2.1 Polymorphism of HgI₂ at ambient conditions

Mercuric iodide (HgI₂) is known to crystallize from various organic solvents and also by sublimation into three concomitant polymorphs whose colours are red, orange, and yellow [6]. At ambient conditions, the red phase is thermodynamically stable, while the orange and yellow phases are metastable and transform into the red phase when touched. The aspects and speeds of these transformations are different for the two forms. In the orange crystals, red nuclei develop into domains of irregular shape, which spread and transform a crystal of diameter 0.2 to 0.5 mm within hours or days while preserving the external shape. In contrast, in the yellow form, a sharp red front propagates from the contact point and consumes all of a crystal of dimensions 0.2 to 0.5 mm within a few seconds while the crystal may warp. From a 2-chloroehtanol solution, the yellow crystals are the first, and the red crystals the last to crystallize, as observed by optical microscopy [7]. Faster evaporation favours the formation of yellow crystals. Above 323 K, only yellow crystals are formed, whereas slow crystallization in a closed vessel gives only red, orange or red-orange composite crystals.

2.2 Crystal structures of the polymorphs at ambient conditions

The structures of the polymorphs of HgI₂ at ambient conditions are shown in Fig. 1; Table 1 lists unit cell parameters and space group symmetries. All the structures of the red and the orange phases have in common a nearly ideal cubic closest packed (ccp) substructure of iodine atoms with 1/4 of the tetrahedral interstices occupied by Hg. The HgI₄ tetrahedra are corner-linked and all Hg–I–Hg and I–Hg–I angles are close to tetrahedral. There is only one kind of Hg atom, i.e. all tetrahedra (but not all corners, i.e. iodine atoms) are symmetry-equivalent. It can be shown that six different structures with these properties, all of them tetragonal, can be derived from a ccp-array of anions (at least two such structures exist for the hexagonal hcp array with space group symmetries Pmc2₁ and Pbam; to our knowledge, they have never been observed experimentally). Four of these six structures are realized by HgI₂; one of remaining structures has not (yet) been identified for HgI₂ but is realized by ZnCl₂; a sixth structure has never been observed. The stable red form consists of layers of corner-linked HgI₄ tetrahedra [8,9], stacked by I···I contacts. The metastable orange crystals comprise three different crystal structures, all of which are built from corner-linked Hg₄I₁₀ supertetrahedra [10,11]. Two of them are end members with the maximum degree of order (MDO) of a polytypic layer structure; the third shows a three-dimensional diamantoid-type linkage (D). The two MDO structures are fractal complications of the stable red form of HgI₂ obtained by replacing single tetrahedra by supertetrahedra. The geometries and anisotropic displacement parameters of the layers in MDO1 and MDO2 and the interlayer contacts are nearly identical in the two stackings [10]. The two MDO structures usually coexist and form a disordered crystal. A quantitative analysis of the stacking disorder by fitting a Markov chain model to the intensities of the corresponding rods of diffuse X-ray scattering suggests the presence of nearly equal volumes of the MDO structures with an average domain thickness of about five layers or 30 Å [10]. The third orange structure D (diamantoid) is also composed of corner-linked Hg₄I₁₀ supertetrahedra. However, the supertetrahedra are not linked into layers but into two interpenetrating four-connected three-dimensional networks [11]. This is the fractal complication of the structure of ZnCl₂. Composite crystals containing all three supertetrahedral structures MDO1, MDO2 and D, twinned in three perpendicular directions resulting in a pseudo-cubic edifice, have also been observed [11].

In contrast to the red and orange forms, the structure of the yellow metastable (yellow_M) polymorph at room temperature is molecular with linear I–Hg–I molecules (angle 180.4° not constrained by symmetry) [7,9]. Not surprisingly, the packing of the iodine atoms is very distorted. Closest-packed layers with six-coordinated spheres are stacked roughly midway between ccp and hcp. Therefore, the iodine substructure may be described as related to the cubic volume-centred bcc structure [7]. The coordination of mercury by iodine is very distorted octahedral 2+4, with two close and four considerably more distant neighbours.

2.3 Mechanisms of crystallization and transformation of HgI₂

When crystallizing HgI₂ at ambient conditions, rapid evaporation of the solvent is essential for the formation of the yellow_M polymorph, indicating that the crystallization of this form is governed by kinetics. Raman spectra of an HgI₂ solution in 2-chloroethanol indicate HgI₂ to exist in the form of molecules [7]. Therefore, the observed crystallization sequence can be understood as a kinetic effect: the pre-existing molecules rapidly pack into the yellow_M form, while the formation of HgI₄ tetrahedra demands the breaking of bonds. On the other hand, it remains unclear why the orange structures are kinetically favoured and formed before the stable red structure. In both structures, there are four shortest Hg–Hg distances, but the Hg distribution in the orange forms is globular and somewhat more homogeneous than the planar distribution in the red form. It seems reasonable that the formation of clusters with limited dimensions is kinetically favoured over the formation of extended sheets. The choice between the different non-layer structures does not seem to be dominated by electrostatic interactions; we would expect the ZnCl₂ structure, which is not realized by HgI₂, to have the largest Madelung energy of the six tetrahedral structures mentioned above and thus to be the most ionic one. We also note that ZnI₂ at ambient conditions forms the supertetrahedral D-structure. The transformations from orange to red presumably involve only movements of Hg atoms in an invariant ccp matrix of I. The thermal displacement parameters of Hg in the tetrahedral structures are indeed larger than those of iodine [10,11]. In contrast, the transformation from yellow_M to red implies a complete rearrangement of the atoms.

2.4 Phase diagram of HgI₂

The phase diagram of HgI₂ has been investigated by many authors with different techniques. The somewhat controversial results are reviewed in [12]. A phase diagram compiled from all the various experimental observations is presented in Fig. 2. Five different phases (I, II, III, IV, VI) are reported in the moderate pressure–temperature range of p < 4 GPa, T < 700 K, the boundaries of which have been established by dilatometry (compression-curve method) [13], differential thermal analysis [14], Raman spectroscopy [15,16] and powder X-ray diffraction [17,18].

At 400 K and ambient pressure, the red form discussed above (phase III) undergoes a reconstructive transition to the yellow high-temperature phase II (yellow_HT), which has been assumed to be identical to the yellow_M polymorph obtained under ambient conditions [12]. However, we have observed the symmetry of the yellow_HT phase to be P2₁, a subgroup of the space group Cmc2₁ of the yellow_M structure. The two structures are closely related; both show similarly packed I–Hg–I molecules. The main difference consists in the molecules in yellow_HT being bent and not linear (Fig. 3), with angles of 163(1)° and 157(1)° for the two molecules in the asymmetric unit. Electron diffraction and vibrational spectroscopy cannot rule out the existence of slightly bent molecules in the gas phase (165–180°); ab initio studies usually assume straight molecules [19]. The mechanisms of the III → II and the reverse II → III transitions are different. Upon heating, the transition occurs through surface nucleation while, upon cooling, it shows bulk nucleation [20]. The kinetics of the II–III transition is highly asymmetric. The growth rate of a newly formed crystal has been measured by the increase of a normalized integrated Bragg intensity with time, (ΔI_Bragg/I_Bragg) (s^–1). For III→II, the growth rates of phase II at 402.9 and 404.0 K are 3.6 × 10^–4 s^–1 and 4.2 ^× 10^–4 s^–1, respectively. In contrast, for II→III, the growth rate at 381.0 K is much smaller, 1.8 × 10^–5 s^–1 [7]. The hysteresis of the transition is also asymmetric in the sense that phase II can be undercooled to room temperature, while the red phase can be overheated by only about 10 K above the transition temperature [7]. The II–III phase boundary presents the particularity of a maximum at about 0.5 GPa (Fig. 2). The volumetric data of Bridgman [13] show that there is no change of volume when crossing the transition line at the maximum of 0.5 GPa. At lower pressures, phase II is less dense than phase III, while at higher pressures phase II is denser. In addition, phase II presents a higher compressibility than phase III over the whole pressure range up to 1 GPa. This peculiar combination of higher compressibility of the denser phase is also observed for ice and water.

Several of the high-pressure structures have been determined, at least qualitatively [17,18,21]. Differential thermal analysis data [14] show the existence of a transition between the red phase III and another red phase IV at moderate temperatures and pressures (Fig. 2). Powder diffraction data collected as a function of pressure at room temperature show that the III–IV transformation is neither associated with a change of space group symmetry nor with a discontinuous change of volume of the unit cell. Instead, it shows two different ‘tetragonalities’ [18]: the slope of the change of the c/a ratio with pressure, and also with temperature, is discontinuous at the phase boundary, while c/a – 22 changes sign. All observations obtained with various experimental techniques [18] may be explained by assuming that the III–IV transformation is an anti-isostructural transition [22].

Upon further increasing the pressure at room temperature, the red phase (III, IV) undergoes a transition to a yellow_HP phase VI at about 1.2 GPa. At room temperature, phase VI is stable up to 7.7 GPa, at which pressure a transformation to phase VII occurs. Based on the first published diffraction data at high pressure [17], obtained with a sealed Mo tube and X-ray film, phase VI was proposed to be identical with phase II, and phase VII a polytypic variant of the CdI₂ structure. Later Raman spectroscopy data disproved the identity of phases II and VI and the corresponding phase boundary was established [16]. The most recent diffraction experiments carried out with synchrotron radiation and area detector showed that the structure of phase VI is another variant of the yellow molecular phase [18,21], showing bent HgI₂ molecules that form dimers (Fig. 3). Powder diagrams of phase VII at 11.0 GPa could be indexed with a hexagonal cell, c/a = 1.43 [18]. The data did not allow to solve the structure; for a CdI₂-type structure c/a ~ 1.6.

3.1 Phase diagram of HgBr₂

At ambient conditions, the stable phase I of HgBr₂ adopts the same structure as the yellow metastable polymorph of HgI₂ [23,24]. At ambient pressure, no phase transitions have been reported, phase I being stable from 90 K, the lowest temperature investigated, to the melting temperature of 509 K. The application of pressure up to 4.5 GPa induces several transitions as revealed by dilatometry (compression-curve method) [25] and Raman spectroscopy [26]. Three phases (II, III, IV) have been reported in this pressure range (Fig. 4). The crystal structure of phase IV has recently been solved making use of synchrotron data, and an additional phase V above 9 GPa was identified [27].

3.2 Phase transition sequence under pressure

The I–II transition was first detected by dilatometry. It is accompanied by a volume change similar to the one observed for the II–III and III–IV transitions [25]. However, its existence was not conclusively confirmed by Raman spectroscopy [26] or by X-ray diffraction [27]. Phase III displays a more complicated diffraction pattern than phase I, additional powder lines suggesting a lower, possibly monoclinic symmetry. Attempts to index the pattern were unsuccessful. In addition, the corresponding Raman spectrum shows an inferior signal-to-noise ratio compared to the spectra from I, II and IV [26]. These difficulties might indicate the formation of a strongly disordered state within the stability range of phase III. The III–IV transition has a destructive character and leads to a well-defined powder pattern of a trigonal phase IV, the structure of which is a variant of the CdI₂-type. It is built from layers of edge-sharing octahedra that are stacked by Br···Br contacts. In contrast to the CdI₂-type structure, some of the coordinated metal atoms are displaced along c from the centroids of the octahedra, leading to three distinct mercury sites (Fig. 5). One of the three Hg atoms adopts a trigonal-pyramidal coordination with three nearest neighbours, whereas the other two show an approximately octahedral coordination with six bonds of roughly the same length [27]. The structure of phase IV may be described as a commensurate modulation of the CdI₂ structure.

4.1 Structure of HgCl₂ at ambient conditions

At ambient conditions, the structure of HgCl₂ (phase I) is molecular with straight Cl–Hg–Cl molecules [28] that form herringbone layers in the mirror plane of the space group Pnma (Fig. 7). This structure is not related to HgI₂-yellow_HT and ambient HgBr₂. The coordination of Hg is 2 + 6 (rather than 2 + 4) forming a distorted rhombohedron. The smallest distance between chlorine atoms of adjacent molecules is less than the sum of the van der Waals radii of Cl suggesting attractive Cl···Cl interactions. The lattice parameters of phase I may be expressed as an orthorhombic distortion of a cubic closest packing of chlorine atoms (Table 1).

4.2 Phase diagram

Upon increasing temperatures at ambient pressure, the lattice constants show a decrease of the orthorhombic distortions of phase I, followed by a transition to a cubic phase V at 572 K, three degrees below the melting temperature [29]. The structure of phase V has not been investigated in detail; one might consider Hg to be disordered over, or diffusing between, voids of the cubic closest packing of Cl atoms.

Infrared, Raman [30] and NQR [31] spectroscopy revealed the existence of two high-pressure phases between 0 and 2.5 GPa at room temperature (Fig. 6). The structures of these two high-pressure phases have recently been determined from X-ray powder diffraction with synchrotron radiation [29]. At ambient temperature, HgCl₂ undergoes a continuous transition to a monoclinic phase IV at 0.7 GPa. The structure of this phase is a slight monoclinic distortion of the orthorhombic phase I, β = 91.01(5)° at 1.7 GPa, whereby the mirror plane is preserved. There are four symmetry independent Cl–Hg–Cl molecules placed on the mirror that are now bent (Fig. 7). Phase IV is stable up to 2.1 GPa, at which pressure a first-order transition occurs to a rhombohedral R3¯m phase II showing straight molecules. Phase II is stable up to 6.0 GPa, at which pressure new powder lines appear in addition to those of the R3¯m phase. These lines have been attributed to phase III. Surprisingly, phases II and III coexist at least up to 20 GPa.