Meissner–Ochsenfeld superconducting anomalies in the Be–Ag–F system

Solid State Communications 130 (2004) 137–142 www.elsevier.com/locate/ssc

Meissner – Ochsenfeld superconducting anomalies in the Be –Ag –F systemq Wojciech Grochalaa,b,*, Adrian Porchc, Peter P. Edwardsd,* a

Schools of Chemistry, and Metallurgy and Materials, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK b Department of Chemistry, University of Warsaw, Pasteur 1, Warsaw 02093, Poland c School of Engineering, University of Cardiff, Cardiff CF1 3TB, UK d Department of Chemistry, University of Oxford, South Parks Road, Oxford OX1 3QR, UK Received 30 October 2003; accepted 20 November 2003 by C.N.R. Rao

This paper is dedicated to Prof. Alex K. Mu¨ller, while commemorating the 175th anniversary of isolation of beryllium metal by Friederich Wo¨hler in 1828

Abstract We report here observations of sudden drops in the magnetic susceptibility of a large number of samples in the Be – Ag– F system, over the temperatures ranging from 8.5 to 64 K. These magnetic anomalies, strongly reminiscent of the Meissner– Ochsenfeld effect, suggest the presence of superconductivity in a small sample fraction of this fluoroargentate system, a cousin of the oxocuprate High-TC materials. q 2003 Elsevier Ltd. All rights reserved. PACS: 74.20 M; 74.25 H; 75.40 C; 81.05 Z Keywords: A. Fluorides; A. Silver; A. Superconductors; D. Magnetic properties

A recent paper has advanced the intriguing prospect of hightemperature superconductivity in the unusual and relativelyunexplored chemical compounds of silver and fluorine, the fluoroargentates [1]. The experimental search for superconductivity in these materials will be far from straightforward; they are distinctly difficult to synthesize, extremely hygroscopic and aggressively oxidizing [2] toward metals, ruling out the detection of superconductivity via d.c. electrical resistivity measurements [3]. We have taken advantage of the intrinsic high sensitivity of magnetic susceptibility measurements to search for signs of possible superconductivity in fluoroargentates and we q Supplementary data associated with this article can be found in the online version at doi: 10.1016/j.ssc.2003.11.046 * Corresponding authors. Address: Department of Chemistry, University of Warsaw, Pasteur 1, Warsaw 02093, Poland. Tel.: þ 48-22-8220211x276; fax: þ 48-22-8222309. E-mail addresses: [email protected] (W. Grochala), peter. [email protected] (P.P. Edwards).

0038-1098/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.ssc.2003.11.046

report here observations of sudden drops in the magnetic susceptibility of a large number of samples in the Be – Ag– F system which we attribute to possible superconductivity— and the attendant Meissner– Ochsenfeld effect—at temperatures ranging from 8.5 to 64 K. The observed magnetic anomalies are even the more dramatic given that they are readily observable in the presence of a substantial (background) ferromagnetic signal from unreacted AgF2 (the latter also renders problematic any mutual a.c. inductance measurements to detect superconductivity). We have deliberately targeted the Be – Ag– F system in an attempt to ‘crystal-engineer’ a fluoroargentate with either [AgFþ] infinite chains or [AgF2] sheets-proposed key ingredients [1] for potential high TC —with a proper count of fluorine ligands, and a presumed high Debye temperature provided by the light chemical element (Be) component of the crystal structure. A hypothetical compound BeAgIIF4 would fall between the LiAgF3 (i.e. presumably Liþ(AgF2 3) [4]) and BAgF5 (i.e. [AgFþ][BF2 4 ]) systems, and this makes the title system a good candidate for high covalency within

138

W. Grochala et al. / Solid State Communications 130 (2004) 137–142

Fig. 1. X-ray diffraction pattern for one of the ‘BeAgF4’ samples. Starlets indicate peaks which could not be assigned to any of the known compounds of Ag and/or Be, and which we tentatively assign to novel superconducting phases.

the Ag– F bonds [5], contained within the [Ag –F2] sheets, and suitably ‘self-doped’ via electronic band crossing. We have synthesized a total of 21, black, strongly oxidizing and highly hygroscopic specimens [6] centred around a target composition of ‘BeAgF4’ [7]. The samples were typically fired at 350– 390 8C for 3 – 24 h in a purified N2 gas flow using the highest available purity binary fluorides BeF2 (99.9%) and AgF2 (99.5%). The starting materials were weighed, mechanically-ground and loaded in a controlled-atmosphere argon glovebox. Powder X-ray diffraction (Bruker D5005 diffractometer) revealed several new phases, as yet unidentified, in addition to the unreacted starting materials, most notably AgF2 and BeF2, and AgF. The various weak diffraction peaks, which could not be assigned to any of the known compounds of F and Ag or Be, are seen at 18.45, 22.47, 22.95, 29.34, 32.13, 33.11, 40.12, 42.93, 49.62 and at 55.668 in the diffraction pattern (see Fig. 1). Our proposal is that these low – abundant unidentified phases are responsible for the magnetic anomalies described below, which we ascribe to superconductivity. We examined all materials by measurements of the temperature-dependent magnetic susceptibility ðxÞ with a Superconducting Quantum Interference Device (SQUID) d.c. magnetometer (Cryogenics S100). Magnetization data corresponding to the zero-field-cooled (ZFC) regime were recorded on heating, after the sample was cooled in zero applied field down to 5 K. Magnetization in the field-cooled (FC) regime were measured as a function of increasing temperature after the sample was cooled in the applied field (typically 10 – 20 G). Plots of the measured magnetic susceptibilities, x; vs. temperature, T; for AgF2 and three representative Be – Ag – F samples are given in Fig. 2. Note that our susceptibility

values are presented in arbitrary units; the original data are recorded in terms of gm susceptibilities, xg (emu g21). To determine the susceptibility in units of emu cm23, x0 ; and potentially the fraction of complete diamagnetism, one would normally proceed via x0 ¼ xg £ density: Clearly, any density determination for the material responsible for superconductivity in these multi-phasic samples would be meaningless. Thus, the fraction of flux expulsion relative to (theoretical) complete diamagnetism ðx0 ðtheorÞ ¼ 21Þ cannot be estimated. Similarly, the presence of a majority, ferromagnetic phase, AgF2, severely complicates any determination (see below). The absolute values of x are overall positive, since the magnetic response in all samples is generally dominated by unreacted AgF2, a spin-canted ferromagnet with a Curie temperature of 163 K [8]. However, a sharp decrease in the measured x was observed for a total of 13 of our samples as the temperature was decreased in the ZFC experiments. Eight samples exhibited similar magnetic behaviour at 8.5– 13.0 K, two samples at 25.6 K [9], while three other samples showed distinct changes at temperatures of 45, 52 and 64 K. Similar magnetic anomalies were also observed below 89 K for one particular sample. Usually the susceptibility change reaches 3% of the (limiting) total low-temperature value, but for one sample having particularly low AgF2 content (labelled d in Fig. 2) it was as high as 60% [10]. Importantly, we have not observed such behaviour when AgF2 itself was thermally processed in a similar synthetic procedure to that outlined here. We note the pronounced difference between ZFC and FC magnetisation curves (Fig. 2). Three physical phenomena might be responsible for such a steep change of x : (i) a Peierls distortion [11]; (ii) a sudden magnetization of

W. Grochala et al. / Solid State Communications 130 (2004) 137–142

139

Fig. 2. The variation of magnetic susceptibility, x; vs. temperature, T; for AgF2 (a), and for three representative samples within the Ag–Be –F system (b, c and d). ZFC is a zero-field-cooled, and FC is a field-cooled measurement.

ferromagnetic domains spontaneously oriented opposite to the external field at low temperatures (recall, ferromagnetic AgF2 is invariably present in the samples); and (iii) a Meissner –Ochsenfeld effect, arising from magnetic flux expulsion from superconducting component(s) of the samples. A Peierls distortion can be ruled out when one compares the FC and ZFC experiments (no x increase was seen in the FC experiments). Instead, sample c exhibited a x decrease below 45 K in a ZFC experiment accompanied by a slight x decrease above 45 K in a FC experiment (Fig. 2; see also b). A superconductor is unique in its magnetic response; screening supercurrents (ZFC regime) and the Meissner –Ochsenfeld effect or magnetic flux expulsion (FC regime) represent distinct magnetic signatures for the occurrence of superconductivity below a superconducting critical temperature [12]. Thus, for example, the x vs. T dependence seen for sample c points to the potential presence of a superconducting phase with TC around

45 K. We therefore, propose the existence of hightemperature superconductivity in the title system arising from minority-phase particles of a new superconducting phase embedded within a (majority) host AgF2 ferromagnetic matrix. The possible coexistence of regions of superconductivity within a strongly ferromagnetic host is highly unusual and certainly not usually encountered in the superconducting cuprates. Much of our data can be accounted for by analysing the expected magnetic behaviour of this system. For example, the difference between the experimental ZFC and FC data can be explained by the high concentration of the applied magnetic field caused by the AgF2 host. We postulate that these local fields within the sample exceed the lower critical field of the superconducting regions, resulting in a highly suppressed diamagnetic transition in the FC situation. The magnitude of the susceptibility change on entering

140

W. Grochala et al. / Solid State Communications 130 (2004) 137–142

the superconducting state is impossible to quantify from the raw SQUID magnetometer data, not just due to uncertainties in superconducting volume fraction, but also (crucially) due to the make-up of the samples. We currently have no knowledge of the normal-state magnetic properties of the Be – Ag– F phase which we associate with the superconductivity. We have carried out detailed calculations [13] on the size of the diamagnetic transition on entering the superconducting, bearing in mind that the superconducting phase in its normal state is also likely to be ferromagnetic with a different susceptibility compared with the AgF2 host, and that the Meissner effect will lead to flux being excluded into a strongly ferromagnetic host. Assuming isolated spherical superconducting grains, we find that the decrease in the stored magnetic energy, DU (proportional to the measured SQUID voltage), on entering the superconducting state is given [13]   a 1 þ x1 DU / ð1 þ x2 Þ ; a¼ aþ2 1 þ x2 where x1 is the magnetic susceptibility of the superconducting Be – Ag – F phase in its normal state and x2 is the susceptibility of the AgF2 host. We predict that even when embedded in a strongly ferromagnetic host, the superconducting grains exhibit flux exclusion in the usual way when measured with a SQUID magnetometer but, crucially, that the size of the diamagnetic transition is suppressed by the factor [13] 3a aþ2 If a ¼ 1 we observe the full magnetic transition, in which case one could quantify the volume of superconducting material in the usual manner. However, we would not expect the superconducting Be – Ag – F phase to be strongly ferromagnetic in its normal state, in which case a p 1 and the size of the diamagnetic transition is strongly suppressed. Hence, even a small measured transition could imply a large superconducting volume fraction. For these reasons we place no emphasis on, and we are unable to analyse further, the magnitude of the raw SQUID voltage data for these samples in terms of volume fraction of superconductivity [14,15]. The question may also naturally arise as to whether the observed susceptibility changes attributed to superconductivity are connected with the presence of copper-based superconducting materials in the specimens (note: Ag compounds are often contaminated by Cu). In the highlyoxidizing synthetic conditions provided by the presence of AgII fluorides, and in contact with small amounts of water vapour, some superconducting oxocuprates might conceivably be formed. In order to eliminate the possibility, we have obtained Inductively Coupled Plasma Mass (ICP-MS) spectra for several samples studied, and also for AgF2 and BeF2 used for synthesis (spectra were obtained using a Agilent Technologies 4500 spectrometer with Cetac LSX-

200 laser ablation in the Center of Analytical Sciences at the University of Sheffield, UK). The ICP-MS data revealed that the samples are contaminated mainly by Si (at most, up to 5 molar%, but typically 1 molar%). This is mainly due to significant contamination of commercial BeF2 with Si, presumably due to using glass fragments during the manufacture of the material. The measured Be/Ag ratio in our samples varied from 0.90 to 1.02. A significant negative deviation from unity for one sample may be attributed to loss of slightly volatile BeF2 during synthesis. The largest contamination with Cu was found at a level of 0.18% (but typically one order of magnitude less). This certainly cannot account for the significantly larger drops in susceptibility observed for over half of our samples. Thus, we conclude that the observed susceptibility behaviour originates from the phases containing Ag, Be, F, and possibly traces of O and C (the two latter elements cannot be detected by ICP-MS spectroscopy). Complementary experimental evidence for superconductivity, along with the magnetic signals of the Meissner– Ochsenfeld effect, might usually be provided by d.c. electric resistivity measurements. However, measuring the electrical conductivity of highly air- and moisture-sensitive microcrystalline powders such as fluoroargentates presents a formidable experimental challenge. Over the past few years, prompted by similar problems in other air- and moisturesensitive materials, we have developed a microwave cavity perturbation technique to measure the a.c. electrical conductivity of powder samples without the need for sample contacts; complete experimental details are given elsewhere [16]. This technique has been recently used to investigate the electric conductivity of KAgF3 [3b]. We applied this non-intrusive technique to determine the conductivity behaviour of the Be – Ag – F phases. Our cavity perturbation technique, unfortunately, is also weakly sensitive to the ferromagnetic ordering (163 K) of AgF2, which is a major magnetic impurity in our samples. However, the variation of the resonance microwave frequency (1.5%) and of the resonance band width (comparable to the noise level) is very small as compared to the respective magnetic susceptibility variation seen by SQUID for AgF2 (over 2 orders of magnitude). Thus, with our present experimental set-up we could not detect any significant conductivity increase at low temperatures [17]. Low sensitivity might potentially be due to relatively high microwave frequency used (3.2 GHz), current facilities available to us operate at this frequency only. An alternative non-contact method at low frequencies is now being developed (requiring construction of a completely new copper hairpin probe) in order to determine the electric conductivity component along the magnetic response detected by SQUID for the Be – Ag –F phases. We note that the behaviour observed here is reminiscent of the recent report of 35 K superconductivity in graphite – sulphur composites [18]; in that system, cooling actually

W. Grochala et al. / Solid State Communications 130 (2004) 137–142

leads to an increase in the measured electrical resistivity, even though magnetization measurements unequivocally demonstrate superconducting behaviour below a the critical temperature. Although the phases responsible for the susceptibility changes detected by our SQUID measurements are not yet established, the clear magnetic anomalies, typical of the Meissner – Ochsenfeld effect, suggest the presence of superconductivity—up to 64 K—in fluoroargentates, cousins of the oxocuprate High-TC materials. The challenge now remains to isolate and fully identify [19] the phases responsible for the magnetic susceptibility anomalies in these new materials at the very limits of extreme chemical reactivity and metastability [20].

Acknowledgements We thank the Royal Society for the award of a Visiting Fellowship to WG and to The Crescendum Est-Polonia Foundation (Poland) for financial support; this author now holds a Fellowship of the Foundation for Polish Science (FNP). Prof. David Jefferson (Cambridge University) is gratefully acknowledged for performing the STM analysis, Dr Simon Kitchin (University of Birmingham) for obtaining of the 19F NMR spectra, Dr Alan Cox (University of Sheffield, UK) for measuring of the ICP-MS spectra, and Dr Andrzej Kudelski (University of Warsaw, Poland) for collecting of the Raman spectra. Mr Terry Green is thanked for his kind technical assistance during the ESR measurements.

References [1] W. Grochala, R. Hoffmann, Angew. Chem., Int. Ed. Engl. 40 (2001) 2743. [2] (a) N. Bartlett et al., J. Fluorine Chem. 71 (1995) 163. (b) B. Zˇemva et al., J. Am. Chem. Soc. 112 (1990) 4846. [3] (a) W.J. Casteel Jr. et al., J. Solid State Chem. 96 (1992) 84. The non-intrusive (non-contact) microwave cavity perturbation technique has recently allowed for the confirmation of metallic a.c. conductivity of KAg(II)F3 perovskite above 50 K: (b).W. Grochala, P.P. Edwards, Phys. Stat. Solidi B 240 (2003) R11. [4] The attempts have failed to obtain LiAgF3 both via the classical solid state synthesis: W. Grochala, P. P. Edwards, unpublished results, and through the controlled thermal decomposition of LiAgIIIF4 in vacuum: Z. Mazej, B. Zˇemva, personal communication. [5] Significant covalency of the Ag– F bonds has recently been confirmed by XPS measurements for higher fluorides of Ag: W. Grochala et al., Chem. Phys. Chem. 4 (2003) 997 and by the independent hybrid HF/DFT spin-polarized calculations for AgF2: N. Harrison, Daresbury Laboratory, personal communication.

141

[6] These highly-reactive compounds should be stored in sealed teflon containers-preferably prefluorinated. [7] (a) The magnetic susceptibility falls described in this paper are also seen (typically at 8 –10 K) for ca. 30% of the samples with formal composition of Ag2BeF6 and AgBe2F6. (b) The role of F-non-stoichiometry is yet unclear, the slight F-deficiency being the most probable scenario. Recollect that F2 is released from higher fluorides of Ag upon heating. It is unlikely that grains of locally hole- (i.e. AgIII) doped system are formed at the reaction conditions. The investigations of mysterious electron-doped Ag9F16 are now conducted at the University of Tennessee: N. Bartlett, personal communication. [8] B.G. Mu¨ller, Naturwissenschaften 66 (1979) 519. [9] One of these samples have shown an evident kink at 26 K in the dependence of the integrated ESR intensity vs. temperature; this shows that free-electron susceptibility suddenly decreases at the same temperature at which the all-electron susceptibility does (as seen in SQUID). [10] This sample has been subject to prolonged spontaneous decomposition in an attempt to remove as much ferromagnetic AgF2 as possible. The varying doping level of Ag1þto the predominantly Ag2þ-containing samples (higher fluorides of silver slowly decompose at elevated temperatures) might account for the large differences in TC for various samples. It also indicates large sensitivity of the Be–Ag–F phases to tiny variations in the conditions of their preparation and handling. [11] The U-shaped susceptibility drops at 63 K were first observed for the AgII compounds by Bartlett and coworkers in 1992 (Ref. [3]), and interpreted as a manifestation of a Peierls distortion. Later this group realized that x drops are independent on the counterion (MF2 6 , M ¼ Sb, As, Au), and they are rather the effect of washing the [AgFþ] salts with anhydrous HF. The material that exhibits the 63 K anomaly, does not produce identifying lines in the X-ray diffraction pattern (the parent materials give sharp strong patterns). A proposal is that some unidentified impurity might be responsible for x anomalies (included as a personal communication in Ref. [1]). Interestingly, the temperature at which the anomaly is seen for the [AgFþ] salts (63 K) is close to the largest temperature at which the magnetic susceptibility drops are seen in our experiments (64 K). [12] Compare with the magnetic response of the superconducting Ta/C multilayers, given in: I.I.S. Suzuki, M. Suzuki, J. Walter, Solid State Commun. 118 (2001) 523. [13] A. Porch, W. Grochala, P.P. Edwards, calculations available as Appendix. [14] Unfortunately, the volume fraction of this novel SC cannot also be easily determined e.g. via comparison with an internal standard such as Nb or Pb; the extreme chemical reactivity leads to instantaneous decomposition of the internal superconductor. We note also that the magnetic response from the SC phase is strongly suppressed (this is especially true for intergranular supercurrents, even at low external fields; see: A.M. Campbell, J. Blunt, Supercond. Sci. Technol. 3 (1990) 450 and may even be eliminated as seen in the FC measurements. [15] It cannot be excluded that our samples exhibit surface SC from a thin surface slab of the new superconducting phase (the analysis discussed in the text also applies). Such slab might be generated via the local crystal structure modification of AgF2 by the attached grains of BeF2 and through the subsequent

142

[16]

[17] [18] [19]

W. Grochala et al. / Solid State Communications 130 (2004) 137–142 influence of the high-frequency phonons of BeF2 on the electronic structure of the surface slab of AgF2. P.A. Anderson et al., J. Phys. Chem. B 101 (1997) 9892 At high frequencies (3.2 GHz), the effects of powder intergrain contacts are negligible, and the microwave method thereby probes the intragrain microwave conductivity. Interestingly, we could also detect a sharp transition at about 210 K (possibly crystalline phase change for BeF2). R.R. da Silva et al., Phys. Rev. Lett. 87 (2001) 147001-1– 147001-4. We have taken up extensive efforts to further characterize our samples. (a) Wishing to determine the crystal structure of the low-abundant components of our samples, we have obtained the electron diffraction pattern for tiny crystals chosen from the reaction products, using the ultra-high resolution STM at the University of Cambridge. However, the Be – Ag – F samples, and the higher fluorides of silver (and at certain experiment conditions also AgIF) show remarkable electronacceptor properties, and they decompose in the electron beam (even at low currents) yielding metallic Ag and sometimes AgF (as shown by the crystal morphology, X-ray composition analysis, and electron diffraction patters). Recollect that AgII is powerful oxidizer, and its electron affinity is significant. In conclusion, the crystal structure of the superconducting impurity is yet unknown. It cannot be excluded that the unidentified phase seen in the diffractograms contains silicon (see the results of the composition analysis via the ICP MASS spectroscopy). (b) Our one-year lasting efforts to optimize the reaction conditions (F2/N2, maximum temperature, time of firing and of cooling, etc.) and to increase the yield of the unidentified phases, in order to better characterize them, have been unsuccessful. We were thus forced to remove the unreacted AgF2 via slow (and mainly photochemical) decomposition of chosen samples in the Ar-filled glovebox (see sample d), but we experienced only limited success. Importantly, when the black Be– Ag– F phases are allowed to decompose in the glass capillaries (to accelerate the decomposition via strong affinity of Si to F), the grayishgreen substance results. On the other hand, when the brown mixture of AgF2 and BeF2 is subject to similar procedure, the yellowish product is formed, containing of AgF and its derivatives (fluoroberrylates and fluorosilicates). This shows that the reaction product is not only a simple mixture of AgF2, AgF and BeF2. (c) We have tried to get insight into the composition of our samples using Raman microscopy. This technique could be particularly efficient in the surface analysis of our samples, via the component-selective resonance effect (recollect the samples are black), and via the possibility of Raman signal collection via optical microscope from the separate crystallites of the (1 mm)2 surface area. However, the resonance Raman spectra (recorded at the Jobin Yvonne T64000 with the 647.1 nm Krypton laser excitation, from the

sample sealed in Ar atmosphere in a teflon container with the quartz optical window), collected from various crystallites, have shown the presence of the bands typical to AgF2, AgF and BeF2, without additional features. (d) We have also tried to determine the composition of our samples via the solid state MAS 19F NMR spectroscopy (using a Chemagnetics CMXInfinity 300 spectrometer, equipped with the 19F probe using chemically resistant ZrO2 rotor at rotation frequency up to 16 kHz, measurement was at room temperature using various echo delays in the pulse sequence, and referenced to solid C6F6 as an external standard). The spectra show presence of a signal at 2313.2 ppm (which may be assigned to AgF, present in the samples due to thermal decomposition of AgF2; this occurs during synthesis and partially due to fast sample rotation in the NMR experiment), a triplet of overlapping signals at about 2191.0, 2201.9 ppm and a shoulder at ca. 2206 ppm, and a complex feature between 2119 and 2152 ppm (this feature may be assigned to the products of the reaction of our samples with the Si-containing agat mortar as confirmed by additional experiments with the samples kept in glass containers; recollect that samples were shortly ground before inserting in the probe). Interestingly, the relaxation times of the signals at 2 191, 2 202 and 2 206 ppm are different for each component, and they are all significantly longer than for the 2192.0 ppm signal of crystalline BeF2. Most probably, these signals originate from fluorine atoms bound to Be, each F in different coordination environment, and some of these possibly belonging to separate crystal phases (a-form, bform, and glassy BeF2, or of its adducts with AgF and/or AgF2). Surprisingly enough, the broad progression centred at þ 150 ppm, exhibiting slight thermal shift (þ 149.6 to þ 152.2 ppm at 12 to 13.5 kHz rotation frequency) typical to paramagnetic compounds, and characteristic of AgF2 (undoubtedly present at large quantities in our samples) is not observed in the spectra of the Be–Ag– F phases. In conclusion, the NMR spectra testify that the composition of our samples is quite complex, and do not allow to unequivocally determine the origin of three important signals about 2200 ppm. [20] The attempts have recently failed to synthesize stoichiometric 1:1 AgF2:BeF2 compound via the methathetic reaction between K2BeF4 and Ag(SbF6)2 solved in anhydrous HF at 230 8C, the reaction product being the mixture of AgF2 and BeF2: Z. Mazej, B. Zˇemva, personal communication. The efforts to obtain AgBeF4 via the low-temperature fluorination of AgI2BeF4, and to remove ferromagnetic AgF2 from the reaction mixture, are now continued in this laboratory. We are also trying to bring one more independent coordinate to the phase diagram, by applying ultra-high pressures on the fluorides of silver (II): V. Struzhkin, Carnegie Institution of Washington, personal communication.