Micro- and nano-size hydrogarnet clusters and proton ordering in calcium silicate garnet: Part I. The quest to understand the nature of “water” in garnet continues

Garnet crystal chemistry, general OH^--stretching mode behavior, and stability of the hydrogarnet substitution

Vibrational spectroscopy, here IR and Raman, measures local vibrations of atoms and atomic groups in molecules and crystals (Nakamoto 2009). The IR and Raman spectra of different silicate garnet species and intermediate composition solid solutions in the energy region of the lattice vibrations (1200 to 100 cm^–1) have been measured and analyzed based on their crystal chemistry (e.g., Moore et al. 1971; Hofmeister and Chopelas 1991; McAloon and Hofmeister 1993; Kolesov and Geiger 1998; Geiger 1998). The crystal-structure properties of various garnets have been investigated and described many times and in great detail, as made possible by a multitude of diffraction studies (e.g., Novak and Gibbs 1971; Armbruster and Geiger 1993; Geiger and Armbruster 1997).

Consider the local crystal chemistry around the crystallographic Wyckoff 24d position, space group Ia3δ, and the cation-oxygen bonding arrangement. The (Z) cation at 24d, site symmetry –4, is Si⁴⁺ for most natural garnets. In garnet containing a hydrogarnet component, a Si⁴⁺ atom is absent and proxied locally by four H⁺ (or D⁺) atoms, as shown by several neutron diffraction investigations (Cohen-Addad et al. 1967; Foreman 1968; Lager et al. 1987; Lager and von Dreele 1996). The protons bisect and lie slightly above the four surfaces of a tetrahedron that is defined by four oxygen atoms to which they are bonded with an OH^- bond length of 0.906(2)/0.925(12) Å (Lager and von Dreele 1996). The crystallographic positions of the four H⁺ cations result from the atomic forces acting in the garnet and the necessity to maintain local charge balance in the vicinity of the 24d site. Crystallographically, there is just a single O atom located at a general position (x,y,z) in garnet. Thus, there are just single OH^- groups that vibrate and there are no (H₄O₄)^4–-type vibrations. There are only individual OH^--stretching vibrations in the energy region around 3600 ± 100 cm^–1 (Kolesov and Geiger 2005; Orlando et al. 2006). There is little reason to consider (H₄O₄)^4–-group-like vibrations (i.e., Harmon et al. 1982), because they should not exist. This is a good reason to express the general formula of hydrogarnets as {X₃}[Y₂](H₁₂)O₁₂. Furthermore, it has been argued that there is very weak, if any, hydrogen bonding present (cf. Harmon et al. 1982; Orlando et al. 2006; Geiger and Rossman 2018). Thus, this potential complication to OH^- vibrational behavior can probably be ignored, based on the existing results.

The relevant local atomic configuration is shown in Figure 2. The H⁺ atom is bonded to a single oxygen atom that is also chemically bonded to a single Y-cation (Wyckoff 16a position) and two X-cations (Wyckoff 24c position) with different X-O chemical bond lengths. Because of this arrangement, there must be OH^--mode coupling and/or mixing with Y- and X-cation-related vibrations. For the common silicate garnets, X can be Ca, Mg, Fe²⁺, and Mn²⁺ and Y can be Al³⁺, Fe³⁺, and Cr³⁺ and they can exchange with one another to various degrees at the X- and Y-sites, respectively, giving a range of compositionally complex substitutional solid solutions (Geiger 2008). There are considerable differences in ionic radii, masses, and electronic states among these various cations and, therefore, the associated chemical bonds must be different in nature.

Figure 2

Local atomic environment involving a single OH^- dipole in the case of calcium hydrogarnet (no Si cation). The oxygen anion (red) is bonded to one Y (Al or Fe³⁺) cation (blue), two X (Ca) cations (yellow) with bond lengths of Ca1-O = 2.462(3) Å and Ca(2)-O = 2.520(3) Å in hydrogrossular (Lager et al. 2002) and a H atom (pink). (Color online.)

Adopting a harmonic oscillator to describe the vibration of an atom pair (A and B), the frequency is given by:

where k is the force constant (i.e., in simple terms the bond strength) and m_AB is the reduced mass of the system [m_AB = (m_Am_B)/(m_A + m_B)]). It is obvious that an O-H stretching mode will have a different vibrational frequency depending on the precise local atomic configuration and the nature of the various chemical bonds involving the common oxygen atom (Fig. 2) and also possibly beyond. Geiger and Rossman (2018) discussed the case for various OH^--bearing “end-member” garnet species. They all show a single OH^- stretching mode at RT whose energy is a function of the atomic masses and strengths of the different chemical bonds of the “extended vibrational system.” The greater the mass, the lower will be the wavenumber of the OH^- stretching mode, ignoring other factors.

This reasoning explains, for example, the lower energy of the OH^- stretching mode in the IR spectra (Albrecht et al. 2019) of the synthetic hydrogarnets Ca3Cr23+H12O12,  Sr3Cr23+H12O12and Ca₃Rh²₃⁺H₁₂O₁₂ (i.e., 3616, 3625, and 3592 cm^–1, respectively) compared to that for Ca₃Al₂H₁₂O₁₂ (i.e., 3662 cm^–1). The OH^- stretching mode, observed in RT spectra, is generally broad in OH^--bearing end-member garnets because of the large vibrational amplitudes of the H⁺ atom (Lager et al. 1987). Cooling the crystal dampens significantly the amplitudes of vibration and the single OH^- band at RT narrows greatly and ultimately at about 80 K or slightly below two modes are observable in IR spectra (Kolesov and Geiger 2005; Geiger and Rossman 2018).

Of course, pure, end-member composition silicate garnets are rare, if they even exist in nature. There is almost always some degree of cation mixing at the X- and Y-sites. This means there can be various local cation configurations around a given oxygen atom in a solid-solution crystal (Fig. 2). Their concentration depends on the garnet bulk composition. The precise local crystal-chemical situation (i.e., atomic order-disorder and structural relaxation) is complex, and it is not fully understood at this time. Suffice it to say, here, that the intensities of the different OH^- stretching modes will depend on the relative amounts of the various local cation configurations. Different vibrational possibilities could exist in a solid-solution garnet. One is that each local atomic configuration will result in a representative OH^- mode with its characteristic energy, as in the respective “end-member” garnets grossular, andradite or katoite, for example. Another possible behavior is that a distribution of relatively closely spaced OH^- modes, each with not too different energies, occurs. This latter situation can give rise to OH^--band broadening and varying the temperature of the IR measurement will not significantly affect OH^- band widths. This effect is called chemical-mode broadening in comparison to pure vibrational-mode broadening. In terms of many natural OH^--bearing garnets both can occur.

Finally, it must be stressed, as a key basis for our analysis, that the substitution (H₄O₄)^4– ↔ (SiO₄)^4– or (H₄)⁴⁺ ↔ (Si)⁴⁺ is energetically favorable for grossular-katoite crystals, as shown by synthesis experiments (Flint et al. 1941; Carlson 1956; Kobayashi and Shoji 1983) and also to a large, but not complete, degree in Ca3Fe23+(SiO4)3-Ca3Fe23+(H4O4)3garnets (Flint et al. 1941; Dilnesa et al. 2014). No other common silicate binary garnet system shows such a complete exchange. Phase equilibrium studies on Ca₃Al₂(SiO₄)₃-Ca₃Al₂(H₄O₄)₃ solid solutions at high P and T under hydrothermal conditions indicate that thermal stability increases with decreasing H₂O in the garnet (Yoder 1950; Pistorius and Kennedy 1960). Finally, empirical pair potential calculations show that the reaction grossular + water → katoite is thermodynamically favored (Wright et al. 1994).

A simple question is then: why are the IR spectra of nominally anhydrous grossular and other calcium silicate garnets so variable and complex compared to those of other natural silicate garnets (e.g., pyrope, almandine, spessartine) in terms of their OH^- incorporation?

IR spectra, OH^- stretching modes and their assignment, hydrogarnet domains and clusters

Katoite and water-rich katoite-grossular solid solutions. We start our analysis with synthetic end-member katoite, Ca₃Al₂H₁₂O₁₂, with a well-known crystal structure (Cohen-Addad et al. 1967; Foreman 1968; Lager et al. 1987) and IR and Raman spectrum (Kolesov and Geiger 2005). At RT, its IR spectrum shows a broad, single asymmetric OH^- stretching mode at 3662 cm^–1. It can be fit with two components with energies of about 3681 and 3666 cm^-l at about 80 K. At even lower temperatures, both the IR and Raman spectra become complex in the OH^- stretching region with the appearance of many new, weak OH^- modes, but this aspect is not considered further here. The OH^- “doublet” at roughly 80 K is asymmetric, whereby the higher energy OH^- band is considerably more intense than the lower energy one. Geiger and Rossman (2018) argued that this spectroscopic behavior of OH^--mode splitting characterizes the hydrogarnet substitution in most “end-member” silicate garnets.

Consider now the IR single-crystal spectra of intermediate Ca₃Al₂Si₃O₁₂-Ca₃Al₂H₁₂O₁₂, both natural and synthetic, garnets (Fig. 3, replotted from Rossman and Aines 1991) containing major H₂O contents. From these five spectra it is immediately apparent that the band at about 3660 cm^–1 does not vary greatly in wavenumber, but it decreases in intensity with an increasing number of Si atoms in the formula unit (or grossular component). We conclude, therefore, that this mode must represent the presence of katoite-like domains in the solid-solution composition garnets. [A strong OH^- band at 3660 cm^–1, as well as a couple of others with high wavenumber, was observed in the IR powder in KBr spectra of a series of synthetic Ca₃Al₂Si₃O₁₂-Ca₃Al₂H₁₂O₁₂ garnets (Kobayashi and Shoji 1983). However, the quality of the published spectra does not permit a precise analysis.] The local crystal-chemical situation is shown in Figure 4a.

Figure 3

IR single-crystal spectra in the energy range of the OH^- stretching vibrations for five natural and synthetic garnets along the join Ca₃Al₂(H₄O₄)₃-Ca₃Al₂(SiO₄)₃ as replotted from Rossman and Aines (1991). The sample labels (Table 1) and the number of Si atoms per formula unit (pfu) are given. (Color online.)

Figure 4

Various possible hydrogarnet-like groups or clusters in grossular or andradite. H atoms are pink, SiO₄ groups are red and YO6 groups with Y = Al³⁺ or Fe³⁺ are blue. Hydrogarnet (H₄O₄) “tetrahedra” are shown in green. The oxygen atoms located at the corners of the polyhedra and the X-cation are not shown for the sake of clarity. (a) Crystal-chemical environment for end-member katoite or a finite-size katoite-like cluster, as given by the OH^- stretching mode at about 3662 cm^–1. (b) A single (H₄O₄)^4– group and the rest SiO₄ groups, as given by the OH^- stretching mode at about 3600 cm^–1. (c) Cluster arrangement involving two (H₄O₄)^4– groups and the rest SiO₄ groups, as given by the mode at about 3611 cm^–1. (d) Cluster arrangement involving three (H₄O₄)^4– groups and the rest SiO₄ groups, as given by the mode at about 3622 cm^–1. (e) Cluster arrangement involving four (H₄O₄)^4– groups and the rest SiO₄ groups, as given by the mode at about 3633 cm^–1. (f) Cluster arrangement involving five (H₄O₄)^4– groups and a single SiO₄ group, as given by the mode at about 3644 cm^–1 and (g) Hypothetical cluster arrangement involving six (H₄O₄)^4– groups, as possibly given by the mode at about 3657 cm^–1. (Color online.)

It can also be observed in a couple of IR spectra (GRR 1329 and GRR 1358) that there is a strong OH^- band between 3595 and 3660 cm^–1 that increases in intensity with increasing grossular component in the solid solution (Fig. 3). Rossman and Aines (1991) wrote in their IR study of grossular garnets, “We assign the band at 3598 cm^–1, which is dominant in the silica-rich hydrogrossular samples, to (O₄H₄)^4– groups, adjacent to SiO₄ groups.” We accept this proposal. The local crystal-chemical situation is illustrated in Figure 4b and it shows a single, isolated (H₄O₄)^4– group imbedded in an anhydrous grossular “matrix.” The wavenumber of the associated OH^- mode is about 60 cm^–1 less than in end-member katoite and could be a function of two effects. The first is related to bond strengths. In end-member grossular the two different Ca-O bond lengths (Fig. 2) are Ca(1)-O(4) = 2.322(1) Å and Ca(2)-O(4) = 2.487(1) Å (Geiger and Armbruster 1997). They are shorter, and thus stronger, than in end-member katoite, where Ca(1)-O(4) = 2.465(2) Å and Ca(2)-O(4) = 2.511(1) Å (Lager et al. 1987). It can be expected that local Ca-O bonds in water-bearing grossular vary in length depending on the precise structural state within a solid solution. Stronger Ca-O bonds translate to weaker O-H bonds (Geiger and Rossman 2018). Second, it could be expected that there is extended mode coupling and/or mixing beyond the fragment shown in Figure 2 that affects O-H bonding. The presence of heavier Si⁴⁺ cations compared to H⁺ atoms in a garnet solid solution could act to lower the OH^- stretching vibration energy.

The IR and Raman spectra indicate, therefore, two “end-member type” (H₄O₄)^4– substitution mechanisms. Obvious questions arise however. One is apparent in the spectrum of garnet GRR 1058 (Fig. 3). The broad OH^- mode envelope has structure and shows a peak maximum at roughly 3620 cm^–1. How is it to be assigned? More generally, published IR spectra of water-poor grossulars show a varying number of OH^- bands located between 3660 and 3600 cm^-l. How are they to be assigned?

“End-member” or nearly end-member grossular. To answer these questions, we consider the IR spectra of several “end-member” or closely end-member composition grossulars, both synthetic and natural. The spectra of a hydrothermally synthesized grossular (Geiger and Armbruster 1997, grown from the oxides CaO from CaCO₃, Al₂O₃, and SiO₂) and a nearly end-member composition natural garnet (GRR 53, Table 1) are shown together in Figure 5a. The two spectra are remarkably similar in appearance and give the first “spectroscopic match” between a synthetic and a natural crystal [We interject, here, that the IR powder spectra on synthetic grossular recorded by Hsu (1980) show weak absorption features at about 3620 and 3660 cm^–1 that could be interpreted as indicating structural OH^-, as well as a broad and intense absorption band centered at 3420 cm^–1 that can be assigned to liquid water. It is our experience, though, that IR powder spectra are normally insufficient for characterizing precisely small amounts of OH^- in minerals.] The spectra of both the synthetic crystal and GRR 53 show three intense OH^--bands at 3666, 3633, and 3623 cm^–1. Consider, furthermore, the IR spectra of the synthetic and natural grossulars shown in Figure 6. They show different OH^- spectral features from those garnets in Figure 5a and have intense modes at 3623 (±1) and 3603 cm^–1. The two garnets give a second spectral “match” between a synthetic and natural end-member composition crystal.

Figure 5

(a) IR single-crystal spectra in the wavenumber range of the OH^- stretching vibrations at RT of synthetic “end-member” grossular normalized to 1 mm thickness (Geiger and Armbruster 1997) and natural nearly end-member grossular GRR 53 (Table 1). (b) Fitted spectrum of synthetic grossular. (Color online.)

Figure 6

IR single-crystal spectra in the energy range of the OH^-stretching vibrations at RT of natural grossular GRR 1285 (intensity 3×) and synthetic grossular Gr83 (normalized to 1 mm thickness) of Withers et al. (1998). Samples described in Table 1. (Color online.)

The spectrum of the synthetic grossular in Figure 5a showsfine structure (e.g., shoulders on the main peaks) indicating the presence of additional OH^- modes and, therefore, it was curve fit to describe them (Fig. 5b). Additional bands at 3688, 3660, 3645, 3613, 3604, 3578, and 3561 cm^–1 result. This gives a total of 10 possible OH^- modes for this synthetic grossular. We analyze and assign them. First, however, we note that the weak modes at 3561, 3579, and 3688 cm^–1 may be experimental artifacts arising from the difficultly in collecting a spectrum with a smooth baseline on the tiny synthetic crystal. On the other hand, they could also be real and possible assignments are discussed below. Summarizing for the present discussion, this gives seven OH^- modes at about 3666, 3660, 3645, 3633, 3623, 3613, and 3604 cm^–1. There appears to be an approximate regularity in the wavenumber spacing between the modes.

It turns out that these OH^– modes, in various combinations and with differing intensities, are observed in most, if not all, published IR spectra of natural grossular crystals. To be sure, the large number of spectra in Rossman and Aines (1991) shows intense OH^- bands at about 3612, 3622, 3633, and 3641 cm^–1 depending on the specific crystal in question. Indeed, the presence or absence and the intensity of these modes (i.e., the spectral pattern) were used to define or to characterize their different classes of OH^--bearing grossulars. Phichaikamjornwut et al. (2012) in their IR study of different natural grossular-andradite garnets also measured intense modes at about 3602, 3612, 3631, and 3641 cm^–1.

All the IR results, analyzed together, indicate a spectral systematics defined by OH^--bands with similar wavenumbers between 3595 and 3670 cm^–1. Not every band is present in the spectrum of every garnet, but systematic OH^--mode behavior is observable. Thus, we propose that OH^- stretching modes at about 3599, 3612, 3622, 3633, 3641, and 3657 cm^–1 could arise from the presence of various, local hydrogrossular-like clusters occurring in an anhydrous grossular “matrix.” Table 2 lists the different modes. (Note: Based on our analysis of many spectra, the mode at about 3657 cm^–1 appears to be the most problematic in the sense that it often does not occur and/or that it is weak in intensity in the spectra of many garnets.) We have no good explanation for this, except to state that this mode is related to a six-membered (H₄O₄)^- cluster—Fig. 4g—that is structurally, and thus probably energetically, similar to a katoite-like cluster—Fig. 4a—having an OH^- mode around 3660 cm^–1. Thus, the latter type of cluster may occur more preferentially). From this analysis, seven types of local (H₄O₄)^4– clusters (and/or domain in the case of water-rich hydrogarnets) can be constructed by having different numbers of (H₄O₄)^4– groups in them and they can be assigned to the different energy OH^- modes. Figures 4a–4g shows the respective atomic configurations for the different cluster types. According to our model analysis, the OH^- mode wavenumber increases with the number of (H₄O₄)^4– groups in a given cluster.

Summarizing our analysis, the observed OH^--stretching modes between 3595 and 3670 cm^–1 in the IR spectra for many different OH^--bearing grossulars and katoites can be interpreted and assigned in a completely new context. Figure 7 shows IR spectra, including one water-rich grossular and six water-poor samples having an intense OH^- mode at one or more of the characteristic wavenumbers. It must be noted, though, that other OH^- bands can be also observed above and below this wavenumber range. This aspect is now considered.

Figure 7

Stacked plot of different spectra of natural grossulars (Table 1) and the synthetic hydrogrossular GRR 1059 showing hydrogrossular-like OH^- modes at about 3599, 3612, 3622, 3633, 3643, 3657, and 3662 cm^–1. They are related to different microscopic- to nano-size hydrogarnet clusters (see text). The mode at 3688 cm^–1 may be due to tiny “serpentine mineral” inclusions (see text). (Color online.)

Other possible OH^-substitution mechanisms in “end-member” grossular and hydrous inclusions. The IR spectra of water-poor natural grossular can often show OH^- bands that are not listed above and lie at energies above 3670 cm^–1. Their assignments are not known and remain to be made. Consider, first, the high-energy OH^- mode at 3684–3688 cm^–1 at RT that has often been observed (Fig. 8 and see spectra in Rossman and Aines 1991; Kurka et al. 2005; Dachs et al. 2012; Phichaikamjornwut et al. 2012). This mode has stronger O-H bonding compared to those OH^- bands between 3595 and 3662 cm^–1 and it cannot be assigned based on the model analysis given above. One speculative possibility is that it arises from the substitution {2H⁺ ↔ Ca²⁺}. A H⁺ atom, as part of an OH^- dipole directed toward the interior of an X-cation-free (site 24d) oxygen coordinated dodecahedron, should have relatively little interaction with other atoms in the garnet structure. It could be expected to be the “freest OH^- dipole” of any in grossular and thereby have the strongest O-H bond.

Figure 8

IR spectrum of grossular GRR 1386 (Table 1) and its curve fit. Note the OH^- modes at 3688 and 3674 cm^–1. They are possibly related to the presence of tiny inclusion phases of hydrous layer silicates and probably a “serpentine mineral” (see text). (Color online.)

Alternatively, this high wavenumber mode might be related to the presence of tiny inclusions of a foreign OH^--bearing phase. High wavenumber OH^- bands (greater than 3670 cm^–1) in the spectra of andradite and pyrope have been argued to be due to the presence of inclusions of minute layer silicates such as talc and phlogopite (Geiger et al. 2018; Geiger and Rossman in review). What about grossular? The spectra of serpentine minerals (i.e., chrysotile, lizardite, antigorite) are characterized by OH^- bands with high energies. Schroeder (2002) gives 3686 cm^–1 for the strong OH^- band for the spectrum of lizardite and Mellini et al. (2002) measured 3684 cm^–1 for lizardite from Elba Island. Band energies depend on the precise compositions of the phases (Heller-Kallai et al. 1975). The absorption features can be broad in nature and show fine structure and are variable depending on the precise serpentine mineral. A band at about 3674–3677 cm^–1 can also be observed in some spectra of grossular (Fig. 8). Antigorite can show a strong OH^- band (asymmetric in shape) with a peak maxima of 3674–3678 cm^–1 (Heller-Kallai et al. 1975; Mellini et al. 2002). Talc, as well, shows an intense OH^- band at 3677 cm^–1 (Schroeder 2002). It is important to note that nearly end-member grossular crystals can often be found in rodingites associated with serpentinites (see the discussion in Part II on grossulars from Asbestos, Canada). Therefore, in light of the evidence, the high energy bands at about 3684–3688 and 3674–3677 cm^–1 in the spectra of grossular are best assigned, at least at this time, to the presence of tiny inclusions of hydrous Mg-rich layer silicates.

In addition to these high wavenumber OH^- modes, other bands with energies less than about 3595 cm^–1 have been observed in the spectra of many natural grossulars (Rossman and Aines 1991; Kurka et al. 2005; Phichaikamjornwut et al. 2012; Dachs et al. 2012; this work). We address them below, but first we consider the IR spectra of synthetic Ca₃Fe²₃⁺H₁₂O₁₂-bearing hydrogarnet solid solutions and natural end-member andradite.

Water-rich Ca₃Fe²³₊Si₃O₁₂-Ca₃Fe₂³₊H₁₂O₁₂-Ca₃Al₂Si₃O₁₂- Ca₃Al₂H₁₂O₁₂ solid solutions. End-member Ca₃Fe²₃⁺H₁₂O₁₂ has yet to be synthesized and, therefore, its OH^- mode energy cannot be measured directly, as was done with Ca₃Al₂H₁₂O₁₂. However, water-rich hydrogarnets in the four-component system Ca₃Fe23+Si₃O₁₂-Ca₃Fe23+H₁₂O₁₂-Ca₃Al₂Si₃O₁₂-Ca₃Al₂H₁₂O₁₂ have been synthesized (Dilnesa et al. 2014). Thus, we made IR ATR powder measurements on five garnets of nominal composition Ca₃Al₂(Si_1.0,H_8.0)O₁₂, Ca₃[Al_1.4,Fe0.63+₂(Si_1.0,H_8.0) O₁₂, Ca₃[Al_1.0,Fe1.03+]₂(Si_1.0,H_8.0)O₁₂, Ca₃[Al_0.6,Fe1.43+]₂(Si_1.0,H_8.0)O₁₂, and Ca₃[Fe2.03+](Si_0.75,H_9.0)O₁₂ (Table 1). The spectra are shown in Figure 9. Two main absorption features occur at 3611 and 3658 cm^–1. We think these two energies describe the presence of Ca₃Fe23+H₁₂O₁₂- and Ca₃Al₂H₁₂O₁₂-like domains in these garnets, respectively. The latter mode energy agrees with the spectroscopic results discussed above. Spectral variations as a function of garnet composition are apparent. The OH^- mode (expressed as a band or a shoulder) at about 3658 cm^–1 decreases in relative intensity with decreasing Ca₃Al₂H₁₂O₁₂ component in the garnet, as expected.

Figure 9

Stacked plot of IR ATR powder spectra of five synthetic hydrogarnets as synthesized and described by Dilnesa et al. (2014). (Color online.)

Concurrently, the OH^- mode at 3611 cm^–1 increases in relative intensity with increasing Ca3Fe23+H12O12component in the garnet (Fig. 9)—there also appears to be a weaker OH^- mode at about 3534 cm^–1 in the spectra of these garnets, but it should have another origin and is not considered in this work.

“End-member” or nearly end-member andradite. The RT IR single-crystal spectra of several “end-member” andradites have been recorded (Amthauer and Rossman 1998; Adamo et al. 2011; Zhang et al. 2015; Geiger et al. 2018). The low-temperature IR spectra of several natural crystals and one synthetic were measured down to 80 K by Geiger and Rossman (2018). The IR spectra of most, but not all, “end-member” andradites are simpler compared to the situation for grossular in terms of the number of observed OH^- modes. Many spectra of andradite contain just one or two OH^- modes at RT. Most natural and hydrothermally synthesized crystals show the most intense OH^- band, which is broad and asymmetric, at or near 3563 cm^–1. It splits into two bands at 3575 and 3557 cm^–1 at 80 K (Geiger and Rossman 2018). Or it possibly splits into three modes, including the one at 3543 cm^-l for andradite GRR 1263 (Fig. 10 and Table 1). At any rate, we consider these bands to represent the presence of single, isolated (H₄O₄)^4– groups (Fig. 4b) in andradite (Table 2).

Figure 10

IR single-crystal spectra in the energy range of the OH^- stretching vibrations at 333 and 80 K of a natural “end-member” andradite GRR 1263 (Table 1).

The lower OH^- stretching energy compared to the analogous situation for single, isolated (H₄O₄)^4– groups in grossular is largely due to the fact that the mass of Fe³⁺ (55.85 amu) is twice that of Al³⁺ (26.98 amu). Equation 1 describes the general situation. The mass effect is clearly demonstrated by the energies of the IR-active translational (T) vibrations of both cations in garnet. The wavenumber range for T(Fe³⁺) modes lies between 130 and 360 cm^–1 for andradite and for T(Al³⁺) modes in grossular between about 240 and 470 cm^–1 (McAloon and Hofmeister 1993).

Other higher wavenumber OH^- bands can be observed in the spectra of some water-poor andradites such as that shown in Figure 9 as given by the mode at 3605 cm^–1 at 333 K. Moreover, an OH^--band with a wavenumber of 3612 (±1) cm^–1 can be observed as part of a very broad OH^- absorption feature in the RT spectra of andradite crystals that contain a significant amount of the Ca3Fe23+H12O12component (see Fig. 2 in Amthauer and Rossman 1998). We propose that the OH^- band at roughly 3612 cm^–1 at RT in andradite-rich garnets (Table 2) could be related to the presence of finite-sized Ca3Fe23+O12H12-like clusters (Fig. 4a).

Finally, we propose, because andradite and grossular are so similar structurally that, in an analogous manner to the situation discussed above for hydrogrossular-like clusters in water-poor grossular, a series of different hydroandradite-like clusters may be present in some andradite crystals. They should be characterized by having OH^- stretching modes with energies between 3563 and 3606–3612 cm^–1. Consider, for example, the RT IR spectrum of “end-member” andradite 4282 (Fig. 11) that contains only 0.02 wt% Al₂O₃ (Geiger et al. 2018, see also a similar spectrum of andradite in Zhang et al. 2015). It is “untypical.” Starting with the intense mode at 3561 cm^–1 and moving to higher wavenumbers, one observes two OH^- modes at 3582 and 3595 cm^–1. We think they could represent the presence of two different hydroandradite-like clusters, whose exact (H₄O₄)^4–-group configuration (Figs. 4c–4f) cannot be presently determined. Finally, in terms of this andradite garnet, the three modes at 3610, 3621, and 3633 cm^–1 can be assigned to three hydrogrossular-like-clusters (Table 2).

Figure 11

IR single-crystal spectrum in the energy range of the OH^- stretching vibrations at RT of natural “end-member” andradite 4282 (Table 1) with fitted bands. (Color online.)

We close this section by noting that many andradites and grossulars are not of “end-member” composition. There is often some degree of solid solution between the two and also to a lesser degree with other common garnet components (e.g., Fe³₂⁺Al₂Si₃O₁₂, Mn₃²₊Al₂Si₃O₁₂, Mg₃Al₂Si₃O₁₂). To address the issue of OH^- incorporation in calcium silicate garnets more fully, we need to consider the IR spectra of water-poor intermediate composition grossular-andradite crystals.

Water-poor intermediate composition grossular-andradite garnets. Many, if not most, Ca-rich silicate garnets in Earth’s crust are grossular-andradite solid solutions that crystallize during metamorphism. There can be complete Al ↔ Fe³⁺ exchange at [Y] across the join and the thermodynamic mixing behavior has been studied (Dachs and Geiger 2019). The cation mixing behavior can be complex because varying partial degrees in the long-range order can occur, decreasing the space group symmetry below cubic (Takéuchi et al. 1982). At any rate, water-poor grossular-andradite garnets offer an additional system for investigation, because their temperatures of crystallization are well above those where many water-rich crystals are stable. What do their IR spectra show?

Various IR spectra have been published (i.e., Rossman and Aines 1991; Phichaikamjornwut et al. 2012) and several garnets have been investigated in this work. Conceivably, it could be expected that spectral complexity could increase compared to the situation for end-member garnet species. The crystal chemistry of solid-solution garnets is more complex because of Al³⁺-Fe³⁺ mixing at [Y]. Local OH^- vibrational behavior could be affected, as discussed above (Fig. 2). Assuming statistically random Al³⁺-Fe³⁺ mixing, there can be five possible local Y-cation configurations around a single, common (H₄O₄)^4– group. They are: Al-Al-Al-Al, Fe³⁺-Al-Al-Al, Fe³⁺-Fe³⁺-Al-Al, Fe³⁺-Fe³⁺-Fe³⁺-Al, and Fe³⁺-Fe³⁺-Fe³⁺-Fe³⁺. For the case of strict random mixing, the probability for each of the five configurations is a function of the bulk composition of the crystal. The situation becomes more complex, if additional (H₄O₄)^4– groups are brought into the analysis. Suffice it to note that different energy OH^--stretching modes could potentially occur.

What is observed in the available IR spectra? To begin, many, water-poor Ca₃(Al_x,Fe³_{1+ –x})₂Si₃O₁₂ garnets show an OH^- mode at 3563 cm^–1 at RT. Following the analysis for the case of “end-member” andradite, it can be assigned to a single, isolated (H₄O₄)^4– group surrounded by Fe³⁺ cations. Clearly, the spectra of a number of Fe³⁺-bearing grossulars in Rossman and Aines (1991) show the presence of an OH^- band at about 3563 cm^–1 (their Figs. 4, 5, 6, 8, 9, and 10 and samples 1125, 936, 1042, 1051, 1430, 938, 1327, 1357, and 1411, for example). The same is true for the IR spectra of several grossular-andradite garnets in Phichaikamjornwut et al. (2012). In addition, OH^- modes at 3581 and 3594 cm^–1 are also observed in the spectra of some of these grossular-andradite garnets with the assignments given above and in Table 2.

Take two specific examples, the first being garnet 6741 (Table 1) of approximate composition Gro₇₃And₂₃. Its IR spectrum was recorded at 273 and 80 K (Fig. 12a). At 273 K three distinct and relatively strong OH^- bands are observable (i.e., 3688, 3666, and 3629 cm^–1) and at 80 K four bands (i.e., 3694, 3672, 3634, and 3615 cm^–1). Weak bands (i.e., 3565 cm^–1 at 273 K and 3562 cm^–1 at 80 K) and shoulders can also be discerned. In the case of this garnet, it is important to bear in mind that, although it has a considerable 25 mol% andradite component, nearly all of the OH^- is contained in various hydrogrossular-like clusters. The simplest interpretation is that the energetics associated with the formation of hydrogrossular-like clusters are more favorable than for hydroandradite-like clusters. This proposal is consistent with synthesis experiment results (Flint et al. 1941; Dilnesa et al. 2014) for garnets in the system Ca₃Al₂(SiO₄)₃-Ca₃Al₂(H₄O₄)₃- Ca₃Fe²₃⁺(SiO₄)₃-Ca₃Fe₂³₊(H₄O₄)₃ as shown in Figure 1. Only the binary Ca₃Al₂(SiO₄)₃-Ca₃Al₂(H₄O₄)₃ shows complete solid solution from end-member to end-member.

Figure 12

IR single-crystal spectra in the energy range of the OH^-stretching vibrations for: (a) sample 6741 of composition grossular73-andradite23 at 273 and 80 K and (b) sample KPK 56-12-9 of composition andradite73-grossular26 at RT. Sample description in Table 1.

Finally, consider the RT spectrum of an andradite garnet of composition And₇₆Gr₂₃ (sample KPK 56-12-9 of Phichaikamjornwut et al. 2012) shown in Figure 12b. It shows three strong OH^- bands with the two lowest wavenumber ones occurring at 3562 and 3582 cm^–1. They are associated with hydroandradite-like clusters (Table 2). The mode at 3611 cm^–1 is difficult to assign and could represent an end-member Ca₃Fe³₂⁺-H₁₂O₁₂-like or a hydrogrossular-like cluster (Table 2). The various weak bands, observable at higher wavenumbers, probably represent hydrogrossular clusters present in small amounts. At any rate, the bulk of the OH^–-like clusters in this solid-solution garnet is partitioned into hydroandradite-like clusters. Research needs to be done to determine the nature of H₂O partitioning and hydrogarnet-cluster formation in the nominally anhydrous system Ca₃(Al_x,Fe³₁⁺_–x)₂Si₃O₁₂. In other words, the role of garnet composition with regards to the nature of structural “water” needs to be investigated as a function of P_H2O and T.