Comparison of various NMR methods for the indirect detection of nitrogen-14 nuclei via protons in solids

Accepted Manuscript Comparison of various NMR methods for the indirect detection of nitrogen-14 via protons in solids Ming Shen, Julien Trébosc, Luke A. O’Dell, Olivier Lafon, Frédérique Pourpoint, Bingwen Hu, Qun Chen, Jean-Paul Amoureux PII: DOI: Reference:

S1090-7807(15)00138-X http://dx.doi.org/10.1016/j.jmr.2015.06.008 YJMRE 5667

To appear in:

Journal of Magnetic Resonance

Received Date: Revised Date:

27 April 2015 11 June 2015

Please cite this article as: M. Shen, J. Trébosc, L.A. O’Dell, O. Lafon, F. Pourpoint, B. Hu, Q. Chen, J-P. Amoureux, Comparison of various NMR methods for the indirect detection of nitrogen-14 via protons in solids, Journal of Magnetic Resonance (2015), doi: http://dx.doi.org/10.1016/j.jmr.2015.06.008

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1

Comparison of various NMR methods for the indirect detection of nitrogen-14 via protons in solids Ming Shen,1,2 Julien Trébosc,1 Luke A. O’Dell,3 Olivier Lafon,1* Frédérique Pourpoint,1 Bingwen Hu,2 Qun Chen,2 Jean-Paul Amoureux1,2* 1

UCCS, CNRS UMR 8181, Univ. Lille, Villeneuve d’Ascq 59652, France.

2

Physics Department & Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai 200062, China. 3

Institute for Frontier Materials, Deakin University, Waurn Ponds Campus, Geelong, Victoria 3220, Australia. * emails: [email protected] [email protected] Abstract: We present an experimental comparison of several through-space Hetero-nuclear Multiple-Quantum Correlation experiments, which allow the indirect observation of homo-nuclear single- (SQ) or double-quantum (DQ) 14N coherences via spy 1H nuclei. These 1H-{14N} D-HMQC sequences differ not only by the order of 14N coherences evolving during the indirect evolution, t1, but also by the radio-frequency (rf) scheme used to excite and reconvert these coherences under Magic-Angle Spinning (MAS). Here, the SQ coherences are created by the application of center-band frequency-selective pulses, i.e. long and low-power rectangular pulses at the 14N Larmor frequency, ν0(14N), whereas the DQ coherences are excited and reconverted using rf irradiation either at ν0(14N) or at the 14N overtone frequency, 2ν0(14N). The overtone excitation is achieved either by constant frequency rectangular pulses or by frequency-swept pulses, specifically Wide-band, Uniform-Rate, and SmoothTruncation (WURST) pulse shapes. The present article compares the performances of four different 1H-{14N} DHMQC sequences, including those with 14N rectangular pulses at ν0(14N) for the indirect detection of homonuclear (i) 14N SQ or (ii) DQ coherences, as well as their overtone variants using (iii) rectangular or (iv) WURST pulses. The compared properties include: (i) the sensitivity, (ii) the spectral resolution in the 14N dimension, (iii) the rf requirements (power and pulse length), as well as the robustness to (iv) rf offset and (v) MAS frequency instabilities. Such experimental comparisons are carried out for γ-glycine and L-histidine.HCl monohydrate, which contain 14N sites subject to moderate quadrupole interactions. We demonstrate that the optimum choice of the 1H-{14N} D-HMQC method depends on the experimental goal. When the sensitivity and/or the robustness to offset are the major concerns, the D-HMQC sequence allowing the indirect detection of 14N SQ coherences should be employed. Conversely, when the highest resolution and/or adjusted indirect spectral width are needed, overtone experiments are the method of choice. The overtone scheme using WURST pulses results in broader excitation bandwidths than that using rectangular pulses, at the expense of reduced sensitivity. Numerically exact simulations also show that the sensitivity of the overtone 1H-{14N} D-HMQC experiment increases for larger quadrupole interactions. Key words. Solid-state NMR; 14N isotope; Indirect detection; Overtone; Through-space HETCOR. I-Introduction Nitrogen atoms are ubiquitous in nature and are present in organic molecules, such as pharmaceuticals or biological molecules, e.g. proteins or DNA, as well as organic, hybrid and inorganic materials, such as polyamides, functionalized silica, metal-organic frameworks or nitrides. Nevertheless, nuclear magnetic resonance (NMR) observation of nitrogen nuclei still remains a difficult task owing to the unfavorable properties 14 15 of the two magnetic isotopes, N and N. The latter is a spin-1/2 nucleus, and hence it can easily lead to high 15 1 resolution NMR spectra, even for solids. However, its low gyromagnetic ratio, γ( N)/γ( H) = 0.1, and its very low 15 natural abundance of 0.35% often prevent its NMR detection in isotopically unmodified samples and N 14 enrichment can be expensive and difficult. Conversely, the N isotope has a very high natural abundance of 1,2 99.65%. However, and as explained in recent review articles, this nucleus presents three main difficulties from the NMR point of view.

2 First, it has a low gyromagnetic ratio, γ( N)/γ( H) = 0.072, that leads to a low Larmor frequency, even at high 14 magnetic fields: e.g. ν0( N) = 57.8 MHz at B0 = 18.8 T. This low gyromagnetic ratio decreases the sensitivity, 3/2 14 which is proportional to γ in the case of N detection. It also results in additional experimental difficulties, including weak maximum radio-frequency (rf) fields and long electronic dead-times, since the rf field and deadtime are roughly proportional to γ and 1/γ, respectively. These two limitations often further decrease the 14 2 sensitivity. Second, N has a relatively large electric quadrupole moment of Q = 2.04 fm and thus it is subject to 2 3 large quadrupole interactions with CQ = e qQ/h of up to 6 MHz. Third, it is a spin-1 nucleus, hence has no 14 SQ central transition, and the signal from the two single quantum transitions, N : ΔmN = ±1, for a powder sample is broadened by the very large first-order quadrupole interaction (HQ1) over a width of 3CQ/2, which reaches 9 MHz 14 SQ for CQ = 6 MHz. The combination of these three unfavorable properties leads to a very weak and broad N NMR signal. 14 SQ Generally, the direct detection of one-dimensional (1D) N signal of solids is a challenging task, which has 4 5 notably been achieved in static samples using piecewise acquisition, or frequency-swept excitation. It must be 14 6 mentioned that N nuclei have also been directly detected under MAS, which improves the resolution. However, this type of experiment is quite experimentally demanding and requires optimization of the spectrometer bandwidth and impedance as well as a highly stable and accurate MAS angle and frequency. 14 Furthermore, given the low rf field delivered at ν0( N) by standard solid-state NMR probes, small flip-angles must then be used to broaden the excitation bandwidth, bringing a corresponding reduction in sensitivity. Trains of rotor-synchronized short rectangular pulses in the manner of Delays Alternating with Nutation for Tailored Excitation (DANTE) can achieve efficient broadband excitation at lower rf fields than short pulses by matching 7a,b the rf excitation profile to the spinning sidebands. However, globally these 1D methods are only applicable to samples containing a small number of different nitrogen species. Recently, in order to increase the sensitivity and resolution, two-dimensional (2D) methods have been 14 SQ proposed, in which the N signal is observed under MAS indirectly via a more sensitive ‘spy’ nucleus, such as 1 8 1 14 SQ H. These hetero-nuclear correlations are denoted H-{ N } in the following. For solid samples, most of these 8 indirect detection methods are based on the hetero-nuclear multiple-quantum coherence (HMQC) scheme. It should be mentioned that the hetero-nuclear single-quantum coherence (HSQC) scheme, has also been used, but 9 this sequence is less sensitive than HMQC. For these HMQC experiments, the coherence transfer has been achieved (i) through a combination of J-couplings and second-order quadrupolar-dipolar cross-terms, also known 8 as residual dipolar splittings (RDS), using the (J+RDS)-HMQC sequence, and (ii) more recently, through 10 1 14 SQ 1 14 dipolar couplings, using the D-HMQC one. In the H-{ N } D-HMQC sequence, the H- N dipolar couplings are reintroduced during the defocusing and refocusing delays by applying hetero-nuclear dipolar recoupling 1 10 14 SQ sequences, such as SR421 , on the H channel. As for the 1D NMR observation of N signals, the challenge for 1 14 14 SQ all H-{ N} HMQC experiments is the excitation of the very wide N spectra. A first approach consists of the 14 SQ broadband excitation of N . Optimal broadband excitation using single rectangular pulses is often not compatible with the rf specifications of the common probes. Conversely, DANTE schemes made of short pulses 7a,b can achieve efficient broadband excitation using a low rf field. A second approach relies on sideband selective 14 SQ excitation of the N signal using a single long and low-power rectangular pulse or a train of those pulses in the 11 14 SQ manner of DANTE. In HMQC experiments used for the indirect detection of N , the indirect evolution 14 period, t1, defined as the interval between the centers of the two N pulses or DANTE schemes, must be rotorsynchronized, i.e. equal to an integer number of rotor periods, and the rotor axis must be precisely set to the magic-angle in order to refocus the first-order quadrupole interaction. 14 An alternative approach for the acquisition of high-resolution N spectra from a powder sample is to observe 14 DQ homo-nuclear double-quantum ( N ) transitions, ΔmN = ±2, since these are not broadened by the first-order (HQ1) but only by the second-order (HQ2) quadrupole interaction. 14 DQ The N coherences can be excited from quadrupolar order, Qz, i.e. the population distribution determined by 11 12 the quadrupole interaction, by applying a selective or broadband excitation pulse. Such coherences can also be 14 DQ 13 excited from the z-magnetization by applying: either a ‘classical’ fundamental DQ excitation, NF , or a DQ 14 14-16 14 The fundamental excitation employs rf irradiation at ν0( N) and ‘forbidden’ overtone DQ excitation, NOT . relies on a second-order perturbation of HQ1 by the rf field. This was first demonstrated for static single-crystals 13 11 and powders, and more recently for powders under MAS conditions. The overtone excitation is performed at 14 2ν0( N) and is based on the admixture of the Zeeman states, e.g. some mN = 0 character in the mN = ±1 states when the magnitudes of Zeeman and quadrupole interactions are comparable. The admixture of Zeeman states 14 DQ 14 also allows the direct observation of the overtone signal, i.e. the NOT transition, at 2ν0( N) in 1D experiment. 14,15 Overtone excitation and observation was first reported for single-crystals and powders under static conditions. Direct observation of overtone signals under MAS remained elusive for a long time. Nevertheless, the first 1D 14 DQ 16a NOT powder patterns were recently reported under MAS, and their widths are significantly narrowed owing to nd the elimination of CSA and the partial averaging of HQ2 terms. MAS causes the preferential selection of the 2 spinning sideband which exhibits a frequency shift of twice the spinning speed, 2νR, with respect to the center14 SQ 14 DQ band. Hence, and contrary to the ‘classical’ N and NF signals, the apparent position of the overtone signal 14

1

3 16c

depends on the spinning speed value and also on the sense of rotation. Drawbacks of overtone NMR include (i) bandwidth limitations owing to the long length of the overtone pulse and (ii) its low sensitivity. The excitation bandwidth of overtone pulses can be increased by using amplitude and frequency modulated Wide-band, 16a Uniform-Rate, and Smooth-Truncation (WURST) pulses, at the expense of the sensitivity. Therefore, 1 14 polarization transfers, including H → N cross-polarization and Dynamic Nuclear Polarization, have been 15 17,18 employed to enhance the sensitivity of this technique under static and MAS conditions. We have chosen to use WURST pulses herein because they achieve more uniform broadband excitation than other frequency-swept pulse shapes, such as hyperbolic secant. 14 DQ The N transition can also be observed indirectly using 2D hetero-nuclear correlation experiments, such as 8,9 HMQC and HSQC. The indirect detection via protons improves sensitivity and HMQC is more sensitive than 9c 14 DQ HSQC. In these experiments, the N coherences evolve during the indirect evolution period, t1. At the end of 14 the defocusing and refocusing periods, the N population distribution for the terms contributing to the detected 14 DQ HMQC signals are a combination of Qz and Iz terms. Therefore, the N coherences can be excited and 14 DQ reconverted by either broadband, or selective irradiation. The indirect detection of N coherences via protons 14 1 14 DQ using fundamental irradiation at ν0( N) is denoted H-{ NF } hereafter. It has been shown that its sensitivity is 1 14 SQ 8,11 lower than that of H-{ N }. Overtone irradiation can also be used for the excitation and the reconversion of 14 DQ 16c,19 1 1 14 DQ N coherences in HMQC experiments. Such experiments with H detection are denoted H-{ NOT } in the 1 14 DQ 1 14 SQ 16c,19 14 following. The sensitivity of H-{ NOT } HMQC is comparable to that of H-{ N }, and its N excitation 16c bandwidth can be increased with WURST pulses, again at the expense of the sensitivity. 1 14 SQ 1 14 DQ 1 14 DQ The present article presents a comprehensive comparison of H-{ N }, H-{ NF } and H-{ NOT } D-HMQC 14 experiments. Using a sample of γ-glycine, we compare their sensitivity, their resolution along the N dimension, and their rf field requirements as well as their robustness to rf offset and MAS instabilities. The performances of these methods are also assessed by numerical simulations of the spin dynamics as well as D-HMQC experiments on L-histidine.HCl monohydrate. II. Pulse Sequences The 1H-{14N} D-HMQC sequence is a variant of the (J+RDS)-HMQC experiment, but hetero-nuclear dipolar 1 recoupling schemes are applied on H channel during the defocusing and refocusing periods to reintroduce the 1 14 H- N dipolar couplings. These interactions are usually larger than the sum of J-couplings and residual dipolar splittings and hence shorter recoupling delays are employed in the D-HMQC experiment compared to the (J+RDS)-HMQC one, which minimizes signal loss. Consequently, D-HMQC experiments are often more 12 sensitive than (J+RDS)-HMQC ones. 2 Here, the SR41 scheme was used as a dipolar recoupling method.20 This sequence is a super-cycled version of the symmetry-based R421 sequence, and it offers several advantages for 1H-{14N} D-HMQC experiments,21 including: (i) the suppression of 1H-1H dipolar interactions in the first-order average Hamiltonian, which limits the signal loss during the recoupling periods, τD, (ii) a high robustness to isotropic and anisotropic 1H chemical shifts as well as to rf inhomogeneities, (iii) the reintroduction of 1H-14N dipolar couplings in the form of the longitudinal two-spin order Hamiltonian, which is unaffected by dipolar truncation, (iv) a low rf field requirement, ν1H = 2νR, which permits its use at high MAS frequencies, and (v) a single adjustable parameter, τD, which simplifies the sequence optimization. 1 14 SQ 1 14 DQ 1 14 DQ H-{ N }, H-{ NF } and H-{ NOT } D-HMQC sequences differ by (i) the coherences selected by the phase 1 14 SQ 1 14 DQ 1 cycling during the t1 period, i.e. homo-nuclear SQ coherences for H-{ N } and DQ ones for H-{ NF } and H14 DQ 14 { NOT }, and (ii) the type of pulses applied on the N channel. The SQ and DQ homo-nuclear coherences are 14 selected by two and four-step phase cycling, respectively, of the excitation pulses on N channel. For all sequences, quadrature detection along the indirect dimension was achieved using the States-TPPI procedure by 14 1 14 SQ 1 14 DQ incrementing the phase of the excitation pulse. The N coherences in the H-{ N } and H-{ NF } D-HMQC 14 experiments were excited using a conventional single rectangular pulse at ν0( N) frequency. This single rectangular pulse can achieve either broadband or selective excitation, depending on the pulse length and the rf 11 1 14 SQ 1 14 DQ field strength. H-{ N } and H-{ NF } D-HMQC sequences can also employ DANTE trains, instead of regular single rectangular pulses. Nevertheless, the comparison of single rectangular pulses and DANTE trains 11 for these methods is beyond the scope of the present article and will be presented elsewhere. We have shown 1 14 SQ for H-{ N } D-HMQC experiments that given the rf field delivered by common solid-state NMR probes, 14 center-band selective single N rectangular long pulses lead to slightly lower sensitivity than DANTE trains, but 11 14 1 14 DQ benefit from higher robustness to MAS instabilities and offset. The N coherences in H-{ NOT } D-HMQC 14 experiment were excited and converted at 2ν0( N) using either two rectangular single pulses or two frequency1 14 DQ 1 14 DQ swept WURST pulses. Those sequences are denoted H-{ NOT,SP} and H-{ NOT,WT} D-HMQC in the following. III. Experiments

4 The 1H-{14N} D-HMQC spectra were recorded at 18.8 T on an Avance-III Bruker spectrometer. Experiments were carried out at a high magnetic field because it decreases the second-order quadrupolar broadening and increases the chemical shift range, thus enhancing the resolution along the F1 (14N) and F2 (1H) dimensions. It also increases the Larmor frequency and hence contributes to an improved sensitivity by increasing the Boltzmann factor and by decreasing the second-order quadrupolar broadening of 14N signals. 14 DQ In the case of the 1H-{ NOT } experiments, the excitation of DQ coherences is a priori less efficient at high magnetic field because the admixture of Zeeman states is inversely proportional to B0. Nevertheless, this decreased transfer efficiency at high field is countered by the increased Boltzmann polarisation, the larger 14 DQ accessible rf-fields and the decreased NOT powder pattern widths, which are dominated by the second-order quadrupole interaction. Globally, the ‘on-resonance’ sensitivity largely increases with the magnetic field in case 14 DQ 14 DQ of 1H-{14NSQ} and 1H-{ NF } experiments and it is almost independent of B0 in the case of 1H-{ NOT } ones (see Fig.3 in ref.[17]). Another limitation for the use of high magnetic fields may stem from the limited excitation 14 DQ bandwidths of these sequences, particularly the 1H-{ NOT } experiment (see section IV.d). However, the resonance frequency is the sum of the isotropic chemical shift, which is proportional to B0, and the quadrupole induced shift (QIS), which is very important in the case of 14N and inversely proportional to B0 (Eq.3). As a result, the 14N frequency range, Δν14N, may increase or decrease with B0, depending on the chemical shift and CQ values of the various species, and the evolution with B0 of Δν14N is therefore sample-dependent. For 1H-{14N} experiments, it has been shown that the fastest possible MAS frequencies are preferable in order to decrease the losses related to 1H-1H interactions.19 Ultra-fast MAS (νR ≥ 60 kHz) also allows (i) the observation of 1H resonances with high-resolution along the F2 dimension, (ii) the suppression of all sidebands along F1 and F2, and (iii) the improvement of the resolution along the 14N dimension by applying decoupling schemes using low rf fields during the t1 period.22 High MAS frequencies require the use of rotors with small diameters and fast MAS probes contain small rf coils, which can deliver large rf fields and thereby improve the sensitivity (see ref [19] and Fig.1). Here, we have used a two-channel 1H/X 1.3 mm probe with a MAS frequency νR = 62.5 kHz (TR = 16 μs). The rf field amplitude on the 1H channel was 188 kHz for the π/2 and πpulses, and for the SR421 recoupling was equal to DQ

ν1H = 2νR = 125 kHz. For H-{ N } and H-{ NF } experiments, the rf field amplitude of the 14N pulses, ν1N, was 1 14 DQ constant during the pulses and smaller than 90 kHz owing to the probe specifications. For the H-{ NOT } 14 DQ experiments, the NOT coherences are excited by the total rf field that is sent by the coil along the rotor axis at 14 twice the N Larmor frequency. Hence, contrary to usual NMR methods, both transverse and longitudinal rf field components act on the nuclear spins. Given the complex nutation behavior of the overtone transition, all rf fields on 14N will be defined in the following by their transverse component, ν1N, which can be measured easily. 1 14 DQ 1 14 DQ For H-{ NOT,SP} experiment, ν1N was constant during the pulse and smaller than 90 kHz, whereas for H-{ NOT,WT 1

14

SQ

1

14

}, the rf field amplitude varies during the pulse and its maximal value, ν1Nmax, was also smaller than 90 kHz. 14 DQ Experimentally, in the case of NOT experiments this transverse component was measured indirectly by nutation 14 DQ experiments on 2H nuclei in D2O, as the 2H and NOT Larmor frequencies are close, being equal to 122.8 and 115.6 MHz respectively at 18.8 T. To determine the optimal 14N rf irradiation parameters (Fig.1), the robustness to offset (Fig.3), and to MAS frequency instabilities (Fig.4), γ-glycine powder was used as a test sample, because this compound only contains one crystallographic nitrogen site, and most of the results could thus be obtained by acquiring 1D D-HMQC spectra, for which the t1 delay was fixed to its minimum value. The achievable resolution along the 14N dimension was determined by acquiring 1H-{14N} D-HMQC 2D spectra of γ-glycine and L-histidine.HCl monohydrate (Fig.2). The dipolar recoupling delay, τD, the recycle delay, τRD, the number of scans, NS, and the experimental time, Texp, of 2D spectra are given in the figure captions. All 2D experiments were performed in a rotor-synchronized way with Δt1 = TR, which is mandatory for 1H-{14NSQ} HMQC experiments, but not for 1H1 14 DQ {14NDQ}. For the H-{ NOT,WT} experiment, the phase dispersion in the acquired signal imparted by the WURST excitation pulse was compensated for by a second-order phase correction made to the spectrum after Fourier transformation.23 No 1H decoupling was applied during t1. The 14N chemical shifts were referenced to the resonance of solid-state NH4Cl at 0 ppm. IV. Experimental results

5 IV.a. Optimum ‘on-resonance’ rf parameters We first optimized the amplitude and the length of the 14N rf pulses in order to maximize the signal in the case of ‘on-resonance’ excitation. Fig.1 shows the 1H signal intensity observed in a 1D 1H-{X} D-HMQC spectrum of γ14 DQ 14 DQ glycine, with X = 14NSQ (a), NF (b), and NOT (c,d) versus the rf field amplitude, ν1N (or ν1Nmax), and the pulse DQ

length, τp. In the case of Fig.1d, the NOT,WT excitation was performed using two WURST80 pulses with same direction for the frequency modulation and a frequency sweep range of 40 kHz chosen to cover the full range of 14 OT N signals. 14 DQ In Fig.1, the experiments were performed ‘on-resonance’. This means that for X = 14NSQ and NF , the 14N carrier frequency was equal to the isotropic resonance frequency of 14N nucleus, which is the sum of the isotropic chemical shift and the quadrupolar induced shift (νQIS). Note that the QIS differs between SQ and DQ 14 DQ DQ 1 coherences and they are denoted νSQ QIS and ν QIS , in the following. For H-{ NOT } experiments, the carrier frequency 14

14

DQ

was on resonance with the +2νR ‘overtone spinning sideband’. Indeed, under MAS, NOT signals are split into five overtone sidebands (0, ±1, ±2) separated by the spinning frequency, νR. It has been shown that with a powder 14

DQ

sample, the +2νR overtone sideband provides the highest signal intensity in directly observed NOT spectra and 14 DQ the highest transfer efficiency in 1H-{ NOT } HMQC experiments. The other four sidebands exhibit much lower intensities and in practice are difficult to detect since their intensities are further decreased by the limited bandwidth of the long overtone excitation pulse.16b,c The five overtone sidebands are related to the fact that the efficiencies of rf excitation and detection are both modulated by the sample rotation through the rotor synchronized motion of the quantization axis.24 These sidebands should not be confused with standard spinning sidebands, which are created by the modulation of the transition frequency. The five overtone sidebands are always present, whatever the spinning speed, whereas standard spinning sidebands decrease in number and intensity with increasing MAS rate. 1 14 SQ Fig.1 demonstrates that for γ-glycine, the highest transfer efficiency is achieved by the H-{ N } sequence. For that sequence, the optimal rf parameters, ν1N ≈ 40 kHz and τp ≈ 27 μs, correspond to the center-band selective excitation regime, which is observed for a broad region centered around ν1N ≈ 2νR/3 and τp ≈ 1.5TR.11 The broad band regime corresponding to short high-power pulses is not observed in Fig.1 as efficient transfer in this regime 1 14 DQ would require ν1N ≥ 150 kHz.11 Optimal efficiency of the H-{ NF } experiment also requires a long pulse (τp ≈ 1.5TR), but is ca. 4 times less efficient than the 14NSQ scheme, even when using a two-fold higher rf field (c.f. Fig.1a and b). 14 DQ 1 14 SQ The 1H-{ NOT,SP} experiment is ca. 40% less efficient on resonance than the H-{ N } one (compare Fig.1a and c). Furthermore, the former experiment requires a much larger rf field (ν1N ≈90 kHz) and a 7-fold longer pulse (τp 1 14 SQ ≈200 μs) than H-{ N }.19 The WURST version in this instance was approximately two-fold less efficient than the on-resonance SP version (compare Fig.1c and d), and employed a slightly longer pulse length (τp ≈ 300 instead of 200 μs). In summary, the relative ‘on resonance’ signal intensities of 1H-{X} D-HMQC experiments on γ-glycine are 14 DQ 14 DQ 14 DQ 1 14 SQ equal to 1, 0.6, 0.3, 0.25 for X = 14NSQ, NOT,SP, NOT,WT and NF , respectively. Furthermore, the H-{ N } experiment requires a moderate rf field, whereas the others require high rf fields. IV.b. Absolute referencing of 14N resonances Here, we directly compared the 14N resonance frequencies in the 1H-{X} D-HMQC powder spectra of γ-glycine DQ

DQ

DQ DQ with X = 14NSQ, NF , and NOT . These frequencies, denoted respectively as νSQ exp , ν F,exp and ν OT,exp, correspond to the 14 maximal intensity along the N dimension and were found to be equal to: νSQ νDQ νDQ (1) exp ≈57,801.8; F,exp ≈115,603.7; and OT,exp ≈115,728.4 kHz 14

14

within ±0.1 kHz. For a perfect excitation of 14NSQ and 14NDQ coherences (and more precisely of the +2νR overtone 14 DQ sideband for 1H-{ NOT } experiment), we postulate the following expression for the resonance frequencies DQ DQ DQ 14 νSQ = ν0(14N) + νSQ = 2ν0(14N) + νDQ (2) QIS, ν F QIS , and ν OT = 2ν 0( N) + 2νR + νQIS DQ 14 14 where ν0( N) is the N Larmor frequency, including the isotropic chemical shift. For nitrogen-14, the νSQ QIS and ν QIS frequencies are given by25

6 ௌொ ߥொூௌ ൌ

ವೂ

ఔೂ಺ೄ ଶ

ൌ

మ మ ஼ೂ ൫ଷାఎೂ ൯

ସ଴ఔబ ൫ భర୒൯

(3)

Therefore, the νSQ, νDQ and νDQ F OT frequencies must be related by SQ = 2νSQ and νDQ + 2νR, νDQ F OT = 2ν

(4)

which is supported by the experimental results: SQ DQ SQ (5) νDQ F,exp −2ν exp = (0.1 ±0.2) kHz and ν OT,exp −2ν exp −2ν R = (0.2 ±0.2) kHz. It must be noted that the QIS values are difficult to define accurately from the maximum signal, especially when taking into account the experimental errors and the line-shape distortions related to imperfect excitation of 14N magnetization. Moreover, note that even when using very short excitation pulses, the overtone HQ2 line-shape is 1 14 SQ 1 14 DQ different from that recorded with the same perfect excitation and H-{ N } or H-{ NF } experiments.16b,c This intrinsic difference arises from the facts that (i) the excitation and detection operators are both rotor modulated, and (ii) the overtone excitation efficiency is strongly dependent on the relative orientation of the quadrupole interaction and the magnetic field, and thus varies across the powder pattern.24 Nonetheless, overtone powder patterns can be used to measure chemical shifts and quadrupolar parameters, as has previously been shown by O’Dell and co-workers.16b,c Note also that the 2νR frequency in Eq. 4 is an algebraic value and its sign depends on the sense of sample spinning. The νR frequency is usually positive for Bruker BioSpin and Agilent probes and negative for JEOL ones. According to Eq. 4, a universal ppm scale can be employed for the 14N spectral dimension of 1H-{X} D-HMQC 14 DQ 14 DQ 14 DQ 2D spectra with X = 14NSQ, NF , and NOT . For NF , the frequency must simply be divided by twice the Larmor frequency. The overtone frequency must be first shifted by 2νR frequency (in algebraic value) and then divided by twice the Larmor frequency. In the following, the overtone spectra are displayed with a frequency scale since (i) the overtone frequency depends on the sense and the frequency of MAS and (ii) most overtone spectra have been reported with a frequency scale. IV.c. Resolution An important parameter to consider for 1H-{14N} D-HMQC 2D spectra is the achievable resolution along the 14N dimension. This parameter was analyzed by acquiring the 1H-{X} D-HMQC 2D spectra of γ-glycine (not shown), and the FWHM values of the 14N signal were, within ±60 Hz each, equal to: 14 DQ 14 DQ 14 DQ (6) (FWHM (Hz), X) = (600, 14NSQ); (800, NF ); (780, NOT,SP); and (700, NOT,WT) As seen in the above equation, the FWHMs for the last three experiments corresponding to the evolution of DQ coherences during t1 period are very close with an average value of 760 ±35 Hz. Moreover, as mentioned above, 14 DQ 14 DQ the isotropic shifts, sum of isotropic chemical shifts and QIS, are doubled for 1H-{ NF } and 1H-{ NOT } 14 DQ 1 14 SQ 26 1 experiments with respect to that of H-{ N } (Eq.3). Therefore, the experimental resolutions of H-{ NF } or 14 DQ 1 H-{ NOT } spectra in indirect spectral dimension are about 1.6 (600/380) times higher than that of 1H-{14NSQ}. At a MAS frequency of νR = 62.5 kHz, the 14N-1H dipolar and 14N CSA interactions are nearly completely averaged. Moreover, it has been shown that the residual broadenings due to these interactions are proportional to the 14N coherence order,27 and hence these two residual interactions affect the spectral resolution for the various 1 H-{14N} D-HMQC 2D spectra in an identical way. This is the same for the broadening produced by the distribution in isotropic chemical shifts due to the distribution of local environments. The second-rank part of HQ2 14 DQ 14 SQ is always averaged out by MAS and the fourth-rank term is also twice larger for N than for N coherences. 1 14 SQ 1 14 DQ So, these interactions lead to the same resolution in H-{ N } and H-{ N } D-HMQC experiments. However, the signals along the indirect dimension of 1H-{14N} D-HMQC 2D spectra are also broadened by the decay of 14N-1H multiple-quantum coherences under 1H-1H dipolar couplings.21b Such broadening is identical in Hz for both 14NSQ and 14NDQ signals, instead of the factor of two for previous interactions. Moreover, the 14NSQ signal is also slightly broadened by an imperfect refocusing of HQ1 owing to spinning speed fluctuations and maladjustment of the magic angle or the wobbling of the rotor axis around the direction defined by the magic angle.28 Conversely, such imperfections do not affect the 14NDQ signals, which are not subject to HQ1. In addition, 14 SQ 14 DQ third-order quadrupole interactions also broaden the N signals but not the N ones.29 Nevertheless, this broadening along the indirect dimension is much smaller than that arising from HQ2 terms. Molecular dynamics 14 SQ 14 DQ may also affect the line-shapes of N and N signals in a different manner.30

7 The 14N resolution was further analyzed by acquiring several 1H-{14N} D-HMQC 2D spectra of L-histidine.HCl monohydrate. This sample contains three distinct nitrogen sites,31 N1 (NH3+), N2 (pros nitrogen atom, close to π the side chain in the imidazole ring, denoted N according to IUPAC) and N3 (tele nitrogen atom, far from the τ

side chain in the imidazole ring, denoted N according to IUPAC), which have similar quadrupolar parameters, (CQ (MHz), ηQ) = (1.25, 0.35), (1.29, 0.94), (1.56, 0.26),16c corresponding at 18.8 T to νSQ QIS (Hz, ppm) = (2148, 37), (2836, 49), (2848, 49), for N1, N2, and N3, respectively.16c The projections along the 14N indirect dimension of 14 DQ 1 H-{14NSQ} and 1H-{ NOT } D-HMQC 2D spectra are shown in Fig.2. The assignment of the 14N signals by increasing resonance frequency are N1, N3 and N2. As the quadrupolar parameters for L-histidine.HCl monohydrate are comparable to those of γ-glycine, the optimal rf conditions are similar to those in Fig.1.16b,c The 14 DQ comparison of the projections along the 14N dimension for 1H-{14NSQ} (Fig.2a) and 1H-{ NOT,WT} (Fig.2b) DHMQC 2D spectra confirms that the overtone experiments benefit from a higher spectral resolution (as explained in section IV.b). IV.d. Robustness to offsets A very important attribute of 1H-{14N} sequences is their sensitivity to 14N offset, as the range of 14N resonances, Δδ14N, can extend over several hundreds of ppm.3,26 Such robustness was quantified by 1H-{14N} 1D experiments on glycine for various offsets of the 14N carrier frequency (Fig.3). When the 14N pulses are applied ‘onresonance’, the signal intensities are maximal and their relative amplitudes agree with the previous results shown 1 14 SQ in Fig.1. The H-{ N } D-HMQC sequence benefits from a broad excitation bandwidth, the full width at half maximum (FWHM) of signal intensity versus offset being equal to 43 kHz, i.e. Δδ14N ≈740 ppm at 18.8 T. This FWHM is approximately equal to the inverse of the 14N pulse length, 27 μs. This large offset robustness allows all 14N species to be observed in a single experiment, even at high magnetic field strengths. 1 14 DQ 1 14 DQ The H-{ NF } and H-{ NOT } experiments indirectly detect the DQ coherences and hence the 14N resonance 1 14 DQ frequencies are ‘doubled’ (see Eq.2). In other words, exciting the same 14N chemical shift range for H-{ NF } 1 14 DQ 1 14 SQ and H-{ NOT } experiments requires twice the excitation bandwidth in Hz than for H-{ N }. The comparison 1 14 DQ 1 14 SQ of Figs.3a and b shows that the offset robustness of H-{ NF } sequence is similar to that of H-{ N }. Nevertheless, as noticed in Fig.1, the sensitivity of the former is four-fold lower than that of the latter. The 1 14 DQ excitation bandwidth of H-{ NOT,SP} sequence is FWHM ≈6 kHz (Fig.3c), which corresponds to Δδ14N ≈50 ppm at 18.8 T. This value, which is approximately the inverse of the overtone pulse length, τp ≈200 μs, is insufficient for most samples. It has been shown previously that this limitation can be overcome by using 14N frequencymodulated pulses.16 Here, we have used WURST pulses with a frequency sweep range of 40 kHz and these resulted in an increase in offset robustness to FWHM ≈30 kHz, Δδ14N ≈250 ppm at 18.8 T, albeit at the expense of a drop in the signal intensity. 1 14 SQ 14 DQ 14 DQ The different excitation bandwidths of the H-{X} D-HMQC sequences with X = N , NOT,SP and NOT,WT can also be observed by comparing the projections along the indirect 14N dimension for experiments on Lhistidine.HCl mono-hydrate (see Fig.2). When the carrier frequency coincides with the middle of the 14N 1 14 SQ 1 14 DQ 14 1 14 DQ spectrum, both H-{ N } and H-{ NOT,WT} sequences are able to excite the three N signals, whereas H-{ NOT,SP } scheme only excites the N3 signal at about 180 ppm. Hence, these results confirm that the excitation 1 14 SQ 1 14 DQ 1 14 DQ bandwidths of H-{ N } and H-{ NOT,WT} sequences greatly exceed that of H-{ NOT,SP}. Furthermore, the 1 14 DQ experiments on this sample exemplify how the H-{ NOT,SP} excitation bandwidth is too narrow to cover the full 14 1 14 DQ 1 14 DQ range of N isotropic shifts. Figs.2c-e show the projections of H-{ NOT,SP} and H-{ NOT,WT} spectra when the 14 14 1 14 DQ N carrier frequency coincides with the three N signals. In that case, the H-{ NOT,SP} signal on resonance with 14 1 14 DQ the N carrier frequency is about twice as intense as that of H-{ NOT,WT} experiment, in agreement with the 1 14 DQ 14 experimental results for γ-glycine (see Figs.1 and 3). The H-{ NOT,WT} projections exhibit the three N signals in 1 14 DQ Figs.2c-e, whereas H-{ NOT,SP} experiment is not able to excite signals with a frequency offset of about 8 kHz, 1 14 DQ in agreement with Fig.3. Therefore, H-{ NOT } D-HMQC experiments should be used with WURST pulses, except (i) when all 14N resonance frequencies are relatively close, or (ii) when selective excitation is desired. IV.e. Robustness to MAS frequency instabilities

8 Another important property is the robustness to spinning speed stability. Indeed, most of the recently introduced solid-state NMR sequences use rotor-synchronized pulses and/or signal acquisition, and rotor frequency instability can lead to signal loss and/or t1-noise in 2D or 3D experiments.32 This may be particularly the case when trying to manipulate or acquire broad spectra, such as those arising from 14N. This stability is more difficult to achieve for small diameter rotors used for experiments at high MAS rate because these rotors have a small inertia. The MAS frequency robustness of 1H-{14N} D-HMQC experiments was quantified by measuring the signal intensity of glycine versus the delay, t1, between the 14N excitation and reconversion schemes (Fig.4). The delay between the end of the first SR421 scheme and the beginning of the second one was kept constant since 1 14 this recoupling scheme is non-γ-encoded. Thus, curves similar to those of Fig.4 should be obtained for H-{ N} (J+RDS)-HMQC sequences. In practice, for 1H-{14N} D-HMQC experiments, the robustness to MAS frequency fluctuations also depends on the choice of the hetero-nuclear recoupling sequence and the length of the defocusing and refocusing periods. 1 14 SQ As seen in Fig.4a, the H-{ N } D-HMQC sequence is sensitive to MAS frequency instability and the t1 delay should be chosen as an integer, or half-integer for SR421 , multiple of the rotor period, TR. It should be noted that peaks at half-integer multiples of TR are due to the use of the SR421 recoupling sequence. At νR = 62.5 kHz, a deviation of t1 delay from an integer or half-integer multiple of the rotor period, TR, by 0.8 μs = 0.05TR suffices to 1 14 SQ produce a two-fold decrease in signal intensity. For the H-{ N } D-HMQC sequence, the maximal evolution period, t1max, is typically about 50TR and hence the MAS rate fluctuations must be smaller than ± 60 Hz at νR = 62.5 kHz in order to maintain the rotor-synchronization of 14N pulses. High MAS frequency stability is required 1 14 SQ 14 SQ for the H-{ N } D-HMQC sequence because N coherences are (i) subject to the large HQ1 term and (ii) excited and reconverted by center-band selective pulses. The rotor-synchronization of these pulses ensures 11 identical phases of the center-band during the excitation and reconversion pulses. Note that the required 1 14 SQ stability for the H-{ N } D-HMQC experiment using center-band selective excitation by long pulses is not as demanding as that for strong short pulses or DANTE trains,7 or for STMAS experiments;28 since the pulses are much shorter in the latter cases. 1 14 DQ The comparison of Figs.4a and b shows that the H-{ NF } D-HMQC sequence benefits from higher robustness 1 14 SQ to MAS instabilities than H-{ N }. At νR = 62.5 kHz, a two-fold decrease in signal intensity is observed when 1 14 DQ the t1 delay deviates from an integer multiple of the rotor period by 0.1TR. Hence, the H-{ NF } D-HMQC 1 14 SQ sequence tolerates two-fold larger MAS frequency fluctuations than the H-{ N } experiment (± 120 Hz at νR = 1 14 DQ 62.5 kHz). The H-{ NF } D-HMQC method benefits from higher robustness to MAS instabilities because, 14 SQ 14 DQ contrary to N , the N transition is not subject to the HQ1 interaction. Nevertheless, the rotor-synchronization 14 14 DQ 11 of N pulses is still required because the N coherences are primarily excited by center-band selective pulses and the center-band must have identical phases during the two pulses. Figs.4c and d show that overtone sequences, using single rectangular pulses or WURST excitations, are completely insensitive to spinning speed fluctuations. This high robustness stems from the facts that (i) DQ coherences are only subject to the small HQ2 terms, and (ii) the ‘forbidden’ excitation of the +2νR ‘overtone 14 SQ 14 DQ spinning sideband’ fundamentally differs from the ‘traditional’ excitation of N and N transitions using center-band selective pulse. In the latter cases, the Hamiltonian describing the rf irradiation depends on the amplitude and the phase of the center-band of 14NSQ spectrum during the pulses, or in other words of the modulation of HQ1 during sample spinning. Conversely, the overtone excitation corresponds to a selective excitation of the +2νR ‘overtone spinning sideband’. It has been shown that the number of rotor-encoded overtone sidebands is always equal to five and that their relative amplitudes are independent of the MAS frequency.16b Thus the amplitude and the phase of the Hamiltonian describing the selective excitation of the +2νR overtone spinning sideband does not depend on the MAS frequency and the initial rotor position. Consequently, the Hamiltonians describing the overtone excitation during the two pulses are identical for any t1 delay. A detailed theoretical description of the selective excitation of the +2νR overtone spinning sideband is beyond the scope of this article and will be presented elsewhere. As overtone excitation does not depend on the spinning speed, the 14 DQ 14 DQ indirect spectral width of 1H-{ NOT } HMQC can be finely adjusted to the width of the NOT spectra. Such a choice permits the minimization of the experimental time, by minimizing the sampling along the indirect spectral dimension. Conversely, the indirect spectral width of 1H-{14NSQ} D-HMQC sequence must be an integer multiple of νR to refocus the dephasing of 14NSQ coherences under HQ1.

9 V. Simulations of 14N overtone experiments under MAS 1

14

DQ

For a better understanding of the H-{ NOT,SP} D-HMQC experiments, exact numerical simulations of a single pulse experiment with overtone excitation were performed at B0 = 18.8 T and νR = 62.5 kHz, using methods previously described.16b Figure 5a maps the intensity of +2νR overtone sideband of the nitrogen site in γ-glycine (CQ = 1.18 MHz and ηQ = 0.53) excited by an ‘on-resonance’ single rectangular overtone pulse as a function of rf field strength and pulse length. The simulations of Fig.5a are in good agreement with the experimental data of Fig.1. In both cases, the optimum pulse length at an rf field strength of ν1N = 90 kHz is approximately τp ≈200 μs. As expected, optimal pulse length decreases for higher ν1N field, which should increase the excitation bandwidth. 14 DQ Small rf coil diameters are therefore highly advantageous for NOT -based experiments. Simulations similar to those of Fig.5a were also performed for a CQ value twice as large as that of γ-glycine (Fig.5b). For overtone pulses, the doubling of the CQ constant produces a doubling of the effective overtone nutation rf field. As a result the optimum pulse length with ν1N = 90 kHz is approximately of τp ≈100 μs. Since the optimal pulse length and rf field both depend on CQ constant, the same rf pulse cannot yield optimal excitation for all 14N sites when they have different CQ values.16c 14 DQ Figure 5c shows the simulated NOT powder pattern of the +2νR overtone sideband. The maximum signal 14 intensity occurs at a shift of νDQ QIS = 4.1 kHz with respect to 2ν 0( N) + 2νR (see Eq.2). Its FWHM is equal to 600 Hz,

which is slightly less than that observed experimentally for H-{ N } D-HMQC spectrum of γ-glycine (Eq.6: 1

14

DQ

760 ±35 Hz). This difference, compared to the decay of N coherences due to HQ2 alone, can be ascribed to the additional decay of 1H-14N multiple-quantum coherences produced by 1H-1H dipolar couplings. Figure 5d shows the simulated intensity of the +2νR overtone sideband of the γ-glycine as function of the offset of 14

DQ

the overtone pulse with ν1N = 90 kHz and τp = 200 μs. These parameters correspond to those used for the experiments of Fig.3 and the simulated excitation bandwidth of approximately 6.5 kHz is in excellent agreement with the experimentally determined value of 6 kHz (see Fig.3). VI. Conclusion DQ

DQ

We have compared three different 1H-{X} D-HMQC experiments with X = 14NSQ, NF and NOT for the indirect detection of 14N isotopes in solids. The 1H-{14NSQ} single-quantum D-HMQC experiment using center-band selective excitation is very efficient and easy to optimize. Its pulses on the 14N channel require a moderate rf field amplitude of ca. 2νR/3 with durations of ca. one to two rotor periods. Its broadband version with single 14N rectangular pulses is typically difficult to use efficiently owing to the rf specifications of standard solid-state NMR MAS probes. Nevertheless, its version with rotor-synchronized DANTE excitation with DKN trains scales up the effective rf-field by the factor K. Compared to center-band selective pulses, DANTE trains using short pulses lead to a slightly better sensitivity but are more sensitive to offset.7 The offset bandwidth of one single DKN DANTE train is equal to FWHM ≈1.35/KTR,7d whereas that of an HMQC sequence using two DKN trains is 14

14

equal to FWHM ≈ 1.09/KTR.33 Therefore, at νR = 62.5 kHz, the excitation bandwidth of 1H-{14NSQ} D-HMQC with DKN trains is FWHM ≈23 and 17 kHz, for K = 3 and 4, respectively, instead of 43 kHz for the center-band selective pulses (Fig.2a). The 1H-{14NDQ} 2D spectra benefit from a higher resolution along the 14N dimension than for 1H-{14NSQ}, since (i) the spinning speed instability and the deviation of the rotor axis from the magic angle result in an incomplete refocusing of HQ1, which broadens the 14NSQ signal, and (ii) the isotropic shifts are doubled in the DQ spectrum compared to the SQ one, whereas the broadening produced by 1H-1H dipolar couplings is identical for all 1H14 DQ {14N} spectra. Furthermore, the 1H-{ NOT } experiment, which employs a selective excitation of the +2νR 14 DQ overtone sideband, is much more robust to MAS instability than the other sequences. Nevertheless, the 1H-{ NF DQ 14 } and 1H-{ NOT } D-HMQC experiments require large rf fields, especially the former which is also the less 14 DQ sensitive method. The observation of 1H-{ NOT,SP} signals is efficient when ‘on-resonance’ but very sensitive to rf offsets, and it should thus be reserved (i) for samples exhibiting a narrow dispersion of 14N signals or (ii) when 14 DQ a selective excitation is targeted. Its version with frequency-swept pulses, 1H-{ NOT,WT}, is more broadband and can thus be used for all samples, but at the expense of reduced sensitivity. As a conclusion, the choice of the 1H-{14N} D-HMQC sequence depends on the targeted criteria.

10 DQ

When the major concern is the highest resolution, 1H-{ NOT } D-HMQC methods should be preferred. Conversely, when the maximum sensitivity is essential, 1H-{14NSQ} D-HMQC sequence with selective long pulses or broadband DANTE schemes should be employed. 14 DQ The 1H-{14NSQ} and 1H-{ NOT } D-HMQC methods are also complementary for the observation of 14N nuclei subject to different quadrupole interactions. Indeed, the nutation field and maximum sensitivity of overtone excitation both increase with the CQ value (Fig.5a,b), whereas it is the reverse case for 14NSQ excitation. The present experimental analysis has been performed on glycine and histidine samples containing 14N sites with moderate CQ values ranging from 1.18 to 1.53 MHz. However, many 14N sites in solids exhibit much larger CQ values,3 for which overtone spectroscopy may become the method of choice. It must also be remembered that HMQC correlation spectra involving quadrupolar nuclei are not directly quantitative. Nevertheless, they may allow a determination of the proportion of the various species and their shielding and quadrupolar parameters by fitting the line-shapes with sophisticated software in the case of 1H{14N}.34,17,16b,c 14

Acknowledgements. This work was supported by: Fédération de Recherche CNRS (FR-3050) Infrastructure de recherche RMN à Très Hauts Champs, French contract ANR-2010-jcjc-0811-01. FP, JT, OL, and JPA are grateful for funding provided by Région Nord/Pas de Calais, European Union (FEDER), CNRS, French Ministry of Scientific Research, USTL, ENSCL, and CortecNet. MS is grateful for the financial support from China Scholarship Council. Authors would like to thank Z. Gan for fruitful discussions. OL is also grateful for the funding supported by Chinese Academy of Sciences President’s international fellowship initiative and “National Natural Science Foundation of China” (Grant No. 21450110412). 1) S. Cavadini, Indirect detection of nitrogen-14 in solid-state NMR spectroscopy, Prog. Nucl. Magn. Reson. Spectrosc. 56 (2010) 46-77. 2) L.A. O’Dell, Direct detection of 14N in solid-state NMR spectroscopy, Prog. Nucl. Magn. Reson. Spectrosc. 59 (2011) 295-318; L.A. O’Dell, 14N Overtone Magic Angle Spinning NMR, Annual Reports on NMR Spectroscopy, doi.org/10.1016/bs. arnmr.2015.04.003 3) J.P. Yesinowski, E.A. Hill, Solid-state NMR spectroscopy of inorganic materials, J. Fitzgerald (Ed.), ACS Symposium Series, American Chemical Society, Washington, DC (1999) 358-376 (Chapter 13). 4) E.A. Hill, J.P. Yesinowski, Wide-line 14N NMR in solids and reorientation-induced redistribution of isochromats, J. Am. Chem. Soc. 118 (1996) 6798-6799. 5) (a) L.A. O’Dell, R.W. Schurko, Fast and simple acquisition of Solid-State 14N NMR spectra with signal enhancement via population transfer, J. Am. Chem. Soc. 131 (2009) 6658-6659; (b) L.A. O’Dell, R.W. Schurko, Static solid-state 14N NMR and computational studies of nitrogen EFG tensors in some crystalline amino acids, Phys. Chem. Chem. Phys. 11 (2009) 7069-7077; (c) L.A. O’Dell, R.W. Schurko, K.J. Harris, J. Autschbach, C.I. Ratcliffe, Interaction tensors and local dynamics in common structural motifs of nitrogen: a Solid-State 14N NMR and DFT study, J. Am. Chem. Soc. 133 (2011) 527-546. 6) (a) H.J. Jakobsen, H. Bildsøe, J. Skibsted, T. Giavani, 14N MAS NMR spectroscopy: the nitrate ion, J. Am. Chem. Soc., 123 (2001) 5098-5099; (b) H.J. Jakobsen, H. Bildsoe, Z. Gan, W.W. Brey, Experimental aspects in acquisition of wide bandwidth solid-state MAS NMR spectra of low-γ nuclei with different opportunities on two commercial NMR spectrometers J. Magn. Reson. 211 (2011) 195-206. 7) (a) V. Vitzthum, M.A. Caporini, S. Ulzega, G. Bodenhausen, Broadband excitation and indirect detection of nitrogen-14 in rotating solids using delays alternating with nutation (DANTE), J. Magn. Reson. 212 (2011) 234-239; (b) V. Vitzthum, M.A. Caporini, S. Ulzega, J. Trébosc, O. Lafon, J.P. Amoureux, G. Bodenhausen, Uniform broadband excitation of crystallites in rotating solids using interleaved sequences of delays alternating with nutation, J. Magn. Reson. 223 (2012) 228-236; (c) D. Carnevale, V. Vitzthum, O. Lafon, J. Trébosc, J.P. Amoureux, G. Bodenhausen, Broadband excitation in solid-state NMR of paramagnetic samples using Delays Alternating with Nutation for Tailored Excitation (‘Para-DANTE’), Chem. Phys. Lett. 553 (2012) 68-76; (d) X. Lu, J. Trébosc, O. Lafon, D. Carnevale, S. Ulzega, G. Bodenhausen, J.P. Amoureux, Broadband excitation in solid-state NMR using interleaved DANTE pulse trains with N pulses per rotor period, J. Magn. Reson. 236 (2013) 105-116. 8) (a) S. Cavadini, A. Lupulescu, S. Antonijevic, G. Bodenhausen, 14N NMR spectroscopy using residual dipolar splittings in solids, J. Am. Chem. Soc. 128 (2006) 7706-7707; (b) Z. Gan, Measuring amide nitrogen quadrupolar coupling by highresolution 14N/13C NMR correlation under magic-angle spinning, J. Am. Chem. Soc. 128 (2006) 6040-6041; (c) S. Cavadini, S. Antonijevic, A. Lupulescu, G. Bodenhausen, Indirect detection of 14N in solids via protons by nuclear magnetic resonance spectroscopy, J. Magn. Reson. 182 (2006) 168-172;; (d) S. Cavadini, S. Antonijevic, A. Lupulescu, G. Bodenhausen, Indirect detection of 14N in solid-state NMR spectroscopy, ChemPhysChem, 8 (2007) 1363-1374. 9) (a) S. Antonijevic, N. Halpern-Manners, Probing amide bond nitrogens in solids using 14N NMR spectroscopy, Solid State NMR, 33 (2008) 82-87; (b) S. Cavadini, A. Abraham, G. Bodenhausen, Coherence transfer between spy nuclei and 14N in solids, J. Magn. Reson. 190 (2008) 160-164. (c) O. Lafon, Q. Wang, B. Hu, F. Vasconcelos, J. Trébosc, S. Cristol, et al., Indirect detection via spin-1/2 nuclei in solid state NMR spectroscopy: application to the observation of proximities between protons and quadrupolar nuclei., J. Phys. Chem. A. 113 (2009) 12864-12878.

11 10) (a) Z. Gan, 13C/14N hetero-nuclear multiple-quantum correlation with rotary resonance and REDOR dipolar recoupling, J. Magn. Reson, 184 (2006) 39-43; (b) Z. Gan, J.P. Amoureux, J. Trébosc, Proton-detected 14N MAS NMR using homo-nuclear decoupled rotary resonance, Chem. Phys. Lett. 435 (2007) 163-169; (c) J.A. Jarvis, I.M. Haies, P.T.F. Williamson, M. Carravetta, Phys. Chem. Chem. Phys., 15 (2013) 7613-7620; (d) Y. Nishiyama, Y. Endo, T. Nemoto, H. Utsumi, K. Yamauchi, K. Hioka, T. Asakura, Very fast magic angle spinning 1H-14N 2D solid-state NMR: sub-micro-liter sample data collection in a few minutes, J. Magn. Reson. 208 (2011) 44-48. 11) M. Shen, J. Trébosc, O. Lafon, Z. Gan, F. Pourpoint, B. Hu, Q. Chen, J.P. Amoureux, Solid-state NMR indirect detection of nuclei experiencing large anisotropic interactions using frequency-selective pulses, Submitted. 12) D. Merlet, M. Sarfati, B. Ancian, J. Courtieu, P. Lesot, Description of natural abundance deuterium 2D-NMR experiments in weakly ordered liquid-crystalline solvents using a tailored cartesian spin-operator formalism, Phys. Chem. Chem. Phys. 2 (2000) 2283-2290. 13) (a) P. Brunner, M. Reinhold, R.R. Ernst, Double quantum cross polarization: hetero-nuclear excitation and detection of NMR double quantum transitions in solids, J. Chem. Phys. 73 (1980) 1086-1094. (b) M. Reinhold, P. Brunner, R.R. Ernst, Double quantum cross polarization via dipolar order, J. Chem. Phys. 74 (1981) 184-188. (c) T.K. Pratum, M.P. Klein, heteronuclear cross-polarization of Solid-State 14N NMR powder patterns, J. Magn. Reson. 55 (1983) 421-437. 14) M. Bloom, M.A. LeGros, Direct detection of two-quantum coherence, Can. J. Phys., 64 (1986) 1522. 15) (a) R. Tycko, S.J. Opella, High-resolution 14N overtone spectroscopy: an approach to natural abundance nitrogen NMR of oriented and polycrystalline systems, J. Am. Chem. Soc., 108 (1986) 3531-3532; (b) R. Tycko, P.L. Stewart, S.J. Opella, Peptide plane orientations determined by fundamental and overtone 14N NMR, J. Am. Chem. Soc. 108 (1986) 5419-5425; (c) R. Tycko, S.J. Opella, Overtone NMR spectroscopy, J. Chem. Phys. 86 (1987) 1761-1774; (d) R. Tycko, Overtone spectroscopy of quadrupolar nuclei, in: R.K. Harris, R.E. Wasylishen (Eds.), Encyclopedia of Magnetic Resonance, John Wiley, Chichester, 3425-3430. 16) (a) L.A. O'Dell, C.I. Ratcliffe, 14N magic angle spinning overtone NMR spectra, Chem. Phys. Lett., 514 (2011) 168-173; (b) L.A. O'Dell, A. Brinkmann, 14N overtone NMR spectra under magic angle spinning: experiments and numerically exact simulations, J. Chem. Phys., 138 (2013) 064201; (c) L.A. O'Dell, R. He, J. Pandohee, Identifying H-N proximities in solidstate NMR using 14N overtone irradiation under fast MAS, CrystEngComm, 15 (2013) 8657-8667. 17) I.M. Haies, J.A. Jarvis, H. Bentley, I. Heinmaa, I. Kuprov, P.T.F. Williamson, M. Carravetta, 14N overtone NMR under MAS: signal enhancement using symmetry-based sequences and novel simulation strategies, Phys. Chem. Chem. Phys, 17 (2015) 6577-6587. 18) A.J. Rossini, L. Emsley, L.A. O’Dell, Dynamic nuclear polarization enhanced 14N overtone MAS NMR spectroscopy, Phys. Chem. Chem. Phys. 16 (2014) 12890-12899. 19) Y. Nishiyama, M. Malon, Z. Gan, Y. Endo, T. Nemoto, Proton-nitrogen-14 overtone two-dimensional correlation NMR spectroscopy of solid-sample at very fast magic angle sample spinning, J. Magn. Reson., 230 (2013) 160-164. 20) A. Brinkmann, A.P.M. Kentgens, Magic-Angle-Spinning solid-state NMR spectroscopy for the determination of hydrogen bond lengths, J. Am. Chem. Soc. 128 (2006) 14758-14759. 21) (a) B. Hu, J. Trébosc, J.P. Amoureux, Comparison of several hetero-nuclear dipolar recoupling NMR methods to be used in MAS HMQC/HSQC., J. Magn. Reson. 192 (2008) 112-122; (b) M. Shen, J. Trébosc, O. Lafon, F. Pourpoint, B. Hu, Q. Chen, J.P. Amoureux, Improving the resolution in proton-detected through-space heteronuclear multiple quantum correlation NMR spectroscopy., J. Magn. Reson. 245 (2014) 38-49. 22) (a) S. Cavadini, A. Abraham, G. Bodenhausen, Proton-detected 14N NMR by recoupling of hetero-nuclear dipolar interactions using symmetry-based sequences, Chem. Phys. Lett., 445 (2007) 1-5; (b) S. Cavadini, V. Vitzthum, S. Ulzega, A. Abraham, G. Bodenhausen, Line-narrowing in proton-detected nitrogen-14 NMR, J. Magn. Reson. 202 (2010) 57–63; (c) M. Shen, J. Trébosc, O. Lafon, F. Pourpoint, B. Hu, Q. Chen, et al., Improving the resolution in proton-detected through-space heteronuclear multiple quantum correlation NMR spectroscopy., J. Magn. Reson. 245 (2014) 38-49. 23) E. Kupce, R.J. Freeman, Adiabatic pulses for wideband inversion and broadband decoupling, J. Mag. Reson. A 115 (1995) 273-276; L.A. O’Dell, The WURST kind of pulses in solid-state NMR, Solid State NMR, 55-56 (2013) 28-41; I. Hung, Z. Gan, On the practical aspects of recording wideline QCPMG NMR spectra, J. Magn. Reson. 204 (2010) 256-265. 24) (a) L. Marinelli, S. Wi, L. Frydman, A density matrix description of 14N overtone nuclear magnetic resonance in static and spinning solids, J. Chem. Phys. 110 (1999) 3100-3112; (b) N.M. Trease, P.J. Grandinetti, J. Chem. Phys. 128 (2008) 052318. 25) A. Jerschow, From nuclear structure to the quadrupolar NMR interaction and high-resolution spectroscopy, Prog. Nucl. Magn. Reson. Spectrosc. 46 (2005) 63-78. 26) J. Mason, Nitrogen NMR, in: R.K. Harris, R.E. Wasylishen (Eds.), Encycl. Magn. Reson., John Wiley & Sons, Ltd, 2007: p. March 15th. 27) M. Pruski, A. Bailly, D.P. Lang, J.P. Amoureux, C. Fernandez, Studies of hetero-nuclear dipolar interactions between spin-1/2 and quadrupolar nuclei using REDOR during MQ evolution, Chem Phys Lett, 307 (1999) 35-40. 28) C. Huguenard,F. Taulelle, B. Knott, Z. Gan, Optimizing STMAS, J. Magn. Reson. 156 (2002) 131-137; J.P. Amoureux, C. Huguenard, F. Engelke, F. Taulelle, Unified representation of MQMAS and STMAS NMR of half-integer quadrupolar nuclei, Chem. Phys. Lett. 356 (2002) 497-504. 29) Z. Gan, P. Srinivasan, J.R. Quine, S. Steuernagel, B. Knott, Third-order effect in solid-state NMR of quadrupolar nuclei, Chem. Phys. Lett. 367 (2003) 163-169. 30) J.M. Griffin, A.J. Miller, A.J. Berry, S. Wimperis, S.E. Ashbrook, Dynamics on the microsecond timescale in hydrous silicates studied by solid-state 2H NMR spectroscopy. Phys. Chem. Chem. Phys, 12 (2010) 2989-2998. 31) H. Fuess, D. Hohlwein, S.A. Mason, Acta Crystallogr. Sect. B: Struct. Crystallogr. Cryst. Chem., 33 (1977) 654.

12 32) (a) T. Giavani, H. Bildsoe, J. Skibsted, H.J. Jakobsen, 14N MAS NMR spectroscopy and quadrupole coupling data in the characterization of the IV-III phase transition in ammonium nitrate, J. Phys. Chem. B 106 (2002) 3026-3032; (b) E. Hughes, T. Gullion, A simple inexpensive and precise magic angle spinning speed controller, Solid State NMR. 26 (2004) 16-21. 33) Y. Li, J. Trébosc, O. Lafon, , B. Hu, F. Pourpoint, Q. Chen, J.P. Amoureux, Indirect detection of broad spectra in solidstate NMR using interleaved DANTE trains, to be published. 34) Mathematica (Wolfram Research, Inc., Champaign, IL) with SpinDynamica 2.5.3, M.H. Levitt, http://www. spindynamica. soton.ac.uk.

Figure caption Fig.1. Experimental ‘on-resonance’ 1H signal (arbitrary unit) of 1H-{X} D-HMQC 1D experiment on γ-glycine for X = (a) DQ DQ DQ NSQ, (b) 14NF , (c) 14NOT,SP and (d) 14NOT,WT, versus ν1N (ν1Nmax in d) and τp values at B0 = 18.8 T and νR = 62.5 kHz. These signals were acquired using τRD = 0.5 s, τD = 200 µs, and NS = 64. In each subfigure, the maximal signal amplitude is indicated by a black cross.

14

Fig.2. Experimental projections along the 14N dimension of the 1H-{X} D-HMQC 2D spectra of histidine-HCl mono-hydrate DQ with X = (a) 14NSQ, (b-e) 14NOT with single-pulse (SP, blue) or WURST (red) excitation. The position of the 14N carrier frequency is indicated by a black arrow. The 2D spectra result from the averaging of NS = 8 transients for each of N1 = 512 t1 increments with Δt1 = 16 ms. The τRD delay is 1s, which corresponds to a total acquisition time of Texp = 70 mn for each 2D spectrum. The ν1N and τp values correspond to the optimal ‘on-resonance’ conditions indicated by a black cross in Fig.1. The other experimental parameters are identical to those of Fig.1. DQ

Fig.3. Experimental 1H signal of 1H-{X} D-HMQC 1D experiment on γ-glycine for X = (a) 14NSQ, (b) 14NDQ, (c) 14NOT,SP and DQ (d) 14NOT,WT versus offset in kHz. The offset step differs between the subfigures. In (b), the amplitude has been multiplied by two. The ν1N and τp values correspond to the optimal on-resonance conditions indicated by a black cross in Fig.1. The other experimental parameters are identical to those of Fig.1. Fig.4. Experimental on-resonance 1H signal of 1H-{X} D-HMQC 1D experiment on γ-glycine for X = (a) 14NSQ, (b) 14NDQ, DQ DQ (c) 14NOT,SP and (d) 14NOT,WT versus the t1/TR ratio. The delay between the two SR421 schemes was kept constant. The other experimental parameters are identical to those of Fig.3. In (a), signals obtained with smaller increment near t1 = TR are also displayed. In (b), the amplitude has been multiplied by two. Fig.5. (a) Simulated signal intensity of the +2νR overtone sideband of glycine (CQ = 1.18 MHz and ηQ = 0.53) versus ν1N (kHz) and τp (μs) values. (b) The same as (a) with CQ = 2.36 MHz. Simulations for glycine with ν1N = 90 kHz and τp = 200 μs, of (c) the powder pattern, and (d) of the intensity versus rf offset.

Figure

35

(a)

200 "2dmapSQHMQC"

30

( s) p

15

150

50

5 0

400

"2dmapDQHMQC" 40

25 20

30

15

20

10 10

5 0

10

20

30

40

50

60

70

80

(c)

0

0

120 400 "2dmapovertoneSP" u 1:2:($3*4)

350

100

300

( s)

50

100 10

p

(b)

30

25 20

35

0

10

20

30

40

50

60

70

80

(d)

0

60 "2dmapovertonewurst" u 1:2:($3*4)

350

50

300 80

250 200

60

150

40

100

40

250 200

30

150

20

100 20

50 0

0 0

10

20

30

40

50

1N

(kHz)

60

70

80

90

10

50 0

0 30

40

50

60

1N

70

(kHz)

80

90

Figure

(a) (b) (c)

(d) WURST SP

(e)

300

200

100

δ14N (ppm)

0

Figure

(a) SQ

(b) DQ I*2

(c) overtone SP

(d) overtone + WURST pulse

-60

-40

-20

0

offset (kHz)

20

40

60

Figure

(a)

(b) (c)

(d) 0

0.5

1.0

t1/TR

1.5

2.0

Relative intensity

Figure

400

40

(a)

32

Pulse length / µs

300

24 16

200

8 0 -8

100

-16 -24

0 25

0

50

75 RF power / kHz

100

125

150 Relative intensity

400

81

(b)

64

Pulse length / µs

300

47 30

200

13 -4 -21

100

-38 -55

0 0

25

50

75 RF power / kHz

(c)

100

Relative intensity

4

125

150

(d)

3 2 1 0

6

5

4 kHz

3

2

-10

-5

0 RF offset / kHz

5

10

13 3

G Grapphical ab bstraact

Exp perim men ntall 1H-{14N} glyc g cinee siggnal, veersu us offseet with w (a) single-- orr (b--d) dou d ublee-qu uanttum m seelecttion n, w with fu undameental (b) or o overrton ne excittatio onss witth (c) hard h d pu ulsees orr (d) WUR W STss. (b b) Th he amp a plitu ude hass been multiplied d byy tw wo.

H Highhlighhts

• • • • •

1

1 H-{ H 14 N} D-HM D MQ QC HET H TCO OR Frrequuency sseleectivve sing s gle-qquaantum exci e itatiion 1 HMQ H QC deteectiion of 14 N Ovvertoone 1 HMQ H QC deteectiion of 14 N singgle-- annd doub d ble-quaantuum Hard H d-puulse andd WUR W RST T ovvertoone exccitaationn