Formation and decomposition of CO2-filled ice (original) (raw)

Skip Nav Destination

Research Article| October 05 2017

B. Massani

;

Institute of Physical Chemistry, University of Innsbruck

, Innrain 52c, 6020 Innsbruck,

Austria

Search for other works by this author on:

C. Mitterdorfer;

Institute of Physical Chemistry, University of Innsbruck

, Innrain 52c, 6020 Innsbruck,

Austria

Search for other works by this author on:

T. Loerting

Institute of Physical Chemistry, University of Innsbruck

, Innrain 52c, 6020 Innsbruck,

Austria

Search for other works by this author on:

J. Chem. Phys. 147, 134503 (2017)

Recently it was shown that CO2-filled ice is formed upon compression of CO2-clathrate hydrate. Here we show two alternative routes of its formation, namely, by decompression of CO2/ice VI mixtures at 250 K and by isobaric heating of CO2/high-density amorphous ice mixtures at 0.5–1.0 GPa above 200 K. Furthermore, we show that filled ice may either transform into the clathrate at an elevated pressure or decompose to “empty” hexagonal ice at ambient pressure and low temperature. This complements the literature studies in which decomposition to ice VI was favoured at high pressures and low temperatures.

I. INTRODUCTION

The investigation of inclusion compounds such as clathrate hydrate or filled ice at high-pressure conditions is of interest for a fundamental understanding of the interactions in guest-host systems. CO2–H2O is of importance in a broad range of fields, including astrophysics, geochemistry, marine biology, food chemistry, and medicine. Under high pressure (e.g., in the interior of planets/moons) or upon impact solid mixtures of CO2–H2O may form clathrate hydrates, filled ice, or carbonic acid.1–6 The formation of H2CO3 was suggested in 2016 by Saleh and Oganov using computer simulation.7 Experimentally, H2CO3 was found by Wang et al. who performed IR-experiments at high pressures above 2.4 GPa.6 This compound is surprisingly stable at low temperatures and high pressures.8

In cold environments, also mixed compounds such as CO2-clathrate hydrates may form. Clathrate hydrates are hydrogen bonded water cages, stabilized by non-covalently bond guest molecules.1,9–11 Depending on the size of the guest molecule, different crystal structures are formed (e.g., cubic structures CS-I and CS-II and hexagonal structure H).2 It is worth remarking that carbon dioxide, as a medium sized molecule with an average diameter of about 5 Å, may occupy the small 512 cages (consisting of 12 pentagons formed by the H2O molecules with oxygen atoms on each corner position) as well as the larger 51262 cages (consisting of 12 pentagons and 2 hexagons formed by the H2O molecules with oxygen atoms on each corner position) of the CS-I structure.2 Clathrates can therefore be used to fix large amounts of CO2. This type of carbon sequestration may be of environmental interest. Presently many different routes to synthesize clathrate hydrates are known. The most common ones are liquid-liquid, liquid-gas, and solid-gas interface methods.2,12–15

Loveday et al. first investigated the high-pressure behavior of clathrate hydrates on the example of methane guests: Cubic structure clathrate hydrates transform with increasing pressure into the hexagonal structure.1 The hexagonal structure changes on pressurization either into the tetragonal structure or into methane hydrate III (MH-III), a filled ice structure.16,17

Hirai et al. first reported on a new crystalline high pressure phase of CO2 and H2O.3 Formed by compressing CO2-clathrate hydrate type I (space group Pm_3_n)2 using diamond anvil cells and a helium-refrigeration cryostat, this phase was revealed at pressures above approximately 0.6 GPa and below 1.0 GPa. Above this pressure, the new phase decomposes to dry ice (space group Pa_3̄⁠)18 and ice VI (space group P42/nmc)19. Tulk et al. solved the structure of this new phase20 in 2014 by fitting diffraction data with computer models (space group Imma). According to their fundamental work, the structure is similar to MH-III, which itself is remarkably similar to the structure of hexagonal ice (ice I_h, space group _P_63/mmc)21: The guest molecules are located in the channels of the H2O network. Due to the guests, the structure of filled ice is—compared to ice I_h_—stretched and shows larger O–O distances and larger lattice constants. In 2013, Bollengier studied the decomposition behavior of carbon dioxide filled ice at temperatures above 250 K.4 They found that filled ice decomposes to ice VI and dry ice upon heating to temperatures above ≈270 K. An open question is that why Bollengier et al.4 did not find filled ice at temperatures lower than 250 K, while Hirai et al.3 did. Filled ice structures were formerly only known in systems consisting of cubic ice or ice II lattices in the presence of very small molecules and atoms such as H2 or noble gases.22–24

Other routes to the formation of CO2 filled ice as well as a careful evaluation of the decomposition of filled ice at different p/T combinations are missing in the literature. Therefore, we here investigate the high pressure behavior of CO2/H2O. We reveal two methods to form CO2 filled ice: from ice VI19 and dry ice25 during decompression and from high-density amorphous (HDA) ice and dry ice by isobaric annealing, using a combination of dilatometry, powder X-ray diffraction, and calorimetry. Using these approaches, we observe filled ice in a broader pressure range than known before. Furthermore, we reveal that filled ice may either transform into the clathrate or decompose to hexagonal ice and CO2, while Bollengier et al. observed only ice VI and CO2.

II. MATERIALS AND METHODS

Filled ice is formed by pressurising a vapor-deposit of CO2–H2O. The vapor co-deposit of H2O and CO2 was produced at 77 K. A scheme of the deposition apparatus is depicted in Fig. 2 in the work of Mayer and Pletzer.26 The process of the co-deposition was described by Mitterdorfer et al. in 2011.27 In brief, water vapour and gaseous CO2 enter a high-vacuum chamber via two fine metering valves through a copper tube with a diameter of 13 mm. The inlet pressures for CO2/H2O were chosen at a ratio of 1:10, namely, the pressure for CO2 is 0.02 mbar and for H2O is 0.2 mbar (background pressure 10−4 mbar). The two gases are deposited on a copper plate kept at 77 K (deposition time ≈12 h). The deposit of roughly 2 mm in thickness is recovered at ambient pressure by scratching it off the copper plate and stored under liquid nitrogen.

The deposit was characterised ex situ by powder X-ray diffraction (XRD). To that end, the samples were cold-loaded under liquid nitrogen to the cold (77 K) sample holder in the XRD-chamber (Anton Paar TTK450). Diffractograms were recorded at 77 K in the range 15_°_ <2Θ< 55_°_ at approximately 10−1 mbar using Cu–K_α_ radiation (1.541 Å), following the standard procedure employed in many studies in the past by our group with the X-ray apparatus Siemens D_5000.28 The phase compositions of crystalline samples containing more than one specimen were determined using the program PowderCell (version 2.4, BAM, Bundesanstalt fuer Materialforschung und -pruefung, Berlin, Germany). An X-ray diffractogram of the co-deposit is shown in Fig. 1(i) and can be compared with Fig. 1(B) in the work of Mitterdorfer et al.27 A broad halo peak between 20_° and 32_°_ and several sharp Bragg peaks are visible: the former indicates microporous amorphous solid water (ASW)29 and is superimposed by the latter indicating crystalline CO2.30

FIG. 1.

(i) XRD scan of ASW-CO2 co-deposit. The black curve shows a sample that was not heated above 77 K. The grey curve shows a sample that was heated in the deposition apparatus to 130 K. Upon this heating, the surface CO2 desorbs. The orange labeled reflections (ice I_h_) indicate condensed water from air. (ii) Sample after the compression to 1.8 GPa. The maximum of the halo peak is shifted to larger angles, compared to ASW, indicating the transformation to HDA. (iii) Sample after compression to 1.8 GPa and annealing to 195 K. (iv′) Sample after compression to 1.8 GPa and annealing to 250 K. The blue squares, orange dashes, and black dashes represent Bragg peaks of dry ice, ice I, and ice VI, respectively.

FIG. 1.

Close modal

After the co-deposition experiment and X-ray characterisation, the ice powder was transferred under liquid nitrogen to our piston-cylinder setup. Figure 1 in the work of Elsaesser et al.31 illustrates the experimental setup of our piston-cylinder apparatus. The ice powder was filled into the 8 mm bore of the hardened steel cylinder, where an indium lining acts as a lubricant.32,33 Without lubricant, pressure drops and shock-wave heating occur in the sample and may lead to crystallisation.33 The loose ice powder was then precompacted to a cylinder shape by applying a pressure of 0.2 GPa30 and used as the starting material for all further experiments carried out. They are marked by (i) in the piston displacement curves shown in Fig. 2; the respective powder diffractogram is shown in Fig. 1(i). These samples were then used to conduct compression/decompression cycles including a high-pressure annealing step. Specifically, the samples were compressed at 77 K to 1.8 GPa and then heated isobarically to 200 or 250 K with a heating rate of 3 K/min. The samples were then quench recovered or isothermally decompressed at 200 or 250 K to selected pressures (0.02–0.8 GPa) before quench recovery.

FIG. 2.

Piston displacement curves of (a) and (b) compression at 77 K, (b) and (c) isobaric heating at 1.8 GPa and decompression at 200 K (a) and 250 K (b), respectively. A piston displacement in the upward direction is equivalent to the densification of the sample. The sample mass cannot be determined accurately based on weighing. However, from the size of the densification steps, we estimate the sample mass to be on the order of 200 mg.

FIG. 2.

Close modal

The quench-recovered specimens were analyzed by X-ray diffraction,30,33 and some selected specimens were additionally investigated by differential scanning calorimetry (DSC). DSC experiments were carried out using DSC8000 (PerkinElmer) at ambient pressure and heating rates between 3 and 30 K min−1. Samples were placed in an aluminium crucible, which was closed with a lid. During this, the samples were kept under liquid nitrogen.

III. RESULTS

A. Formation of the filled ice—Decompression of ice VI and solid CO2

Figure 2 reports a volumetric experiment on the CO2/H2O co-deposit. Point (i) in panels (a) and (b) represents the vapor co-deposit of carbon dioxide and amorphous solid water (ASW + crystalline CO2 with the space group _Pa_3̄⁠)18 after the transfer to the piston-cylinder setup at 77 K and precompaction. The respective diffractogram is shown in Fig. 1(i). Upon compression, a step in density is seen between 0.4 GPa and 1.1 GPa. This corresponds to the well known polyamorphic transformation of ASW to high-density amorphous (HDA) ice. This was reported first by Mishima in 1994 on the example of pure low-density amorphous (LDA) ice.34 As compared to pure LDA ice, the step is about 0.1 GPa broader for CO2/H2O.35 Thus, at point (ii), the samples consist of CO2(s) and HDA ice. The transformation can be recognised in Fig. 1(ii), in which a shift of the broad halo peak to higher angles is seen. For comparison, the halo peak before compression is shown in light colour in Fig. 1(ii). From the step height, we can estimate in hindsight that we had loaded approximately 200 mg of ASW, not considering the additional mass of crystalline CO2. This corresponds roughly to a cylinder of 8 mm in diameter and 10 mm in height.

Isobaric heating at 1.8 GPa to 200 K leads to crystallization of HDA ice to ice VI,36–38 which can be observed in the decrease of volume around 195 K in Figs. 2(c) and 2(d). Therefore, the diffractogram of the quench-recovered specimen in Fig. 1(iii) reveals a mixture of crystalline CO2 and ice VI. Upon continued heating, a second step is observed at about 230 K in Fig. 2(d). The positive value of the piston displacement indicates a loss of volume.

The height of this second step is dependent on the amount of carbon dioxide in the sample: The grey curve in Fig. 2(d) shows an experiment performed with a co-deposit that was heated in the deposition apparatus to 130 K. At that temperature, the excess CO2 adsorbed on the external surface of the sample desorbs, whereas it remains on the internal pore surfaces. A diffractogram of such a co-deposit after heating _in vacuu_m is shown in Fig. 1(i) (grey diffractogram). Since the height of this step is dependent on the amount of CO2 in the starting material, this decrease of volume is assigned to the thermal desorption of CO2 from the ice surface: carbon dioxide above 230 K shows an increased mobility. The desorbed CO2 then leaves the piston cylinder instrument, which is not gas tight.

The samples in Figs. 1(iii) and 1(iv′) consist therefore of ice VI and solid carbon dioxide entrapped in an ice matrix. The respective diffractograms are shown in Figs. 1(iii) and 1(iv′), respectively.

Also during the decompression, the piston displacement curves exhibit steps, which are indicative of phase changes to lower dense phases. For the decompression at 200 K [starting from point (iii) in Fig. 2(a)], a single step is found between 0.4 and 0.3 GPa. In the case of the 250 K experiment (starting from (iv′), in Fig. 2(b), there are two steps located between 0.8 and 0.6 GPa as well as 0.6 and 0.4 GPa. A final step at <0.2 GPa is seen both in Figs. 2(a) and 2(b). The XRD scans after quench-recovery from the points marked in Figs. 2(a) and 2(b) are given in Fig. 3 for the decompression at 200 K and 250 K, respectively.

FIG. 3.

$FIG. 3. Top: Powder X-ray diffractograms for isothermal decompression at 200 K and quench-recovery at 0.02 GPa, 0.3 GPa, 0.4 GPa, and 0.6 GPa [curves (a) and (c) in Fig. 2]. Bottom: Powder X-ray diffractograms for isothermal decompression at 250 K and quench-recovery at 0.02 GPa, 0.4 GPa, 0.6 GPa, and 0.8 GPa [curves (b) and (d) in Fig. 2].$

Top: Powder X-ray diffractograms for isothermal decompression at 200 K and quench-recovery at 0.02 GPa, 0.3 GPa, 0.4 GPa, and 0.6 GPa [curves (a) and (c) in Fig. 2]. Bottom: Powder X-ray diffractograms for isothermal decompression at 250 K and quench-recovery at 0.02 GPa, 0.4 GPa, 0.6 GPa, and 0.8 GPa [curves (b) and (d) in Fig. 2].

FIG. 3.

Close modal

At 200 K and 0.6 GPa, see Fig. 3(iv), crystalline CO2 and ice VI are still the two dominant phases. Decompression to <0.4 GPa, Fig. 3(v), leads to the formation of the cubic clathrate hydrate type I (CS-I), with crystalline carbon dioxide and ice VI still being present. The most prominent Bragg peak of ice VI at 35.6_°_ vanishes, whereas the Bragg peak of CS-I hydrate at 28.5_°_ becomes the most intense one. Thus, the CO2-ice VI mixture transforms directly into the clathrate hydrate,

Ice VI+CO2(s)→<0.6 GPa200 KCS−I.

Both CO2 and ice VI were not completely transformed and remain partly in the sample. However, whereas large amounts of solid carbon dioxide remain, only small amounts of ice VI remain.

At 200 K and 0.3 GPa, Fig. 3(vi), besides the clathrate, ice II is also observed. That is, the remainder of ice VI in (v) has transformed to ice II in Fig. 3(vi), e.g., peaks at 24_°_, 37_°_, and 38_°_ grow. Further decompression to 0.02 GPa results in the observation of hexagonal ice in Fig. 3(vii) due to the transformation of ice II to hexagonal ice. The remaining ice VI undergoes phase transitions upon decompressing according to the phase diagram of water,

Ice VI→<0.4 GPa200 KIce II→<0.1 GPa200 KIce I.

Both the CO2-clathrate and CO2(s) remain stable down to 0.02 GPa and 200 K.

At 250 K the evolution of phases is slightly different: At 0.8 GPa and 250 K, Fig. 3(v′), the Bragg peaks originate mainly from ice VI and crystalline CO2 just like at 200 K. However, after the first step in Fig. 2(b), a new phase emerges in Fig. 3(vi′): Reflections attributed to CO2-filled ice (FI) appear, namely, 40.4_°_, 31.7_°_, 29.5_°_, 22.4_°_, 21.5_°_, 18.2_°_, and 16.2_°_3 at 0.6 GPa and 250 K, Fig. 3(vi′). Additionally, crystalline CO2 and ice VI remain, and a small amount of clathrate hydrate cubic structure I is formed,

Ice VI+CO2(s)→<0.7 GPa250 KFI.

Further decompression beyond the second step in Fig. 2(b) to 0.4 GPa and 250 K, Fig. 3(vii′), leads mainly to the formation of clathrate hydrate CS-I and some ice V, whereby filled ice has completely vanished,

Quench recovery at 0.02 GPa features clathrate hydrate CS-I and solid CO2 as the main components, Fig. 3(viii′). However, the CS-I hydrate reflections shrink at the expense of the CO2(s) reflections: The clathrate hydrate decomposes to ice I and solid CO2.39 The presence of ice V and ice I is a result of the transformation of excess ice VI during the decompression at 250 K according to the phase diagram of water,

Ice VI→<0.7 GPa250 KIce V→<0.2 GPa250 KIce I.

The most important difference between these two decompression experiments is the presence of filled ice in the case of 250 K, whereas at 200 K, it is not emerging. The thermal desorption of surface CO2 and the increased mobility of carbon dioxide, if heated to 250 K, seem to be essential for the formation of filled ice.

Note that the FI transforms to the clathrate for decompression at 250 K. This is the reverse of the formation mechanism reported by Hirai et al.3

B. Formation of the filled ice—Heating HDA and solid CO2

In our experiments above, we gave evidence that filled ice is formed by decompression at 250 K but not at 200 K from CO2–H2O. Hirai et al. showed that filled ice may be formed by compression of CS-I at lower temperatures,3 down to 100 K.

Besides those two ways, we formed filled ice from solid carbon dioxide and HDA by isobaric heating: Here we compressed the co-deposit of carbon dioxide and water to 1.2 GPa to form a mixture of HDA ice and solid carbon dioxide in our piston and cylinder setup at a temperature of 77 K. The sample was then decompressed to various pressures between 0.35 GPa and 1.0 GPa and heated to 230 K. In Fig. 4, the dilatometric curves for the heating steps are depicted.

FIG. 4.

FIG. 4. Dilatometric curves of isobaric annealing of HDA ice and CO2(s) in the pressure range between 0.35 GPa and 1.0 GPa.

Dilatometric curves of isobaric annealing of HDA ice and CO2(s) in the pressure range between 0.35 GPa and 1.0 GPa.

FIG. 4.

FIG. 4. Dilatometric curves of isobaric annealing of HDA ice and CO2(s) in the pressure range between 0.35 GPa and 1.0 GPa.

Dilatometric curves of isobaric annealing of HDA ice and CO2(s) in the pressure range between 0.35 GPa and 1.0 GPa.

Close modal

At 0.35 GPa, only two phase transitions were observed: one at 141 K and the other at 160 K. The sharp step at 141 K is assigned to the crystallisation of HDA ice according to the phase diagram of water.40 After the second step, the quench-recovered sample consists solely of clathrate hydrate—no filled ice forms at 0.35 GPa,

HDA+CO2→0.35 GPa>160 KCS−I.

At higher pressures, more steps are observed, the first one at roughly 160 K. Right after this step, the sample was quench recovered and analyzed by X-ray diffraction. A diffractogram at this temperature is depicted in Fig. 5(a). A broad halo peak at 30_°_ gives evidence that HDA ice is still present. Furthermore, reflections of carbon dioxide are visible as well as emerging Bragg peaks of CO2-clathrate hydrate and hexagonal ice.

FIG. 5.

FIG. 5. Examples of the XRD scans of the quench-recovered sample at 165 K (a) and at 230 K (b).

Examples of the XRD scans of the quench-recovered sample at 165 K (a) and at 230 K (b).

FIG. 5.

FIG. 5. Examples of the XRD scans of the quench-recovered sample at 165 K (a) and at 230 K (b).

Examples of the XRD scans of the quench-recovered sample at 165 K (a) and at 230 K (b).

Close modal

Upon further heating, a sharp step with an onset at 167 K, 174 K, and 179 K occurred at 0.5 GPa, 0.75 GPa, and 1.00 GPa, respectively. This step is assigned to the crystallisation of HDA ice (cf. Stern and Loerting41). Compared to pure water, this step is shifted to slightly higher temperatures by about 7 to 8 K. After this step, the clathrate has been formed,

HDA+CO2→0.5−1.0 GPa>180 KCS−I.

At a temperature of 198 K, 203 K, and 207 K, at 0.50 GPa, 0.75 GPa, and 1.00 GPa, respectively, another step can be observed in the piston displacement curves in Fig. 4. This step is missing in the experiment at 0.35 GPa.

Figure 5(b) shows a respective XRD pattern of the quench-recovered samples at 230 K. Here, the sample consists mainly of filled ice (red *) together with large amounts of ice VI and solid carbon dioxide. At this temperature, no clathrate hydrate can be observed. The step at ≈200 K can therefore be assigned to the formation of filled ice from the clathrate hydrate and CO2(s),

CS−I+CO2→0.5−1.0 GPa>198 KFI.

Here, no direct transition from an amorphous solid to the filled ice is observed. In fact, filled ice is formed at the expense of the clathrate hydrate.

These experiments reveal one major thing: At temperatures above 230 K and 0.5 GPa, all pathways lead to the formation of filled ice. However, at 200 K, it depends on the pathway whether or not filled ice forms, e.g., at 0.5 GPa. In other experiments, it forms in the isobaric heating experiments, but not in the isothermal decompression experiments.

C. Decomposition of filled ice at 1 bar

The decomposition of CO2-filled ice at high pressures has been previously studied by Hirai et al. and by Bollengier et al.3,4 They observed decomposition to ice VI/dry ice and to CS-I. In Sec. III A, we have shown that filled ice transforms to CS-I upon decompression at 250 K. In this subsection, we are going to describe the stability of filled ice using DSC and XRD experiments at ambient pressure (heating rates 3 K min−1 and 30 K min−1, respectively).

The first heating scan in Fig. 6(a) shows two peaks in the calorimetric experiment: a very broad, smeared exotherm between about 105 K and 145 K and a sharp peak with an onset temperature of about 150 K (at 3 K min−1). We attribute the first exotherm to the decomposition of CO2-filled ice. The very broad and also smeared features indicate a very slow transition. The second exothermic peak at 150 K is the well known formation of ice I_c_ from high-pressure ice phases, which was first observed calorimetrically by Handa et al. in 1988.42

FIG. 6.

FIG. 6. Top: DSC scan of the quench-recovered specimen with a heating rate of 3 K min−1. Bottom: The quench-recovered specimen was heated to 140 K, cooled to 90 K, and heated once again with a heating rate of 30 K min−1.

Top: DSC scan of the quench-recovered specimen with a heating rate of 3 K min−1. Bottom: The quench-recovered specimen was heated to 140 K, cooled to 90 K, and heated once again with a heating rate of 30 K min−1.

FIG. 6.

Close modal

In order to determine whether or not the transition is reversible, a second DSC experiment was performed: Fig. 6(b) shows two subsequent heating scans with a heating rate of 30 K min−1: first to 140 K and then from 90 K to 160 K. Also in this case, the broad exothermic peak is clearly visible between 100 K and 140 K in the first scan. However, the exotherm disappears to a large extent in the second heating cycle, i.e., the transition is not reversible. The shift of the onset temperature from 109 K at 3 K/min to 100 K at 30 K/min is unexpected. Two samples from two different co-depositions were used and so different degrees of filling might be at its origin.

Possible decomposition products for filled ice at ambient pressure are CS-I hydrate (as noticed above for the decompression experiment at 250 K), ice I_c_, and I_h_ (as well as solid CO2). To investigate this, X-ray diffraction (XRD) experiments were performed in this temperature range: The quench-recovered filled ice was transferred onto a sample holder in a vacuum chamber. This sample holder was slowly heated with a heating rate of 3 K min−1 to match the rate used for the DSC experiments [Fig. 6(a)]. Bragg peaks were recorded in three different intervals, namely, 22.0-26.5_°_, 30.0-33.5_°_, and 35.0-36.5_°_. The first interval contains the three most intense ice I_h_ Bragg peaks [see Fig. 7(d)]. The second contains the most intense reflections of ice VI, CO2-clathrate hydrate, and CO2-filled ice. Crystalline CO2 is found in the third interval. The XRD-patterns plotted against the temperature are given in Fig. 8.

FIG. 7.

FIG. 7. (a)–(c) The (101), (002), and (001) Bragg peak heights of ice Ih plotted against the temperature. (d) The corresponding powder X-ray scans between 22.0° and 26.5°. (For the experiment, a heating rate of 3 K min−1 was used.)

(a)–(c) The (101), (002), and (001) Bragg peak heights of ice I_h_ plotted against the temperature. (d) The corresponding powder X-ray scans between 22.0_°_ and 26.5_°_. (For the experiment, a heating rate of 3 K min−1 was used.)

FIG. 7.

Close modal

FIG. 8.

FIG. 8. Top: XRD scan between 30.0° and 33.5° with the most intense reflections of ice VI, CO2-clathrate hydrate, and CO2-filled ice. Bottom: In situ XRD scan of the interval of 35.0°–36.5°, containing a reflection of crystalline CO2. A heating rate of 3 K min−1 was used.

Top: XRD scan between 30.0_°_ and 33.5_°_ with the most intense reflections of ice VI, CO2-clathrate hydrate, and CO2-filled ice. Bottom: In situ XRD scan of the interval of 35.0_°–36.5_°, containing a reflection of crystalline CO2. A heating rate of 3 K min−1 was used.

FIG. 8.

Close modal

It is clearly visible that the Bragg-peak intensities of ice VI, crystalline carbon dioxide, and filled ice decrease with temperature, whereas the intensities of ice I_h_’s Bragg peaks increase steadily. Compared to that, the intensity of the Bragg peak of cubic type I clathrate hydrate remains relatively constant; hence, the possibility of clathrate hydrate formation during the decomposition of filled ice can be excluded. The respective relative Bragg-peak intensities of ice I [relative to the (002)-peak intensity of hexagonal ice at 230 K] plotted against the temperature are shown in Fig. 7.

Hereby, the Bragg peak at 26.0_°_ reflects the lattice plane (101) and can be found in hexagonal ice only. The most intense of those three reflections is the (002) peak in ice I_h_ at 24.2_°_. It occurs in the diffraction patterns of hexagonal as well as the cubic type ice I.43 In contrast, the Bragg peak of the lattice plane (001) at 22.7_°_ is more complicated. It is also experimentally observed in patterns of ice I_c_ and represents a measure of hexagonal stacking faults. The more hexagonal stacking faults44–46 are incorporated in cubic ice, and the more “hexagonal” the ice gets, the more intense this peak becomes. The transition from mainly cubic ice to hexagonal ice usually occurs at around 200 K.44

The (002) reflection grows between 100 K and 160 K where it reaches its maximum intensity [see Fig. 7(b)]. After that point, no more ice I is formed. One may even notice two steps, namely, between 100 K and 140 K and between 140 K and 160 K, consistent with the two exotherms in Fig. 6(a). That is, ice I forms both from filled ice and ice VI. Whether ice I_c_ or ice I_h_ forms from filled ice is answered in Fig. 7(a): There we clearly see that the intensity of the (101)-peak grows between 90 K and 140 K. This indicates that hexagonal ice is formed from filled ice. This is an unusual behavior since high pressure ices usually form cubic ice upon heating.

At last, the Bragg-peak intensities in Fig. 7(c) provide us with the (001) reflections. In this plot, we see two increases of the intensity: the first between 90 K and 140 K and the other one between 200 K and 220 K. The first increase is identical with the rise given in Fig. 7(a), which also indicates that hexagonal ice is formed from filled ice. The second plateau is reached after the remaining cubic ice (formed from ice VI) is transformed into I_h_.

The quench-recovered sample gets “more hexagonal” in exactly the same temperature range as in which filled ice is decomposed. Additionally to that, we see in Fig. 8 that the crystalline CO2 is removed from the sample in the very exact temperature range. Hence we conclude that filled ice directly transforms into hexagonal ice by the release of CO2 from the filled channels,

FI→1 bar100−140 KIh+CO2(s).

This slow process of CO2 release also explains the slow transition observed in our DSC experiments.

Additionally to that, we have the formation of cubic ice from ice VI, which itself is transformed into hexagonal ice at 220 K,

Ice VI→1 bar150 KIc→1 bar200−220 KIh.

The temperature dependence of the Bragg peak positions as a measure of ice lattice constants is depicted in Fig. 9. In this figure, the filled symbols were measured by us, whereas the open symbols were measured by Roettger et al. in 1994.21 According to Roettger’s studies, water’s lattice constant usually becomes larger upon heating due to the expansion of its primitive cell at temperatures >75 K. The behavior of our samples at temperatures >140 K fits with Roettger’s data quite well. (The larger standard deviation of our data is due to Roettger using single crystals rather than a powder.) However, in the temperature range between 80 K and 140 K, we clearly see that our data points in Fig. 9 shifted to larger angles for the (001) and (101) reflections. This indicates a shift in the lattice constant in the precise temperature range in which filled ice decomposes and therefore also fortifies our explanations. We attribute the deviation from the hexagonal ice data by Rettger et al.21 to the transformation from filled ice (space group Imma)20 to hexagonal ice (space group _P_63/mmc)21.

FIG. 9.

FIG. 9. The filled symbols give the positions of the most intense Bragg peaks observed in the course of the decomposition of CO2-filled ice at ambient pressure plotted against the temperature. The open symbols as well as the dashed lines shows the data measured by Roettger et al. in 1994 for the lattice panes (100), (002), and (101) of hexagonal ice.21

The filled symbols give the positions of the most intense Bragg peaks observed in the course of the decomposition of CO2-filled ice at ambient pressure plotted against the temperature. The open symbols as well as the dashed lines shows the data measured by Roettger et al. in 1994 for the lattice panes (100), (002), and (101) of hexagonal ice.21

FIG. 9.

Close modal

IV. CONCLUSIONS

The existence of CO2-filled ice was only discovered recently. Hirai et al. were the first to report on a novel but uncharacterized high-pressure phase in the CO2/H2O system in 2010.3 Only in 2014 it was recognized by Tulk et al. that this high-pressure phase is in fact a filled ice based on the hexagonal ice structure.20 Bollengier et al. investigated the stability of filled ice at temperatures >250 K in 2013.4

In this work, we here show two alternate routes to form CO2-filled ice. Furthermore, we investigate the decomposition of filled ice at elevated and ambient pressures.

Carbon dioxide filled ice can be formed directly by the decompression of ice VI and solid CO2 from 1.8 GPa at 250 K to 0.6 GPa,

Ice VI+CO2(s)→<0.7 GPa250 KFI→<0.5 GPa250 KCS−I.

Upon further decompression, filled ice forms clathrate hydrate type I.

However, decompression at a temperature of 200 K leads to the formation of clathrate hydrate with no filled ice being formed,

Ice VI+CO2(s)→<0.6 GPa200 KCS−I.

That is, a previously unknown route towards filled ice has been revealed in our work. We speculate an increased mobility of the carbon dioxide molecules at 250 K which start to desorb from the ice at temperatures >230 K to be at the origin of the filled ice formation.

Besides that, filled ice was formed by isobaric heating of high-density amorphous ice and solid carbon dioxide. Upon annealing, first HDA-CO2 transforms to clathrate hydrate,

HDA+CO2(s)→0.35−1.00 GPa>160 KCS−I.

Upon further heating, the remaining HDA ice crystallises according to the phase diagram of water. At temperatures above ≈200 K, once again filled ice is formed with the clathrate hydrate vanishing from the sample. A transformation

CS−I+CO2→0.5−1.0 GPa>198 KFI

is the reason for this. In the decompression experiments, at this very temperature, no filled ice was observed. Also, Bollengier et al.4 did not find filled ice at that temperature. That is, the formation of filled ice is strongly dependent on the pathway. It forms much more readily from the CO2 clathrate than from ice VI/CO2.

All experiments carried out are summarised in Fig. 10. Also selected data points from Hirai et al.3 and Bollengier et al.4 are depicted.

FIG. 10.

Comparison of CO2/H2O phases observed in different types of experiments. The grey data points are selected experiments of Hirai et al.3 and Bollengier et al.4 ↑⁠: isobaric annealing; →⁠: isothermal compression of CS-I; ←⁠: isothermal decompression of ice VI and dry ice. Blue, brown, and red arrows represent CO2(s)+ice VI, clathrate hydrate, and filled ice, respectively, and indicate the most abundant component found at this point in the p/T plane.

FIG. 10.

Close modal

In this figure, the grey line corresponds to the crystallisation temperatures of the amorphous water phases. The orange line is the melting curve of clathrate hydrate. The dotted red line gives the proposed stability region of carbon dioxide filled ice as given by Hirai et al.

We found filled ice at lower as well as higher pressures. This may be due to us using a macroscopic amount of the sample (200 mg), whereas Hirai et al.3 only studied filled ice in diamond anvil cells with much smaller samples. Furthermore, this may also be due to different mixing ratios between CO2 and H2O in different sets of experiments. Since the mixing ratios are not known from our experiments, they are not explicitly included in Fig. 10.

At ambient pressure, filled ice does not transform to CS-I as it does at high-pressure but releases gas and directly transforms into hexagonal ice in a temperature range between 100 K and 140 K. High-pressure ice is usually converted to cubic ice not hexagonal ice first at ambient pressure. Here, the structural similarity between I_h_ and FI leads to a direct transformation to hexagonal ice,

During this process, the CO2 is released from the channels, resulting in solid CO2. These results are not only novel from the experimental perspective but also represent a benchmark for theoretical studies. For example, the formation of the CS-I clathrate from filled ice as well as the absence of carbonic acid near 1 GPa provides strict constraints on the intermolecular potentials governing the CO2/H2O systems.

Note added in proof. Very recently a publication by Amos et al. appeared [Daniel M. Amos, Mary-Ellen Donnelly, Pattanasak Teeratchanan, Craig L. Bull, Andrzej Falenty, Werner F. Kuhs, Andreas Hermann, and John S. Loveday, A Chiral Gas-Hydrate Structure Common to the Carbon Dioxide-Water and Hydrogen-Water Systems, J. Phys. Chem. Lett. 8, 4295–4299 (2017)], in which the filled ice was actually shown to be of a more complex, chiral structure based on neither a known hydrate nor ice structure but instead related to two Zintl phases. Therefore, our finding is surprising that this high pressure hydrate of CO2 transforms into hexagonal ice rather than cubic ice at ambient pressure.

ACKNOWLEDGMENTS

We gratefully acknowledge funding of the study from the Austrian Science Fund (FWF), Project Nos. P18187 and I1392.

The authors declare no competing financial interest.

REFERENCES

E. D.

Sloan

Nature

426

353

(

2003

Sloan

and

Kohl

Clathrate Hydrates of Natural Gases

(

CRC Press

2008

Hirai

Komatsu

Honda

Kawamura

Yamamoto

, and

Yagi

J. Chem. Phys.

133

124511

(

2010

Bollengier

Choukrounr

Grasset

E. L.

Menn

Bellino

Morizet

Bezacier

Oancea

, and

Tobie

Geochim. Cosmochim. Acta

119

322

(

2013

Bernard

Seidl

Mayer

, and

Loerting

Chem. Phys. Chem.

3087

(

2012

Wang

Zeuschner

Eremets

Troyan

, and

Willams

Sci. Rep.

19902

(

2016

Saleh

and

A. R.

Oganov

Sci. Rep.

32486

(

2016

Abramson

Bollengier

, and

Brown

Sci. Rep.

821

(

2017

Mahajan

C. E.

Taylor

, and

G. A.

Mansoori

J. Pet. Sci. Eng.

(

2007

Pellenbarg

and

Max

J. Chem. Ed.

896

(

2001

B. A.

Buffett

Annu. Rev. Earth Planet. Sci.

477

(

2000

Takeya

Hori

Hondoh

, and

Udachin

J. Phys. Chem. B

104

4164

–

4168

(

2000

Sakemoto

Sakamoto

Shiraiwa

Ohmura

, and

Uchida

Cryst. Growth Des.

1296

(

2010

Saito

Kishimoto

Tanaka

, and

Ohmura

Cryst. Growth Des.

295

(

2011

D. K.

Staykova

W. F.

Kuhs

A. N.

Salamatin

, and

Hansen

J. Phys. Chem. B

107

10299

(

2003

J. S.

Loveday

R. J.

Nelmes

Guthrie

D. D.

Klug

, and

J. S.

Tse

Phys. Rev. Lett.

215501

(

2001

J. S.

Loveday

and

Nelmes

Phys. Chem. Chem. Phys.

937

(

2007

R. T.

Downst

and

M. S.

Somayazulu

Acta Crystallogr., Sect. C: Cryst. Struct. Commun.

897

(

1998

Kamb

Science

150

205

(

1965

C. A.

Tulk

Machida

D. D.

Klug

Guthrie

, and

Molaison

J. Chem. Phys.

141

174503

(

2014

Roettger

Endriss

, and

Ihringer

Acta Cryst. B

644

–

648

(

1994

Londono

W. F.

Kuhs

, and

J. L.

Finney

Nature

332

141

(

1988

Machida

Hisako

Taro

Yoshitaka

, and

Takehiko

J. Chem. Phys.

129

224505

(

2008

Lukman

Kenichiro

, and

Hideki

Phys. Rev. B

144105

(

2010

Simon

and

Peters

Acta Crystallogr., Sect. B: Struct. Sci.

2750

(

1980

Mayer

and

Pletzer

J. Chem. Phys.

2939

(

1984

Mitterdorfer

Bauer

, and

Loerting

Phys. Chem. Chem. Phys.

19765

(

2011

Mitterdorfer

, “

Phase changes in CO2/H2O ices

,” Ph.D. thesis,

University of Innsbruck

2014

E. F.

Burton

and

W. F.

Oliver

Nature

135

505

(

1935

Loerting

Mitterdorfer

Bauer

Winkel

Seidl

P. H.

Handle

C. G.

Salzmann

Mayer

D. T.

Bowron

, and

Finney

Phys. Chem. Chem. Phys.

8783

(

2011

M. S.

Elsaesser

Kohl

Mayer

, and

Loerting

J. Phys. Chem. B

111

8038

(

2007

Mishima

L. D.

Calvert

, and

Whalley

Nature

310

393

(

1984

Kohl

Mayer

, and

Hallbrucker

Phys. Chem. Chem. Phys.

602

(

2001

Mishima

J. Chem. Phys.

100

5910

(

1994

Winkel

M. S.

Elsaesser

Mayer

, and

Loerting

J. Chem. Phys.

128

044510

(

2008

Bernstein

Cruikshank

, and

Sandford

Icarus

179

527

(

2005

Loerting

Kohl

Salzmann

Mayer

, and

Hallbrucker

J. Chem. Phys.

116

3171

(

2002

C. G.

Salzmann

Loerting

Kohl

Mayer

, and

Hallbrucker

J. Phys. Chem. B

106

5587

(

2002

The same process is visible in the 200 K case. However, due to a lower temperature this transition is slower.

Salzmann

Loerting

Klotz

Mirwald

Hallbrucker

, and

Mayer

Phys. Chem. Chem. Phys.

386

(

2005

Stern

and

Loerting

Sci. Rep.

3995

(

2002

Y. P.

Handa

Klug

, and

Whalley

Can. J. Chem.

919

(

1988

Traditionally, ice I was considered to exist in two well-defined crystalline forms at ambient pressure: stable hexagonal ice (ice Ih⁠) and metastable cubic ice (ice Ic⁠). However, it is becoming increasingly evident that what has been called cubic ice in the past does not have a structure consistent with the cubic crystal system. Instead, it is a stacking-disordered material containing cubic sequences interlaced with hexagonal sequences.46

T. C.

Hansen

M. M.

Koza

, and

W. F.

Kuhs

J. Phys.: Condens. Matter

285104

(

2008

Malkin

Murray

Salzmann

Molinero

Pickering

, and

Whale

Phys. Chem. Chem. Phys.

(

2015

Kuhs

Bliss

, and

Finney

J. Phys. Colloq.

(

C1-631

(

1987