From Operando Raman Mechanochemistry to “NMR Crystallography”: Understanding the Structures and Interconversion of Zn-Terephthalate Networks Using Selective ¹⁷O-Labeling

Abstract

The description of the formation, structure, and reactivity of coordination networks and metal–organic frameworks (MOFs) remains a real challenge in a number of cases. This is notably true for compounds composed of Zn²⁺ ions and terephthalate ligands (benzene-1,4-dicarboxylate, BDC) because of the difficulties in isolating them as pure phases and/or because of the presence of structural defects. Here, using mechanochemistry in combination with operando Raman spectroscopy, the observation of the formation of various zinc terephthalate compounds was rendered possible, allowing the distinction and isolation of three intermediates during the ball-milling synthesis of Zn₃(OH)₄(BDC). An “NMR crystallography” approach was then used, combining solid-state NMR (¹H, ¹³C, and ¹⁷O) and density functional theory (DFT) calculations to refine the poorly described crystallographic structures of these phases. Particularly noteworthy are the high-resolution ¹⁷O NMR analyses, which were made possible in a highly efficient and cost-effective way, thanks to the selective ¹⁷O-enrichment of either hydroxyl or terephthalate groups by ball-milling. This allowed the presence of defect sites to be identified for the first time in one of the phases, and the nature of the H-bonding network of the hydroxyls to be established in another. Lastly, the possibility of using deuterated precursors (e.g., D₂O and d₄-BDC) during ball-milling is also introduced as a means for observing specific transformations during operando Raman spectroscopy studies, which would not have been possible with hydrogenated equivalents. Overall, the synthetic and spectroscopic approaches developed herein are expected to push forward the understanding of the structure and reactivity of other complex coordination networks and MOFs.

Introduction

In recent years, the use of mechanochemistry in material synthesis has undergone an exponential development. Ball-milling (BM) syntheses have notably proved to be of great interest in pharmaceutical materials screening, battery development, or even catalysis.^1,2 More than offering a drastic reduction of bulk-solvent use in the syntheses, it offers a precise control over the stoichiometry of the reaction, as well as simple and straightforward synthetic procedures.^3,4 However, because of the closed environment of mechanochemical reactions (especially when conducted in stainless steel jars), these syntheses are still considered as “black boxes”. In this context, the last decade has seen the emergence of in situ time-resolved analysis techniques applied to BM syntheses, which allow the evolution of the reaction medium to be monitored without needing to interrupt the milling process and open the jar. Among them, two are of particular interest: powder X-ray diffraction (pXRD) and Raman spectroscopy.⁵⁻¹¹ While the first requires the use of a synchrotron beamline, the second can be set up in standard laboratories thanks to bench-top Raman spectrometers and probes. Such systems permit reaction rates to be followed, complex reaction schemes to be understood, and can also shed light on the appearance of yet unknown/unobserved reaction intermediates.^12,13 Furthermore, these experimental installations can also be coupled with thermal imaging cameras to have an even more precise description of the course of the chemical reactions.^14,15 It is worth mentioning that meticulous kinetic studies from operando Raman spectroscopy experiments have been recently proposed.^16,17

Metal–organic frameworks (MOFs) represent a major family of porous materials, with porosities that can exceed 50% of their volume. The impressive range of their applications extends from fuel/gas storage to catalysis or even cancer therapy.¹⁸⁻²⁰ Various synthetic routes have been developed for obtaining and/or shaping these materials, like solvothermal or microwave syntheses, and also mechanochemistry.²¹⁻²⁴ The numerous properties of MOFs being strongly and directly related to their crystallographic structures, it is then of primary importance to precisely control their synthesis to obtain pure phases through meticulous and reproducible procedures.²⁵ In this context, in situ monitoring experiments can help observe or even trap interesting intermediates. For example, recently, Jones et al. presented the use of in situ¹H liquid-state NMR for observing the formation of a Ni-MOF.²⁶ Insights into the nucleation and crystal growth processes of this MOF were obtained through the kinetic studies carried out at different temperatures. Concerning MOFs formed by mechanosynthesis, a recent example by Karadeniz et al. revealed the interplay between MOF-525 and its polymorph PCN-223, two porphyrinic zirconium MOFs. In their case, in situ pXRD monitoring allowed the MOF-525 to be detected as an intermediate during the formation of PCN-223, depending on the synthetic conditions used.²⁷ A similar experimental procedure also allowed the presence of metastable MOFs in the formation of dia-Zn(MeIm)₂ (MeIm: 2-methylimidazole) to be demonstrated.⁶

Among MOFs, those based on zinc (Zn-MOFs) have been the focus of much interest because of their high porosity and/or interesting applications. In particular, because zinc is a biocompatible metal, zinc-containing materials such as the porous scaffold Zn-xWE43 (WE43 being a Mg alloy) or the Zn₂(BDC)₂(diazabicyclooctane) MOF (with BDC standing for 1,4-benzenedicarboxylate) have been studied as candidates for implants and drug delivery.^28,29 Furthermore, Zn-MOFs have shown great potential for applications as catalysts, anodes for lithium batteries, or materials for uranium adsorption/detection.³⁰⁻³² Among the numerous Zn-MOFs described to date, coordination complexes based on Zn²⁺-dicarboxylate motifs have been found to exist as a very broad diversity of structures.³³ For instance, the well-known structure of MOF-5 (Zn₄O(BDC)₃) exhibits 4-fold-coordinated zinc sites with bridging carboxylate functions, while MOF2 (Zn₂(BDC)₂) presents one 5-fold-coordinated zinc ion with four bridging oxygens from carboxylate ligands and one additional coordinated water molecule.^34,35 Interconversion between these two structures has also been demonstrated in specific synthetic conditions (high pressure, temperatures up to 120 °C).³⁶ Furthermore, the different investigations performed so far on coordination networks involving Zn²⁺ ions and terephthalate ligands (noted Zn-BDC from hereon) have enabled many other structures to be described, e.g., Zn(BDC)(H₂O),³⁷ Zn₃(OH)₄(BDC)·6H₂O,³⁸ and (Zn₃(BDC)₄·4H₂O)_n,³⁹ just to name a few. Here, we decided to focus our attention on four other Zn-BDC structures: a dihydrated zinc terephthalate coordination polymer, Zn(BDC)(H₂O)₂,⁴⁰ and three lamellar phases involving terephthalate and hydroxyl ligands (see Table 1).

Table 1. Zn-BDC Compounds Studied in This Contribution, Together with Structural Details, as Initially Reported in the Corresponding Refs (41−44).

The Zn(BDC)(H₂O)₂ phase (noted here ZTA1) consists of a “zig-zag” arrangement of chains formed by distorted tetrahedral zinc environments, in which each Zn²⁺ ion is linked to two monodentate BDC ligands and two water molecules (Table 1). Regarding the three lamellar structures, two are polymorphs of general formula Zn₂(OH)₂(BDC) (noted here as ZTA2a and ZTA2b). The structure of the latter polymorph (ZTA2b) has been more thoroughly described so far:⁴³ it exhibits only one penta-coordinated zinc environment, with well-ordered π-stacked BDC units. The structure of the other polymorph (ZTA2a), a contrario, reveals more uncertainties. In fact, although several articles have reported the presence of this phase (by looking at pXRD data), no consensus has been found in the literature about its structure. Rodríguez et al. attributed this compound to a distorted MOF-5 phase caused by a default in one of the binding modes of carboxylate groups to Zn²⁺.⁴⁵ Thirumurugan and Rao observed this phase as an intermediate in their hydrothermal reactions when using a Zn/BDC 1/1 molar ratio but were not able to propose an exact formula for it.⁴⁶ Hirai et al. were unable to isolate a pure phase of this material but were able to detect it in a microwave synthesis when using a Zn/BDC ratio of 2/1 and working at specific pH (between 5.7 and 4.9).³⁸ In 2016, a crystal structure was reported in the Cambridge crystallography database (CCDC ECATIO), which matches the pXRD pattern of this same phase and corresponds to the Zn₂(OH)₂(BDC) chemical formula (with a Zn/BDC ratio of 2/1). Unfortunately, no information on its synthesis could be found.⁴² When looking at the crystallographic file, it appears that it corresponds to a lamellar structure with two distinct zinc sites in octahedral geometry, which are separated by one bridging oxygen belonging to a hydroxyl (Table 1). Between two layers, the BDC ligands exhibit a “zig-zag” configuration of the aromatic rings. Hence, both polymorphs differ by subtle modifications of the BDC arrangements, resulting in different coordination modes of the metal centers, and the ZTA2a phase displays a slightly smaller interlamellar spacing than ZTA2b (about 10.1 vs 10.8 Å, respectively) (see Table 1). Lastly, the fourth phase of interest in this work is Zn₃(OH)₄(BDC) (noted ZTA3).⁴⁴ This other lamellar Zn-hydroxide terephthalate is formed in a 3/1 Zn/BDC stoichiometry. Three distinct zinc environments were found to be present in the Zn–O polyhedra layers (with tetra/penta/hexacoordination, according to the initially reported crystallographic file), which are connected to each other through terephthalate and hydroxyl anions.

Diffraction techniques remain, at the moment, the most common analytical tool to obtain structural information on Zn-BDC MOFs. However, preparing good-quality crystals can be very challenging. Moreover, providing exact positions for hydrogen atoms from pXRD data remains an issue, and only two out of the four reported structures in Table 1 fully report hydrogen positions (aromatics and hydroxyls/water). In complement to XRD, chemical and structural information on these phases can also be obtained through FTIR spectroscopy. For example, the O–H stretching modes of hydroxyls and water molecules are easily observed above 3000 cm^–1 and can inform on the H-bond networks in presence. Moreover, the C–O stretching modes of the carboxylate groups are identified by the presence of antisymmetric and symmetric vibrations (noted ν_as(COO^–) and ν_s(COO^–)) between 1650–1540 and 1450–1360 cm^–1, respectively, and information on the binding mode of the carboxylate ligands to zinc metal centers can a priori be proposed on the basis of the difference in wavenumbers between these two frequencies (Δ = ν_as(COO^–) – ν_s(COO^–)).^47,48 However, very few articles have used solid-state NMR (ssNMR) techniques to probe the local environment of the atoms in Zn-BDC structures, even for some of the more readily observable isotopes such as ¹H and ¹³C.^37,49−51 This can be seen as surprising considering that the use of ssNMR alongside XRD analyses is more and more frequent to elucidate the structure of MOFs and coordination polymers, including phases prepared using mechanochemistry,⁵²⁻⁵⁴ such as MOF crystal-glass composites (MIL-53/ZIF-62),⁵⁵ cadmium-imidazole ZIFs,⁵⁴ or even paramagnetic MOFs (e.g. Ni(II)-MOF-74, bimetallic ZnCu-MOF-74).^56,57 More generally speaking, such techniques could help numerous debates, some of which will be discussed more extensively later in this contribution, which still remain about such zinc structures and their intermediates.

To further elucidate the details of the structure of these phases, ¹⁷O ssNMR is a very attractive tool. Oxygen-17 is a spin-5/2 quadrupolar nucleus of low natural abundance (0.04%), which presents a very wide chemical shift range (exceeding 1000 ppm) and also a very wide variation of quadrupolar interaction (QI) parameters (C_Q, η_Q). Moreover, recent investigations have shown that ¹⁷O ssNMR spectroscopy can be of high interest for studying the structure of coordination polymers and MOFs.⁵⁸⁻⁶¹ For example, Martins et al. have shown the powerful use of ultra-high-field ¹⁷O ssNMR experiments (35.2 T) applied on ¹⁷O-enriched Mg and Al MOFs (α-Mg₃(HCOO)₆ and MIL-53(Al)). Impressively, they were able to distinguish up to 12 inequivalent oxygen sites for as-made and activated phases of α-Mg₃(HCOO)₆.⁶² While ¹⁷O ssNMR is still not routinely used to investigate the structure of MOFs, our recent developments of cost-efficient ¹⁷O-labeling schemes using mechanochemistry,^60,63,64 including for carboxylic ligands like terephthalic acid,⁶⁵ imply that such nonroutine NMR now become conceivable for studying Zn-BDC structures like those in Table 1.

In this contribution, we report a detailed investigation of the structure and interconversion between the four aforementioned Zn-BDC systems (ZTA1, ZTA2a, ZTA2b, and ZTA3) synthesized using mechanochemistry. We first show how by modifying the initial stoichiometry of the reaction, and carefully monitoring the course of the BM reactions using operando Raman spectroscopy and thermal analysis of the milling jar, each of these different phases can be easily isolated in a pure form and subsequently studied by ex situ methods (FTIR, powder X-ray diffraction, multinuclear ssNMR), allowing further information about their structures to be gained. Moreover, we then demonstrate for the first time how, by using this synthetic methodology, it becomes possible to perform selective ¹⁷O isotopic labeling of the different types of oxygen environments in such MOF structures (by enriching either the water/hydroxyl ligands or the carboxylate ones), thereby opening the way to a level of structural insight, which had not been achieved before. Lastly, we show how the combination of high-resolution ¹⁷O NMR analysis of these compounds and ab initio calculations of NMR parameters using the gauge-including projector-augmented wave method (GIPAW) can improve and/or correct the structural models of these phases, thereby paving the way to other investigations on the structure/property relationships in complex coordination networks and MOFs.

Materials and Methods

Reagents

The following commercial precursors were used as received without further purification: terephthalic acid (C₈H₆O₄, Janssen Chemicals, 98% purity, noted here H₂BDC), deuterated terephthalic acid (C₈D₄H₂O₄, with deuteration on the aromatic H positions only, Cambridge Isotope Laboratories, 98% purity, noted here d₄-BDC), ZnO (particle size < 5 μm, Sigma-Aldrich, 99.9% purity), and D₂O (Sigma-Aldrich, 99.9% purity).¹⁷O-labeled water (with ∼40, 70, or ∼90% ¹⁷O-enrichment) was purchased from CortecNet.

¹⁷O-enriched H₂BDC was prepared using two different mixer mills: either a vibratory mixer mill (Restch MM400 apparatus, as published elsewhere)⁶⁵ or by scaling-up the procedure using a planetary ball mill (Fritsch P7 apparatus), as described below.

H₂BDC-Enrichment in a Planetary Mixer Mill (Fritsch P7)

H₂BDC (330 mg, 1.98 mmol, 1.0 equiv) and CDI (708.6 mg, 4.37 mmol, 2.2 equiv) were introduced into the stainless steel grinding jar (20 mL inner volume), to which 80 stainless steel beads (5 mm diameter) were then added. The jar was closed and subjected to grinding for two cycles of 10 min at 500 rpm, with a break of 30 min between each cycle in the P7 planetary ball mill. ¹⁷O-labeled (91.5%, 106 μL, 5.96 mmol, 3.0 equiv) water was then added into the jar, and the mixture was subjected to further grinding for two cycles of 10 min at 500 rpm, with a break of 30 min between each cycle. To help recover the product, nonlabeled water (2 mL) was added into the jar, and the content was subjected to grinding for 1 min at 500 rpm. Then, the suspension was transferred to a beaker (together with 9 mL of nonlabeled water used here to rinse the jar). The medium was acidified under stirring to pH ∼ 1 with an aqueous solution of HCl (6 M, 44 drops). The white precipitate was immediately filtered on a glass frit, washed (3 × 2 mL of 1 M HCl and then 3 × 2 mL of ultra-pure water), and then dried under vacuum. Yield: 312.7 mg (∼95%). Enrichment yield: 84%, enrichment level: 38.7%, as determined by the MS.

Mechanochemical Syntheses of Zn-BDC Networks

The Fritsch Pulverisette-23 (P23) vertical mixer mill was used in all reactions. BM syntheses were conducted in 10 mL inner-volume Perspex jars,⁷ with two 10 mm diameter zirconia beads.

A typical mechanochemical synthesis was performed as follows. First, ZnO was introduced into a Perspex jar followed by H₂BDC (see Table 2 for exact masses). Both powders were gently mixed with a spatula before adding the two zirconia beads. Then, the appropriate amount of water was added. The jar was quickly closed, and parafilm was added to cover the junction of top/bottom parts of the jar to avoid any leaks as well as to increase air/water tightness. The system was then subjected to grinding for milling times ranging from 5 to 90 min in the P23 mixer mill operating at 30 or 50 Hz. The reactor was then opened, and the resulting compound was left to dry in air for several hours. The solid powder was subsequently recovered by scraping the edges of the reactor and beads with a spatula. All samples were obtained as white powders. After several uses, reactors can exhibit signs of erosion; however, no trace of Perspex was detected in the final compounds by ¹³C ssNMR and ex situ Raman spectroscopy.

Table 2. Synthetic Details for BM Experiments of Zn-BDC Compounds^a.

					milling parameters
code name	formula	mZnO (mg)	mH2BDC (mg)	VH2O (μL)	frequency (Hz)	duration (min)
ZTA1	Zn(BDC)(H2O)2	129.9	265.1	115	30	30
		1.59 mmol	1.59 mmol	6.38 mmol
		1 equiv	1 equiv	4 equiv
ZTA2a	Zn2(OH)2(BDC)	178.4	182.1	40	50	20
		2.19 mmol	1.09 mmol	2.22 mmol
		2 equiv	1 equiv	2 equiv
ZTA2b	Zn2(OH)2(BDC)	197.8	201.9	350	50	90
		2.43 mmol	1.21 mmol	19.43 mmol
		2 equiv	1 equiv	16 equiv
ZTA3	Zn3(OH)4(BDC)	262.1	178.4	310	50	90
		3.22 mmol	1.07 mmol	17.21 mmol
		3 equiv	1 equiv	16 equiv
17O-Labeled Compounds
ZTA1-*BDC	Zn(BDC)(H2O)2	34.0	71.0*	17	50	5
		0.42 mmol	0.42 mmol	0.94 mmol
		1 equiv	1 equiv	2 equiv
ZTA2a-*BDC	Zn2(OH)2(BDC)	60.9	63.6*	15	50	12
		0.75 mmol	0.37 mmol	0.83 mmol
		2 equiv	1 equiv	2 equiv
ZTA2a-*OH		178.4	182.1	40*	50	20
		2.19 mmol	1.10 mmol	2.20 mmol
		2 equiv	1 equiv	2 equiv
ZTA2b-*OH	Zn2(OH)2(BDC)	171.2	174.7	40*	50	90
		2.10 mmol	1.05 mmol	2.20 mmol
		2 equiv	1 equiv	2 equiv
ZTA3-*BDC	Zn3(OH)4(BDC)	121.1	84.3*	50	50	90
		1.49 mmol	0.49 mmol	2.77 mmol
		3 equiv	1 equiv	5 equiv
ZTA3-*OH		119.6	81.4	50*
		1.47 mmol	0.48 mmol	2.77 mmol
		3 equiv	1 equiv	5 equiv
Deuterated Compounds
ZTA1-d4BDC	Zn(d4BDC)(H2O)2	125.0	262.0d	315	30	15
		1.54 mmol	1.54 mmol	17.48 mmol
		1 equiv	1 equiv	11.2 equiv
ZTA2a-d4BDC	Zn2(OH)2(d4BDC)	197.4	207.5d	350	50	90
		2.42 mmol	1.22 mmol	19.43 mmol
		2 equiv	1 equiv	16 equiv
ZTA2b-OD	Zn2(OD)2(BDC)	188.1	192.8	335d	50	90
		2.31 mmol	1.16 mmol	18.47 mmol
		2 equiv	1 equiv	16 equiv
ZTA3-d4BDC	Zn3(OH)4(d4BDC)	242.0	168.7d	90	50	90
		2.97 mmol	0.99 mmol	5.00 mmol
		3 equiv	1 equiv	5 equiv
ZTA3-OD	Zn3(OD)4(BDC)	250.7	171.1	350d
		3.08 mmol	1.03 mmol	19.06 mmol
		3 equiv	1 equiv	19 equiv

^a“*” is used for ¹⁷O-labeled materials and “d” is used for deuterated materials (prepared from D₂O or d₄-BDC precursors).

Further details on the amounts of reagents and milling conditions used to isolate each of the four phases, in their nonlabeled, ¹⁷O-labeled, or deuterated forms, are provided in Table 2. It is worth noting that for the synthesis of ¹⁷O-labeled compounds, amounts of reagents were reduced to further decrease the cost of obtaining an enriched material for ssNMR analyses.

“Routine” Characterizations of the Different Phases

Powder XRD analyses were performed on an X’Pert MPD diffractometer using Cu Kα₁ radiation (λ = 1.5406 Å), with the operation voltage and current maintained at 40 kV and 25 mA, respectively. Diffractograms were recorded between 2θ = 5 and 50° in the Bragg–Brentano configuration, with a step size of 0.050° and a time per step of 60 s.

Infrared (IR) spectra were recorded on a Perkin Elmer Spectrum 2 FTIR spectrometer, using the ATR measurement mode (4000–400 cm^–1 range, four repetitions).

Operando Raman Spectroscopy and Thermal Analysis

In this manuscript, we use the term operando to insist on the fact that spectroscopic measurements were acquired during the milling (and without stopping the milling), as opposed to in situ analyses, which can be performed on a medium when the milling is stopped for a few seconds/minutes.⁶⁶

The Raman measurements were performed with a WP 785 ER Raman spectrometer (Wasatch Photonics) using an IPS (Innovative Photonic Solutions) laser source (working at ∼300 mW, wavelength of λ = 785 nm) and a contactless probe head (the probe’s focal length is 11 mm for a 1 mm diameter spot). The positioning of the Raman probe was optimized such that the focus of the laser was localized at the inner jar wall. A laser exposure time of 5 s with five accumulations was chosen, and spectrum collection was repeated every 30 s during the milling time. Raman spectra were collected across a range from 220 to 3120 cm^–1, and a dark spectrum (laser OFF) was also recorded before starting the experiment and directly subtracted from the recorded spectra. No degradation of the jar and/or the compound by the laser was observed during the operando Raman experiments. The thermal imaging camera showed that during the milling, the laser beam did not prompt any additional heat around its focal point on the Perspex jar walls (compared to the jar’s overall temperature). Moreover, no noticeable degradation of the reagents and products was observed under the operating conditions used here for operando Raman analyses. The obtained Raman spectra were processed with the OriginPro 2021 program. Baseline correction (asymmetric least-squares smoothing) was first applied to each spectrum, and then normalization of the signal of the Perspex jar (at 1728 cm^–1 where no overlapping from reagents nor product is expected) was performed, prior to the subtraction of the signal of the vibrating empty jar (see Supporting Information Figure S1). The reproducibility of the kinetic measurements performed is illustrated in Figure S2.

Temperature measurements were carried out using an OPTRIS PI450i thermal imaging camera. The optical resolution of the device is 382 × 380 pixels. The temperature accuracy measured using this system is estimated to be about ±2 °C. Four different temperatures were recorded every 2 s during the milling, focusing on three to four different areas, corresponding to the top, center, and bottom of the jar, and also to the “hot spot” at each time point.

Cautionary remarks: the laser source used here is designed as a class IV during all procedures of operation, maintenance, and service, meaning that safety measures must be taken accordingly. The use of laser-specific safety goggles is required when operating this laser, as failure to wear appropriate eye protection may result in permanent eye damage. Caution must be taken to never look directly into the laser beam and make sure the laser direction points toward the wall. The power range for the laser is 0–580 mW. Measurements presented in this contribution were performed at ∼300 mW in a dedicated isolated dark room with proper safety signs on the door, and operators received specific security training before using this equipment.

Solid-State NMR

¹H ssNMR

¹H ssNMR experiments were performed on a Varian VNMRS 600 MHz (14.1 T) NMR spectrometer, using a Varian 3.2 mm probe tuned to ¹H (599.82 MHz). Spectra were recorded under magic angle spinning (MAS) conditions, with a spinning speed of 16 kHz and under a regulated airflow (0 °C). A ¹H direct excitation pulse of 3.3 μs was used. Recycle delays used for the Zn-BDC compounds were 4 and 15 s for pure H₂BDC. Additional ¹H MAS NMR experiments were performed on a Bruker Avance NEO 850 MHz (20.0 T) NMR spectrometer, using a Bruker 1.3 mm probe, tuned to ¹H (850.23 MHz), with a spinning speed of 60 kHz. Adamantane was used as a secondary reference for setting ¹H chemical shifts (δ_iso = 1.8 ppm with respect to tetramethylsilane).

¹³C ssNMR

¹³C ssNMR experiments were performed on a VNMRS 600 MHz (14.1 T) NMR spectrometer, using a Varian 3.2 mm HX probe tuned to ¹H (599.82 MHz) and ¹³C (150.81 MHz). All one-dimensional (1D) ¹³C NMR spectra consist of cross-polarization experiments (CP) under MAS conditions, with spinning frequencies from 4 to 15 kHz. A ¹H excitation pulse of 3.3 μs was used prior to a ramped spin-lock pulse of 5 ms contact time. Acquisition was performed under spinal-64 ¹H-decoupling (∼75 kHz RF). Recycle delays used for the Zn-BDC compounds were 4 and 15 s for pure H₂BDC. ¹³C chemical shifts were referenced to adamantane used as a secondary reference (high-frequency peak at 38.5 ppm with respect to tetramethylsilane).

¹⁷O ssNMR

¹⁷O ssNMR experiments were first performed on all samples on a Varian VNMRS 600 MHz (14.1 T) NMR spectrometer, using either Varian 3.2 mm HX or HXY probes, or a Phoenix 3.2 mm probe, tuned to ¹H (599.82 MHz) and ¹⁷O (81.31 MHz). Spectra were recorded under MAS conditions under a regulated airflow (0 °C), with spinning speeds ranging from 16 to 20 kHz, depending on the sample. Additional ¹⁷O MAS NMR experiments were performed at lower field on a VNMRS 400 MHz (9.4 T) NMR spectrometer, using a Varian 3.2 mm HXY probe tuned to ¹H (399.92 MHz) and ¹⁷O (54.21 MHz). Moreover, higher-field ¹⁷O NMR analyses were performed first on a Bruker Avance NEO 850 MHz (20.0 T) NMR spectrometer, using Bruker 1.3 or 4 mm probes tuned to ¹H (850.23 MHz) and ¹⁷O (115.26 MHz) or on a Bruker Avance III HD 800 MHz (18.8 T) NMR spectrometer using a 3.2 mm Low-E probe spinning at 16 kHz (ν_L(¹H) = 800.12 MHz and ν_L(¹⁷O) = 108.47 MHz) to perform multimagnetic field data fitting (see Supporting Information Figures S14–S17). A double frequency sweep (DFS) excitation scheme was used consisting of a 1 ms pulse with a sweep between 1 MHz and 50 kHz⁶⁷⁻⁶⁹ for ¹⁷O-signal enhancement (when possible) prior to the excitation pulse, and spinal-64 ¹H-decoupling was applied during acquisition. The acquisition parameters used for each sample (including number of scans and recycle delay D1) can be found in Table 3. The D-HMQC (dipolar-heteronuclear multiple-quantum coherence) sequence used an SR4₁² recoupling scheme on the ¹H channel with a radio-frequency field of twice the spinning speed; the detection is performed on the ¹⁷O channel.⁷⁰¹⁷O chemical shifts were referenced externally to D₂O at −2.7 ppm (or tap water at 0.0 ppm).

Table 3. Main Acquisition Parameters for ¹⁷O NMR Experiments.

sample	field (T)	Ørotor (mm)	expt	νrot (kHz)	D1 (s)	NS (/#t1)	νRF (kHz)	pulses (μs)	dec {1H} (kHz)
ZTA1-*BDC	14.1	3.2	DFS-One pulse	18	2	1024	∼40	2	75
	20.0	3.2	DFS-One pulse	14.286	4	256	10	8.33	60
ZTA2a-*BDC	9.4	3.2	DFS-One pulse	18	1	20480	∼40	2	50
	14.1	3.2	DFS-One pulse	18	1	6144	∼40	2	75
	18.8	3.2	MQMAS	16	5	768/22	∼16	3/1/5	62.5
	20.0	4	DFS-One pulse	14.286	1	4096	10	8.33	60
ZTA2a-*OH	9.4	3.2	DFS-One pulse	18	2	512	∼40	2	50
	14.1	3.2	DFS-One pulse	18	2	3072	∼40	2	75
	20.0	4	DFS-One pulse	14.286	2	256	10	8.33	60
ZTA2b-*OH	9.4	3.2	DFS-One pulse	18	4	3072	∼40	2	50
	14.1	3.2	DFS-One pulse	18	4	1024	∼40	2	75
	20.0	4	DFS-One pulse	14.286	4	512	10	8.33	60
ZTA3-*BDC	9.4	3.2	DFS-One pulse	18	2	6144	∼40	2	50
	14.1	3.2	DFS-One pulse	18	2	12288	∼40	2	75
	18.8	3.2	MQMAS	16	5	768/32	∼16	3/1/5	62.5
	20.0	4	DFS-One pulse	14.286	2	1024	10	8.33	60
		1.3	(DFS) D-HMQC	60	4	9216/24	20	4.167/8.33	75
ZTA3-*OH	9.4	3.2	DFS-One pulse	18	4	10240	∼40	2	50
	14.1	3.2	DFS-One pulse	18	4	1024	∼40	2	75
	18.8	3.2	MQMAS	16	5	480/32	∼16	3/1/5	62.5
	20.0	4	DFS-One pulse	14.286	4	256	10	8.33	60
		1.3	(DFS) D-HMQC	60	2	1536/23	20	4.167/8.33	75

The NMR parameters were obtained by fitting the spectra using DMfit.⁷¹

NMR Conventions

In this article, the “Herzfeld–Berger” convention is used to describe the magnitude of the chemical shift anisotropy (CSA) arising from the magnetic shielding interaction.⁷² The isotropic chemical shift (δ_iso), span (Ω), and skew (κ) are given as follows

1

2

3

The principal components of the chemical shift tensor are ordered as follows: δ₁₁ ≥ δ₂₂ ≥ δ₃₃.

The quadrupolar interaction is described by two parameters, the quadrupolar coupling constant (C_Q) and the asymmetry parameter (η_Q)

4

5

Here, V₁₁, V₂₂, and V₃₃ are the principal components of the traceless electric field gradient (EFG) tensor, with |V₃₃| ≥ |V₂₂| ≥ |V₁₁|, e is the fundamental electronic charge, Q is the nuclear quadrupole moment, and h is Planck’s constant.

GIPAW-DFT Calculations

The unit cell parameters were set to the X-ray diffraction parameters and kept fixed during geometry optimizations to ensure consistency between experimental and optimized structures. Missing protons (in CCDC PUCYAO and PEKGAO structures) were initially positioned to be consistent with the expected structure of the system, and successively, only protons and then all atomic positions were relaxed with the VASP (Vienna ab initio simulation package) code⁷³ based on the Kohn–Sham density functional theory (DFT) and using a plane-wave pseudopotential approach. For the two structures mentioned above, several initial OH proton positions were tested. The NMR parameters were then calculated within the Kohn–Sham DFT using the QUANTUM-ESPRESSO code,^74,75 keeping the atomic positions equal to the values previously calculated with VASP. The PBE generalized gradient approximation⁷⁶ was used, and the valence electrons were described by norm-conserving pseudopotentials⁷⁷ in the Kleinman–Bylander form.⁷⁸ The shielding tensor was computed using the gauge-including projector-augmented wave (GIPAW) approach,⁷⁹ which enables the reproduction of the results of a fully converged all-electron calculation.⁸⁰ The isotropic chemical shift δ_iso is defined as δ_iso = −[σ – σ^ref], where σ is the isotropic shielding and σ ^ref is the isotropic shielding of the same nucleus in a reference system. For ¹H, ¹³C, and ¹⁷O, previously used referencing was adopted using glycine,⁸¹ calcite,⁸² and a series of silicates,⁶³ respectively. Diagonalization of the symmetrical part of the calculated tensor then provides its principal components σ₁₁, σ₂₂, and σ₃₃ from which the chemical shift components δ₁₁, δ₂₂, and δ₃₃ can be calculated. The principal components V_xx, V_yy, and V_zz of the electric field gradient (EFG) tensor are obtained by diagonalization of the tensor. The quadrupolar interaction can then be characterized by the quadrupolar coupling constant C_Q and the asymmetry parameter η_Q, which are defined previously. The experimental value of the quadrupole moment of ¹⁷O (Q = −25 × 10^–30 m²) was used to calculate C_Q.⁸³

Results and Discussion

Given the rich variety of structures involving Zn²⁺ and BDC ligands, in terms of Zn/BDC ratios, H₂O or OH species linked to the Zn²⁺, and also carboxylate coordination modes, we decided to investigate the formation of these compounds using mechanochemistry, a synthetic method that had not yet been tested in a systematic way for these phases. The general idea was to see if conditions allowing each of these phases to be isolated in a pure form and high yield could be achieved using mechanochemistry. Coupled with the use of ¹⁷O-enriched precursors, this approach would allow all of the different forms of Zn-BDC phases to be analyzed with an unprecedented level of detail, thereby providing novel information in the debate related to the structures of these materials.

Formation and Isolation of Zinc (Hydroxyl-) Terephthalate Coordination Complexes

Observation of Several Intermediate Phases Using Operando Raman Analyses

The three reagents, namely, H₂BDC, ZnO, and H₂O, were introduced in a Perspex milling jar to enable the monitoring of different vibration bands by Raman spectroscopy. An illustrative example of the formation of various Zn-BDC phases in a one-pot mechanochemistry synthesis is shown in Figure 1 for a ZnO/H₂BDC/H₂O stoichiometry of 2/1/16. In Figure 1a, the evolution of the Raman signals during the milling for two selected regions of the spectra is presented after data processing (see the Materials and Methods Section for more information). It is worth noting that water plays multiple roles in this system: (i) it is directly involved in the reaction, as a reagent, by forming complexes with the Zn²⁺ cations and in some cases leads to the formation of coordinated hydroxyls, (ii) it plays the role of a liquid-assisted grinding (LAG) agent, allowing a better mixing of the solid reagents and thus a better efficiency of the milling process,^84,85 (iii) it guarantees the formation of paste-like materials during the milling, which will be homogeneously spread over the inner surfaces of the milling jar. The latter point is of great importance for the operando Raman measurements, justifying the use of an excess amount of water in this particular example (≥16 equiv).

Figure 1.

(a) Top: starting compounds used for the operando experiment with their respective equivalents and milling conditions. Bottom: time-resolved operando Raman spectra of two selected areas (gray and orange, see panel (b)): 820–900 cm^–1 range on the left and 1380–1480 cm^–1 range on the right (see Figure S2 for repetitions of the analysis in this zone); two-dimensional (2D) plots have been overlapped with spectra from panel (b). (b) Experimental ex situ Raman spectra of H₂BDC (in black), ZTA1 (in green), ZTA2a (in red), and ZTA2b (in blue) when isolated under the milling conditions reported in Table 2. (c) Experimental ex situ pXRD diffractograms of the corresponding “pure” compounds when isolated under the milling conditions reported in Table 2 (in black), and comparisons with the simulated pXRD data for these Zn-BDC phases (colored lines; pXRD simulated from available CIF files in the CCDC database, as reported in Table 1).

The system undergoes several structural transformations during the milling process, as illustrated by the shifts of specific vibration modes in the time-resolved experiment (Figure 1a). Between 800 and 900 cm^–1 (Figure 1b, gray box), we focus on the aromatic breathing and out-of-plane (γCH) deformation region.⁸⁶⁻⁸⁸ The signal from the starting H₂BDC molecule quickly shifts from 832 to 878 cm^–1, and after ∼8 min of milling, it evolves to 862 cm^–1, and finally moves to 865 cm^–1 in the final compound. Accordingly, corresponding modifications of the ν_s(COO^–) band can also be easily followed between 1380 and 1480 cm^–1.

The products successively formed were identified as ZTA1, ZTA2a, and ZTA2b. Their Raman spectra and corresponding pXRD patterns are presented in Figure 1b,c in green, red, and blue, respectively. No formation of ZTA3 occurred when using an initial Zn/BDC molar ratio of 2/1 in the milling.

Kinetic Study of the Transformations

In view of describing further the reactions occurring during the milling, the kinetic evolution of the reaction medium established by Raman spectroscopy (as determined by the integration of the previously mentioned vibration modes between 800 and 900 cm^–1, for example) was complemented with the thermal imaging of the milling jar (Figure 2). Simultaneous recording of temperatures at three different areas of the jar (referred to as top, center, and bottom) led to the plots displayed in Figure 2b. The inset (dashed gray area in Figure 2a) presents the first 500 s of the experiment, corresponding to the transformation of H₂BDC into ZTA1 (from 832 to 878 cm^–1), which has a 1/1 Zn/BDC molar ratio.⁴⁰ This reaction is then followed by the introduction of an additional Zn²⁺ center in the structure and the deprotonation of two water molecules to form, after less than 8 min (<480 s) of milling, the ZTA2a phase (with characteristic bands at 862 and 1424 cm^–1 in the two aforementioned regions), which has a Zn/BDC ratio of 2/1 (red curve). The reaction is probably exothermic, as shown by the increase of the temperature up to ∼40 °C (brown curve in Figure 2b). The final transformation observed in Figure 1a corresponds to the structural modification of ZTA2a to form the polymorph ZTA2b (865 and 1449 cm^–1), with a second increase of the temperature up to ∼47 °C (for the center temperature). Such temperature behaviors have been observed in various other systems during milling syntheses.^14,89Figure 2b exhibits, for comparison, the milling temperature profile of a mixture of H₂BDC (400 mg) and H₂O (350 μL) (without ZnO but using similar milling conditions as in Figure 1a), allowing a difference of about 10 °C to be highlighted after ∼15 min of milling (and >15 °C after 1 h) between the Zn-BDC synthesis and the control experiment (brown vs gray curves).

Figure 2.

(a) Reaction profiles corresponding to signals highlighted in Figure 1b. Empty symbols (ZTA2a and ZTA2b at 1424 and 1449 cm^–1, respectively) are shown only for the last half of the milling time for clarity. The inset on the right-hand side corresponds to a zoom of the first 500 s of the time-resolved Raman experiment. (b) Temperature profiles of three different areas of the jar (top, center, and bottom); the light gray curve corresponds to a control experiment in which the temperature evolution caused by the milling of pure H₂BDC with H₂O in the same experimental conditions as in Figure 1a was followed. (c) Picture obtained with the thermal imaging camera of the milling jar, highlighting the three measurement zones. The shaking direction is emphasized by the orange arrows.

The fact that the apparent proportion of ZTA2a is not null at the end of the reaction in Figure 2a (despite that its resonances have disappeared from the Raman data) can be explained by the way in which these kinetic plots were fitted here in the 860 cm^–1 region using simple Gaussian signals centered at the different maximum frequencies of the intermediates, with intensities that were then left to freely evolve at each time point. This created ambiguities in the fitted intensities due to the closely overlapping signals of ZTA2a and ZTA2b (only ∼3 cm^–1 difference in this spectral region). More accurate kinetic plots related to the transformation of ZTA2a into ZTA2b could be obtained by fitting the 1380–1480 cm^–1 region, where the Raman resonances of the two phases are more distinctly resolved. This is shown in Figure 2a, where the Raman signal from ZTA2a (red empty circles) decreases close to 0% at the end of the reaction (in line with pXRD analyses of the final compound). Because the purpose of this manuscript was not to establish a full kinetic model of the formation of these Zn-BDC phases, no attempt was made to further refine or analyze this operando Raman data. Yet, it is worth noting that this example illustrates the importance of having access to different regions of the Raman spectra to evaluate the quality of such kinetic fits and potentially resolve overlapping resonances. Other possibilities along this line (e.g. with deuterated precursors) will be discussed at the end of this article.

In the frame of this work, the clear identification of each transformation thanks to the operando setup involving the combined use of the Raman instrument and the thermal camera allowed each of these structures to be isolated as a pure phase for further ex situ structural analyses (using the optimized reaction conditions reported in Table 2). Similarly to ZTA1, ZTA2a, and ZTA2b, the formation of ZTA3 could also be followed in the same fashion by operando Raman spectroscopy when starting from a 3/1 ratio between the ZnO and H₂BDC precursors. Hence, by choosing the adequate stoichiometry and by stopping the milling at carefully selected times, the different Zn-BDC phases could be isolated pure, without them changing with time (see Supporting Information Figure S3). It then became conceivable to refine their structures, especially the proton positions, and to gain more insights into the terephthalate binding modes.

Refinement of the Zn-BDC Structures Using an “NMR Crystallography” Approach

When the limits of XRD methods are reached, for example, for phases that cannot be isolated as single crystals, in which some atoms cannot be positioned (e.g. protons) or which contain defects, ssNMR methods can be particularly useful when used in conjunction with computational modeling to gain precise information on the crystallographic structures. Such an approach, referred to as “NMR crystallography”,^90,91 implies that the structures of interest possess NMR-active isotopes, which can be readily analyzed by high-resolution NMR.

In the case of Zn-BDC structures, the vicinity of the Zn²⁺ cations and more specifically the precise coordination modes of the BDC and hydroxyl ligands to Zn²⁺ are of great importance. From an NMR perspective, all atoms present in these phases have at least one NMR-active isotope, which can be studied by NMR, making them potentially attractive targets of analysis: ¹H (I = 1/2, 100% natural abundance), ¹³C (I = 1/2, 1.07% natural abundance), ¹⁷O (I = 5/2, 0.04% natural abundance), and ⁶⁷Zn (I = 5/2, 4.11% natural abundance). Even though ⁶⁷Zn NMR has recently shown promising results for providing a detailed description of zinc local environments, it is still considered as very challenging⁹² because of the quadrupolar nature of zinc-67 (which can lead to very broad lines)⁹³ and its poor receptivity (low natural abundance and very low resonance frequency (ν_L(⁶⁷Zn) < 10% ν_L(¹H))).^92,94 In the present work, we thus set our focus on the other three nuclei (¹³C, ¹⁷O, and ¹H) to reach a more detailed description of the binding mode of the dicarboxylate ligands and for positioning the H-atoms of the hydroxyls within the structures.

¹³C NMR Analyses of Carboxylate Binding Modes in Zn-BDC Structures

As briefly explained earlier (vide supra), IR spectroscopy has been used for decades as a first approach for proposing binding modes of carboxylates to various metal ions. The difference in wavenumbers between the antisymmetric and symmetric stretching modes, defined by the parameter Δ (=ν_as(COO^–) – ν_s(COO^–)), has been shown to change depending on the binding mode of carboxylates, allowing, as a first approximation, the following tendency to be derived upon coordination to divalent metal ions: Δ(Chelating) < Δ(Bridging) < Δ(Ionic) < Δ(Monodentate) (see Figure 3a).⁴⁷ However, this method can be ambiguous as Δ can be affected by other parameters, depending on the types of ligands involved and how they interact with each other within the structures. Moreover, it relies on the clear attribution of the vibration modes (ν_as(COO^–) and ν_s(COO^–)), which actually remains uncertain in many experimental cases, due to numerous overlapping vibration bands (e.g., ν_as(COO^–), ν_s(COO^–), ν(CC), β(CCH), etc.) on the IR spectra (see Supporting Information Figure S4). In the present case, when looking at the FTIR spectrum of ZTA2a, the Δ = 210 cm^–1 splitting is significantly higher than expected for bridging ligands. Indeed, the other phases with bridging ligands, namely, ZTA2b and ZTA3, display values at Δ = 171 and 180 cm^–1, respectively. Moreover, the monodentate bound ligand in ZTA1 shows a splitting of Δ = 192 cm^–1. Overall, this tends to show the limitations of IR spectroscopy in providing direct information on the terephthalate binding modes, but may also hint that ZTA2a actually exhibits more complex binding modes than suggested by the X-ray diffraction data.

Figure 3.

(a) Different possible binding modes of carboxylate ligands to Zn²⁺ ions. (b) Left: ¹³C CP MAS NMR spectra recorded at 14.1 T with a spinning speed of 15 kHz under a regulated temperature (0 °C) (in black). Right: ¹³C CP MAS NMR spectra recorded at 14.1 T with a spinning speed of 4.0 or 4.2 kHz under a regulated temperature (0 °C) (in black). Black-filled squares indicate the isotropic chemical shifts. The parameters used for fitting the spectra (in red) are gathered in Table 4.

Here, we decided to perform ¹³C ssNMR analyses to obtain more information on the binding mode of the carboxylates. The ¹³C ssNMR spectra of the four samples obtained by mechanochemistry are presented in Figure 3b (see Supporting Information Figure S5 for data on H₂BDC). Spectra were recorded with different spinning speeds, such as to focus on the isotropic chemical shifts (Figure 3b, left) or to derive the chemical shift anisotropy (CSA) parameters (Figure 3b, right).

Based on the reported crystal structures (see Table 1), two main binding modes were expected for the Zn-BDC compounds of interest: monodentate and bridging. For ZTA1, BDC ligands exhibit a monodentate configuration, and only one crystallographically inequivalent carboxylate carbon is present in the structure. For this phase, a single resonance was observed in the carboxylate region of the ¹³C ssNMR spectrum (Figure 3b), with δ_iso(¹³COO) = 176.6 ppm. On the other hand, ZTA2b and ZTA3 structures have different types of bridging terephthalate ligands based on the reported X-ray diffraction data. While the former clearly exhibits a single carboxylate environment (δ_iso(¹³COO) = 172.1 ppm), as expected from the crystal structure, the latter shows two main carboxylate resonances centered at 174.0 and 172.1 ppm (with an additional shoulder in the low-frequency area, which could perhaps be due to the presence of domains within the structure, in which terephthalate ligands exhibit very slight changes in symmetry/local environment). Finally, for the ZTA2a phase, while only one carbon signal was expected from the crystallographic structure, two carboxylate resonances were observed, centered at 175.7 and 172.5 ppm, and present in an ∼5/1 ratio. The weaker low-frequency resonance was systematically observed for this phase (for different synthetic batches) and with the same approximate ratio. Although its ¹³C chemical shift is relatively close to that of the H₂BDC precursor (173.5 vs 172.5 ppm), no traces of residual H₂BDC could be detected by IR spectroscopy in the C=O stretching region and no acidic proton signal was observed by ¹H ssNMR, meaning that this small signal cannot arise from the presence of residual H₂BDC (see Supporting Information Figures S6 and S7). Moreover, given that no diffraction peaks ascribable to known impurities could be detected on the X-ray diffraction pattern of ZTA2a (see Figure 1c and Supporting Information Figure S3), the weak resonance must correspond to systematic defects in the crystal structure, which could not be resolved nor identified by X-ray diffraction.⁹⁶

In the literature, a few articles report ¹³C ssNMR data for zinc-carboxylates and more specifically Zn-BDC phases.^37,49−51 In all cases, δ_iso(¹³COO) was found to vary between 171.8 and 177.7 ppm. Moreover, when comparing the reported δ_iso(¹³C) values of bridging carboxylates versus monodentate ones, a trend can be extracted. Indeed, the δ_iso(¹³COO) values of monodentate ligands are nearly systematically more deshielded than the δ_iso(¹³COO) of bridging ones. For example, Yuan et al. reported values of δ_iso(¹³COO) = 176.7 and 177.3 ppm for two Zn-BDC phases exhibiting monodentate binding modes (the first phase was actually the same as ZTA1, while the second one was Zn(BDC)(H₂O)), and for a Zn-BDC·DMF MOF with bridging terephthalates, they reported δ_iso(¹³COO) = 172.7 ppm.⁵⁰ Similarly, Julien et al. showed that the Zn(H₂O)₂(2,5-dihydroxyterephthalate) structure, which displays a monodentate configuration, exhibits a ¹³C resonance at δ_iso(¹³COO) = 176.0 ppm, while MOF-74, with bridging 2,5-dihydroxyterephthalate ligands, shows signals at δ_iso(¹³COO) = 173.2 and 173.9 ppm (for compounds obtained by LAG using either H₂O or DMF, respectively).⁵¹ Habib et al. also studied two Zn-dicarboxylate complexes, [Zn₂(μ₂-BDC)₂(μ₄-btre)] and [Zn₂(μ₃-ip)₂(μ₂-btre)(H₂O)₂]·2H₂O (with btre = 1,2-bis(1,2,4- triazol-4-yl)ethane, ip = benzene-1,3-dicarboxylate). Each structure exhibits two distinct δ_iso(¹³COO) by ¹³C ssNMR, corresponding to two monodentate binding modes (δ_iso(¹³COO) = 174.4 and 173.3 ppm) for the first one, while the second complex has a bridging and a monodentate ligands (δ_iso(¹³COO) = 174.6 and 176.7 ppm).⁴⁹ In line with this literature survey, which shows that the δ_iso(¹³COO) values for monodentate binding modes are globally deshielded compared to those for bidentate ones, it was found that the isotropic ¹³C shifts of the carboxylates in ZTA1, ZTA2b, and ZTA3 also appeared to follow this trend. In contrast, for ZTA2a, there seemed to be discrepancies with what one would have expected from the reported crystal structure, not only in the number of ¹³C resonances (due to the additional small peak—vide supra) but also in the ¹³C isotropic shift of the main resonance, which is rather high for a “bridging mode”. Hence, overall, this points once again to the issues related to the X-ray data available for this phase.

To go further in the study of these, we conducted various DFT optimizations of the different structures studied here. Starting from the reported crystal structures, H-atoms were positioned (when missing). Then, keeping cell parameters constant, geometry optimizations were performed by optimizing either H positions only or all atomic positions. The H-relaxed structures did not lead to satisfactory results after the computation of NMR parameters, except for ZTA1 (as illustrated for δ_iso(¹³COO) in Table 4). A more complete geometry optimization was thus needed to refine the various crystal data. Moreover, for some of the structures, it was necessary to test various initial H orientations for the hydroxyls before geometry optimization to converge toward a model that matches best the experimental data; this will be further detailed for ZTA3 later in this article. In all cases, after these additional geometry optimization steps, the GIPAW-calculated NMR parameters for ¹³C were found to be in relatively good agreement with the experimentally extracted values. Notably, δ_iso(¹³COO)_calc of the monodentate ZTA1 ligand was the most deshielded of the series, with a calculated ¹³C chemical shift at 181.8 ppm, while δ_iso(¹³COO)_calc from bridging ligands were ∼6–10 ppm lower in frequency. Moreover, the calculated chemical shift anisotropy (CSA) parameters after full geometry optimization were also found to be in relatively good agreement with the experimental values, as further discussed below. These DFT-optimized structures can thus be considered as more reliable structural models of the four Zn-BDC phases studied here (see Supporting Information Figure S8).

Table 4. ¹³C NMR Parameters of Carboxylates Extracted from Fitting Spectra Recorded at Multiple Spinning Speeds and Calculated Ones Using DFT from Optimized Structural Models of the Phases.

	experimental			calculated (GIPAW-DFT)a
				H-relaxed	fully optimized structures
compound	δiso (ppm)	Ω (ppm)	κ	δiso (ppm)	δiso (ppm)	Ω (ppm)	κ
ZTA3	172.1 ± 0.3	136.9 ± 0.4	0.05 ± 0.01	167.6	172.4	130.7	0.15
	174.0 ± 0.1	139.9 ± 2.4	0.13 ± 0.01	168.0	175.6	132.3	0.31
ZTA2b	172.1 ± 0.1	137.4 ± 0.2	0.22 ± 0.01	168.0	173.4	136.2	0.15
ZTA2a	172.5 ± 0.2	140.0 ± 8.2	0.34 ± 0.15
	175.7 ± 0.1	137.5 ± 0.4	0.20 ± 0.01	199.4	177.6	141.7	0.23
ZTA1	176.6 ± 0.1	154.7 ± 6.3	0.03 ± 0.02	176.5	181.8	146.4	0.35

^aA systematic study of ¹³C chemical shifts was performed in 2009 on 14 molecules (aromatics and carbohydrates) whose crystal structures have been precisely determined by neutron diffraction. This study showed that the GIPAW ¹³C predictions are more accurate than a cluster approach, with an error on the isotropic shift of ∼1 ppm.⁹⁵

Following this, the structures after full relaxation of atom positions were analyzed in detail, and relationships between calculated ¹³C NMR parameters and local environments of the carboxylates (in the geometry-optimized structures) were looked into (see Supporting Information Figure S9). Significant structural modifications were observed upon full geometry optimization of some of the phases. For example, the ZTA3 crystallographic structure was found to undergo various changes, the major one being the modification of one zinc coordination from a 5-fold O-coordination environment (Zn2 in Table 1) to a 4-fold one. In the case of ZTA2a, significant variations in the carboxylate binding modes were observed during the relaxation of all atom positions, which can explain the ∼20 ppm decrease in the δ_iso(¹³COO)_calc value between H-relaxed and fully relaxed models (see Table 4). Still for this phase, the calculated ¹³C NMR parameters were found to be in good agreement with those of the most intense ¹³C resonance (i.e., δ_iso(¹³COO)_exp = 176.4 ppm vs δ_iso(¹³COO)_calc = 175.7 ppm), again with a relatively high chemical shift compared to what one may have expected for bridging carboxylates. Yet, when looking more closely at the local environment of the carboxylate after full geometry optimization, the coordination mode was actually found to be closer to a bridging syn,syn/anti configuration (see Figure 3a) than to monodentate or bridging syn,syn ones. In this configuration, one of the d_Zn···O distances was found to be shorter (in the range of d_Zn···O for syn,syn bridging modes, i.e., ∼2.0 Å) than the two others (which were measured between 2.2 and 2.3 Å). This may be one of the reasons for the higher ¹³C shift of this phase, compared to bridging modes involving only shorter Zn···O bonds.

To go even further in trying to relate ¹³C NMR parameters to structural/geometric information about carboxylate bindings, the calculated data collected from the numerous geometry optimizations on Zn-BDC phases (especially the ZTA3 structure) were compiled. The most obvious trend we found was between the δ_iso(¹³COO)_calc and the longest C–O distance of the carboxylate, (d_C–O)_l: the highest chemical shifts are clearly related to structures with the longest C–O distances (d_C–O)_l (see Supporting Information Figure S9). A contrario, no unambiguous relationship relating δ_iso(¹³COO) and the type of carboxylate coordination could be extracted, showing that this parameter actually depends on several factors.

As ¹³C CSA parameters can also be used to describe the local environment of carbon atoms, these were extracted from the experimental ¹³C NMR spectra recorded at two different spinning speeds (Table 4). The most noticeable difference between compounds was in the CSA span (Ω), which was found to be experimentally higher by approximately 15–20 ppm for ZTA1 (monodentate binding) than for the carboxylates of the three other compounds (155 vs 135–140 ppm, respectively) (see Supporting Information Figure S10 and Table S1). DFT-calculated values of spans also followed the observed trend, the highest value being for ZTA1, with, however, the calculated span of ZTA2a now only about 5 ppm lower than that calculated for ZTA1. Using the data library obtained by optimizing the Zn-BDC structures and also different structural models of ZTA3, attempts were made to relate the various CSA parameters stemming from the calculations to carboxylate binding modes. Here, a general trend was also observed between the longest C–O distance (d_C–O)_l and Ω_calc (see Supporting Information Figure S11), but no direct relationship with the carboxylate coordination.

Overall, although hints about carboxylate local environments could be derived from ¹³C ssNMR spectra, no complete insight into the carboxylate coordination mode could be obtained. Hence, we then considered analyzing local environments of oxygen atoms because they are directly coordinated to the Zn²⁺ ions in the Zn-BDC structures and may help gain insights into the various binding modes.

¹⁷O NMR Analyses of Carboxylate and Hydroxyl Binding Modes in Zn-BDC Structures

With the aim of directly probing the binding modes of the terephthalate ligands by ¹⁷O ssNMR, ¹⁷O-labeled terephthalic acid was used as a starting material to prepare Zn-BDC phases enriched on the carboxylate groups. Concerning the bridging hydroxyls, a direct enrichment during the mechanochemical synthesis of the Zn-BDC phases was envisaged using ¹⁷O-enriched water as a starting reactant (instead of normal water).

Both labeling strategies were tested for each compound (ZTA1, 2a, 2b, and 3) to isolate Zn-BDC phases selectively enriched on the carboxylates or on the hydroxyls/water and then perform high-resolution NMR spectroscopy. All samples could be prepared by mechanochemistry, with selective enrichment on the carboxylate part (phases noted ZTA-*BDC) or the hydroxyl/water part (phases noted ZTA-*OH), with the exception of ZTA2b, for which the carboxylate-enriched phase (ZTA2b-*BDC) could not be obtained with sufficient purity to be analyzed by ¹⁷O ssNMR. In total, six different enriched samples were thus isolated with good purity, using the BM synthetic procedure developed above thanks to the operando Raman setup. In all cases, preliminary evidence of the success of the enrichment could be assessed by FTIR, with the slight shifts of vibration modes such as ν_as(COO^–) and ν_s(COO^–) (see Supporting Information Figure S4c). The corresponding 1D ¹⁷O MAS NMR spectra are presented in Figure 4, which could all be recorded with good sensitivity in just a few hours.

Figure 4.

¹⁷O MAS NMR spectra of ¹⁷O-labeled Zn-BDC compounds prepared by mechanochemistry, B₀ = 14.1 T, ν_rot = 18 kHz, and T = 0 °C. (a) ZTA-*OH samples corresponding to the enrichment of the hydroxyl/water groups. (b) ZTA-*BDC samples corresponding to the enrichment of the BDC moiety. For all spectra, experimental spectra are in black, and the red dotted line corresponds to the results of the fitting process.

Because oxygen-17 is a quadrupolar nucleus, the NMR resonances recorded on the 1D spectra show characteristic second-order quadrupolar line shapes. To extract the ¹⁷O NMR parameters associated with each oxygen environment (δ_iso, C_Q, and η_Q), spectra were fitted at multiple fields (see Supporting Information Figures S14–S17), since the second-order quadrupolar broadening decreases as the magnetic field increases. Moreover, additional 2D high-resolution MQMAS (multiple-quantum MAS) experiments were also recorded for some of the phases, allowing the fits to be further refined (see Supporting Information Figures S12 and S13).

Regarding the ZTA-*OH phases labeled on the hydroxyl/water positions, the NMR spectra were found to be generally easier to analyze (Figure 4a). Only one *OH site was observed for both ZTA2a-*OH and ZTA2b-*OH, as expected from the published crystal structures. In contrast, ZTA3-*OH was found to exhibit a more complex line shape consisting of several inequivalent sites (Figure 4a). These were resolved using a ¹⁷O MQMAS NMR experiment (see Supporting Information Figure S12). Moreover, the OH network of this compound was further investigated through a ¹H–¹⁷O dipolar correlation experiment, as detailed below (see Figure 5).

Figure 5.

(a) ¹H–¹⁷O D-HMQC of ZTA3-*OH recorded at 20.0 T under a spinning speed of 60 kHz. The ¹⁷O and ¹H 1D spectra recorded in the same condition as the D-HMQC are shown in blue. Extracted slices (in black) are fitted (red lines) with the experimental parameters presented in Table 5. The inset displays the FTIR spectrum centered on the OH stretching region. H-bonded OH groups are highlighted in orange, and isolated OH groups are shown in green. (b) Close-up views of the DFT-optimized ZTA3 structure of the different hydroxyl environments; color code: brown for carbon, red for oxygen, light pink for hydrogen, and purple for zinc. (c) Comparison of experimental ¹H MAS spectrum (recorded at B₀ = 20.0 T and ν_rot = 60 kHz, in black) and simulated spectrum from calculated ¹H NMR parameters in dotted gray.

The D-HMQC experiment allows ¹H···¹⁷O proximities to be probed and turned out to be highly useful to help position the hydroxyl protons (which were absent from the initial X-ray diffraction data). Here, a short recoupling time was used (τ_rec = 100 μs) to observe the shortest O···H contacts (i.e., the O–H bonds). In these conditions, no correlation between hydroxyl oxygen atoms and aromatic protons was observed (blue-shaded zone) but four environments could be resolved for the hydroxyls, one per inequivalent OH, in line with the published crystal structure (three being μ₃-OH and one being μ₂-OH). More interestingly, when looking at the FTIR spectrum of ZTA3 (see the inset of Figure 5a), four environments of the hydroxyls could also be detected and separated into two groups: two for isolated OH species (green-shaded zone) and two for OH species involved in H-bonding (orange-shaded zone). A similar distinction can be made using the D-HMQC data. Indeed, the most shielded oxygen site (δ_iso(¹⁷O) = −27.9 ppm) and the most deshielded one (δ_iso(¹⁷O) = 12.8 ppm) are correlated to ¹H resonances with low chemical shift (close to 0 ppm), such ¹H shifts being characteristic of isolated OH groups (zone shaded in green).^97,98 On the other hand, the two remaining OH sites are correlated to ¹H resonances, with δ_iso(¹H) close to 5 ppm, implying that they are involved in hydrogen bonds (orange-shaded zone).⁸⁷⁻⁹⁹

Using these high-resolution ¹H–¹⁷O experiments as a guideline, different configurations were tested to position the four hydroxyl hydrogens in the ZTA3 structure, which were then geometry-optimized by DFT. The resulting GIPAW-DFT-computed ¹H ssNMR spectrum of the best model is highly comparable to the high spinning speed ¹H NMR spectrum (Figure 5c). The resulting refined ZTA3 structure exhibits two isolated hydroxyls: one (O5-H13) is shared between Zn3–Zn2–Zn3 (μ₃-OH), while the second one (O7–H10) is located between Zn1–Zn2 (μ₂-OH); both display a low δ_iso(¹H) chemical shift (1.34 and 0.80 ppm, respectively). In contrast, the hydroxyl protons denoted H11 and H12 have a higher chemical shift, indicative of the presence of a hydrogen-bonding network, as can be seen in the geometry-optimized structure (see Figure 5b): the H11···O7 and H12···O2 distances were found to be only ∼1.7 Å.

Generally speaking, the calculated ¹⁷O NMR parameters were found to be in satisfactory agreement with those extracted from experimental data. The overestimation of C_Q, often observed in Table 5, is consistent with what has been recently reported in the ¹⁷O NMR literature on MOFs.^59,100 However, unlike the calculated ¹³C NMR parameters, no clear trend between ¹⁷O NMR parameters and structural geometric parameters (d_C–O, d_Zn···O, θ_{Zn···O–C}...) could be derived (see Supporting Information Figures S18 and S19). This is probably due to the multifactorial dependence between the ¹⁷O NMR parameters and the local environment and geometry around the oxygen in these structures.

Table 5. ¹⁷O ssNMR Parameters Extracted from Experimental Spectra (Figure 4) Recorded at Multiple Magnetic Fields and Calculated Using DFT from Structural Models of the Phases.

		experimental				calculated
enrichment	compound	δiso(17O) (ppm)	CQa (MHz)	ηQ	Int (%)	δiso(17O) (ppm)	CQ (MHz)	ηQ
*OH	ZTA3	12.8 ± 0.6	5.59 ± 0.02	0.30 ± 0.02	27.2	6.1	–7.88	0.64
		–1.9 ± 0.1	4.71 ± 0.04	0.43 ± 0.10	17.9	3.7	6.81	0.98
		–8.5 ± 0.1	5.22 ± 0.01	0.23 ± 0.02	23.6	0.4	–4.31	0.53
		–27.9 ± 0.2	5.92 ± 0.02	0.47 ± 0.01	31.3	–27.3	–7.98	0.51
	ZTA2b	–5.6 ± 0.4	5.10 ± 0.10	0.36 ± 0.01	100	–9.9	–5.92	0.33
	ZTA2a	5.5 ± 1.3	5.90 ± 0.12	0.25 ± 0.02	100	1.8	–6.89	0.34
	ZTA1	–3.2 ± 3.0	7.66 ± 0.20	0.61 ± 0.04	100	–15.8	9.32	0.49
*BDC	ZTA3	258.4 ± 0.1	7.23 ± 0.01	0.56 ± 0.01	28.5	276.3	7.73	0.80
		249.0 ± 0.7	6.97 ± 0.02	0.74 ± 0.03	25.3	273.4	7.93	0.84
		237.3 ± 0.5	7.06 ± 0.01	0.75 ± 0.06	19.9	254.7	–6.77	0.97
		205.2 ± 0.8	6.39 ± 0.05	0.90 ± 0.10	26.3	210.2	–7.26	0.63
	ZTA2a	278.3 ± 1.6	7.07 ± 0.10	0.51 ± 0.11	32.3	298.1	7.37	0.69
		261.3 ± 1.0	7.04 ± 0.02	0.56 ± 0.13	10.5
		255.7 ± 1.5	6.70 ± 0.21	0.88 ± 0.12	5.2
		240.3 ± 5.3	6.74 ± 0.30	0.85 ± 0.15	8.7
		223.0 ± 1.1	7.85 ± 0.05	0.60 ± 0.08	43.3	253.0	8.79	0.62
	ZTA1	286.6 ± 1.3	6.90 ± 0.05	0.46 ± 0.02	50.9	320.6	7.46	0.69
		221.7 ± 2.1	5.83 ± 0.49	0.85 ± 0.01	49.1	251.3	–6.71	0.67

^aExperimental C_Q values are given in absolute values.

Concerning the phases enriched on the carboxylates, the experimental δ_iso(¹⁷O) arising from the BDC* ligands of the different structures were found to span between 205.2 and 286.6 ppm (see Figure 4b). ZTA1-*BDC exhibits two clear inequivalent oxygen resonances, as expected from the crystal structure. On the contrary, ZTA2a-*BDC and ZTA3-*BDC gave more complex signatures due to the presence of several overlapping resonances. The ¹⁷O MAS NMR spectra of ZTA3-*BDC could, for example, be fitted considering the presence of four different ¹⁷O signals, as expected from the crystal structure and resolved with the MQMAS experiments (see Supporting Information S12). Concerning the ZTA2a-*BDC phase, only two carboxylate oxygen resonances were expected according to the reported crystal structure. However, it was not possible to obtain a satisfactorily simulated spectrum with only two signals. This result supports the observation made previously by ¹³C NMR for ZTA2a, where an additional ¹³C carboxylate resonance of weak intensity was detected at 172.5 ppm (see Figure 3b).

To learn more about the minor carboxylate environments in ZTA2a, the ¹⁷O MQMAS spectrum of ZTA2a-*BDC was recorded (Figure 6). The two most intense resonances, which are centered at δ_iso(¹⁷O) = 278.3 and 223.0 ppm, belong to the main carboxylate group (i.e., with the ¹³C signal at 175.7 ppm). Yet, several signals of weaker intensity can also be observed (red-shaded circle), corresponding to multiple and slightly different carboxylate local environments (such regions of signals with lower intensity have been observed on two different batches of ZTA2a-*BDC with two different NMR spectrometers). Three examples of these additional oxygen environments were extracted from this area for further study (Figure 6, right).

Figure 6.

¹⁷O MQMAS spectrum of ZTA2a-*BDC recorded at B₀ = 18.8 T under ν_rot = 16 kHz. Extracted slices (in black) are fitted (red dotted lines) with experimental parameters presented in Table 5. The red-shaded circle enlightens some of the additional oxygen sites corresponding to structural defects.

Despite their low signal-to-noise ratio, the signals from the three different slices could be tentatively fitted using quadrupolar line shapes, suggesting that they correspond to carboxylate oxygens in a relatively well-ordered local environment. Using the NMR parameters derived from the different ¹⁷O environments resolved by MQMAS, it was then possible to propose a fit of the ¹⁷O NMR spectra of ZTA2a-*BDC in the carboxylate region (see Figure 4b and Table 5). However, it should be kept in mind that additional weak-intensity carboxylate resonances may also be overlapping with the two main resonances (which remained unresolved at this stage). Interestingly, the ¹⁷O NMR data also show that the main resonances for ZTA2a-*BDC (δ_iso(¹⁷O) = 278.3 and 223.0 ppm) are relatively close in shift to those of ZTA1-*BDC (δ_iso(¹⁷O) = 286.6 and 221.7 ppm), further underscoring the similarities between ¹⁷O ssNMR and ¹³C ssNMR parameters (δ_iso(¹⁷O), δ_iso(¹³COO), and ¹³C CSA) for the two different binding modes, i.e., syn,syn/anti bridging (ZTA2a) and monodentate (ZTA1). Regarding the weaker signals (red circle in Figure 6), they are likely to arise from the presence of the systematic defects inside the crystallographic structure, as already hinted from the ¹³C NMR data.

Thanks to the selective ¹⁷O-labeling of the various Zn-BDC compounds by BM, it was possible to gain novel insights into the local environment of the oxygen-containing ligands, as well as refine one of the crystallographic structures (ZTA3) by DFT calculations. Although no straightforward trend relating ¹⁷O NMR parameters to carboxylate binding modes could be derived, most probably because of the influence of several geometrical features (Zn···O and O–C distances, bond angles, positioning of neighboring ligands, etc.), high-resolution ¹⁷O ssNMR spectroscopy, in conjunction with DFT calculations and GIPAW calculations of NMR parameters, clearly appears as a crucial asset to help confirm the structural models of these types of MOFs. In particular, they will be useful in the future to help elucidate the nature of the defect sites in ZTA2a in an NMR crystallography type of approach.¹⁰¹

Outlook: Isotope Labeling for Simplified In Situ Observations of Evolutions in the Reaction Media

Because structural rearrangements occur around the water molecules/hydroxyls during the Zn-BDC transformations described above, being able to follow this by Raman spectroscopy appeared to us as an important goal to increase our insight into the transformations taking place. Unfortunately, it was not possible to follow the evolution of the O–H stretching modes upon synthesis by BM with our Raman setup because it only allowed spectra to be recorded from 220 to 3120 cm^–1, while ν(OH) bands most often appears at higher wavenumbers.^102−104

To circumvent this issue, we thus decided to perform isotope labeling of Zn-BDC phases using deuterium oxide rather than normal water as a starting material in the syntheses. In doing so, the stretching modes arising from −OH groups would correspond to -OD groups, which can be observed between 2400 and 2800 cm^–1. Figure 7 presents as an example the result of a synthesis followed by operando Raman spectroscopy and performed using an initial Zn/BDC ratio of 3/1 (i.e., one additional equivalent of ZnO compared to the experiment presented in Figure 1) and D₂O instead of H₂O. The corresponding temperature measurement through the thermal imaging camera is placed on the side of the Raman spectra (Figure 7a), and the focus of the Raman data is set in the zone between 2600 and 2700 cm^–1, where O–D stretching vibrations appear.

Figure 7.

(a) Top: starting compounds used for the operando experiment with their respective equivalents. Bottom: time-resolved operando Raman spectra of two selected areas: 2600–2700 cm^–1 range on the left and 1380–1480 cm^–1 range on the right, following the Raman spectroscopy convention for displaying the data (high wavenumbers on the right). The graph on the far right corresponds to the temperature measurement recorded using the thermal imaging camera. (b) Experimental ex situ FTIR spectra of ZTA2b (in blue), ZTA2b-OD (in light blue), ZTA3 (in purple), and ZTA3-OD (in light purple), following the FTIR spectroscopy convention for displaying the data (high wavenumbers on the left).

With the starting experimental conditions used here, it is expected to obtain a different final product (i.e., ZTA3) compared to the experiment presented in Figure 1 (final product was ZTA2b). The formation and progressive transformation of the intermediates into ZTA3 are indeed noticeable through the change in temperature, as well as the modifications of the Raman spectra. It should be noted that no signal from ZTA1 was observed in the 2600–2700 cm^–1 area (Figure 7) because of the lack of isolated OH environments in this phase, the water molecules being involved in hydrogen bonds. The broad signal arising from these water molecules is indeed too weak to be detected during the operando Raman experiment. Hence, in the 2600–2700 cm^–1 range, when working with a 3/1 Zn/BDC ratio in D₂O, we observed first the formation of the ZTA2a-OD, through the presence of the ν(OD) band (2660 cm^–1). As with the synthesis presented in Figure 1, it then transforms into the ZTA2b-OD structure (2618 cm^–1). Then, the excess of zinc oxide allows the formation of the ZTA3-OD compound as the final form (ν(OD) = 2630 and 2671 cm^–1). This final phase has a Zn/BDC stoichiometry of 3/1, consistent with the one between the reagents introduced in the jar. The transformation from H₂BDC to ZTA3 can also be observed looking at ν_s(COO^–) band between 1380 and 1480 cm^–1, the only difference with Figure 1 being the final transformation into the ZTA3 phase (ZTA3-OD here) in the later synthesis. It is worth noting that we were able to observe the two distinct sharp O–D bands in the final compound, corresponding to the expected two isolated hydroxyls (Zn–O(H)–Zn).

The interest of using deuterated precursors for operando Raman mechanochemistry was further demonstrated here, using ring-d₄ terephthalic acid as a starting reagent. In doing so, the ν(CH) Raman signals become ν(CD) signals, which, in the case of ZTA1, for example, corresponds to a shift of the stretching vibrations from 3064 and 3105 cm^–1 to 2287 and 2321 cm^–1 (see Supporting Information Figure S20).¹⁰⁵ Therefore, isotope labeling by deuteration can be seen as a way of shifting Raman vibrations in an area of the spectrum where no signal from the jar is expected, allowing a more straightforward observation of the evolving vibration modes as well as an easier postsynthesis data treatment. Both methods, using D₂O or d₄-BDC, can also be combined to study the change of kinetics for such reactions. Overall, this demonstrates, beyond the study of reaction mechanisms in ball-milling,¹⁰⁶ how isotopic labeling by deuteration can be particularly useful also for following the kinetics in mechanochemical reactions by shifting vibration bands into regions where they become detectable (e.g., OD vs OH) and/or to avoid overlaps with other vibration frequencies (including from the jar). Such strategies may also turn out to be particularly valuable to refine kinetic data, which may be ambiguous in other regions (as highlighted at the beginning of this manuscript) and thereby help elucidate complex mechanochemical reactions.

Conclusions

In this contribution, we have performed the first in-depth study of the formation and structure of four coordination polymers involving Zn²⁺ cations and terephthalate ligands, noted ZTA1, ZTA2a, ZTA2b, and ZTA3. Although each of these phases had been previously reported independently (for instance, as part of investigations aiming at studying the formation of the well-known MOF-5 structure), the details of their structure, and in the case of ZTA2a, the possibility to isolate it as phase-pure, were missing. Here, it is shown for the first time how using mechanochemistry, in conjunction with operando Raman spectroscopy and thermal imaging, the evolution of the reaction medium could be followed. Each of the phases could be isolated as pure, either by adapting the stoichiometry of the Zn and BDC precursors involved in the milling, or by stopping the milling synthesis at a specific time point based on knowledge of the kinetics of the reactions. The study of the formation of ZTA3 by mechanochemistry was shown to proceed through the successive formation of the ZTA1, ZTA2a, and ZTA2b intermediates (see Figure 8).

Figure 8.

Summary of the structural evolution of the Zn-BDC compounds presented in this article upon milling synthesis. The molecular schemes arise from the DFT-optimized structures obtained in this article.

For each of the phases, an NMR crystallography approach was then used, combining high-resolution NMR spectroscopy with ab initio DFT calculations of NMR parameters, to help refine the crystal structures, as much information was lacking in the crystallographic data available to date. Regarding ¹³C ssNMR, a systematically lower δ_iso(¹³COO) was observed for syn,syn bridging ligands as opposed to monodentate and weaker syn,syn/anti bridging modes (∼4 ppm difference in shifts). Moreover, δ_iso(¹³COO) and the CSA span (Ω) parameter were found to correlate reasonably well with the longest C–O distance of the carboxylate, making the measurement of these parameters valuable for extracting structural information in future studies of other unknown Zn-terephthalate coordination networks. Furthermore, the ¹³C ssNMR study allowed an additional carboxylate resonance to be detected for the ZTA2a compound reflecting the likely presence of “defect” in that structure.

Regarding ¹⁷O ssNMR, the key point to highlight is that using the protocols developed here by mechanochemistry, it was possible to produce, at low cost, selectively ¹⁷O-labeled phases in high yield, with either enriched hydroxyls (Zn–O*(H)–Zn) or enriched carboxylates (CO*O*–Zn) and no isotopic scrambling during the milling (as shown in Figures S14 and S15). Such selective labeling was possible because reactions were performed here by BM under “ambient” temperature and pressure, and in short times (less than 3 h). Moreover, it is important to highlight that the syntheses of these ¹⁷O-enriched compounds would have been far too expensive if not for the BM procedures optimized through the use of real-time monitoring Raman spectroscopy. Indeed, the extent of enriched H₂O* employed here was very low (for ZTA-*OH, ∼45 μL for ∼300 mg of sample recovered and ∼20 μL for ∼110 mg of sample recovered for ZTA-*BDC) and could not have been matched for syntheses carried out using the hydrothermal or microwave procedures proposed so far to prepare such compounds. Thanks to this isotopic labeling, high-resolution ¹⁷O ssNMR spectra could be recorded for each of the phases in just a few hours, allowing different carboxylate oxygen and hydroxyl environments to be resolved. This enabled, for example, a very precise positioning of hydroxyl hydrogens in the case of the ZTA3 phase. Regarding terephthalate ligands, however, no straightforward correlation between ¹⁷O NMR parameters and geometric features around the carboxylate oxygens could be derived, showing that the ¹⁷O data in these phases are very sensitive to the local environment of oxygen and depend on many different factors (e.g., bond distances, angles). This implies that combined experimental–computational approaches involving various NMR-active nuclei, including ¹⁷O, are a very valuable tool for validating structural models (and notably carboxylate binding modes) of these materials.

Among the four phases studied, the ZTA2a compound was proven to present a more complex structure than the one reported, with the probable presence of “defect” sites, as evidenced from both ¹³C and ¹⁷O ssNMR analyses. The information derived from these spectra can serve as the basis for proposing structural models of these defects, which can then be validated using DFT calculations of NMR parameters. More generally speaking, considering the increasing number of studies on MOF structures, which highlight the importance of further refining structural data (beyond what is possible by using powder X-ray diffraction) and to establish the nature and role of defects on reactivity, the possibility of using approaches such as the ones proposed herein, including selective ¹⁷O isotopic enrichment, appears very valuable.

Lastly, through the study of the mechanochemical formation of the four Zn-BDC phases using operando Raman spectroscopy, we also demonstrate here for the first time how the use of deuterated precursors can be useful to help follow the course of the reactions by enabling the observation of specific vibration modes (e.g., ν(OD) and/or ν(CD)) and thereby avoiding cutoff of high wavenumbers of the ν(OH) and/or ν(CH) modes due to instrumentation, and/or avoiding their overlap with signals arising from the milling jar in the Raman spectra. We foresee that on a more general perspective, the study of deuterated compounds in operando Raman investigations by BM will be particularly useful to refine the understanding of the reactions occurring in the jars and may enable the more straightforward investigation and optimization of reactions, which not only concern the formation of MOFs but may actually more specifically involve the formation of C–D (instead of C–H) bonds, as it is the case in reduction reactions in organic chemistry.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.chemmater.1c04132.

• Experimental details and additional supporting tables and figures: selected ssNMR data (¹H, ¹³C, and ¹⁷O NMR spectra, together with their fits); pXRD, FTIR, and Raman data; and graphical plots related to DFT calculations and structural parameters (PDF)

Author Contributions

The project was conducted by C.L. in close interaction with D.L., T.-X.M., and C.G. T.-X.M., C.L., and D.L. developed together the operando Raman spectroscopy setup coupled to mechanochemistry. C.L. performed all BM syntheses, all characterizations by IR, Raman, and pXRD, and the vast majority of ssNMR analyses. I.H. and Z.G. performed part of the high-field ¹⁷O NMR experiments reported herein (MagLab). C.G. carried out the computational simulations and GIPAW-DFT calculations. C.L. performed the data analyses and prepared all figures and tables. C.L. and D.L. wrote the initial draft of the manuscript. All authors discussed the results and contributed to the final preparation of the manuscript.

The authors declare no competing financial interest.