hydrogenating a target molecule in just a few seconds without the use of high H 2 pressure. 91 In PHIP reactions, the presence of paramagnetic impurities, catalyst degradation products, resting states, or transient intermediates are likely to relax hyperpolarised molecules and destroy the PHIP effect. As soon as para-H 2 is ligated to a metal centre its symmetry is broken and the spin order can then be lost before further transfer of the para-H 2 originating protons. Therefore, unwanted polarisation loss will depend on relaxation and the lifetimes of catalystbound para-H 2 species, of which there may be many in the catalytic cycle. These factors will play a role in the polarisation transfer efficiency of PHIP and are additional factors that are not typically considered in traditional hydrogenation catalysis with thermally equilibrated H 2 .
Therefore, the roles of catalyst structure and concentration on hyperpolarisation efficiency were studied for [Rh(COD)(BINAP)] [BF 4 ] and [Rh(COD)(dppb)][BF 4 ] catalysts in butyl acrylate hydrogenation using para-H 2 . 92 The authors dened a parameter called polarisation transfer efficiency (PTE) which quantied the ratio of the experimental and the idealised hyperpolarisation that could be achieved for the product (butyl propanoate) in an ensemble of product molecules; it depended on measurables such as relaxation and hydrogenation rate. For [Rh(COD)(BI-NAP)][BF 4 ], PTE increased as the catalyst concentration is lowered, and generally increased as reaction time is increased, which was related to a drop in hydrogen pressure as the reaction proceeds. The authors suggested that this effect is likely related to a shortened residence time of the para-H 2 molecule in the coordinated state on the metal centre when the alkene/hydrogen ratio decreases. 92 The shorter residence can favour high PTE since the time for relaxation of the para-H 2 spin order is reduced. The PHIP efficiency was stronger at higher magnetic elds (14 T compared to 7 T) 92 where the T 1 relaxation time of dihydrides is typically longer. 93 The change of PTE as a function of hydrogenation reaction time appeared different demonstrating rise and fall behavior when the [Rh(COD)(dppb)][BF 4 ] catalyst was used which highlights the importance the catalyst can have on the efficiency of the PHIP process. The reaction time, magnetic eld, catalyst identity and concentration, all inuence the efficiency of PHIP and must be carefully considered in conjunction with the rate, turnover and yield usually addressed in studies of traditional hydrogenation systems.
Some studies have sought to improve the Rh complexes used in PHIP by variation of ligands in the coordination sphere. A series of [Rh(norbornadiene)(bisphosphine)][BF 4 ] catalysts (1-4) were used to hydrogenate the example substrate vinyl acetate. 90 A variety of different bisphosphines ligands were screened and their effect on hydrogenation kinetics and the magnitude of PHIP NMR signal enhancements were recorded (Fig. 4c). This study demonstrated that highly electron-donating bisphosphines with dicyclohexylphosphine groups (2-4) gave higher catalytic activity with 1.8-2.0 times higher turnover frequencies compared to the standard commercially available [Rh(norbornadiene)(dppb)][BF 4 ] (1). 90 The initial precatalyst activation involves the loss of the norbornadiene ligand and its replacement with two solvent molecules (in this case, methanol). This activation step was faster using 2-4 as the rate of norbornadiene ligand loss was 3.3-4.8 times faster than the standard Rh system. 90 Similarly, hydrogenation was also faster in the case of the modied catalysts which in turn gave rise to higher NMR signal enhancements in PHIP experiments. The reason for this higher catalytic activity is yet unconrmed with the mechanism of this transformation not yet completely understood. The active [Rh(solvent) 2 (diphosphine)][BF 4 ] catalyst can coordinate substrate before irreversible oxidative addition of hydrogen. 94 However, routes in which hydrogen activation occurs before substrate coordination have also been proposed. 95 Further studies of this nature are likely to discover catalysts that can produce greater PHIP than those currently used. Investigation into hydrogenation catalytic cycles focussed on PHIP are likely to stimulate new and improved systems that retain the para-H 2 spin state more efficiently during catalytic transformations, and give rise to larger NMR signal enhancements for hydrogenated products. PHIP catalysis performed in coordinating solvents including methanol, ethanol or acetone can give high hydrogenation efficiencies and high NMR signal enhancements. 96,97 This has been linked to the relatively high para-H 2 solubility in these solvents, and is also closely related to the role of metal-solvent adducts in catalysis. Displacement of norbornadiene in [Rh(norbornadiene)(dppb)][BF 4 ] to form active [Rh(solvent) 2 4 ] can be favoured in such solvents. Readily-coordinating solvents also favour rapid elimination of hydrogenated products from the metal coordination sphere, which limits the destructive inuence of nuclear spin relaxation on PHIP of the products while bound to the metal centre. 96 Commonly-used PHIP catalysts are typically most soluble in organic solvents (such as methanol, ethanol, acetone, chloroform, dichloromethane, toluene, etc.) but exhibit limited solubility in aqueous solutions. Many have sought to increase catalyst solubility in water by modication of spectator ligands (Fig. 4d). 44 4 ] catalysts that contain monodentate tris(hydroxymethyl) phosphine (THP) ligands are also water-soluble and have been used to hydrogenate 2-hydroxyethyl 1-13 C-acrylate-d 2,3,3 yielding hyperpolarised 2-hydroxyethyl 1-13 C-propionate-d 2,3,3 . 98 Also ionic liquids based on rhodium salts have been used as hydrogenation catalysts to achieve PHIP in aqueous solvents. 73 Further development of highly active water-soluble PHIP catalysts are likely to facilitate hyperpolarisation of biologically relevant molecules, which may have applications in NMR imaging. 4,6 The use of cheaper and more readily available metals as catalysts in symmetry-breaking hydrogenation reactions has also been explored (Fig. 4e). Consequently, PHIP has been used to study the hydrogenation of molecules such as phenylacetylene 53 or diphenylacetylene 47,54 by palladium phosphine 47,54 or palladium bisimino 53 complexes. In the last few years catalysts based on cobalt such as [Co(bis(mesityl-benzimidazol-2ylidene)phenyl)(pyridine)] have been used to hydrogenate ethyl acrylate with 13 C signal enhancements of ca. 1130-fold at 11.7 T for the terminal site of the hydrogenated product in acetone-d 6 . 59 This is signicantly larger than the ca. 250-fold achieved under analogous conditions using a traditional [Rh(COD)(dppb)][BF 4 ] catalyst. 59 The scope of PHIP studies will expand as catalysts are improved to optimise their functional group tolerance, performance, speed, and many other factors. We also note that other metals such as V, Zr, Ta, Ce and Mo have been used to demonstrate PHIP with heterogeneous catalysts, 99-101 and therefore the list of possible metal-based homogeneous catalysts may be extended signicantly in the future.

Ruthenium catalysts for trans and geminal alkene hydrogenations
Most transition metal-catalysed alkyne hydrogenations involve the formation of Z (cis) alkenes and is a consequence of concerted hydrogen activation in which the introduced hydride ligands are located in a cis relationship. 40,102 This typically restrains the types of molecules that can be produced from the homogeneous hydrogenation of alkynes in a hyperpolarised state to Z alkenes. E (trans) alkenes can form from homogeneous alkyne hydrogenation, although in many cases this process is actually a cis hydrogenation followed by isomerisation from the Z to E alkene mediated by the metal hydrogenation catalyst. 103 Such trans hydrogenations can occur when heterogeneous solid surface catalysts are employed. In recent years, examples of homogeneous water-soluble ruthenium pentamethylcyclopentadiene (Cp*) catalysts have arisen that are able to trans-hydrogenate internal alkynes 104 in a process that typically involves carbene intermediates (Fig. 5a). 105,106 These [Ru(Cp*)(L) n ] systems have been reported to react with alkyne and H 2 to form a [Ru(Cp*)(H) 2 (alkyne)] intermediate that breaks para-H 2 symmetry by oxidative addition. 105 Subsequent hydride migration and rearrangement can form a ruthenium carbene intermediate which can react either to give cis or trans hydrogenated products depending on the geometry of this intermediate. Hydrogenation in a trans fashion is driven by steric factors in these intermediates 105 and can be favoured by parameters such as temperature and solvent. 104,107 These novel catalysts have expanded the scope of PHIP by allowing the preparation of E-alkenes in a hyperpolarised state (Fig. 5b). The formation of E-alkenes from alkynes via ruthenium carbene intermediates can also produce geminal hydrogenated products in which both 1 H nuclei of the starting para-H 2 are transferred to a single (rather than adjacent) carbon atom (Fig. 5a). 105,106,108 While the production of hyperpolarised CH 2 groups via geminal PHIP is possible (Fig. 5b), a challenge of this approach is the slower reaction times compared to traditional PHIP. 109 Fast relaxation of the newly introduced CH 2 protons can limit signal enhancements of these sites. Nevertheless, an appropriate resonant radiofrequency eld can preserve the singlet spin order within the CH 2 group which decays with the relaxation time constant for the long-lived state, T LLS , which can be longer than T 1 . 109 In these systems, trans and geminal hydrogenations can both occur as competing pathways. 53,109 Examples of hyperpolarised products resulting from both trans and geminal PHIP catalysis are expected to increase in the future as more catalytic systems that can perform such chemistry are developed. This has been particularly important for the production of hyperpolarised fumarate which can be used as a metabolic imaging probe and is discussed in more detail in Section 8.1.

Metal-catalysed vicinal and geminal proton exchange leading to PHIP
Exchange of vicinal and geminal alkene protons is oen a feature of metal-catalysed alkyne or alkene hydrogenation. 53,[110][111][112] In this process, protons on an alkene can become scrambled via metal hydride intermediates. In the context of PHIP experiments, these kinds of metal-catalysed proton exchange events provide a route to introduce parahydrogen-derived protons into alkenes (Fig. 5c). This can occur following a series of hydride migration steps, with bond rotation in the intermediate facilitating the scrambling of these sites. Such reactions can typically be revealed by analogous reaction of D 2 gas which will exchange with proton sites on the alkene to form H 2 gas and deuterated alkene. 53 Similarly, PHIP NMR experiments utilising para-H 2 can also reveal the occurrence of these reactions from the appearance of hyperpolarised 1 H NMR spectra. For example, reaction of styrene and para-H 2 or D 2 with Rh and Pd catalysts result in scrambling of the para-H 2 or D 2 labels with the geminal alkene protons. 53 Similar exchange of para-H 2 protons with vicinal or geminal alkene protons has been reported for Rh systems. 112 Consequently, 1 H NMR signals for these sites can become enhanced using PHIP. It is important to note that vicinal or geminal exchange of a single H or D by itself is usually insufficient to give rise to the PHIP effect as both para-H 2 atoms must be exchanged for the transfer to be pairwise. 53 The catalysis underlying these NMR signal enhancements does not necessarily require hydrogenation of the alkene, or formation of an alkene from alkyne hydrogenation. Therefore, while geminal and vicinal exchange oen occur in parallel to hydrogenation chemistry, they are reversible and do not necessitate any net changes to the chemical structure of hyperpolarised molecules. Examples of this reactivity in PHIP studies is unusual, likely due to competing hydrogenation that occurs via the same metal intermediates. Nevertheless, they provide unforeseen opportunities for hyperpolarisation at CC double-bonds and analysis of hyperpolarised line shapes can give valuable information on these processes, which may oen go unidentied. 112

Homogeneous hydrogenation of CN groups
The hydrogenation of unsaturated CC bonds, such as alkynes and alkenes, represents the most common example of substrates used in PHIP studies. However, the hydrogenation of other unsaturated functionalities with para-H 2 has also been reported, although examples of these are rare. The hydrogenation of nitriles to imines and primary amines using a cobalt bis(carbene) pincer catalyst has been demonstrated (Fig. 6). 58 This system involves reduction of the cobalt precursor with NaHBEt 3 before the hydrogenation of CN triple bonds via a Co(I/ III) redox process that includes an oxidative addition step. 58 This process is facilitated by the presence of a Lewis acid (such as BEt 3 ) which promotes the h 2 coordination of the nitrile important for initial transfer of para-H 2 to form imine. In these examples, hyperpolarised NMR signals for the imine are observed which suggest it is formed from nitrile hydrogenation in a pairwise fashion. However, no hyperpolarised signals for the amine products are observed which is indicative either of a non-pairwise hydrogenation of imine to form amine, or that the pairwise formation of amine is slow compared to the rate of relaxation. These catalysts have been used to hydrogenate a wide range of nitriles 58 (>20) and indicates that a greater variety of functionalities may become amenable to hyperpolarisation using para-H 2 as new catalysts to perform these transformations are developed. Fig. 6 PHIP effects can be observed when nitriles are hydrogenated using para-H 2 . The example 1 H NMR spectrum (600 MHz) has been recorded for the reaction mixture shown using a 45 pulse (above) with enhanced imine signals (inset) and a 1 H-OPSY sequence (below) that confirms these protons are derived from para-H 2 . Adapted from ref. 58.

Metal-free alkyne hydrogenation
As the range of metal-based systems for PHIP is increasing, so too are examples of metal-free systems. Very recently, ansaaminoboranes have been shown to hydrogenate alkynes with para-H 2 leading to hyperpolarisation of alkene products. 28 In the reported examples, a frustrated Lewis pair (HCAT) reacts with an alkyne to form a HCAT-alkyne adduct (Fig. 7a). This adduct performs heterolytic splitting of para-H 2 as the key step in the catalytic cycle to form a HCAT-alkyne-H 2 species. Hyperpolarised 1 H, 15 N and 11 B NMR signals for the NH and BH sites are observed which is consistent with similar para-H 2 activation by FLPs (vide supra). However, HCAT-alkyne-H 2 species can eliminate alkene products with enhanced NMR signals (Fig. 7b). This is the rst example of a metal-free catalytic system capable of producing hyperpolarised products other than para-H 2 activation adducts. Interestingly, only one proton originating from para-H 2 is transferred into the alkene product, with the other remaining within the HCAT catalyst. This highlights that in some cases both protons originating from the same para-H 2 molecule need not end up in the same product molecule. Similar effects have been reported for reversible iridium-catalysed exchange of para-H 2 and H 2 O, 113 and metal-catalysed hydroformylation reactions (oneH-PHIP as discussed in the next section), although the catalysis and/or polarisation mechanisms involved in these processes are different in each case.

Hydroformylation reactions
Hydroformylation reactions, in which alkenes are converted into aldehydes, present a different type of reaction that can break para-H 2 symmetry leading to PHIP. These reactions lead to the so-called oneH-PHIP effect as only a single 1 H nucleus originating from para-H 2 ends up in the aldehyde product but nonetheless can exhibit a net hyperpolarisation. 23 Despite the fact that the full catalytic cycle in this case is non-pairwise, a pairwise process must occur for PHIP effects to be observed at initial reaction steps. This is achieved by an initial oxidative addition reaction in which both 1 H nuclei within para-H 2 are transferred to an intermediary metal dihydride complex, before only one of these hydrides is incorporated into the nal product (Fig. 8). One of the rst examples of oneH-PHIP was in the hydroformylation of ethene in C 6 D 6 using [Ir(H)(CO) 2 (dppe)]. 23 Upon reaction of this precursor with 3 bar para-H 2 at 80 C, [Ir(COEt)(H) 2 (CO)(dppe)] is formed in which the two 1 H nuclei from para-H 2 are strongly coupled and thereby can gain a signicant level of single-spin net polarisations of opposite signs. Reductive elimination of this acyl dihydride intermediate yields propanal with an enhanced aldehyde CHO proton due to oneH-PHIP. Hyperpolarised signals for the metal product of reductive elimination, [Ir(H)(CO)(dppe)], were not observed which was attributed to either short relaxation times for this intermediate, or a short lifetime due to rapid formation of [Ir(H)(CO) 2 (dppe)]. 23 Hyperpolarised signals for a [IrH 3 (-CO)(dppe)] species are also observed which presumably forms from activation of para-H 2 by [Ir(H)(CO)(dppe)]. 23 These initial studies also utilised Pt-based systems and since then the range of catalysts used for oneH-PHIP have increased. For example, [Ir(CO)(PPh 3 ) 2 (h 3 -C 3 H 5 )] 25,114 or [IrI(CO) 2 (xantphos)] 24 hydroformylation catalysts have been employed and consequently, several PHIP-hyperpolarised iridium acyl and alkyl dihydride intermediates could be detected using NMR. The use of less well-dened cobalt systems to perform such oneH-PHIP reactions has also been studied. 26,27 A wider range of carbonylation reactions, in which CO is incorporated into a substrate, are also accessible to oneH-PHIP. For example, a palladium bis-phosphine complex was demonstrated to convert diphenylacetylene into methyl-2,3diphenyl acrylate in the presence of CO and H 2 . 55 This transformation was accompanied by the formation of cis-and transstilbene and the 1 H NMR signals for all three products were enhanced when this process was performed using para-H 2 . In the case of stilbene, enhanced signals are due to hydrogenative PHIP, although in methyl-2,3-diphenyl acrylate the enhanced 1 H NMR signals were attributed to a oneH-PHIP effect in which a single proton from para-H 2 was incorporated into the substrate molecule. 55 The 1 H NMR signal enhancements observed for aldehyde protons using oneH-PHIP are signicantly lower than can be achieved for alkene and alkanes produced using standard PHIP hydrogenation reactions. Nevertheless, strongly enhanced NMR signals for metal hydride intermediates has been of great use in mapping hydroformylation reaction pathways. [23][24][25][26][27]114 OneH-PHIP has expanded the types of reactions that can be examined using para-H 2 and has been of tremendous value in probing inorganic reaction mechanisms.

Reversible reactions with para-H 2 and polarisation transfer to other molecules (SABRE)
The reactions discussed so far all result in functionalisation of a to-be-hyperpolarised molecule with protons previously located in the para-H 2 feedstock. It was rst reported in 2009 that target substrates can become hyperpolarised without incorporation of para-H 2 originating atoms. 29 The method has been called signal amplication by reversible exchange (SABRE) and it works on the basis of reversible interactions of both para-H 2 and a to-behyperpolarised molecule with a metal catalyst. 16,17,29 The symmetry of para-H 2 is initially broken in a reversible oxidative addition reaction to form an iridium dihydride. The difference is that polarisation is then transferred from the hydride ligands to other 1 H sites or even heteronuclei of ligands bound transiently within the metal complex. Dissociation of the active SABRE catalyst generates hyperpolarised ligand free in solution without hydrogenation of the ligand (Fig. 9). A key advantage of SABRE compared to hydrogenative PHIP is that substrates are not chemically altered. Therefore, SABRE removes the requirement for target molecules to contain unsaturated functionality, although they must now contain motifs that can ligate to the SABRE catalyst. An added benet is that SABRE is reversible: once the hyperpolarised signals have decayed due to relaxation, they can be regained by shaking or bubbling with fresh para-H 2 . This is in contrast to hydrogenative PHIP which is only visible if the unsaturated substrate is not consumed completely in the reaction.
The spin dynamics of polarisation transfer within SABRE catalysts can be complex and there have been several excellent treatments of these effects. 17,19,115,116 Briey, polarisation transfer from the para-H 2 originating proton pair to target sites is enabled by spin-spin couplings connecting the two hydrides to each other and to the target ligands in the SABRE complex. 17,19,116,117 Typically, most efficient transfer to 1 H sites is achieved at 0-10 mT magnetic elds 118,119 which are oen produced in the stray elds of strong NMR magnets (>7 T). It is possible to transfer magnetisation from hydride ligands directly to heteronuclei including 15  follows the same principles as 1 H-SABRE but must be performed at much lower magnetic elds (0.5-20 mT) in order to meet resonance conditions. 39,121 Polarisation transfer within the SABRE catalyst is spontaneous at low magnetic elds (ALTA-DENA conditions), but it can also occur via radiofrequency excitation at high eld (Tesla) (PASADENA conditions). [133][134][135][136][137][138] In addition, feasible, but less efficient, polarisation transfer can occur spontaneously at high magnetic elds, giving rise to high-eld SABRE effects. 133 9 Generic depiction of reversible polarisation transfer using SABRE. The symmetry of para-H 2 is broken using a reversible oxidative addition reaction to form a hyperpolarised metal dihydride complex. This catalyses spontaneous polarisation transfer to other ligands (L 1 ) bound within the complex at low magnetic field (>milliTesla). Auxiliary ligands (L 2 ), which can be non-exchanging, often play a role in this process where x + y ¼ 4. The iridium complex may be neutral or positively charged depending on the identity of L 1 and L 2 and is often formed in situ.
Studies of ligand exchange within these complexes using exchange spectroscopy (EXSY) has revealed that substrate ligands located trans to hydride are typically in reversible exchange with their counterpart free in solution. 140 (Fig. 11). The 5-coordinate [Ir(H) 2 (L)(substrate) 2 ]Cl exists in a square-based pyramidal geometry immediately aer substrate dissociation. 141 However, formation of a trigonal bipyramidal structure is possible if the complex has a sufficiently long lifetime for ligand reorientation. Isomerisation between different square-based pyramidal structures via a trigonal bipyramidal form, or reaction of a trigonal bipyramidal intermediate with either para-H 2 or substrate on either face, is responsible for scrambling of the hydride ligand sites (Fig. 11). 142 This becomes more important for complexes that contain inequivalent hydride ligands. 38 For catalytic polarisation transfer to be observed, [Ir(H) 2 (h 2para-H 2 )(L)(substrate) 2 ]Cl must be short-lived to limit the conversion of para-H 2 into ortho-H 2 . 143 Instead, [Ir(H) 2 (h 2 -para-H 2 )(L)(substrate) 2 ]Cl serves to refresh the para-H 2 -derived singlet spin order within the iridium dihydride catalyst. The rate of H 2 loss typically decreases as the substrate concentration is increased, but increases as hydrogen pressure is increased. 38,48,141 These observations have been used to deduce that hydrogen and substrate exchange occur through a common pathway in which both hydrogen and substrate bind competitively to intermediates such as [Ir(H) 2 (L)(substrate) 2 ]Cl. In other words, pathways involving hydrogen loss directly from [Ir(H) 2 (L)(pyridine) 3 ]Cl to form [Ir(L)(pyridine) 3 ]Cl with changes in oxidation state are unlikely. 38,141 The transient intermediates involved in SABRE catalysis are oen too short-lived to be observed directly by NMR methods, although Density Functional Theory (DFT) calculations have provided supporting evidence for these mechanistic studies. 38,141,144 These studies have conrmed that direct reductive elimination of hydrogen is accompanied with a large energy barrier, favouring a pathway involving species with constant oxidation states.

Variation of the auxiliary ligand to ne tune substrate exchange and relaxation
The SABRE catalyst plays a crucial role in determining the substrates that can become hyperpolarised, and the magnitude of the NMR signal enhancements that can be achieved. The lifetime of the active SABRE catalyst is of vital importance: if it is too short then ligand exchange will occur before there has been sufficient time for a substrate to become hyperpolarised, if it is too long then the hyperpolarisation of substrate will decay due to the enhanced spin relaxation while bound to the metal centre. The active catalyst lifetime is closely linked to the substrate exchange rate, which can be altered by variation of auxiliary ligands on the catalyst, although factors such as temperature 122,145,146 or even solvent will also inuence this. 142,147 The steric and electronic properties of the auxiliary ligand can also inuence the types of substrates that can bind to the SABRE catalyst. Therefore, careful catalyst optimisation can play a signicant role in tuning ligand exchange which in turn provides a valuable route to improving the substrate scope and hyperpolarisation levels achieved using SABRE. Novel pulse sequences have also been used as a route to achieve high polarisation by compensating for suboptimal ligand exchange processes. 148 Catalysts involved in early SABRE studies contained an auxiliary phosphine ligand and were of the form [Ir(H) 2 (-PCy 3 )(substrate) 3 ][BF 4 ], where Cy is cyclohexyl. It was soon noted that the identity of the auxiliary phosphine ligand was one of many important factors that determine the efficiency of SABRE catalysis and the resulting NMR signal enhancements that can be achieved. For example, catalysts containing the phosphine ligand PCy 2 Ph yielded higher 1 H NMR enhancements for pyridine compared to the use of catalysts derived from other phosphines including Pcy 3 and PPh 3 . 149 Sterically large electron-rich phosphines are preferred which is attributed to more favourable substrate dissociation.
It was soon discovered that the use of carbene ligands, instead of phosphines, has a large effect in increasing the efficiency of SABRE catalysis. Their greater SABRE efficiency has been rationalised in terms of optimised substrate exchange within the active catalytic cycle. The rate constant of pyridine dissociation from [Ir(H) 2 (IMes)(pyridine) 3 ]Cl was determined by EXSY to be 23 s À1 at 300 K (ref. 48) which is ca. > 50 times faster than pyridine dissociation from [Ir(H) 2 (Pcy 3 )(pyridine) 3 ] + (0.45 s À1 at 295 K). 149 As a consequence, 1 H NMR signal enhancements of 8100-fold could be achieved for pyridine in methanold 4 at 3 T using an [Ir(H) 2 (IMes)(pyridine) 3 ]Cl active catalyst 141 which is much higher than those achieved using phosphinebased catalysts (<100-fold at 9.4 T). 149 The NHC ligand used to make [IrCl(COD)(NHC)] SABRE precatalysts can easily be modied to exhibit a variety of steric, and to a lesser extent, electronic properties. 151,152 Consequently, a large variety of SABRE catalysts with different NHC auxiliary ligands have been reported (Fig. 13b, also see ESI Table S1 †) and their effect on SABRE efficiency has been investigated. 48,[153][154][155][156] Generally, it is those catalysts containing an NHC ligand based on IMes that provide some of the highest NMR signal gains for substrates of interest using SABRE. When comparing IMesderivatives with variation in the substituent at the para position, R 1 , there is a good correlation between electron-rich carbenes (low TEP) and fast rates of substrate dissociation from [Ir(H) 2 NHC(substrate) 3 ]Cl (Fig. 12c). 153 This is generally benecial for refreshing para-H 2 spin order and facilitating catalytic build-up of substrate polarisation, although rationalisation of catalyst efficiencies can be challenging. Relaxation within the active catalyst is also an important factor that determines the magnitude of hyperpolarised NMR signals that SABRE can produce. Therefore, selected 1 H sites within the NHC ligand can be replaced with deuterium ( 2 H) labels. 153 This reduces any wastage of para-H 2 derived spin order by preventing unwanted transfer to 1 H sites of the auxiliary ligand. It also serves to reduce relaxation of the substrate when bound to the iridium catalyst and therefore substrate relaxation times in the presence of the catalyst are oen longer when deuterium-labelled catalysts are used compared to their protium counterparts. 153

Variation of the auxiliary ligand for water-soluble catalysts
The effect of the auxiliary ligand can also extend beyond subtle steric, electronic, and relaxation effects. For example, catalyst solubility can be altered by changing the auxiliary ligand. Ir-NHC catalysts are typically highly soluble in organic solvents and therefore polarisation transfer reactions are typically performed in methanol-d 4 (ref. 29, 38, 146, 157 and 158) or even dichloromethane-d 2 , 37,38 chloroform-d 159 or ethanol-d 6 . 146 The efficiency of SABRE catalysis in other solvents can be improved by alteration of the auxiliary NHC ligand. For example, a neutral iridium carbene complex where the NHC was 3-(2-methylene-4-nitrophenolate)-1-(2,4,6-trimethylphenyl)imidazolylidene, which contained a pendant alkoxide ligand, has been utilised in low polarity solvents including benzene-d 6 , and THF-d 8 . 160 Water-soluble SABRE catalysts allow the entire hyperpolarisation process to be performed in aqueous solvents. Specically, auxiliary PPh 3 ligands were functionalised with SO 2 Na or SO 3 K groups to produce water-soluble catalysts. 161 A variety of water-soluble IMes-derived ligands have also been Fig. 13 Reported examples of water-soluble SABRE catalysts. developed (Fig. 13). [161][162][163] Catalyst 51 was able to efficiently catalyse SABRE polarisation transfer to pyridine in methanol-d 4 giving total 1 H NMR signal enhancements of 750-fold. 161 However, when this was repeated in 70 : 30 D 2 O : ethanol-d 6 this value dropped to just 22-fold. Even though 47, 50-51 were soluble in water they failed to yield appreciable NMR signal enhancements for pyridine in water which was attributed to low para-H 2 solubility. 161 In contrast, 48 could achieve 1 H NMR signal enhancements of ca. 30-fold for pyridine in water. 162 Similar signal enhancements were observed for pyridine using 49, with up to 42-fold 1 H NMR signal enhancement for the ortho site of nicotinamide also reported. 163 These studies have shown that variation of the auxiliary ligand can alter catalyst solubility and they have been successful in allowing reversible polarisation transfer reactions to be performed directly in aqueous solvent. Nevertheless, they have not yet been reported to achieve NMR signal enhancements comparable to those that can be attained in organic solvents. Further alterations of the auxiliary ligand may yield higher-performing water-soluble catalysts that can give more signicant NMR signal enhancements in the future.  3 ]Cl is not formed due to sterically large, or weakly-ligating, substrates is the use of coligand-supported catalysts. 158,166 For example, coordination of pyridine-derived substrates with functionality in the ortho position that hinders iridium ligation can be facilitated by a coligand. 150,166 To demonstrate, 1 H NMR signal enhancements up to 1442 AE 84fold were recorded for 2,5-lutidine using catalysts of the form [IrCl(H) 2 (NHC)(sulfoxide)(substrate)]. This approach is notable as substrates of this type are not amenable to SABRE using [Ir(H) 2 (IMes)(substrate) 3 ]Cl due to steric constraints. These catalysts could also improve upon NMR signal enhancements for molecules such as 2-picoline compared to those achieved using typical [Ir(H) 2 (IMes)(substrate) 3 ]Cl systems. 166 The inclusion of a coligand can in some cases have a benecial effect on increasing substrate signals enhancements for systems in which [Ir(H) 2 (NHC)(substrate) 3 ]Cl is readily formed. 123,[167][168][169][170] This is possible if relaxation or ligand exchange is more favourable for [Ir(H) 2 (IMes)(substrate) 2 Fig. 14b and c). Since these initial studies, the efficiency of these catalysts have been improved further by variation of temperature to optimise ligand exchange processes and ca. 11% 13 C polarisation for sodium pyruvate-1-[ 13 C] has been reported. 171 The mechanism of formation, and action, of these sulfoxidecontaining catalysts contains several key differences from that of traditional [Ir(H) 2 147 There are other species also present in solution and those of the form [IrCl(H) 2 (sulfoxide) 2 (NHC)] are known to undergo fast para-H 2 exchange. 147 They are therefore likely to play a role in refreshing para-H 2 spin order within the catalytic system. The role of these species is also highlighted by the fact that the identity of the ligand, X, in [IrX(H) 2 (sulfoxide) 2 125,146,171 It is those species containing substrate ligated in the same plane as the hydride ligands that have a spin topology most appropriate for efficient polarisation transfer to substrate. 125 The proportion of active isomer can be maximised by alteration of the steric properties of the NHC. 146 Such systems have allowed the polarisation of previously inaccessible molecules: continued development of novel classes of coligands with unique reactivity will unlock the ability to hyperpolarise an ever-greater range of molecules using SABRE.  48,173 can occur. These adducts are typically of low concentration and are short-lived due to the weak binding of solvent methanol. The role that these adducts play can be revealed by recording SABRE efficiency as a function of polarisation transfer eld (PTF). 173 These PTF plots can include multiple maxima and this has been rationalised in terms of different optimal magnetic elds for polarisation transfer to substrate catalysed by [Ir(H) 2 (NHC)(substrate) 2 (CD 3 OD)]Cl compared to [Ir(H) 2 (-NHC)(substrate) 3 ]Cl. 173 These differences are a result of the different J coupling spin topology formed in the solvent adduct compared to the traditional tris-substrate catalyst. These studies have shown that the role of undiscerned species in SABRE catalysis can be challenging to determine and may be important for those seeking to model polarisation transfer during SABRE. 19,115,116,174,175 Other adducts such as [IrCl(H) 2 (NHC)(substrate) 2 ]Cl have been reported 35,173 and they are typically detected via distinctive hydride resonances or indirect methods based on CEST. 35 [IrCl(H) 2 (-NHC)(substrate) 2 ] has been reported to have a detrimental effect on SABRE efficiency as it results in a loss of para-H 2 spin order. This loss of correlation, or singlet-triplet leakage, is typically larger at higher magnetic eld and it can be reduced by performing SABRE experiments at lower elds 176 or by applying a spin lock pulse during para-H 2 bubbling to increase the lifetime of para-H 2 spin order. 35

Inefficient Ir-NHC SABRE catalysts: multinuclear decomposition products
The SABRE-active catalysts discussed so far are all formed over differing activation time periods which can range from several minutes 146 to several days. 160 While SABRE can easily be repeated multiple times by adding fresh para-H 2 and repeating the polarisation step, this cannot occur indenitely and, as with all catalytic processes, there is catalyst degradation over longer timescales. 131,157 Different systems will remain active for different time windows, which are usually sufficient to allow many repeated hyperpolarisation measurements to take place. For some catalysts, activity is observed for several hours aer catalyst formation. For others, deactivation can be more rapid. 146 Formation of catalyst decomposition products, which are usually dimers or higher order oligomers, can be responsible for a drop in SABRE efficiency at longer reaction times. 146,177 The formation of a wide range of dimer or trimer products, which can be revealed using X-ray crystallography, have been reported (Fig. 15). 131,147,157,159,172,177 These oen contain iridium-iridium single bonds 146,157,159,172 and may also contain bridging hydride, 157,159,172 chloride, 159 substrate, 131,147 or in the case of sulfoxide-containing catalysts, bridging S 2À or SR À ligands. 146,157,172 These species are typically no longer able to undergo the ligand exchange processes necessary for reversible polarisation transfer and therefore do not act as SABRE catalysts. 177 Hence, their formation should be minimised by modi-cation of the substrate, auxiliary ligand or coligand (if applicable). 146

Novel catalysts based on cobalt for SABRE-like hyperpolarisation
Almost all currently reported SABRE catalysts are based on iridium. In the future, completely new systems that are not based on Ir-phosphine or Ir-NHC catalysts are likely to become available that may be more efficient than the current state-of-the-art catalysts. This includes discovery of catalysts that are based on other transition metals. Rhodium complexes have been the focus of attempts to create non-Ir SABRE catalysts 178 as it also resides in group 9 of the periodic table, exhibits similar chemistry to iridium, and is used for hydrogenative PHIP. Despite this, no signicant breakthroughs in the development of such Rh-based systems have yet been reported. However, in recent years a novel cobalt dihydride system containing a pincer bis(carbene) ligand has been developed that is capable of hydrogenating alkenes. 59,103,179 When these reactions were studied using PHIP, NMR signal enhancements for the hydrogenated products were observed, as is typical of the PHIP effect. However, enhanced signals for the alkene reagents were also visible which were attributed to a reversible SABRE-like effect. This is possible due to reaction of [Co( Mes CCC)(N 2 )(PR 3 2 (alkene)] can hydrogenate the alkene to form an alkane, which exists in a hyperpolarised state due to PHIP. As alkene binding must be reversible, it can dissociate chemically unchanged. It also exists in a non-Boltzmann spin state which is a consequence of SABRE-like polarisation transfer within [Co( Mes CCC)(H) 2 (alkene)] (Fig. 16). 179 Recently, these effects have been extended to a wider range of substrates (>15) including styrenes, terpenes, acrylates, and pentenoates with 1 H and 13 C signals enhanced by up to 150-fold. 57 Other mechanisms for these observations, such as insertion-elimination (PHIP-IE) of the alkene to effectively exchange its protons with those derived from para-H 2 , have also been proposed (Fig. 16). 57 In these cobalt systems, substrates can undergo competing PHIP or PHIP-IE/SABRE 103 and additional studies are anticipated to further understand the different polarisation transfer mechanisms in action in these systems. Nevertheless, these ongoing studies have utilised a more widely available, cheaper, transition metal (compared to iridium or rhodium), and have demonstrated that reversible polarisation transfer is likely not exclusive to iridium catalysts. It is also expected that these breakthroughs will stimulate further design of novel systems that may exhibit vastly different catalysis to currently used systems.

Applications of novel PHIP catalysts in hyperpolarisation
In this nal section we highlight just two recent applications of PHIP hyperpolarisation in NMR that would not have been possible without the development of novel catalysts that can break para-H 2 symmetry.
8.1 Water-soluble ruthenium catalysts for trans hydrogenation: in vivo molecular imaging of hyperpolarised fumarate The NMR signal gains that can be provided by hyperpolarisation have provided exciting opportunities for biomolecular imaging. 4,6,180 MRI is currently used for routine diagnosis of tissue structure abnormalities, but its low sensitivity necessitates detecting a signal for bulk water at tens of molar concentrations in the body. Boosting MRI signal intensity allows molecules at much lower (i.e., millimolar) concentrations such as disease markers, drugs, and many other biologically relevant molecules, to be detected in vivo. This dramatically improves the diagnostic value of MRI as it can now give information about tissue biochemistry that are not available in thermally polarised MRI measurements. There are many key metabolic intermediates that are the focus of hyperpolarisation studies as their injection and imaging allows specic metabolic processes to be interrogated in vivo in real time. 9 One of these molecules is fumarate, a key intermediate of the Krebs cycle. Hyperpolarised [1,4-13 C 2 ]-fumarate can be used to probe necrotic cell death by observation of enhanced signals for [1-13 C] and [4-13 C]-malate products (Fig. 17). 107,181 The advent of water-soluble ruthenium catalysts that can perform trans hydrogenation reactions unlock the potential for PHIP to produce fumarate in a hyperpolarised state (Fig. 17a). 181,182 Transfer of polarisation from introduced protons to a 13 Clabelled site can occur using a magnetic eld cycling (MFC) approach and was reported to yield 13 C signals for [1,4-13 C 2 ]fumarate with a polarisation level of up to 24%. 182 It has since been reported that this has been increased to 30-45% by optimisation of the equipment, reaction conditions, and purication procedure (precipitation as a solid before redissolving). 183,184 The great benet of this approach is that it is highly compatible with biological imaging studies as it can produce aqueous solutions of hyperpolarised fumarate without contaminant side products or toxic catalyst. 107,183 Production and injection of PHIP-hyperpolarised [1-13 C]-fumarate into mice containing acetaminophen-induced hepatitis has allowed fumarate biochemistry to be examined in vivo using 2D chemical shi imaging at 1.5 T (Fig. 17c). 107 This feat would not be possible without development of catalytic systems that can hydrogenate starting unsaturated precursors in a trans fashion. Notably, 13 C polarisation levels for fumarate are comparable to those achieved using alternative hyperpolarisation techniques such as dissolution dynamic nuclear polarisation (d-DNP). 185,186 However, PHIP provides several advantages compared to d-DNP as it is rapid (seconds-minutes) and does not require the technically demanding equipment associated with d-DNP. 9,187 Further advances in catalyst design and agent purication steps are expected to increase signal enhancements, biocompatibility, and consequently, further progress in this direction is expected. Optimisation of current catalysts is expected to improve the NMR signal enhancements that can be achieved, which in turn facilitates MRI detection of molecules at lower concentrations. Continued development of novel catalysts that can break the symmetry of para-H 2 is likely to stimulate hyperpolarisation of a greater variety of metabolic imaging probes for disease diagnosis.  46 Reaction of this catalyst with the analytes pyridine, benzimidazole, purine, or adenine, forms complexes of the form [IrCl(H) 2 (PPh 3 ) 2 (analyte)] which also display enhanced hydride NMR resonances with distinctive chemical shis that depend on the identity of the analyte. The improved NMR sensitivity provided by PHIP allows picomolar concentrations of these analytes to be detected via hyperpolarised hydride signals.
Similar effects can be achieved using SABRE catalysts, which also display hydride NMR chemical shis sensitive to the identity of the substrate molecule trans to them. 38,51 In particular, catalysts with inequivalent hydride ligands have been highly benecial as their NMR signals display a wider chemical shi dispersion in a spectrally uncrowded region. 51 The enhanced hydride signals of these [Ir(H) 2 (IMes)(Nsub 1 )(-Nsub 2 )(Nsub 3 )]Cl catalysts have been used to detect specic analytes present in complex mixtures and has been applied to study metabolomics and other systems that contain a great variety of molecules. 50,51 In these examples, peak overlap, even within the hydride region of 1 H NMR spectra, can oen become a problem due to the large number of SABRE-amenable analytes in the mixture. It is therefore oen advantageous to use 2D NMR methods such as COSY, 188,189 DOSY 190,191 or others 49,51 to provide increased chemical resolution between analytes.
A further way to increase chemical shi resolution of these reporter complexes is through novel catalyst design. Recently, the enhanced hydride signals of [Ir(H) 2 (IMes)(k 2 -N,O-imine)(sub)] catalysts have been reported as sensitive markers of particular analytes. 38 They can achieve much wider chemical shi dispersion for their hydride signals than [Ir(H) 2 (IMes)(Nsub 1 )(Nsub 2 )(-Nsub 3 )]Cl systems. The chemical inequivalence of these hydrides, which are located trans to the O-donor site of the imine and trans to the analyte which can bind through an N-, S-or even C-donor site (Fig. 18a) is advantageous in preventing NMR signal overlap. 37,38 As a consequence, the hydride resonance trans to the analyte can display signals over a 15.5 ppm chemical shi range depending on the identity of the analyte bound within the catalyst. 38 Aliphatic and aromatic NMR signals of the analyte itself can become enhanced via SABRE, but for complex mixtures these signals are poorly resolved. The enhanced hydride signals of closely related [Ir(H) 2 (IMes)(k 2 -N,O-amino acids)(pyridine)] catalysts have been used to distinguish different amino acids (Fig. 18b) in complex bio-mixtures. 49 The applications of these catalysts in NMR mixture analysis and biouid analysis has been clearly demonstrated and further examples are expected in the years ahead.

Conclusions
The use of para-H 2 to enhance NMR signal intensity is a versatile platform that allows the hyperpolarisation of a range of different functionalities including metal complexes, 14,30,36,38,45 alkenes, 14,30,57 alkanes, 14,30 esters, 88 aldehydes, [23][24][25][26][27]114 carboxylic acids, 126,192 N-heterocycles, 17,29 amines, 58,193 diazirines, 124,194,195 nitriles, 167 silanes, 132 sulfoxides, 147,172 a-ketoacids, 125,146,147,157,196 amino acids, 197 peptides 78,80,82,198 and many others. Enhanced NMR signals can be achieved most commonly for 1 H, 13  signals have also been reported. The preparation of these molecules in a hyperpolarised state allows exciting applications in the areas of mechanistic elucidations, 11 reaction monitoring, 139,196,202 mixture analysis, 51,188,189,203,204 biomedical imaging, 6,107,182,205,206 and many others. In recent years this list of applications has even grown to include novel uses in magnetooptics, 207 RASER physics, 208 micron-scale NMR, 209 SQUIDbased NMR, 210 and zero-eld NMR. 211,212 Development of novel catalyst systems has spurred some of the greatest advances in this area of science by expanding the scope of molecules that can be hyperpolarised. Many other advances in hyperpolarisation are occurring simultaneously: proton exchange effects have allowed polarisation to be relayed to molecules that might not contain any unsaturated (PHIP-X) 213 or iridium ligating (SABRE-Relay) 214 sites such as alcohols, 215 sugars, 216 silanols 132 and many others. 213,214,217 It is expected that further catalyst development will be a large driving force behind further extension of substrate scope, increased efficiency, and enhanced biocompatibility. In particular, novel metal-free, 28,64-67 trans hydrogenation, 106,109,181,182 sulfoxide-containing SABRE catalysts 125,146,147,157 and cobalt-based SABRE-like systems 59,179 are likely to lead to exciting new results in this area in the years ahead. Fig. 18 (a) The enhanced hydride NMR signals of PHIP and SABRE catalysts can provide significant chemical shift resolution which is of great use for the indirect detection of analytes in mixtures. 38

Conflicts of interest
There are no conicts to declare.