Dynamic Nuclear Polarization Enables NMR of Surface Passivating Agents on Hybrid Perovskite Thin Films.

Surface and bulk molecular modulators are the key to improving the efficiency and stability of hybrid perovskite solar cells. However, due to their low concentration, heterogeneous environments, and low sample mass, it remains challenging to characterize their structure and dynamics at the atomic level, as required to establish structure-activity relationships. Nuclear magnetic resonance (NMR) spectroscopy has revealed a wealth of information on the atomic-level structure of hybrid perovskites, but the inherent insensitivity of NMR severely limits its utility to characterize thin-film samples. Dynamic nuclear polarization (DNP) can enhance NMR sensitivity by orders of magnitude, but DNP methods for perovskite materials have so far been limited. Here, we determined the factors that limit the efficiency of DNP NMR for perovskite samples by systematically studying layered hybrid perovskite analogues. We find that the fast-relaxing dynamic cation is the major impediment to higher DNP efficiency, while microwave absorption and particle morphology play a secondary role. We then show that the former can be mitigated by deuteration, enabling 1H DNP enhancement factors of up to 100, which can be harnessed to enhance signals from dopants or additives present in very low concentrations. Specifically, using this new DNP methodology at a high magnetic field and with small sample volumes, we have recorded the NMR spectrum of the 20 nm (6 μg) passivating layer on a single perovskite thin film, revealing a two-dimensional (2D) layered perovskite structure at the surface that resembles the n = 1 homologue but which has greater disorder than in bulk layered perovskites.

Introduction

Hybrid organic–inorganic perovskites have emerged as a promising alternative to existing solar cell technologies owing to their bandgap tunability,[1−3] facile processing methods,[4,5] and competitive performance.[6−11] The perovskite crystal structure consists of an A-site cation (e.g., formamidinium, FA+, CH(NH2)2+; methylammonium, MA+, CH3NH3+) in a three-dimensional (3D) network of lead halide octahedra as shown in Figure a. As the photoactive phases of hybrid perovskite materials are not intrinsically stable under ambient conditions,[12] methods to stabilize them are of intense current interest, with strategies based today either on elemental doping with inorganic cations (Cs+, K+, Rb+, Mn2+, Co2+, Sb3+, In3+)[13−15] or on passivation by a surface treatment of organic molecules or salts.[14−19] When bulky molecular cations are used, two-dimensional (2D) layered perovskites can form where inorganic perovskite slabs are separated by layers of organic cations (Figure a).[20−24] Bulk layered perovskites are more stable than their 3D counterparts but typically have lower photoconversion efficiencies.[24,25] 2D/3D heterostructures, where the layered perovskite forms at the surface of the bulk perovskite, combine the higher stability provided by the 2D phase and the superior optoelectronic properties of the 3D perovskite, while further passivating interfacial vacancies to reduce nonradiative recombination.[26−29] The detailed manner in which organic moieties interact with the perovskite structure has been thought to modify the energy landscape of the material and thereby template the photoactive α-FAPbI3 phase.[6,30,31]

Figure 1

(a) Schematic representation of the layered Ruddlesden–Popper perovskite structures. (b–d) DNP-enhanced 1H→13C CP spectra with and without microwave (μw) irradiation of (b) FAPbI3 (n = ∞) at 133 K, (c) PEA2PbI4 (n = 1) at 119 K, and (d) PEA2FAPb2I7 (n = 2) at 130 K. All temperatures are given within ±2 K. Asterisks indicate spinning sidebands. The μw on spectra shown in the figure were recorded with 8 (n = ∞), 64 (n = 1), and 4 (n = 2) scans each and a polarization delay of 100 s between scans. The full details are given in the Experimental Section.

Establishing structure–activity relationships to optimize perovskite stability requires a detailed knowledge of the atomic-level local structure for these dopants and surface treatments when applied to perovskite materials. Although diffraction-based techniques can be used to study the inorganic lattice, due to the polycrystalline nature of solution-processed thin films and the lack of long-range order for surface and/or bulk dopants and additives, it is difficult to determine structural information for these minor species using diffraction-based methods.

Nuclear magnetic resonance (NMR) is well suited to establish an atomic-level understanding of both the organic and inorganic components of hybrid perovskites, as well as dopants and additives.[32−35] Specifically, magic angle spinning (MAS) NMR has previously been applied to study cation incorporation,[36−41] phase segregation,[38,42−47] cation dynamics,[48−51] passivating layers,[31,52] degradation,[52,53] and phase transitions.[48−51] NMR has also been successful in probing bulk layered perovskites,[47,55,56] bulk 2D/3D perovskite heterostructures,[57−60] and the surfaces of perovskite nanocrystals.[61−63] However, the intrinsic insensitivity of NMR limits its utility to study thin-film samples as used in perovskite devices, owing to the very low sample mass. This insensitivity is compounded for the study of dopants and additives, which comprise only a small fraction of the sample. Consequently, in the majority of previous studies, either several thick films or mechanosynthesized powders were used to enable NMR experiments to be performed. Thus, key challenges relating to NMR sensitivity in hybrid perovskites remain,[32] with spectra from the surface and interface species in thin films still largely out of reach.

To overcome the intrinsic sensitivity problem associated with NMR and solve structural problems in materials science, dynamic nuclear polarization (DNP) methods have been developed.[64−68] In these experiments, the target material is wetted with a solution of stable biradicals and cooled to ∼100 K.[66] Continuous microwave irradiation at or near the electron spin resonance (ESR) frequency transfers the high electron polarization from the radical to solvent nuclei, which can then be transferred to surface species on the target material or propagate into the bulk by spin diffusion, enhancing the corresponding NMR signal in both cases.[66] This DNP approach has been successfully applied to study surfaces and interfaces in materials for batteries,[69,70] catalysis,[71−76] and optoelectronics.[77]

Sensitivity-enhancing protocols for hybrid perovskites have been investigated by Hanrahan et al.[78] For the detection of 207Pb spectra, they found the most promising technique to be proton detection, while MAS DNP was surprisingly inefficient for both the organic and inorganic components: signal enhancements were typically <10, especially in the case of the most technologically relevant iodide perovskite.[78] A clear understanding of the underlying challenges in applying DNP to organic–inorganic perovskites, and methodologies to overcome them, are still lacking.

Here, we demonstrate efficient DNP for the organic moieties in layered hybrid organic–inorganic perovskites. Specifically, we systematically investigate a class of 2D perovskite homologues to determine that the dynamic A-site cations are the main cause of poor DNP performance. We then demonstrate how to overcome this with deuteration strategies, enabling 1H DNP enhancement factors of up to a factor 100. Finally, using the DNP methodology developed for the layered perovskite combined with a high magnetic field and small sample volumes, we have successfully recorded the NMR spectrum of the surface coating (∼20 nm) deposited on a single FAPbI3 perovskite thin-film sample (2 × 2 cm2 area). This opens the door to detailed atomic-level structural studies of surface treatments on technologically relevant samples.

Results and Discussion

First, we determined the DNP enhancements that can be achieved for the organic FA+ cation in bulk FAPbI3 using a conventional DNP formulation, i.e., impregnation of the materials with 16 mM TEKPol[79] in tetrachloroethane (TCE) at ∼100 K[65,80] (note that FAPbI3 undergoes phase transitions at ∼140 and ∼285 K).[49,81] As shown in Figure b, although the enhancement factor upon microwave irradiation for the 1H of the TCE is 114 (as measured through the 13C signal intensity in a 1H→13C cross-polarization (CP) spectrum), the FA+ cation signal exhibits an enhancement of only ∼3. While this does correspond to an experimental time saving by an order of magnitude, for typical surfaces or even for bulk microcrystalline solids, 1H DNP enhancement factors >30 can routinely be achieved today.[82−84] Furthermore, the solvent enhancement is also lower than the expected factor of ∼200 for this formulation. Overall, DNP experiments do not work as well as expected in this hybrid perovskite.

There are several factors that could explain the poor DNP performance in α-FAPbI3. It is well established that the organic cation that occupies the cuboctahedral cavity in organic–inorganic perovskites is highly dynamic.[48−50] At 100 K, this motion occurs on a similar timescale to the 1H Larmor frequency (∼1 ns), resulting in fast 1H relaxation (T1 ≈ 1 s). As a result, the cations can act as polarization sinks, causing the hyperpolarization to relax before it can diffuse significantly into the particles. Alternatively, the perovskite material could absorb the microwave radiation itself due to the contribution of the dynamic cation to the dielectric properties.[85] This can have two main consequences: (i) an overall reduction in the available microwave energy to drive the ESR transitions and/or (ii) a significant increase in the sample temperature, resulting in shorter electron spin relaxation times and less-efficient DNP (note that while the increased temperature does affect the cation dynamics, there is little change in the 1H T1 due to the proximity to the T1 minimum; Figure S6). Each of these factors could play a significant role in determining the overall enhancement in the perovskite materials, and in the following, we disentangle their relative contributions.

To distinguish between these factors, we performed experiments on samples of two-dimensional Ruddlesden–Popper layered perovskites containing phenylethylammonium (PEA+) spacers,[86,87] which allow the dynamic FA+ cation to be removed and systematically reintroduced in the different homologues. These low-dimensional perovskites are characterized by the number of inorganic layers per organic layer (n), as shown in Figure a. In this convention, α-FAPbI3 can be considered as n = ∞. Notably, the n = 1 member is unique in that it does not contain an A-site cation and therefore offers the advantage to selectively investigate the effects of the mobile FA+ cation.

Figure c shows the DNP enhanced 1H→13C CP spectrum of n = 1 (PEA2PbI4) where we observe enhancement factors of 215 and 43 for the solvent and PEA+ signals, respectively (see Figure S1 for the spectral assignment). This is around the maximum solvent enhancement that can be achieved for this formulation, and the perovskite enhancement is far higher than for pure α-FAPbI3 (see the summary in Figure a). The lower enhancement of PEA+ compared to the solvent is expected since the high polarization from the wetting phase has to diffuse into the particle.[64,88,89] For the n = 2 (PEA2FAPb2I7) composition (Figure d), the solvent enhancement is significantly lower (107), whereas the perovskite enhancements are still appreciable (around 25). Notably, the FA+ and PEA+ moieties show very similar enhancement factors within error since they are in atomic-scale proximity and fast 1H–1H spin diffusion equalizes the nuclear hyperpolarization.[88,89] These large DNP enhancements for both the n = 1 and n = 2 perovskites further allow 1H→15N CP spectra to be measured within minutes (Figure S2).

Figure 2

Comparison of (a) 1H DNP enhancement factors and (b) TDNP values measured via 1H→13C CP spectra of n = ∞ (FAPbI3), n = 1 (PEA2PbI4), and n = 2 (PEA2FAPb2I7) at the specified temperatures. Temperatures are given within ±2 K.

The observation that the solvent enhancements for the n = 2 and n = ∞ samples are similar, while that for n = 1 is higher, is consistent with the presence of dynamic FA+ cations in the former; we now examine the mechanism for this. First, we consider microwave heating: for n = 2 and n = ∞, the microwaves induce heating by 30 ± 2 and 33 ± 2 K, respectively, resulting in a minimum achievable temperature of 133 ± 2 K, whereas for n = 1 the heating is only 19 ± 2 K, resulting in a minimum achievable temperature of 119 ± 2 K. Clearly, the presence of the dynamic FA+ cation does induce greater sample heating. However, when the sample temperature is deliberately increased to 130 ± 2 K for the n = 1 sample by heating the incident gas flows, the solvent enhancement decreases only to 177, still far higher than for the n = 2 and n = ∞ perovskites. Therefore, while sample heating does play a role, it is not the major cause of the differing DNP performance.

Further details on the DNP behavior can be inferred from the buildup of nuclear polarization under microwave irradiation, as characterized by the time constant TDNP (Figure b). The long TDNP for the n = 1 sample is consistent with the high enhancement, since there is more time to accumulate polarization, which relaxes more slowly. TDNP is significantly shorter for the n = 2 and n = ∞ perovskites, consistent with the presence of fast-relaxing FA+ cations. However, despite the similar build-up behavior, the enhancement is far higher for n = 2 than for n = ∞. This difference can be explained by considering another crucial factor for DNP of impregnated solids, the particle size and morphology;[88,89] therefore, scanning electron microscopy (SEM) images were obtained as shown in Figure .

Figure 3

SEM images of mechanosynthesized (a) n = ∞ (FAPbI3), (b) n = 1 (PEA2PbI4), and (c) n = 2 (PEA2FAPb2I7) perovskites.

The n = ∞ and n = 1 compositions show loosely agglomerated secondary particles composed of approximately spherical primary particles with an average particle size of 0.7 ± 0.2 μm (Figure a,b). In contrast, the n = 2 sample (Figure c) exhibits platelet-like primary particles, as commonly observed for layered materials,[60] with a thickness of 0.13 ± 0.05 μm. The morphology in the n = 2 sample favors the propagation of polarization to the bulk owing to the increased surface-to-volume ratio and the shorter diffusion lengths required.[88,89] Therefore, although the FA+ cation reduces the T1 and therefore the spin-diffusion length for the n = 2 homologue, the favorable thin particle size nevertheless allows appreciable hyperpolarization to be relayed to the perovskite.

To determine, and mitigate, the effect of fast 1H relaxation by the dynamic FA+ cation, we synthesized the n = 2 perovskite having deuterated the FA+ cation to different degrees. Note that since the −NH3 hydrogens in PEA+ could exchange with FA+ during synthesis, these were also deuterated. Deuteration of the DNP matrix to decrease the 1H heat capacity is well established,[89−91] while deuteration of fast-relaxing methyl groups in proteins[92] or on the surfaces of heterogeneous catalysts[93] has also previously been shown to improve DNP performance. Figure a shows how with increasing deuteration of FA+, the enhancement of the solvent increases progressively, to reach levels similar to the factors of 177–215 achieved for n = 1. This indicates that dissipation of hyperpolarization in the fast-relaxing perovskite phase is the primary factor limiting the solvent enhancement in the n = 2 formulations. The concomitant increase in the TDNP constants for the perovskite corroborates this observation (Figure b), since the ratio of fast-relaxing FA+ protons to slower relaxing PEA+ protons decreases. For the sample with medium deuteration, the enhancement of the perovskite signals increases by a factor of 4 to over 100 for the PEA+, as compared to the nondeuterated sample. This arises from the lower proportion of the fast-relaxing cations, which slows the average T1 relaxation, increasing the spin-diffusion length and the amount of hyperpolarization that can accumulate both in the solvent and the perovskites. Interestingly, the highly deuterated sample exhibits even higher solvent enhancements, but lower enhancements for the perovskite, comparable to the nondeuterated n = 2 sample. This can be explained by a reduction in 1H spin-diffusion efficiency at the highest deuteration level due to the low 1H concentration, which hampers the propagation of polarization into the perovskite. In particular, when the FA+ cation is highly deuterated, the perovskite layer acts as a barrier to 1H spin diffusion. Furthermore, since the layers are likely to be aligned with the platelet morphology, we speculate that polarization from the majority of the hyperpolarized solvent must pass through FA+ layers. Therefore, although the high deuteration of FA+ moieties increases the perovskite TDNP and reduces polarization loss by relaxation, the overall effect with diminished spin diffusion gives similar enhancements to those of the nondeuterated n = 2 composition.

Figure 4

1H DNP enhancement factors and polarization build-up constants (TDNP) as a function of deuteration level for (a, b) n = 2 PEA2FAPb2I7 and (c, d) n = ∞ FAPbI3. Medium deuteration for n = 2 is achieved using 60% d3-PEAI and 50% d4-FAI. High deuteration for n = 2 is achieved using 80% d3-PEAI and 85% d5-FAI. High deuteration for FAPbI3 is achieved using 85% d5-FAI.

Applying a similar deuteration strategy to pure FAPbI3 is less successful. The solvent enhancement is increased to some extent but not as much as for n = 2 (Figure c); this is ascribed to the fact that the T1 relaxation of the perovskite remains fast (TDNP = 4 s, Figure d), and therefore, although the concentration of the relaxation sinks is reduced, the perovskite does still act as a sink for the solvent hyperpolarization. Greater microwave absorption by n = ∞ than by n = 2 reducing the efficiency of saturating the ESR transitions could also play a role, although the sample heating remains similar. The low enhancement for the perovskite itself is expected because only the fast-relaxing cations with residual protonation can be observed.

Although these DNP experiments significantly improve the NMR sensitivity, they were performed with conventional DNP hardware using 3.2 mm outer diameter sample containers (rotors) that require a high sample mass of at least 50 mg, whereas a typical thin-film device contains only ∼1 mg of perovskite. Very recently, state-of-the-art DNP hardware has been developed using 0.7 mm outer diameter rotors that require only 1–2 mg of sample, while enabling faster magic-angle spinning that improves resolution;[94] sensitivity and resolution are further improved by the use of a high magnetic field of 21 T, compared to 9.4 T used above. We first tested the DNP using this system on a mechanosynthesized n = 1 perovskite sample, as studied above, wetted with a 32 mM solution of the HyTEK-2 biradical[95] in TCE, which has been shown to perform significantly better than TEKPol at a high field with fast spinning.[94] As shown in Figure a, the TCE and PEA+ signals exhibit enhancement factors of 125 and 35, respectively. Although solvent enhancements up to 200 have been demonstrated with this formulation in pure frozen solutions,[94] the sample enhancement is nevertheless similar to that obtained in the more conventional DNP experiments at 9.4 T above, validating the effectiveness of the experimental setup. We note that a slightly lower sample temperature of 111 K was observed (cf. 119 K at 9.4 T), possibly due to different microwave absorption at higher frequency and/or more efficient cooling of the smaller rotor.

Figure 5

DNP-enhanced 1H→13C spectra at ∼110 K and 21 T of (a) mechanosynthesized n = 1 PEA2PbI4 recorded for 2 min (20 min for microwave off) at 50 kHz MAS and (b) a PEAI-treated single thin film of deuterated FAPbI3 recorded for one day at 40 kHz MAS.

We then turned to the challenge of studying a technologically relevant surface coating on a thin film. Phenylethylammonium iodide (PEAI) is one of the most widely used passivating agents[24,96−98] for perovskite thin films owing to its bulky nature that facilitates the formation of 2D perovskites while providing high stability to the material.[98,99] Here, an ∼500 nm layer of deuterated FAPbI3 was spin-coated onto a glass slide (2 × 2 cm2) and passivated with a thin (∼20 nm) surface layer of PEAI, followed by annealing at 100 °C. The film was then scraped off (to yield about 0.8 mg) and packed into a 0.7 mm rotor to which the radical-containing solution was then added by centrifugation. The sample is estimated to contain a total of about 6 μg (50 nmol) of PEA+. Figure b shows that the PEA+ signals can clearly be seen in the resulting DNP-enhanced 1H→13C spectrum obtained in 27 h, whereas even the most intense peak can barely be seen without DNP, giving an estimated enhancement for the PEA+ layer by a factor of ∼10. (We therefore estimate that it would have taken ∼100 days to acquire this spectrum with the same signal-to-noise ratio without DNP.) The FA+ enhancement is minimal, consistent with expectations from the experiments on deuterated mechanosynthesized FAPbI3. The high PEA+ enhancement and low FA+ enhancement are also evident in the 1H NMR spectrum (Figure S5); however, the resolution is insufficient to interpret the structure, even at 40 kHz MAS.

Comparing the observed PEA+ carbon-13 shifts (Figures and S3), we note that there is a significant change between the PEAI salt and the n = 1 perovskite, reflecting the substantial structural transformation.[47] The PEA+ signals observed in the spectra of the surface-coated thin film, as shown in Figure b, align most closely with those of the layered perovskites, indicating that after the surface treatment, the PEA+ adopts the layered perovskite structure, in line with previous reports.[96,98−100] However, the peaks are significantly broader, indicating a broader distribution of local environments and disorder in the thin-film coating. This is consistent with a mixture of n = 1 and higher-order homologues,[55] as well as the nanoscale thickness of the passivating layer, meaning that most of the material is within ∼10 nm of the surface or the bulk perovskite, which can slightly modify the local structure. In addition, a slight shift of ∼1 ppm is observed for the carbon closest to the perovskite layer (Figure S3). This may correspond to a slightly different binding of the organic spacers to the inorganic layers, or a slight difference in the electronic structure of the inorganic layers, in the nanoscale-coating layer as compared to the bulk layered perovskite.

Conclusions

In conclusion, we have identified the factors that previously limited DNP efficiency in hybrid perovskites. In particular, we find that the primary impediment to hyperpolarization is the fast 1H relaxation of the dynamic cation, while microwave absorption by the sample is also detrimental due to both sample heating and reduced saturation efficiency. We have shown that this can be mitigated by deuterating the cation, which then enables large DNP enhancements to develop on both the DNP matrix and target organic species present at low concentrations in the sample. We have further demonstrated that DNP can be favorably applied to two-dimensional layered perovskites, with the high surface-to-volume ratio that can arise from platelet morphologies being particularly beneficial.

There is plenty of further scope to improve the enhancement by optimizing the impregnation protocols, which could be further combined with isotopic labeling to achieve ultimate sensitivity.[74,101,102] This will then enable, for example, multidimensional correlation experiments to provide information about connectivity and spatial proximities.

Here, efficient DNP with deuterated samples enabled the observation of the 13C PEA+ signals from the approximately 6 μg of surface coating on a single FAPbI3 perovskite thin film, by combining with cutting-edge DNP hardware for small-diameter rotors at a high field. The spectra of the thin film immediately reveal that a 2D PEA+ layered perovskite structure forms on the bulk perovskite surface that most strongly resembles that in the n = 1 layered perovskite homologue, rather than remaining as PEAI, but with a greater distribution of local environments. This work extends DNP-enhanced solid-state methods to a new class of semiconducting photovoltaic materials and, therefore, paves the way to characterize the local structure of additives and surface treatments for technologically relevant perovskite thin-film compositions.

Experimental Section

Materials

The following materials were used: formamidinium iodide (Sigma, >99%), PbI2 (Sigma, 99%), phenylethylammonium iodide (Greatcell Solar Materials), anhydrous dimethylformamide (99.8%, Acros), dimethylsulfoxide (99.7%, Acros), acetone (99.6%, Acros), chlorobenzene (99.8%, Acros), isopropanol (99.5%, Acros), and d5-FAI (Cortecnet, 85% CD deuterated, 90% ND2 deuterated). d4-FAI and d3-PEAI were prepared by dissolving in heavy water (1:40 mol/mol ratio), followed by evaporation. This yielded ∼50 and ∼70% deuterium on the FAI and PEAI, respectively.

Bulk Sample Preparation

The perovskite materials were prepared using mechanosynthesis following a previously published protocol.[48] The precursors (PbI2, PEAI, and FAI, ∼200 mg total) were mixed in the appropriate molar ratio and ground in an electric ball mill (Retsch MM 400) using an agate grinding jar (10 mL) and agate ball (⌀10 mm) for 60 min at 25 Hz. The resulting materials were then annealed at 150 °C for 20 min.

Surface-Coated Thin-Film Fabrication

A 1.4 M solution of d4-FAPbI3 in 4:1 (v:v) DMF/DMSO ratio was spin-coated onto a glass slide (2 × 2 cm2) with a two-step program at 1000 and 4000 r.p.m. for 10 and 30 s, respectively. During the second step, 200 μL of chlorobenzene (99.8%, Acros) was dropped onto the spinning substrate 10 s prior to the end of the program and the film was annealed at 150 °C for 20 min. After cooling down the substrate to room temperature, a thin surface layer of d3-PEAI (5 mg in 1 mL of acetone) was spin-coated on the perovskite layer at 5000 rpm for 30 s, followed by annealing at 100 °C for 5 min.

DNP-Enhanced Solid-State NMR Measurements

DNP formulations, except for the 0.7 mm DNP samples, were prepared according to the standard protocols for impregnation DNP[64−66] by wetting ∼50 mg of the perovskite materials with ∼50 μL of 16 mM TEKPol[79] in tetrachloroethane (TCE). 1% d6-EtOH ethanol was included to improve glass formation.[103] For the thin-film formulation, powder was scraped off from the thin film, packed into the 0.7 mm rotor, and then a 32 mM solution of HyTEK-2[95] in TCE was added to the sample by centrifugation at 6000 rpm for a minute.

The majority of the DNP-enhanced spectra were acquired on a commercial Bruker Avance III 400 MHz (9.4 T) NMR spectrometer equipped with a 263 GHz gyrotron microwave source using a 3.2 mm triple resonance low-temperature magic angle spinning (LTMAS) probe with sapphire rotors spinning at 8 kHz. A few grains of KBr were mixed into the material before the impregnation step and the temperature was measured from the 79Br T1 constant.[104] Before measuring the DNP-enhanced NMR spectra, samples were degassed by performing three insert–eject cycles, waiting for ∼1 min at each step. The DNP enhancement factors were measured from the intensity ratio of the 1H→13C cross-polarization (CP)[105] spectra acquired under microwave-on and -off conditions with 100 kHz of SPINAL-64[106] decoupling. A microwave power of ∼12 W was used for all DNP experiments, as measured by a calorimeter halfway along the waveguide. 13C chemical shifts were referenced to the TCE peak at 74 ppm at ∼100 K.[64,107] The TCE enhancement is reported for a recycle delay of 5 s and the PEA+/FA+ enhancements for a recycle delay of 100 s. The 1H build-up time constants (TDNP) were measured using a saturation-recovery 1H→13C experiment. Errors in the time constants were calculated using Monte-Carlo analysis based on the experimental noise level.

High-field fast-spinning DNP solid-state NMR experiments were performed on a 900 MHz (21.1 T) Avance Neo Bruker NMR spectrometer. The spectrometer was equipped with a LTMAS 0.7 mm triple resonance probe coupled with a Bruker gyrotron producing 593 GHz continuous microwaves. The frequency of the gyrotron was tuned to give the previously observed maximum enhancement for cross-effect DNP with HyTEK-2[95] by modifying the gyrotron cavity temperature to 28°C. The DNP enhancement factors were measured from the intensity ratio of the 1H→13C cross-polarization (CP) spectra acquired under microwave-on and -off conditions with 110 kHz of SPINAL-64 decoupling. Details specific to particular spectra are given in Table S2 and the figure captions.

X-ray Diffraction (XRD) Measurements

Powder XRD patterns of mechanosynthesized layered and 3D perovskites were recorded with a Bruker D8 Discover Vario diffractometer with a Cu Kα1 monochromator (1.5406 Å) from 2θ = 2–50°.

Scanning Electron Microscopy (SEM) Measurements

For the SEM images, mechanosynthesized powders were deposited on a standard SEM sample stub with conductive carbon adhesive tabs. A Zeiss Merlin scanning electron microscope was used, and images were acquired at 0.8 kV beam energy using low currents (20–40 pA) detecting secondary electrons with an in-lens detector. SEM images were analyzed using ImageJ software.