I. INTRODUCTION
Initial accounts of the mineral middlebackite, CuII2C2O4(OH)2 from the Iron Monarch Quarry, Iron Knob, South Australia and from Passo di San Lugano, near Trento, Italy have been described by Elliott (Reference Elliott2016, Reference Elliott2018) and Demartin et al. (Reference Demartin, Campostrini, Ferretti and Rocchetti2017), respectively. The current publication reports crystallographic data for the independent discovery of the mineral at Mooloo Downs Station in Western Australia, some 2200 km WNW of the Iron Knob discovery site for the type mineral, where it was co-located with the chemically- and structurally-related mineral moolooite, CuIIC2O4·nH2O reported by Clarke and Williams (Reference Clarke and Williams1986). Significantly, the presence of moolooite was also reported for the Passo di Lugano material but not reported for the Iron Knob discovery.
The initial comparison of an X-ray diffraction pattern of middlebackite by the authors with that of moolooite indicated likely structural similarities, but, importantly, marked differences. While the complex crystallography of moolooite has been extensively studied, neither the chemical composition of the new mineral nor its crystal structure had been described at the time of its extraction at Mooloo Downs. Accordingly, the crystal structure of middlebackite was determined by the authors, with powder diffraction data being used, as a suitable single crystal was not available for Mooloo Downs material. SRD was preferred to laboratory XRD (LXRD) data as SRD pattern definition was expected to be superior. In parallel with the SRD structural solution of the natural mineral, a chemical synthesis procedure was developed, and comparison of the SRD data for natural and LXRD data for synthetic materials showed that the two forms were likely to have the same crystal structure.
As a background to the present study, there is an extensive list of reports in the literature on the crystallography of moolooite based on diffraction data for natural and synthetic forms. This literature has been reviewed by O'Connor et al. (Reference O'Connor, Clarke and Kimpton2019) in describing a recent SRD study of moolooite. Importantly, Schmittler (Reference Schmittler1968) observed sample-to-sample differences involving XRD line shifts, line broadening differences and the presence of additional lines in some samples. The differences were interpreted in terms of order-disorder (OD) theory (Fichtner-Schmittler, Reference Fichtner-Schmittler1979). Recently, Christensen et al. (Reference Christensen, Lebech, Andersen and Grivel2014) described the use of LXRD, SRD, and neutron diffraction (ND) data for synthetic copper oxalate powders in developing a disordered monoclinic structural model for synthetic Cu oxalate. By contrast with moolooite, the middlebackite material described here is fully ordered.
The discovery of middlebackite at three distant locations is of particular interest in terms of any differences in the crystallography of the mineral at the discovery sites, noting that the current powder diffraction study provided crystal data for an ensemble of crystallites, whereas those reported for the Iron Knob and Passo di San Lugano discoveries were based on single crystal data.
The occurrences of middlebackite should be seen with regard to the literature on the formation of metal-organo complexes from chemical reactions between mineral species and biota organics. With reference to the formation of natural copper oxalates, the literature in the subject includes the papers: Purvis (Reference Purvis1984), Chisholm et al. (Reference Chisholm, Jones and Purvis1987), and Fomina et al. (Reference Fomina, Hillier, Charnock, Melville, Alexander and Gadd2005). Middlebackite provides an additional indicator mineral for signs of early life on earth and in the earth planetary system, following speculation by Frost et al. (Reference Frost, Jing and Ding2003) on the potential value of oxalate minerals, such as moolooite, as candidate minerals.
The present X-ray diffraction publication is supported by the broader mineralogical paper of Clarke and O'Connor (Reference Clarke and O'Connor2019) on the discovery of middlebackite at Mooloo Downs Station.
II. EXPERIMENTAL
A. Characterisation: elemental composition, Raman, and FTIR spectroscopy
The summary information provided here is taken from a more detailed description provided by Clarke and O'Connor (Reference Clarke and O'Connor2019). Importantly in relation to crystal structure determination, the following data were reported:
• Photomicrograph imaging
• Measured atomic ratios for C : O : Cu natural and synthetic specimens = 19 : 58 : 21 and 16 : 62 : 21, respectively, compared with the theoretical values for Cu2C2O4(OH)2, 20 : 60 : 20; and
• Presence of functional groups C2O4 and OH from Raman/FTIR analysis.
It is clear from the data that middlebackite, while co-located with moolooite at Mooloo Downs, cannot be a moolooite polymorph.
B. Samples for diffraction analysis
Three specimens of topotype moolooite were re-examined to locate an unidentified mineral, previously reported by Clarke and Williams (Reference Clarke and Williams1986) as a suspected oxalate mineral. A composite sample of ~1 mg was prepared from six discrete subsamples containing the unidentified mineral (Mineral “A”, subsequently described as middlebackite) associated with minor amounts of moolooite and traces of quartz and libethenite.
The strategy developed for the synthesis of middlebackite was consistent with the hypothesis for its natural formation, and possibly that of moolooite, by the reaction of oxidised copper minerals with biogenic oxalate under circumneutral or slightly alkaline conditions. Synthetic malachite, Cu2CO3(OH)2, was prepared by adding excess sodium bicarbonate solution to a solution of copper nitrate trihydrate (1.208 g, 5 mmol) in ~20 mL of high purity water. The sodium bicarbonate solution was added dropwise over 20 min and the reaction mixture allowed to stand at room temperature for a further 30 min to complete out-gassing of carbon dioxide. The blue precipitate of synthetic malachite was separated from the reaction mixture by low speed centrifugation (3500 rpm for 10 min). Soluble salts were removed by washing the precipitate with ~50 mL of distilled water and further centrifugation. The composition of the precipitate as synthetic malachite was confirmed by LXRD and SEM/EDX analysis. The synthetic malachite was transferred by washing with ~30 mL of water into a clean plastic vial. A solution containing oxalic acid dihydrate (0.315 g, 2.5 mmol, Cu:oxalate mole ratio of 2 : 1) in 10 mL of water, was added dropwise with stirring over 30 min. The reaction with synthetic malachite commenced immediately as indicated by gentle effervescence, presumably as carbon dioxide produced by the reaction of ligand carbonate in synthetic malachite with oxalic acid. The reaction mixture was allowed to stand for up to 14 days at 50 °C in a laboratory oven. Progress of the reaction was monitored by frequent LXRD analysis.
C. Diffraction data measurement
An initial LXRD pattern (Figure 1) for the natural mineral was acquired at the ChemCentre with a PANalytical PW1820 Bragg-Brentano laboratory diffractometer (radius = 200 mm) using ca. 100 µg of the natural mineral. The measurement settings were (i) Co-anode X-ray tube with post-diffraction monochromator and a point detector, providing predominantly CoKα radiation and (ii) automatic divergence slit (irradiated length = 12 mm), 2θ range = 14–93°, 2θ step size = 0.02°, and counting time/step = 48 s. Similar settings were used to measure XRD patterns for synthetic middlebackite.
Comparison of a low-angle portion of the SRD pattern for natural middlebackite with the LXRD patterns for natural middlebackite and the synthetic material with hkl indices for middlebackite. Letters indicate “impurity” lines attributed to H peaks from the sample holder; L = libethenite, Cu2PO4OH; (iii) Q = quartz, SiO2; M = moolooite.
The SRD pattern used for structure analysis was measured using the Powder Diffraction Beamline at the Australian Synchrotron (AS) which was set to provide a photon energy 18.0068 keV (λ = 0.68816 Å). The sample had been manually ground between glass slides to a granularity ≪10 µm. Approximately 1 mg of material was packed in a 0.2 mm diameter glass capillary and then mounted on a crystallographic goniometer head which was rotated at 15 rpm. Patterns were measured using Debye–Scherrer geometry with a goniometer diameter = 152 cm, over the 2θ range 5.5087–85.3236°, and with an effective step size = 0.003 75°. The SRD incident and diffracted beams were fully polarised. The incident beam dimensions at the capillary were ~3 mm × 1 mm.
The pattern was measured with a 16 module Mythen II microstrip X-ray detector system (Schmitt et al., Reference Schmitt, Bronnimann, Eikenberry, Gozzo, Horrmann, Horisberger and Patterson2003), each module covering ca. 4.8° in 2θ with there being a gap of about ca. 0.2° between the segments covered by adjacent detector modules. Each pattern was measured twice using 600 s acquisitions, with the Mythen detector set being off-set by 0.5° for the second data acquisition to collect the data between the gaps. The two off-set patterns were merged using the program PDViPeR. An additional pattern was measured for a NIST SRM-660b LaB6 powder under exactly the same conditions in order to determine the instrument wavelength and the 2θ 0 correction, and the instrument full-width-half-maximum (FWHM) widths for the Bragg peaks. A 2θ 0 value = 0.0036° was determined by Rietveld fitting to LaB6 data measured for NIST SRM 660b. The SRD peak widths for the LaB6 pattern gave the following fitting parameters for a pseudo-Voigt function which were applied to correct the specimen peak widths for instrument broadening: U = 0.001 007, V = 0.0015, W = 0.000 133, and γ = 0.218°.
D. Diffraction pattern features
Peak positions, above-background intensities, and FWHM peak widths were determined with the PANalytical HighScore Plus (HSP) [Version 3.0] function PROFIT. The numerical diffraction data for the natural and synthetic materials, together with the pattern indexing results, are given in Table I.
Middlebackite (Natural) SRD pattern and synthetic LXRD pattern line position, peak intensity and width (FWHM).
SRD: λ = 0.688 161 Å, 2θ 0 = −0.0036°; LXRD (CoKα): λ = 1.790 31 Å, 2θ 0 = +0.042°.
SRD lattice parameters from indexing: SG P21/c, a = 7.2659(18) Å, b = 5.7460(11) Å, c = 5.6806(11) Å, β = 104.5883(31)°; Vc = 229.52.
LXRD lattice parameters from Rietveld analysis: a = 7.272(2) Å, b = 5.757(2) Å, c = 5.679(2) Å, β = 104.77(2)°; V c = 229.9.
SRD impurities: libethenite [L], Cu2(PO4)(OH): PDF 00-036-0404; quartz [Q], SiO2: 00-046-1045; moolooite [M], CuC2O4·xH2O: 00-021-0297.
Figure 1 compares a low-angle segment of the natural middlebackite SRD pattern and the corresponding LXRD patterns for the natural and synthetic materials. It is seen that:
• The statistical quality (signal : noise) of the middlebackite SRD pattern is superior in terms of (i) the definition of the Bragg peaks and (ii) the detectability of the weak peaks.
• The SRD and XRD peak widths for middlebackite are comparable as there is substantial sample-related peak broadening (see below).
• The “impurity” features in the patterns were identified as (i) for the LXRD pattern: lines from the quartz sample holder; and (ii) for the SRD pattern: libethenite, Cu2PO4(OH), quartz (SiO2), and moolooite. The level of moolooite in the natural sample amounts to ca. 11% by weight.
The XRD pattern for the synthetic form shows excellent agreement with the SRD pattern for the natural mineral, notwithstanding its significantly diminished angular resolution.
Examination of the lattice parameter data for the natural and synthetic materials, discussed below under Lattice Parameters and Space Group Determination, shows that the unit cell data for the synthetic material agree with the data for the natural specimen within the limits of experimental error. Importantly, the unit cell volume for the synthetic agrees with that for the natural material within ca. 0.2%. On this basis, it is evident that synthetic and natural materials have the same structure and stoichiometry.
E. Line broadening
An analysis of line broadening was conducted using a Williamson–Hall plot to extract volume-averaged estimates of crystallite size and microstrain. Figure 2 shows the plot of FWHM peak width, corrected for instrument broadening (Table I). There are clear indications of strain-size anisotropy according to the spread of data points about the linear line of best fit. The pronounced anisotropic broadening is similar to that reported for moolooite (O'Connor et al., Reference O'Connor, Clarke and Kimpton2019).
Above: plot of SRD FWHM peak widths, corrected for instrument broadening; below: Williamson–Hall plot.
Analysis of the instrument corrected peak widths gave the pseudo-Voigt fit parameters, U = 0, V = 0.0424, W = −0.000 901, γ 1 = 0, and γ 2 = 0.014°. The corresponding Williamson–Hall plot provided the following volume-averaged estimates: crystallite size = 200 nm, microstrain = 0.15%. These values are comparable with the SRD results for moolooite (200 nm and 0.10%), noting that moolooite comprises highly strained submicron crystals (O'Connor et al., Reference O'Connor, Clarke and Kimpton2019). Rietveld analysis with a double-Voigt function for line shape fitting (see below) subsequently indicated a microstrain value = 0.35%, with the crystallite size being indeterminate.
F. Lattice parameters and space group determination
Pattern indexing and space group assessments were conducted with the Panalytical High Score Plus (HSP) program using the SRD data. The HSP peak search function was used for ab-initio indexing with the algorithms: DICVOL04 (Bolti and Louer algorithm) and TREOR90 (Werner algorithm). The same excellent solution was obtained with both algorithms. The preferred monoclinic fit from indexing after least squares fitting, assuming prospective space group P21/c, gave reasonable agreement between the measured and calculated line positions, with a Smith–Snyder figure-of-merit F 57 = 34.7 (0.0072, 222). The lattice parameters from the fit – see Table I – were a = 7.2659(18) Å, b = 5.7460(11) Å, c = 5.6806(11) Å; β = 104.5883(31)°; V c = 229.52Å3. The Smith–Snyder FOM and unit cell dimensions for the synthetic material from LXRD data (Table I) were F 18 = 30.2 (0.017, 35) and a = 7.272(2) Å, b = 5.757(2) Å, c = 5.679(2) Å; β = 104.77(2)°; V c = 229.9 Å3. The middlebackite lattice parameters result from the present study are compared with those from the Elliott (Reference Elliott2018) and Demartin et al. (Reference Demartin, Campostrini, Ferretti and Rocchetti2017) publications on the Iron Knob and Passo di San Lugano middlebackite materials, respectively – Table II. While there is an approximate agreement between the five sets of parameters, there are some discernible differences.
Middlebackite lattice parameters from the present study compared with values from Demartin et al. (Reference Demartin, Campostrini, Ferretti and Rocchetti2017) and Elliott (Reference Elliott2018). Symbol Δ/σ signifies the difference between the SRD parameters from the present study and the corresponding parameters from the other studies.
Analysis of the systematic absences in the indexed pattern with HSP provided the list of prospective space groups – see assigned indices in Table I. Based on the condition “h0l, l = 2n”, the space group is one of Pc (#7), P2/c (#13) or P21/c (#14). The preferred space group is P21/c according to the conditions “h0l, l = 2n” and “0k0, k = 2n’. Condition “h0l, l = 2n” is unequivocally observed, whereas “0k0, k = 2n” is probably observed and 002 is the only 00l line measured.
Structure solution trials and subsequent Rietveld refinements were also run with space group P2/c. While the best model for this space group had plausible structural features, the interatomic distances and angles were not credible.
G. Structure solution with the SRD data
Structure analysis was conducted with (i) the charge flipping (CF) technique (Palatinus and Chapuis, Reference Palatinus and Chapuis2007) using both the HSP (V3) program and the program Topas (V5); and (ii) the simulated annealing (SA) method (Coelho, Reference Coelho2000) with program Topas. Importantly for this study, CF does not require elemental concentrations, whereas SA requires the stoichiometric formula to be specified (or guessed). The calculations were conducted after performing both Le Bail and Pawley fits to the SRD pattern. The CF symmetry analysis for the assumed space group showed excellent figures-of-merit for both 21 and c symmetry operators (0.60 and 0. 89, respectively) which is consistent with the reflection conditions.
The stoichiometry was not assumed for the CF trials, whereas, for SA, simulations were conducted with trial models in which a Cu ion position was optimised along with sets of three and then four “light atoms” with the scattering power of O. The Cu + 3xO configuration, which might represent the moolooite structure, failed to produce a chemically plausible arrangement of atoms. By contrast, the Cu + 4xO model provided chemical sensible packing consistent with there being two molecules of Cu2(OH)2C2O4 in the unit cell. This “best” model showed molecular ribbons packed similarly to those found for moolooite, with (i) the expected octahedral bonding of Cu atoms to oxygens, (ii) an unexpected Cu–O–Cu–O planar ring consistent with OH− groups, and (iii) an oxalate group linking the Cu–O(H)–Cu–O(H) rings. Table III compares the atom positions from the trials, and also the optimised site positions from Rietveld analysis and Figure 3 shows the disposition of the atoms for the optimised model. In terms of model validity, the CF Fourier map density of atom O2 for the CF model was very weak whereas the “sensible” thermal parameter obtained for the atom after Rietveld optimisation confirmed the position of O2 from the two methods.
(Colour online) Structural model from the CF and SA trials after Rietveld optimisation of the atom positions. View down the b-axis of the unit cell. Program Balls & Sticks. Atom codes: Cu blue, O red, and C white.
Atom coordinates from CF and SA trials, with relative densities of atoms from CF electron density Fourier map. Comparison with coordinates after Rietveld optimisation. CF: charge flipping, SA: synthetic annealing.
a Translated by (−1/2, −1/2, 0) relative to the origin for the coordinates reported by Demartin et al. (Reference Demartin, Campostrini, Ferretti and Rocchetti2017) and Elliott (Reference Elliott2018) and also for the coordinates in TABLE IV.
H. Rietveld optimisation of structural model
Figure 4 gives the low-angle section of the Rietveld residual plot for the optimised model.
(Colour online) Rietveld residual plot for the fit of the model in Table IV to the SRD data. The upper set of bar markers is for middlebackite, and the lower set is for the moolooite minor phase.
Table IV provides the final structural model and figures-of-merit from Rietveld optimisation, with the inclusion of the impurity phase moolooite for which the relative phase concentration was ca. 11.2% by weight. No chemical constraints were placed on the atoms. Importantly, we note from Table IV that the Durbin–Watson statistic (Hill and Flack, Reference Hill and Flack1987), DW, is 0.34 which indicates pronounced serial correlation between some parameters, in this case probably due to the planar groups and infinite linear molecular chains in the structure. A consequence of the relatively high serial correlations is examined in Table V, which shows selected correlation coefficients for relevant atom position parameters. In this regard, we would normally expect that positional parameter correlations to be smaller. Therefore for the bond lengths reported in Tables IV and VI, the bond length esds will have been underestimated.
Crystal structure model for middlebackite, following CF analysis, SA analysis and Rietveld optimisation. Moolooite component mineral (ca. 11.2% by weight) included in calculations.
Esd values given in parentheses – esds for the interatomic distances, notably involving C, are underestimated due to the pronounced correlation coefficients for most atom positional parameters not being taken into account.
Selected correlation coefficients between atom positional parameters. Values less than 0.30 not shown.
Comparison of middlebackite positional parameters from the present study with values from Demartin et al. (Reference Demartin, Campostrini, Ferretti and Rocchetti2017) and Elliott (Reference Elliott2018). Esd (σ) values in parentheses. Δ/σ = ratio of parameter difference Δ/difference σ. The Δ/σ (DE) values refer to the Δ/σ values for corresponding parameters from the Demartin et al. and Elliott studies.
The following correlations are noted in terms of their influence on interatomic distance uncertainties.
Cu: Cu(z) and C(z)
O1: O1(x) and O1(z); O1(x) and O2(x); O1(y) and O2(y); O1(y) and O3(y); O1(z) and O3(y)
O2: O2(x) and O3(z)
O3: O3(x) and O3(z); O3(x) and C(x)
C: C(x) and C(z)
The unique hydrogen atom position has been inferred from the hydrogen bond linking O1 and O3 in neighbouring ribbons – see Table IV. The O(H)–H distance was set to 1.0 Å.
I. Comparison of structural parameters from powder study with single crystal values from Elliott (Reference Elliott2018) and Demartin et al. (Reference Demartin, Campostrini, Ferretti and Rocchetti2017)
Table VI provides a comparison. Each parameter comparison is made in terms of the ratio Δ/σ so that values in excess of ca. 3 may be real. The following differences are noted – Cu and O2: y/b and z/c values; and O3 z/c values. While a referee has suggested that the differences may be due to the space group assignment being incorrect, alternate space groups were eliminated – see “Lattice parameters and space group determination”. It is also noted that the single crystal study by Elliott (Reference Elliott2018) proposed the same space group. Possible explanations for the differences are (i) chemical substitution effects for the samples from the three studies, (ii) the well-known tendency for underestimation of uncertainties in Rietveld calculations, (iii) the high parameter correlations for the present powder study, and (iv) the possibility of disorder effects. Also concerning apparent structural parameter differences between the samples from the three sources of middlebackite, we note from Table II that there are significant sample-to-sample differences for some b and c values.
The differences point to the need for further structural research on middlebackite, with neutron diffraction studies using synthetic material (powder and single crystal) being prospective.
III. DISCUSSION
A. Structural overview
The approach taken here, focusing on chemical bonding, within and between the molecular ribbons, complements that followed by Demartin et al. (Reference Demartin, Campostrini, Ferretti and Rocchetti2017) and Elliott (Reference Elliott2018) both of whom mainly described the structure in terms of polyhedra distributions.
Figure 5 depicts the stacking of CuII2C2O4(OH)2 chains, or “ribbons". Oxygen O1 and the inversion-centre related oxygen in the Cu–O(H)–Cu–O(H) 4-membered oxybridged ring have hydrogens attached. End-on views of ribbon stacking are shown in Figure 5, with the inter-ribbon apical Cu-O bonds being (i) the Cu–O1 link (2.45 Å) to hydroxyl oxygen in a neighbouring ribbon and (ii) the substantially longer Cu–O3 bond (2.79 Å) to an oxalate ligand. The Cu ion is therefore described as Cu(II) with distorted 4 + 1 + 1 octahedral bonding. There is also the weak inter-ribbon O1–H…O3 hydrogen bond (2.94 Å) between each hydroxyl group and neighbouring oxalate oxygen. We note, in particular, that the inter-ribbon bonding results in the bonding environments of the two oxalate oxygens, O2 and O3, being distinctly different. Whereas O2 is not involved in inter-ribbon Cu–O bonding or O–H…O hydrogen bonding, O3 participates in both types of inter-ribbon linkages.
(Colour online) Comparison of ribbon stacking in middlebackite with moolooite arrangement. Atom sizes in decreasing order Cu, O, C, and H. H atom and bonds in red. Middlebackite: left view down unique b-axis for P21/c; right view normal to b. Moolooite: left view down unique b-axis for P21/n, right view normal to b. Program Balls & Sticks.
B. Comparison of ribbon stacking in middlebackite with stacking in moolooite
Figure 5 shows a crystal structural comparison using the moolooite structural model optimised with the synthetic CSIRO moolooite SRD data from O’Connor et al. (Reference O'Connor, Clarke and Kimpton2019). It is seen that the middlebackite ribbons compared with those in moolooite have Cu–OH–Cu–OH bridging rings linking adjacent oxalate ions as well as the hydrogen bonds from O1(H) to O3. The ribbon packing in middlebackite is distorted from that seen in moolooite, with the ribbon plane subtending an angle of 62° with the plane for neighbouring ribbons. This distortion is attributed to the influence of the inter-ribbon hydrogen bond which is absent in moolooite. The hydrogen bond results in ribbons being tilted towards the neighbouring ribbon by ca. 28°, whereas in moolooite neighbouring ribbons are tilted from each other by ~90°. In effect, the hydrogen bond sheets act as “guy-wires” between neighbouring ribbons.
C. Bond lengths and angles/coordination
Figure 6 shows the key distances and angles for the intra-ribbon bonds from the present study. In considering the results, caution should be exercised to note the possibility of the positional parameters being biased by correlation effects in the least squares calculations.
Intra-ribbon distances (Å) and angles (°) from this study of middlebackite. Program Balls & Sticks. Inter-ribbon hydrogen bonds not to scale.
Table VII provides a comparative analysis of the bond lengths and angles for the three middlebackite studies and makes comparisons with the recently published structure of the related dihydrate, fiemmeite, Cu2(C2O4)(OH)2·2H2O (Demartin et al., Reference Demartin, Campostrini, Ferretti and Rocchetti2018). The table includes representative weighted average (WA) values for key distances and angles according to the inverse variances. The following points are made on the analysis.
Bond distances (Å) and angles (°) for middlebackite from the SRD data compared with other studies.
a Calculated for this publication.
Hydroxy-bridged ring Cu–O(H)–Cu–O(H):
• The WA middlebackite equatorial Cu–O(H) distance, 1.946 Å, agrees very closely with the value for fiemmeite, 1.944 Å, whereas there is a contraction of the Cu–Cu distance in middlebackite, 2.845 Å, relative to the value for fiemmeite, 2.920 Å. The middlebackite Cu–Cu contraction is mirrored by the more open O(H)–Cu–O(H) angle 86.1° in middlebackite compared with 83.1° for fiemmeite. Oswald et al. (Reference Oswald, Reller, Schmalle and Dubler1990) reported similar Cu–O(H) values for spertiniite, Cu(OH)2 – 1.95 and 1.97 Å – and a Cu–Cu distance of 2.957 which is considerably longer than that for middlebackite.
Cu–oxalate ring Cu–O–C–C–O–Cu:
• The WA middlebackite equatorial Cu–O distance, 1.992 Å, and that for fiemmeite, 1.979 Å are discernibly longer than the Cu–O(H) equatorial average distance in the Cu–O(H) ring, 1.945 Å.
• The oxalate ligand values for the C–O bonds – mean value = 1.253 Å for middlebackite and fiemmeite – is significantly shorter than the mean 1.279 Å reported by International Tables for X-ray Crystallography (Reference Wilson1995) for many studies of Cu-chelating oxalate ions, η 2(O2C)2; whereas the mean WA C–C bond length for middlebackite and fiemmeite, 1.545 Å, agrees with the mean C–C single bond value from the In Tab Cryst compilation, 1.546 Å.
Inter-ribbon contacts:
• The Cu–O octahedra have one short apical bond and one longer apical bond in both middlebackite and fiemmeite – short apical bond = 2.416 Å for middlebackite and slightly contracted for fiemmeite = 2.354 Å; and long apical bond = 2.749 Å for middlebackite and longer for fiemmeite = 2.835 Å.
• The O(H)…O hydrogen bond in middlebackite, 2.944 Å, indicates a relatively weak hydrogen bond compared with that for fiemmeite, 2.882 Å.
D. Crystallochemical validation from the synthesis of the mineral
The unit cell parameters for the synthetic material are provided in the Experimental section where they are compared with the cell parameters for the natural mineral. The close agreement between the diffraction data sets for the natural and synthetic materials gives strong credence to the structural model reported here.
E. Comparison of the d-I data with values from Elliott (Reference Elliott2018)
Elliott (Reference Elliott2018) has quoted LXRD d-I diffraction data. These are compared with SRD d-I data from the current study in Table VIII, which shows sound agreement as expected from the close agreement between the lattice parameters. The considerably superior quality of the SRD d-I data is evident.
Comparison of SRD d-I data for Mooloo Downs middlebackite from the present study with LXRD data from the Elliott (Reference Elliott2018) study of Iron Knob middlebackite.
a Measured with Rigaku Hiflux Homelab diffractometer (CuKα, 1.541 870 Å).
IV. CONCLUSION
Whereas metal-hydroxyl minerals occur commonly (e.g. brucite, Mg(OH)2; gibbsite, Al(OH)3), the presence of hydroxyl groups in metal-oxalate minerals had not been reported prior to the discovery of middlebackite. Of the minerals found in association with moolooite at Mooloo Downs, four other copper hydroxy minerals have been observed in addition to middlebackite which points to the likelihood of circumneutral to moderately high pH conditions applying at the times the site was exposed to weathering (Clarke and Williams, Reference Clarke and Williams1986): brochantite, Cu4SO4(OH)6; antlerite, Cu3(SO4)(OH)4; atacamite, Cu2Cl(OH)3; and libethenite, Cu2PO4(OH). Regarding the observation of atacamite, this mineral is known to be an alteration product of chalcocite, Cu2S, which has been reported to have been found in association with spertiniite, Cu(OH)2, under conditions which favour oxidation in an alkaline environment (Grice and Gasparrini, Reference Grice and Gasparrini1981).
Based on phase evolution observed in the middlebackite synthesis experiment, it is evident that the mineral is an alteration product of moolooite , which is represented by:
$$\lsqb {{\rm C}{\rm u}^{{\rm II}}{\rm C}_2{\rm O}_4} \rsqb _{\rm n} \, + \,{\rm C}{\rm u}^{2 +} \, + \,2{\rm O}{\rm H} \rightarrow \Box\Box\lsqb {{\rm C}{\rm u}^{{\rm II}}{_2} {\rm C}_2{\rm O}_4{\lpar {{\rm OH}} \rpar }_2} \rsqb _{\rm n}$$The observation of a copper oxalate hydroxide mineral is consistent with the moolooite forming from reactions of oxalic acid from biota with mineral Cu. The formation of middlebackite or moolooite under natural conditions may depend on the ratio of cupric mineral substrates to biogenic oxalate. Excess oxalate would appear to favour the formation of moolooite, while oxalate-limiting conditions favour middlebackite. The finding of middlebackite at Iron Monarch without it being associated with moolooite points to a deficiency of oxalate relative to copper ions.
It will be valuable if the reporting of the mineral for three global sites will lead to other such discoveries of middlebackite in association with moolooite. This would include other geological sites at which moolooite has been reported and sites where moolooite has been attributed to mineralogical Cu uptake, as has been proposed as a mechanism for in-plant formation (Chisholm et al., Reference Chisholm, Jones and Purvis1987; Fomina et al., Reference Fomina, Hillier, Charnock, Melville, Alexander and Gadd2005).
Finally, it would be of value to consider (i) carbon dating of copper oxalate minerals from different sources on that basis that copper oxalate minerals formed in the geological environment may be much older than material found in plants; and (ii) stable isotope mass spectrometry.
SUPPLEMENTARY MATERIAL
The supplementary material for this article, comprising the three patterns, can be found at https://doi.org/10.1017/S0885715619000630.
ACKNOWLEDGEMENTS
The authors acknowledge support from their host institutions and from the Australian Synchrotron for access to the Powder Diffraction Beamline. In particular, we acknowledge the ChemCentre for material support in conducting the chemical synthesis experiments, Curtin University for access to software and data bases and W. van Bronswijk of Curtin University for the Raman and FTIR measurements and interpretation.
