Introduction

Previous mass balance calculations have revealed an excess of chlorine (Cl) inputs over outputs in certain subduction zones (Barnes and Straub 2010; John et al. 2011), suggesting a dynamic balance in the Cl geochemical cycle. To gain a deeper understanding of Cl storage, distribution, and budget on Earth, it is essential to analyze Cl in various Earth reservoirs (Frezzotti and Ferrando 2018). Serpentinites play a crucial role in the recycling of hydrogen (H) and Cl through slab subduction (Scambelluri and Philippot 2001). Notably, unusual nano-inclusions containing NaCl, CaCl2, KCl, and PbCl2 chlorides have been discovered in diamonds originating from the transition zone (Wirth et al. 2009). Roberge et al. (2017) demonstrated that hydrous wadsleyite and ringwoodite can contain approximately 100 ppm Cl, suggesting that the transition zone could be a significant Cl reservoir. Recent high pressure-temperature experimental studies have demonstrated that chemical reactions can occur in Fe-, Cl-containing systems, including both wet Mg0.6Fe0.4SiO3‒NaCl‒H2O and FeO2H‒NaCl and dry FeO‒KCl systems, producing a HP-PdF2-type (space group: \(P{a_{\overline 3 }}\), Z = 4) FeCl2 phase (Koemets et al. 2020; Yin et al. 2022; Yuan et al. 2022). Ab initio calculations also predict the formation of this FeCl2 phase under extreme conditions (Du et al. 2018; Yuan et al. 2022). These findings indicate that the HP-PdF2-type FeCl2 may serve as a potential host for Cl in the lower mantle and even the outer core.

Rozenberg et al. (2009) showed that FeCl2 exhibits two electronic phase transitions upon cold compression, maintaining a hexagonal CdI2-type (space group: \(P{_{\overline 3 }}\)m1, Z = 1) structure from ∼ 0.6 GPa to 65 GPa. The transitions first start at a low-pressure phase (LP), then pass through an intermediate phase (IP) between 30 and 57 GPa, and eventually reach a high-pressure phase (HP) at pressures greater than 32 GPa. The LP-IP interface shows no volume change, but exhibits a discontinuous shrinkage of a axis and an increase of c axis, which is attributed to the quenching of the orbital term. The subsequent IP-HP transition shows a decrease in Fe-Cl bond length and an abrupt volume decrease of ΔV / V = ∼ 3.5%, accompanying with a breakdown of the charge-transfer correlation. Note that no spin transitions are involved in these two pressure-induced electronic transitions.

The spin transition of iron holds fundamental significance for mantle physics, as it impacts the density, elasticity, element partitioning and transport properties of major mantle minerals (Lin et al. 2013; Badro 2014). Recent studies demonstrated that FeO2 and FeO2Hx (x ≤ 1) phases might exist in the deep lower mantle, potentially accounting for the global oxygen-hydrogen cycles (Hu et al. 2016; Liu et al. 2017; Mao et al. 2017; Nishi et al. 2017; Yuan et al. 2018). These unique iron oxide-hydroxide compounds were initially thought to be related to the pyrite structure, while the latest single-crystal X-ray diffraction data indicated they align more closely with the HP-PdF2-type structure (Koemets et al. 2021). X-ray emission spectroscopy experiments indicated that HP-PdF2-type FeO2 and FeO2Hx (x ≤ 1) undergo high- to low- spin transitions at around 50‒65 GPa (Jang et al. 2019; Liu et al. 2019). Determining the volume reduction caused by the spin transition in FeO2Hx (x ≤ 1) is challenging, primarily because of the pressure-dependent dehydrogenation contribution and the varying H content x, which can result in a volume difference of up to 10.7% (Hu et al. 2017). As an Fe analogue to FeO2 and FeO2Hx (x ≤ 1) phases, the HP-PdF2-type FeCl2 may provide new insight into this issue. In this study, we have explored the high-pressure behavior of HP-PdF2-type FeCl2 by X-ray diffraction (XRD) and Raman measurements.

Experimental methods

Starting materials and sample preparation

Symmetric diamond anvil cells (DAC) with culet diameters of 250 μm were used for the high pressure and temperature experiments. A cavity with a diameter of ∼ 100 μm was laser-drilled into a rhenium gasket indented to 30 μm thickness to create the sample chamber. We used anhydrous FeCl2 (99% purity; Aladdin chemicals) as the starting material. Sample preparation was carried out in a glove box due to the high deliquescence of the FeCl2. A thin FeCl2 foil, approximately 10 μm in thickness, was carefully situated on pre-installed silica spacers to avoid the direct contact with the diamond surface. The sample assembly was sealed in DAC in the glove box and transferred to a gas-loading system (HPSTAR). Ne was then loaded as a thermal insulator, pressure medium and pressure calibrant.

Laser heating

Samples were compressed to target pressures at room temperature. Then we heated the samples to target temperatures for about 15 min, using a double-sided 1070-nm Ytterbium fiber laser heating system at HPSTAR (Liu et al. 2022; Yang et al. 2023). The heating temperature was determined by fitting the visible portion of the gray-body radiation from both sides of the heated sample to the Planck radiation function. The accuracy of these fittings was precise, yielding errors within a narrow range of 5 to 10 K. The discrepancy in temperature between the two sides of the sample did not exceed 50 K. To ensure stability throughout the heating experiment, the laser power on each side of DAC was meticulously adjusted, thereby restricting temperature fluctuations to ± 100 K. Hereafter, we use this value to describe the temperature uncertainty.

Synchrotron X-ray diffraction (XRD)

Synchrotron XRD experiments at high pressure and room temperature were carried out at the 15U1 beamline of Shanghai Synchrotron Radiation Facility (Zhang et al. 2015). The wavelength of the incident X-ray beam was 0.6199 Å and the beam size was about 5 × 5 µm2. Pressures were determined by the equation of state (EoS) of Ne medium (Fei et al. 2007). Diffraction data were collected in about 1–4 GPa steps on decompression. An exposure time of 10 s was used for every pattern at each pressure. The collected XRD images were calibrated with a CeO2 standard and then integrated to 2θ-intensity patterns using the program DIOPTAS (Prescher and Prakapenka 2015). Unit-cell parameters were refined by nonlinear regression methods using the software UNITCELL (Holland and Redfern 1997) using 6 most intense diffraction peaks.

High-pressure Raman spectroscopy

Raman spectra were collected using a Renishaw Raman spectrometer. Samples were excited by a 532 nm wavelength laser with a holographic notch filter. The matching grating was 2400 l/mm and the maximum output laser power was 50 mW. The system was calibrated before our high-pressure measurements using the Raman peak of a reference Si crystal at 520 cm− 1. The spectra of our samples were collected in the wavenumber ranges of 50–1000 cm− 1 with an exposure time of 100 s and 4 accumulations. Pressures were determined based on the Raman shift of diamond edge (Akahama and Kawamura 2006).

Results and discussion

Synthesis of HP-PdF2-type FeCl2

By comparing the Raman spectra of the sample before and after laser heating, we can verify whether a possible chemical reaction or a phase transition has occurred. As suggested by Rozenberg et al. (2009), the sample above 57 GPa transforms into a high-pressure hexagonal (HP) phase upon cold compression. As shown in Fig. 1a, two Raman modes can be clearly observed at pressures ranging from 59.8 to 65.9 GPa (i.e., ∼ 417 and ∼ 508 cm− 1 at ∼ 65.9 GPa), which can be attributed to the CdI2-type HP phase according to the study of Rozenberg et al. (2009). At 59.8 GPa after heating up the sample at 1650(100) K, we observed the appearance of a new peak around 318 cm− 1, while the Raman modes of the HP phase were still the dominant. Upon increasing the pressure to 61.2 GPa and heating the sample to 1850(100) K, we could only detect this new Raman peak, indicating the formation of a new phase at the expense of the HP phase. This phenomenon was reproduced in another run at 65.9 GPa after annealing. For this run, we collected a two-dimensional (2D) Raman mapping data in a 50 × 40 µm2 area with a 2 μm step size. As shown in Fig. 1b, the new peak was located in the heating center, while the characteristic peaks of the HP phase were distributed in the unheated region. Furthermore, the Raman spectra of the CdI2-type HP phase at 55 GPa remained consistent before and after undergoing comparable laser heating annealing, indicating that a threshold pressure of approximately 60 GPa is necessary to initiate the transition to HP-PdF2-type FeCl2 phase.

Fig. 1
figure 1

(a) Raman spectra before and after laser heating at typical pressures. Gray lines: spectra before laser heating. Red lines: spectra after laser heating. (b) Raman modes distribution at 65.9 GPa based on two-dimensional Raman mapping results

Full size image

Synchrotron XRD measurements were further conducted on the sample at 65.9 GPa after annealing. As the laser spot size of ∼ 20 μm in diameter is larger than the X-ray spot size of ∼ 5 μm in diameter, we can collect the XRD data of the heated center area without the HP phase. As seen in Fig. 2, except for the diffractions of Ne medium and Re gasket, all peaks in the XRD pattern can be well indexed by a cubic \(P{a_{\overline 3 }}\) lattice with a = 5.1353(6) Å and V = 135.43(5) Å3, indicating the formation of pure HP-PdF2-type FeCl2. The synthesis pressure is much lower than megabar pressures in previous studies (Koemets et al. 2020; Yin et al. 2022; Yuan et al. 2022).

Fig. 2
figure 2

Typical high-pressure X-ray diffraction (XRD) patterns of HP-PdF2-type FeCl2 at room temperature. (a) A two-dimensional XRD image at 67.0 GPa. (b) Integrated 2θ-intensity XRD patterns during decompression. Blue and red lines denote the low-spin and high-spin states, respectively. Small vertical bars beneath the relevant patterns signify the indexing of the diffraction peaks associated with the HP-PdF2-type FeCl2. Grey dashed lines serve as visual aids for tracking the evolution of the diffraction peaks for neon (Ne) and rhenium (Re) under varying pressures

Full size image

Decompression XRD experiments

We performed decompression XRD experiments on HP-PdF2-type FeCl2 at room temperatures (Fig. 2–4). Even though non-hydrostatic conditions arising during cold decompression can cause broadening of the diffraction peaks, we successfully indexed the HP-PdF2-type structure across a wide pressure range of 19–67 GPa in all diffraction patterns. As will be discussed later in Raman results, HP-PdF2-type FeCl2 is stable down to a pressure of ∼ 10 GPa. When the sample is decompressed to about 6 GPa, its diffraction peaks can be indexed to a hexagonal lattice (a = 3.5563(7) Å, c = 5.547(4) Å), which is consistent with the CdI2-type LP phase in the cold compression experiments (Rozenberg et al. 2009).

Notably, between 40 and 36.5 GPa during decompression, we found an abrupt shift in the diffraction peaks, suggestive of an isostructural phase transition (Fig. 3). Further analysis of the pressure-volume relationship revealed a sharp volume discontinuity with a pronounced change amounting to 14.3% (Table 1; Fig. 4). Such a significant change in volume—characterized by a marked collapse upon compression or a substantial expansion upon decompression—is likely due to a pressure-induced spin transition of Fe2+ from a high-spin (HS) to a low-spin (LS) state, or vice versa. Spin transition of iron has been observed in other Fe-bearing minerals, such as siderite (FeCO3) and CaFe2O4, where 6-fold coordinated iron is Fe2+ and Fe3+, respectively (Lavina et al. 2009; Merlini et al. 2010). Specifically, FeCO3 and CaFe2O4 undergo pressure-induced isostructural HS-LS transitions at 43 and 50 GPa, with volume decreases of ΔV / V = 10% and 8.4%, respectively (Lavina et al. 2009; Merlini et al. 2010). We denoted the HP-PdF2-type FeCl2 phase before the volume collapse as the HS state, and that after the volume collapse as the LS state. It can be expected that future high-pressure X-ray emission spectroscopy and Mössbauer spectroscopy studies will help to characterize the spin configuration of iron.

Fig. 3
figure 3

Indexing of the diffraction peaks associated with the HP-PdF2-type FeCl2 at 40.0 and 36.5 GPa. Blue and red lines denote the low-spin and high-spin states, respectively

Full size image
Table 1 The cell parameters of HP-PdF2-type FeCl2 phase at high pressures during decompression
Full size table
Fig. 4
figure 4

Cell volumes of HP-PdF2-type FeCl2 at high pressures. Red circles: data of the high-spin state, this study. Blue squares: data of the low-spin state, this study. Light blue squares: data from Yuan et al. (2022). Solid lines: fitted results using the second-order Birch-Murnaghan equation of state (BM-EoS). The pressures in this study were determined by the EoS of Ne (Fei et al. 2007), and the pressures in Yuan et al. (2022) were determined by the EoS of NaCl or Au (Fei et al. 2007)

Full size image

We fitted the second order Birch-Murnaghan equation to the pressure-volume data. We yielded K0 = 67(4) GPa and V0 = 236(3) Å3 for the HS state. For the LS state, data from Yuan et al. (2022) were also included. As a result, we yielded K0 = 69(3) GPa and V0 = 205(2) Å3 for the LS state. Therefore, the compressibilities of the HS and LS states are comparable in HP-PdF2-type FeCl2.

Raman spectroscopy

We have demonstrated the effective utilization of Raman spectra for identifying CdI2-type HP phase and HP-PdF2-type phase of FeCl2 (Fig. 1). After the synthesis of HP-PdF2-type FeCl2, the Raman spectra during decompression have been recorded as well (Fig. 5).

Fig. 5
figure 5

Representative Raman spectra of HP-PdF2-type FeCl2 and CdI2-type LP phase during decompression at 300 K. HS: high-spin. LS: low-spin. The phase below ∼ 7 GPa (black spectra) is assigned to CdI2-type LP phase based on XRD results (Fig. 6)

Full size image

The LS state of HP-PdF2-type FeCl2 has a characteristic Raman mode centered at ∼ 322 cm− 1 (denoted as VLS) at 65.9 GPa, its frequency progressively decreases with decreasing pressure to 34.8 GPa. After further decompression to 32.7 GPa, four new Raman modes emerge, which are centered at 251 cm− 1, 269 cm− 1, 350 cm− 1 and 385 cm− 1 (denoted as VHS1, VHS2, VHS3, and VHS4, respectively). It is intriguing to note that the onset pressure (∼ 33–35 GPa) of the dramatic changes in Raman spectra coincides with the onset pressure (∼ 36–40 GPa) of the abrupt volume collapse in XRD experiments. The slight pressure difference might be caused by the usage of different pressure calibrants. As there is no symmetry change in XRD results, it is reasonable to infer that the abrupt changes in Raman modes reflect the electronic spin transition.

Factor group analysis predicts five active Raman modes (Ag +Eg +3Tg) for HP-PdF2-type structure. To our best knowledge, Raman spectroscopy data is available for only two HP-PdF2-type AX2 compounds: ZnF2 (Kurzydłowski et al. 2020) and CoF2 (Barreda-Argüeso et al. 2013). For HP-PdF2-type ZnF2, all five predicted Raman modes have been observed and assigned by density functional theory (DFT) calculations (Kurzydłowski et al. 2020). In contrast, HP-PdF2-type CoF2 reveals only one Raman active mode, with two additional modes from the coexisting CaCl2-type CoF2 (Barreda-Argüeso et al. 2013). The discrepancy in the number of observed Raman modes for these two compounds may arise from variations in the AX6 octahedra (such as compaction, deformation and distortion), which can affect the frequency and intensity of corresponding Raman modes. In the case of HP-PdF2-type FeCl2 examined in this study, a substantial change in the effective ionic radius of Fe2+ is expected as a result of transition from LS to HS state during decompression. These changes are likely to significantly alter the FeCl6 octahedra, consequently impacting the frequency and intensity of the associated Raman modes. Advanced DFT calculations are expected to shed further light on this issue.

Indeed, significant changes in the Raman spectra have also been observed in FeCO3 compound across its isostructural HS-LS transition, including the disappearance of the lattice mode L, discontinuous jumps of the lattice mode T and the internal vibration modes v1 and v4 to higher frequencies (Cerantola et al. 2015; Müller et al. 2016). In particular, the v1 mode, which corresponds to the stretching vibration of the CO32− group, exhibits a high sensitivity to spin transitions. The change in the C‒O bond lengths is directly influenced by the significant alteration in the radius of neighboring Fe atoms as they undergo a spin transition. The abrupt splitting or shifting of the v1 mode therefore serves as a reliable indicator for detecting either a mixed-spin state or a sharply defined complete transition. For HP-PdF2-type FeCl2, the presence of mixed high-spin HS and LS states within a narrow pressure range of ∼ 33–35 GPa is also possible based on Raman spectroscopy results.

Furthermore, the characteristic Raman spectra of the HS state of HP-PdF2-type FeCl2 can be traced back to approximately 10 GPa, suggesting the lower limit of pressure stability of the HS state at room temperature (Fig. 5). Below 10 GPa, it transforms to a pure CdI2-type LP phase (Fig. 6), which has two distinctive Raman modes centered at 184 cm− 1 and 299 cm− 1 at 7.3 GPa. The pressure dependences of Raman shifts are linearly fitted for the CdI2-type LP phase, HP-PdF2-type HS and LS states, respectively (Fig. 7).

Fig. 6
figure 6

A representative XRD pattern of the CdI2-type low-pressure phase (LP) of FeCl2 obtained by decompression of HP-PdF2-type FeCl2 to ∼ 6 GPa

Full size image
Fig. 7
figure 7

Pressure dependences of the Raman mode frequencies at room temperature. The black, red and blue lines are linear fittings for the CdI2-type LP phase, HS state of HP-PdF2-type FeCl2 and LS state of HP-PdF2-type FeCl2, respectively. The pressure coefficients (dω / dP) are also shown for the corresponding modes

Full size image

Comparison of HP-PdF2-type FeCl2, FeO2 and FeO2Hx (x ≤ 1)

The valence of Fe and O in HP-PdF2-type FeO2 and FeO2Hx (x ≤ 1) continues to be a point of contention in both theoretical and experimental works (Streltsov et al. 2017; Boulard et al. 2019; Jang et al. 2019; Liu et al. 2019; Koemets et al. 2021). The main experimental challenge stems from the fact that the interpretation of high-pressure spectroscopic data relies predominantly on qualitative comparisons with standards that are either under ambient conditions or are sensitive to the structural environments (Boulard et al. 2019; Jang et al. 2019; Liu et al. 2019). For instance, Liu et al. (2019) interpreted their Mössbauer spectra of FeO2Hx (x ≤ 1) and concluded that the iron was ferrous based on the standard samples of ferrous pyrite FeS2, ferrous Fe0.48Mg0.52O, and ferric Fe2O3 under ambient conditions. On the contrary, Koemets et al. (2021) interpreted their Mössbauer spectra of the Fe25O32, Fe and FeO2 mixture, identifying low-spin ferric iron in FeO2. We suggest that ferrous FeCl2 can serve as an appropriate standard to help solve this issue, because it adopts the same HP-PdF2-type crystallographic structure at high pressures, excludes mixed-valence substitutions, and potentially undergoes a pressure-induced spin transition.

Conclusions

We have synthesized HP-PdF2-type FeCl2 at 60‒67 GPa and 1650‒1850 K in laser-heated DACs. Importantly, the decompression XRD results show an abrupt volume collapse of ΔV / V = ∼ 14% in HP-PdF2-type FeCl2 between 36 and 40 GPa, similar to previous observations in FeCO3, suggesting a pressure-induced isostructural HS-LS transition. Meanwhile, dramatic changes in Raman modes can be observed at comparable pressures. Regarding the complex bonding nature and electronic structures in HP-PdF2-type FeO2Hx (x ≤ 1) and FeO2, we suggest that HP-PdF2-type FeCl2 could be an effective standard for interpreting the high-pressure spectroscopic data of these compounds. Combined with previous studies, HP-PdF2-type FeCl2 phase is stable over a wide pressure range from 10 to 160 GPa at room temperature, and can serve as a stable Cl-host under pressure-temperature conditions of the middle lower mantle to the outer core.