X-ray and neutron powder diffraction analyses of Gly·MgSO₄·5H₂O and Gly·MgSO₄·3H₂O, and their deuterated counterparts

The smallest of the amino acids, glycine, is of inter­est by virtue of its widespread occurrence in biological and pharmaceutical systems; the fact of its achirality, minimal number of side chains and overall rigidity mean that it is an ideal model for understanding the structural organization of larger amino acids and their polymers. Similarly, coordination complexes between metallic ions and organic ligands are fundamental building blocks of complex biomolecules, such as chloro­phyll and haemoglobin, and an appreciation of the inter­action and structural organization of these inorganic and organic components is crucial to our understanding of certain living processes and perhaps the origins of life itself. Furthermore, the study of these compounds is worthwhile for their intrinsically inter­esting (and often commercially valuable) properties, such as ferroelectricity, pyroelectricity, magnetism, and possible optoelectronic applications; for example, noncentrosymmetric metal–organic crystals often exhibit substantial nonlinear optical properties, including second harmonic generation (Fleck & Petrosyan, 2010; El-Fadl & Abdulwahab, 2010; Murugan & Ramasamy, 2011).

There is an extensive literature concerning metal coordination compounds with glycine, including halides (Fleck & Bohatý, 2005a), perchlorates (Wang et al., 1998), nitrates (Rao & Viswamitra, 1972; Krishnakumar et al., 2001; Fleck & Bohatý, 2005b; Choudhury et al., 2013), chromates, molybdates and tellurates (Fleck et al., 2006; Tran Qui et al., 1984), to name but a few. Our inter­est, however, lies specifically with the divalent metal sulfates.

Amongst the glycine coordination compounds of divalent metal sulfates, several anhydrous species are known, including 2Glyc·CoSO₄ (Kydyrmishev, 1972) and 2Gly·ZnSO₄ (Moldobaev & Nogoev, 1970), as well as a range of manganese-bearing crystals, e.g. nGlyc·MnSO₄ (n = 2, 4 or 6) (Weng et al., 2009; Tepavitcharova et al., 2012). The hydrates are largely restricted to a series of penta­hydrates and trihydrates. [Although the chemical composition of each corresponds to Glyc·MgSO₄·5H₂O and Glyc·MgSO₄·3H₂O, the occurrence of symmetry inequivalent MgO₆ o­cta­hedra in both structures with differing local coordination environments means that the `true' structural formula units are [Glyc·MgSO₄·5H₂O]₂ and [Glyc·MgSO₄·3H₂O]₂. However, the molecular site symmetry results in the crystallographic formula units being [Glyc·MgSO₄·5H₂O]₂ with Z = 1 and Glyc·MgSO₄·3H₂O with Z = 4, respectively.] Compounds with the general formula Glyc·M²⁺SO₄·5H₂O occur as isotypic triclinic crystals for M²⁺ = Mg, Mn, Fe, Co and Zn (Lindqvist & Rosenstein, 1960; Elayaraja et al., 2007; Fleck & Bohatý, 2006; Tepavitcharova et al., 2012); unit-cell parameters for each of these are given in Table 1. The variation in unit-cell volumes follows the expected trend based on the ionic radii of each cation, with the exception of Glyc·FeSO₄·5H₂O, which ought to have a unit-cell volume very much closer to that of the Mg-bearing species. The compounds of general formula Glyc·M²⁺SO₄·3H₂O were, until very recently, known only as an orthorhombic crystal for M²⁺ = Zn and a monoclinic crystal for M²⁺ = Co (Table 1). Whilst preparing this report, we became aware that Mg, Zn and Fe analogues of monoclinic Glyc·CoSO₄·3H₂O had been prepared (Oguey et al., 2013a,b, 2014a) and the data deposited with the Cambridge Structural Database (Groom & Allen, 2014), although, to the best of our knowledge, the only peer-reviewed references to these materials are private communications in a paper by Stoeckli-Evans et al. (2014).

Solubility data for Glyc·NiSO₄·nH₂O have been reported, which reveal the existence of a trihydrate and a penta­hydrate (Moldobaev et al., 1970; Alymkulova & Salyeva, 1987), although no crystallographic data exist as far as we know. A crystal with n = 6 is also known and the structure of this was determined by Peterková et al. (1991).

With regard to copper, two compounds, namely Glyc·CuSO₄ and Glyc·CuSO₄·2H₂O, are known (Stoeckli-Evans et al., 2014). The existence of Glyc·CuSO₄·5H₂O has been incorrectly reported by Thilagavathi et al. (2012). In fact, their paper provides X-ray powder diffraction data [measured `in the range 283–343 K' (sic)] and unit-cell parameters that correspond to CuSO₄·5H₂O rather than a possible Glyc·CuSO₄·5H₂O. Furthermore, their FT–IR absorption spectra contain no significant features due to inter­nal vibrational modes of the glycine zwitterion [compare their Fig. 4 with Fig. 5 of Tepavitcharova et al. (2012)], even though they proceed to tabulate the frequencies of these modes. Finally, their DSC/TG (differential scanning calorimetry/thermogravimetric) data are inter­preted incorrectly; the transition temperatures and fractional mass losses are clearly due to the dehydration sequence CuSO₄·5H₂O → CuSO₄·H₂O → CuSO₄ [see Fig. 1 of Gadalla (1984), for example]. A comparable lack of care in reporting the properties of many other amino acid-bearing structures has been detailed by Fleck & Petrosyan (2010).

Our particular inter­est in compounds of sulfates (and Mg sulfate especially) with glycine stems from a planetary background, founded upon a long-standing study of water ice (sensu stricto) and other highly hydrated inorganic and organic substances termed `ices' by planetary scientists (cf. Fortes & Choukroun, 2010). Glycine is a molecule of planetary and astronomical significance since it has been found in outgassed dust particles from Comet 81P/Wild 2 (Elsila et al., 2009) and in carbonaceous meteorites such as Murchison, Orgeuil and others (Kvenvolden et al., 1971; Engel & Macko, 1997; Ehrenfreund et al., 2001; Burton et al., 2014). Detection of glycine in the inter­stellar medium by Kuan et al. (2003) was later rebutted by Snyder et al. (2005) although the precursors of glycine have been detected. Laboratory experiments have produced glycine in ices under conditions relevant to the inter­stellar medium (Muñoz-Caro et al., 2002) and there are good theoretical models to suggest that it should be synthesized by both gas phase and solid-phase reactions (e.g. Holtom et al., 2005; Pilling et al., 2011; Garrod, 2013).

If it is the case that the building blocks of life on Earth, amino acids, were delivered by meteorites and comets direct from the inter­stellar `factory floor' during the first 500 Myr of Earth history then the fossil record of this has been permanently destroyed. However, there is a fluvial/lacustrine sedimentary record covering this time frame on Mars, where water-lain sedimentary rocks in Gale crater have been radiometrically dated in situ by the Curiosity rover to within 350 Myr of planet formation (Farley et al., 2014). Since the martian environment is presently subjected to intense UV radiation and strongly oxidizing substances in the regolith, preservation of amino acids requires a protection mechanism, such as storage in minerals (Aubrey et al., 2006; Martinez-Frias et al., 2006; Aerts et al., 2014).

By far the most plausible storage medium on Mars is in highly water-soluble sulfates; abundant Fe³⁺ and Mg²⁺ sulfates have been identified, principally in the form of the minerals jarosite (sensu lato) and kieserite, although other hydration states of MgSO₄ are likely (see Wang et al., 2006; Wendt et al., 2011, and references therein). Martian brine solutions, both today and in the distant past, are believed to be Mg²⁺-rich and [SO₄²⁻]/[SO₄²⁻ + Cl⁻]-rich (King et al., 2004; Möhlmann & Thomsen, 2011). Uptake of glycine in natural (terrestrial) and synthetic jarosites has been investigated by Kotler et al. (2008, 2009) and our objective here is to characterize compounds that may form by inter­action of martian kieserite with glycine-bearing aqueous solutions. These efforts will provide the means to detect such compounds on the martian surface, whether by in situ X-ray diffraction (as on the Curiosity rover) or by Raman spectroscopy (on the future ExoMars rover).

The present work is part of a broader investigation of Glyc·M²⁺XO₄²⁻·nH₂O compounds crystallized at temperatures in the range 260–450 K with the specific aim of identifying new hydration states, complementing related studies by our group of novel M²⁺XO₄²⁻ hydrates in general (e.g. Fortes et al., 2012a,b, 2013; Fortes, 2015).

Glyc·MgSO₄·5H₂O was crystallized by evaporation at room temperature of an equimolar aqueous solution of α-glycine (Alfa Aesar A13816) and MgSO₄·7H₂O (Sigma Aldrich M1880) in deionized water (Alfa Aesar 36645). It was found that the protonated isotopologue of the trihydrate could be prepared in one of three ways: (i) by heating the penta­hydrate to 360 K for 24 h with periodic grinding; (ii) by evaporation of an equimolar solution of glycine and MgSO₄, saturated at room temperature with respect to the penta­hydrate, in a glass vial at σim 348 K down to a volume approximately one-third of the initial volume over a period of 2.5 d, whereafter the resultant viscous syrup crystallized in a matter of minutes when stirred or otherwise agitated; (iii) by the same procedure as described for (ii), but with seeding of the warm solution using fine trihydrate grains once evaporation had reached approximately half of the starting volume, causing well faceted single crystals 1–2 mm in length to grow on the walls of the vial.

Deuterated samples of the penta­hydrate were synthesized from an equimolar mixture of α-glycine-d₅ (Aldrich 175838, 98 atom% D) and β-MgSO₄ (Sigma–Aldrich M7506) dissolved to a concentration of 35 wt% in D₂O (Aldrich 151882, 99.9 atom% D). Crystallization was induced in a Pyrex beaker sealed into a plastic bag with an open container of anhydrous MgSO₄ desiccant. The deuterated trihydrate was prepared in essentially the same manner as described in (i) above. An aliquot of the penta­deuterate in a narrow glass vial was placed inside a larger glass vessel with a separate vial of D₂O, the larger outer container then being closed with a pierced screw-lid. The sample and D₂O were heated to 363 K for 36 h, with a gentle grinding after 24 h. Complete deuteration of the synthesized compounds was verified by Raman spectroscopy. The phase identity of all materials used to synthesize the hydrates, and of the hydrates themselves, was verified by X-ray powder diffraction.

Crystal data, data collection and structure refinement details are summarized in Table 2. X-ray powder diffraction data were collected on a PANalytical X'Pert Pro multipurpose powder diffractometer (using Ge monochromated Co Kα₁ radiation, λ = 1.788996 Å, and an X'Celerator multi-strip detector). Data were measured with variable divergence and receiving slits, converted to fixed-slit geometry with the proprietary X'Pert Pro HighScore Plus software package, and exported in an appropriate format for analysis in the GSAS/Expgui package (Larsen & Von Dreele, 2000; Toby, 2001).

The room-temperature precipitate from an equimolar glycine–MgSO₄ aqueous solution, consisting of block-shaped well faceted crystalline aggregates, was dried on filter paper and ground to a fine powder for X-ray powder diffraction analysis. This was loaded into an Anton–Paar HTK1200N oven mounted on the X-ray diffractometer and data were collected over the 2θ range 5–90° in 10 K increments from 298 to 478 K, counting for 1 h at each temperature (Fig. 1).

Between 298 and 318 K, the specimen was found to consist of Glyc·MgSO₄·5H₂O and accessory α-glycine, these being identified by the HighScore Plus ICDD search-match software, specifically ICDD (2002) entries 00–048-2320 and 00–032-1702, respectively. The α-glycine (space group P2₁/n; Jönsson & Kvick, 1972; Power et al., 1976) presumably originated from adhering mother liquor or blebs of liquid entrained inside the crystals prior to grinding. In the 328 K data set, however, new peaks appear. Above 348 K the diffraction pattern changes significantly in that the penta­hydrate Bragg peaks disappear entirely and are replaced by the nascent peaks observed in the 328 K diffraction pattern. α-Glycine persists through this phase transition.

This new phase survived to 398 K above which temperature further heating yielded only a broad amorphous feature, centred around 27° 2θ, superimposed upon which are the persistent α-glycine Bragg peaks. The signature of α-glycine in the diffraction data finally disappears above 448 K. Upon removal from the diffractometer, the amorphous specimen was found to have transformed into a very dark brown, almost black, semirigid disk. We speculate that this material is the so-called `thermo-melanoid' described by Heyns & Pavel (1957a,b), which is produced by heat treatment of various amino acids, including glycine.

The phase observed from σim 330–400 K was indexed using DICVOL06 (Boultif & Louër, 2004) after identification and elimination of peaks from α-glycine. A single monoclinic solution with a high figure of merit was obtained; furthermore, systematic absences limited the range of possible space groups to P2₁, Pn or P2₁/n. At this point it became apparent that the unit-cell metric and probable symmetry were a close match to a compound of composition Glyc·CoSO₄·3H₂O (space group P2₁/n), first described by Tepavitcharova et al. (2012), leading to the conclusion was that we had obtained a compound isotypic with this, namely Glyc·MgSO₄·3H₂O.

Successful attempts were made subsequently to obtain the trihydrate by equilibrium growth from aqueous solutions at temperatures of σim 360 K, as outlined in Section 2.1. These were characterized by X-ray powder diffraction at room temperature on a standard back-loaded spinner stage.

In the absence of neutron diffraction data (either powder or single-crystal) for any compound of general composition Glyc·M²⁺SO₄·nH₂O, we decided to carry out a low-temperature neutron powder diffraction study of both Glyc·MgSO₄·5H₂O and Glyc·MgSO₄·3H₂O in order to obtain accurate structural parameters for all atoms, including H atoms, for the first time. By virtue of the large incoherent neutron scattering length of ¹H, deuterated specimens were prepared so as to minimize the background signal. After preparation of these, as described in the preceding section, the phase identity of the two deuterated samples was confirmed by X-ray powder diffraction. Using initial structural data reported for Glyc·MgSO₄·5H₂O and Glyc·CoSO₄·3H₂O, Rietveld refinements of Glyc(d₅)·MgSO₄·five-dimensional₂O (measured 5–90° 2θ) and Glyc(d₅)·MgSO₄·three-dimensional₂O (measured 5–140° 2θ) are shown in Figs. 2(a) and 2(b). The inset to Fig. 2(a) reveals small parasitic Bragg peaks from α-glycine (space group P2₁; Iitaka, 1960) and a lesser amount of α-glycine (space group P3₁; Iitaka, 1961; Kvick et al., 1980) in the penta­deuterate sample. After warming to transform the penta­deuterate to the trideuterate, it is apparent that α-glycine has fully transformed to α-glycine, this being the only (very) minor accessory phase evident in Fig. 2(b). The excellent fit of the Co-analogue structure to the Glyc(d₅)·MgSO₄·three-dimensional₂O powder data confirms that the two compounds are isostructural.

Time-of-flight neutron diffraction measurements were carried out in the high resolution powder diffractometer, HRPD, at the ISIS neutron spallation source, Rutherford Appleton Laboratory, UK. This instrument offers unprecedented d-spacing resolution, which is virtually constant across the measured time frame, whilst the liquid CH₄ moderator and upgraded super-mirror guide ensure a neutron spectrum and overall flux optimized to materials of the kind described in this work.

The powder samples of Glyc(d₅)·MgSO₄·five-dimensional₂O and Glyc(d₅)·MgSO₄·three-dimensional₂O were loaded into separate thin-walled vanadium tubes, sealed and then rapidly chilled in liquid nitro­gen before being transferred into a closed-cycle-refrigerator (CCR) on the HRPD beamline at σim 170 K.

Beginning with the trideuterate, the specimen was cooled to 10 K and data were collected in the standard 30–130 ms t-o-f window for σim 4.3 h (corresponding to an integrated proton-beam current of 150 µA h). Once this was completed, the penta­deuterated specimen was mounted in the CCR and cooled to 10 K. Due to the lower symmetry of the penta­deuterate and the anti­cipated difficulty of indexing the triclinic cell far from the known room-temperature cell, data were collected in both the 30–130 ms and 100–200 ms t-o-f windows for 300 and 178 µA h, respectively. The longer flight-time window permits collection of data, albeit with a lower incident neutron flux, in the highest resolution back-scattering geometry to d-spacings of σim 4.0 Å (compared with 2.6 Å in the shorter time window).

Data from each detector bank were focused to a common scattering angle, normalized to the incident spectrum using a vanadium–niobium standard and exported in a format suitable for analysis using the GSAS/Expgui package (Larsen & Von Dreele, 2000; Toby, 2001).

Laser-stimulated Raman spectra were measured using a portable B&WTek i-Raman Plus spectrometer equipped with a 532 nm laser (P_max = 37 mW at the probe tip) that records spectra over the range 171–4002 cm⁻¹ with an optimal resolution of σim 3 cm⁻¹. Measurements were carried out on powdered specimens of Glyc·MgSO₄·3H₂O and Glyc·MgSO₄·5H₂O, both in their protonated and deuterated forms, the latter being used in the neutron powder diffraction experiment. Additionally, protonated and deuterated forms of the starting materials (MgSO₄·7H₂O and α-glycine) were measured so as to aid in identification of the observed vibrational features. All samples were measured in thin-walled glass vials using the BC100 fibre-optic coupled Raman probe; the total integration time and laser power for each sample is provided with the tabulated results.

The neutron powder diffraction data were refined by the Rietveld method using the GSAS/Expgui package (Larsen & Von Dreele, 2000; Toby, 2001) starting from the previously published atomic coordinates of Gly·MgSO₄·5H₂O (Elayaraja et al., 2007) and Gly·CoSO₄·3H₂O (Tepavitcharova et al., 2012). Stiff bond-distance restraints were applied initially to the S—O, Mg—O, C—C, C—O, C—N, O–D, C—D and N—D contacts, with all isotropic displacement parameters, U_iso, for like atoms constrained to be equal. As the refinement proceeded, the restraint weighting was reduced and eventually removed altogether. It was found necessary to refine a wavelength-dependent absorption coefficient (GSAS absorption model #1) in order to avoid U_iso becoming negative for several atoms, including O atoms. Ultimately, the value of U_iso for sulfur in the trihydrate and for both S and Mg in the penta­hydrate were obliged to be fixed at small positive values. The tendency of these atoms to refine to slightly negative U_iso reflects the fact that they are the heaviest atoms in structure and the temperature is very low, thus the displacement amplitudes are extremely small. Furthermore, sulfur has much the smallest neutron scattering cross section of the atoms in these compounds. Nevertheless, these issues do not affect the other geometric parameters and the final fit to the data is excellent in both cases, R_P being better than 2%. Tables 3 and 4 report the Rietveld powder statistics and graphical depictions of the fit to the neutron powder data are given in Fig. 3.

The refined unit-cell parameters and atomic positions for both Gly·MgSO₄·five-dimensional₂O and Gly·MgSO₄·three-dimensional₃O can be found in Table 5 and in the CIF file in the Supporting information. Covalent/ionic bond lengths inter­nal to the three polyhedral/molecular units are detailed in Table 6, along with comparable data for the protonated compounds at various other temperatures (note that the atom labels adopted by other authors may differ from those used in this report, but we refer to equivalent inter­atomic contacts). Details of covalent O—D, N—D and C—D bond lengths and of the hydrogen-bond geometry appear in Tables 7 and 8.

The crystal structures of Gly·MgSO₄·3H₂O and Gly·MgSO₄·5H₂O share some fundamental structural similarities, being built from two types of MgO₆ o­cta­hedra located on inversion centres, one type of sulfate tetra­hedron on a site of symmetry 1, and glycine in its zwitterionic form. Since neutral water molecules coordinate to the metallic cation, the difference in hydration state leads ultimately to modifications in the way the structural elements are connected to one another.

In Gly·MgSO₄·5H₂O, the Mg1 coordination o­cta­hedron consists of tetra­aqua­diglycinemagnesium(II), with glycine being coordinated to Mg by one of the carboxyl­ate O atoms; the inversion centre inevitably results in an all-trans conformation for these units, whereas the Mg2 o­cta­hedron has the form hexa­aqua­magnesium(II). The sulfate tetra­hedra are isolated, accepting hydrogen bonds primarily (but not exclusively) from Mg-coordinated water (Fig. 4).

By contrast to the monomeric structure of the penta­hydrate, in Gly·MgSO₄·3H₂O, the two MgO₆ o­cta­hedra are bridged in a binuclear bidentate fashion by the carboxyl­ate group of the glycine zwitterion, resulting in a chain-like polymer extending infinitely along the unique axis of the crystal (Fig. 5) with the form catena[di-µglycine[tetra­aqua-magnesium(II)][di­aqua-di­sulfate-magnesium(II)]]. As before, the inversion centre obliges all Mg-coordinated units to adopt a trans conformation.

The occurrence of corner sharing between MgO₆ o­cta­hedra and SO₄ tetra­hedra (i.e. coupled ion pairing) is typical of crystals with a limited number of hydrogen-bond donors, which is to say crystals with lower hydration states. The shift from monodentate to bidentate bridging coordination is not, however, a requirement of the reduced hydration state; for example, in orthorhombic Glyc·ZnSO₄·3H₂O (Fleck & Bohatý, 2004; Balakrishnan & Ramamurthi, 2007), there are only tetra­aqua­diglycinezinc(II) monomers, with one carboxyl­ate O atom of the glycine zwitterion engaged in monodentate coordination with the Zn atom.

The pattern of S—O bond lengths in the penta­hydrate agree very well with those found in other sulfates (e.g. Baur, 1964; Zalkin et al., 1964; Ferraris et al., 1973; Fortes et al., 2006), notably in the occurrence of longer S–O contacts where the O atom accepts two hydrogen bonds and shorter S—O contacts when the O atom accepts three hydrogen bonds. The same phenomenon has also been observed in recent high-precision single-crystal neutron diffraction studies of hydrated Mg selenates (Kolitsch, 2002; Fortes & Gutmann, 2014; Fortes et al., 2015). Note in Table 6 that the two single-crystal X-ray diffraction studies (Elayaraja et al., 2007; Tepavitcharova et al., 2012) report a much narrower spread of S—O bond lengths with no obvious systematic pattern, admittedly at much higher temperatures. In the trihydrate, the pattern is less clear, except to state that the Mg-coordinated sulfate O atom has the longest S—O contact, although this is barely statistically significant; nevertheless, Oguey et al. (2014) also report the Mg-coordinated S—O bond to be the longest. Inter­estingly, the mean S—O lengths in the present study for both the penta- and trihydrate are the same, i.e. 1.482 Å.

The pattern of Mg—O bond lengths in both compounds are in good agreement with the available X-ray refinements (Elayaraja et al., 2007; Tepavitcharova et al., 2012; Oguey et al., 2014). In the penta­hydrate, the Mg—O(carboxyl­ate) distance is comparatively large [2.121 (3) Å], whereas the other two Mg—O contacts are short in comparison with similar MgO₆ o­cta­hedra in compounds such as MgSO₄·7H₂O (Baur, 1964). This distortion cannot be entirely due to coordination of the Mg1 o­cta­hedron by theFortes glycine zwitterion, since the Mg2 displays a similar distortion without such coordination. The converse is true of the trihydrate, where the Mg—O(carboxyl­ate) distance (Mg1—O7) is the shortest of the Mg—O contacts. The equivalent bond in the second o­cta­hedron (Mg—O8) is probably longer because it also accepts a hydrogen bond. However, the presence of Mg bonding to sulfate O atoms, as well as to carboxyl­ate O atoms, affects the distribution of charge in the Mg coordination o­cta­hedron.

The geometry of the glycine zwitterion agrees reasonably well with other determinations in the same (protonated) compounds made by X-ray single-crystal diffraction methods at higher temperatures (Tables 8, 9 and 10) and extremely well with the determinations in α-glycine at room temperature by neutron single-crystal diffraction methods (Jönsson & Kvick, 1972; Power et al., 1976), particularly in respect of their mean N—D bond lengths (1.039 Å) and mean C—D bond lengths (1.090 Å). Differences in the carboxyl­ate C—O bond lengths between the two hydrates are attributable to the mono- versuss bidentate coordination found in the penta- and trihydrate, respectively. Note that the disparity in length is even more pronounced in the X-ray refinements of Elayaraja et al. (2007) and Tepavitcharova et al. (2012).

Both structures are characterized by a wide range of hydrogen bonds (Tables 7 and 8, and Fig. 6). The majority are O—H···O hydrogen bonds of medium strength (1.7 < H···O < 2.3 Å) and high linearity (∠ O—H···O > 150°), typical of two-centred inter­actions. Between the two compounds there are some subtle differences; the O—H···O hydrogen bonds in the trihydrate are longer than in the penta­hydrate, overlapping in length with the weaker N—H···O hydrogen bonds. In the penta­hydrate, by contrast, there is a clear delineation between the O—H···O and N—H···O hydrogen bonds at σim 1.9 Å.

As is typical of N—H···O hydrogen bonds, these contacts are longer and more strained than the O—H···O hydrogen bonds, with angles in the range 140–165°, comparable to those found in solid ammonia and in ammonia hydrates (Hewat & Riekel, 1979; Loveday et al., 1996; Fortes et al., 2003; Fortes, Wood & Knight (2009); OR Fortes, Suard et al. (2009); Griffiths et al., 2012).

In both compounds, the methyl groups appear to participate in weak C—H···O hydrogen bonds (cf. Steiner & Desiraju, 1998), with H···O lengths greater than 2.3 Å and a greater propensity to form bifurcated bonds (i.e. three-centred inter­actions) with C—H···O angles much smaller than 150°. However, both compounds have one C—H···O hydrogen bond that forms a two-centred hydrogen bond and the trihydrate has one methyl hydrogen (D2B) that is not involved in a hydrogen bond at all.

The Raman spectra of MgSO₄·7H₂O, α-glycine, Glyc·MgSO₄·3H₂O and Glyc·MgSO₄·5H₂O at 298 K are shown in Fig. 7; the sharpest and strongest vibrational feature, the symmetric stretch of the sulfate ion, has been truncated in order to display the many weaker bands to best effect. The Raman spectra of MgSO₄·7D₂O, α-glycine-d₅, Glyc(d₅)·MgSO₄·three-dimensional₂O and Glyc(d₅)·MgSO₄·five-dimensional₂O are similarly depicted in Fig. 8. The band positions and mode assignments for all eight compounds are given in Tables 9 and 10. Of the two hydrated glycine salts, only the Raman spectrum of the penta­hydrate has been measured previously (Tepavitcharova et al., 2012); their plotted spectrum and tabulated peak positions closely resemble ours (see Fig. 9 and Table 9).

Evidently, the vibrational spectra of the two glycine MgSO₄ hydrates are not simple superpositions of the starting materials, there being substantial changes in both Raman shift and intensity for various bands, which are diagnostic of the differences in structure between the two hydrates.

In the low frequency portion of the spectrum, the highly symmetric ν₁(SO₄²⁻) mode is the sharpest and most intense feature, occurring at 991.6 cm⁻¹ in the trihydrate crystal and at 983.8 cm⁻¹ in the penta­hydrate; compare this with 983.5 cm⁻¹ for the isolated SO₄²⁻ ion in MgSO₄·7H₂O. The asymmetric stretching and bending modes are similarly blue-shifted in the trihydrate (by 15–30 cm⁻¹) with respect to both the penta­hydrate and MgSO₄·7H₂O; the symmetric bending mode appears not be shifted but is split into two quite distinct components in the trihydrate. As one would expect, the vibrational frequencies of the SO₄²⁻ ions in the deuterated species, follow the same pattern of changes as in the protonated isotopologues.

These differences result from the dramatically different coordination environments of the sulfate tetra­hedra. In the penta­hydrate, the SO₄²⁻ tetra­hedra are isolated, with two apices accepting two hydrogen bonds each and the other two apices accepting three hydrogen bonds each – a total of ten hydrogen bonds, of which one is a weaker N—H···O type. In the trihydrate, one of the apical O atoms is directly coordinated to the Mg atom, forming a corner of the MgO₆ o­cta­hedron, whilst the remaining three apices accept two hydrogen bonds each – a total of six hydrogen bonds, of which two are of the N—H···O type. A reduction in the number of hydrogen bonds accepted by the apical O atoms of the XO₄²⁻ ion, due to limited availability of hydrogen-bond donors (e.g. water) and/or formation of coupled-ion pairs by direct cation coordination, leads to a shortening of the X—O bond length and an increase in the stretching vibration frequency; this has been found in a range of materials, both experimentally and computationally (Wang et al., 2006; Chio et al., 2007; Zhang et al., 2009; Pye & Walker, 2011).

The inter­nal vibrations of the glycine zwitterion span almost the whole range of our observations, from the lowest frequency deformation and rocking of the N—C—C—O skeleton at σim 350 cm⁻¹ through to the high-frequency C—H and N—H stretching of the methyl and amine groups, respectively, around 3000 cm⁻¹. The spectrum of our recrystallized α-glycine is an exemplary match in terms of peak position and shape, intensity and resolution to that reported by Shantakumari (1953), differing significantly from that of β- and γ-glycine, where relatively small differences in packing and inter­molecular hydrogen bonding cause quite large changes in vibrational frequencies (e.g. Yang et al., 2008).

With regard to the methyl group we observe: (i) an increasing blue-shift of the stretching frequencies with greater hydration; (ii) the two very sharp neighbouring peaks of the symmetric bending mode in α-glycine, coalesce and are blue-shifted in the trihydrate but red-shifted in the penta­hydrate; (iii) the virtually degenerate twisting and wagging modes at 1325 cm⁻¹ in α-glycine become separated, the splitting being slightly greater in the trihydrate; (iv) the very weak CH₂ rocking mode is not apparent in the spectra of either of the two protonated hydrates. Broadly speaking, the deuterated crystals manifest the same shifts: (i) the CD₂ stretching bands blue-shift, but by a smaller amount than the protonated species; (ii) the CD₂ bending modes are systematically red-shifted with higher hydration; (iii) the twisting mode of CD₂ shifts very little with hydration whereas the wagging mode is red-shifted by σim 10 cm⁻¹ in the penta­deuterate and σim 20 cm⁻¹ in the trideuterate; (iv) the rocking mode of CD₂ is clearly visible and shifts to lower frequencies (by 15 cm⁻¹) only in the penta­deuterate. The vibrational spectra indicate that C—H···O hydrogen bonds of similar strength to those in α-glycine occur in the trihydrate but that the inter­actions in the penta­hydrate are weaker and the H···O distances must be greater. This inter­pretation accords with the crystallographic determination that only one of the methyl H atoms is involved in a hydrogen-bond inter­action of the three-centred variety with two contacts of 2.45 and 2.63 Å, respectively, in the trihydrate, but that both methyl H atoms participate in hydrogen bonding (one of which is also a three-centred inter­action) of length 2.64–2.75 Å in the penta­hydrate.

With regard to the amine group we observe: (i) the N—H stretching mode is difficult to identify since it overlaps with stronger O—H stretching and bending modes from water, in both the protonated and deuterated forms; (ii) the NH³⁺ deformation modes are systematically red-shifted in both isotopologues, the strongest peak due to this vibration being shifted by −59 cm⁻¹ in the trihydrate, −13 cm⁻¹ in the trideuterate, −81 cm⁻¹ in the penta­hydrate and −34 cm⁻¹ in the penta­deuterate, relative to α-glycine and α-glycine-d₅, respectively; (iii) the rocking mode of NH₃⁺ is not visible in the protonated crystal, being masked by ν₃(SO₄²⁻), but is clearly blue-shifted by σim 20 cm⁻¹ in the trideuterate and slightly red-shifted in the penta­deuterate. In the trihydrate, we see that the amine group donates N—H···O hydrogen bonds of 1.77–1.99 Å length to neighbouring sulfate O atoms, whereas in the penta­hydrate the nearest plausible hydrogen-bond acceptors are 1.91–2.29 Å distant from the amine H atoms and the N—H···O hydrogen-bond angles are substanti­ally more strained (nearer 140 than 160°); indeed the longer of these contacts may not be a true hydrogen bond.

In respect of the N—C—C—O skeleton of the glycine zwitterion, there are only slight shifts in N—C and C—C vibrational frequencies; however, there some highly significant changes in the carboxyl­ate (COO⁻) vibrational modes reflecting the change from bidentate to monodentate Mg coordination in the trihydrate and penta­hydrate, respectively. The change is most readily apparent in the spectra of the deuterated materials. In α-glycine, the symmetric COO⁻ stretch occurs as a quite strong doublet (due to a Fermi resonance; see Machida et al., 1979) at 1392 and 1409 cm⁻¹, whereas the asymmetric stretching mode produces a rather weak feature near 1546 cm⁻¹. In Glyc(d₅)·MgSO₄·three-dimensional₂O, the splitting of the symmetric stretch appears to be eliminated (a weak shoulder is present, however) and the peak is blue-shifted by σim 50 cm⁻¹. The asymmetric COO⁻ stretch now appears as a weak doublet that is blue-shifted by σim 80 cm⁻¹ relative to α-glycine. In Glyc(d₅)·MgSO₄·five-dimensional₂O the splitting of the symmetric stretch is still absent (now with weak shoulders on the high- and low-frequency side) and the band centre corresponds roughly with the position observed in α-glycine. The asymmetric stretch, however, is blue-shifted by σim 70 cm⁻¹ relative to α-glycine. Whilst the frequency shifts are attributable to the change from hydrogen bonded to Mg-coordinated carboxyl­ate O atoms, the dramatic increase in intensity of the asymmetric stretch is presumably due to a resonant inter­action. The picture in the protonated crystals is rather unclear, since the relevant peaks are much weaker, but the behaviour of the symmetric stretch is essentially the same as in the deuterated compounds whereas the asymmetric stretch becomes systematically red-shifted in going from α-glycine to the trihydrate (−33 cm⁻¹) to the penta­hydrate (−39 cm⁻¹).

The Raman signature of bound water in both crystals is largely unremarkable. In the protonated crystals, two very broad features occur at σim 3250 and σim 3400 cm⁻¹ in both hydrates, although the intensity is inevitably greater in the penta­hydrate. In the deuterated crystals, the stretching vibrations of water occur in a narrower and more highly structured band exhibiting at least four distinct peaks.

This work is the first to refine the complete structure of Glyc·MgSO₄·nH₂O (n = 3 and 5) using neutron diffraction methods, giving accurate H-atom positions and bond lengths. Both hydration states are built of sulfate tetra­hedra, two types of Mg o­cta­hedra and glycine zwitterions. Differences in the connectivity of these three building blocks is clearly manifested in the vibrational spectra, particularly in relation to the symmetric S—O stretch and the COO⁻ stretching modes. These differences allow unambiguous distinction of the two compounds by Raman spectroscopy should they be discovered `in the field', whether that is on Earth or elsewhere in the cosmos.