Anomalous temperature-induced volume contraction in GeTe

Tapan Chatterji1 , C. M. N. Kumar2,3 and Urszula D. Wdowik4

1
Institut Laue-Langevin, B.P. 156, 38042 Grenoble Cedex 9, France
2
Jülich Centre for Neutron Science, Forschungszentrum Jülich, D-52425 Jülich, Germany
3
Neutron Sciences Directorate, Spallation Neutron Source,
Oak Ridge National Laboratory, Oak Ridge, TN 37831-6475, USA
4
Institute of Technology, Pedagogical University, Podchorazych 2, PL-30084 Kraków, Poland

The recent surge of interest in phase change materials GeTe, Ge2 Sb2 Te5 , and related compounds
motivated us to revisit the structural phase transition in GeTe in more details than was done before.
Rhombohedral-to-cubic ferroelectric phase transition in GeTe has been studied by high resolution
arXiv:1502.02848v1 [[Link]-sci] 10 Feb 2015

neutron powder diffraction on a spallation neutron source. We determined the temperature de-
pendence of the structural parameters in a wide temperature range extending from 309 to 973 K.
Results of our studies clearly show an anomalous volume contraction of 0.6% at the phase transition
from the rhombohedral to cubic phase. In order to better understand the phase transition and the
associated anomalous volume decrease in GeTe we have performed phonon calculations based on
the density functional theory. Results of the present investigations are also discussed with respect
to the experimental data obtained for single crystals of GeTe.

PACS numbers:

I. INTRODUCTION perature Tc depends on the sample stoichiometry and

carrier concentration [8]. Ferroelectric phase transition
Phase change materials possess unique properties that in GeTe was considered to be displacive in its origin
hold considerable promise for applications in data stor- [7, 8, 10, 11]. Recently, however, the displacive character
age [1]. They can be rapidly and reversibly switched be- of the rhombohedral-to-cubic phase transition in GeTe
tween amorphous and crystalline states, which differ sub- has been contested by Fons at al. [12] and Matsunaga
stantially in their properties. Recently materials such as et al. [13]. According to their studies the displacive na-
Ge2 Sb2 Te5 and Ag and In doped Sb2 Te have been discov- ture of this phase transition was due to the misinterpre-
ered to crystallize rapidly enough to enable competitive tation of the Bragg diffraction results as the structure
solutions for rewritable optical data storage. determination based only on the Bragg intensities gives
information about the average structure, but not about
The related binary material GeTe has drawn consid-
the system local structure. Information about the lo-
erable interest because of its higher crystalline temper-
cal structure can be obtained from the total scattering
atures and better data retention at high temperatures
data including the diffuse scattering up to a very high
compared to Ge2 Sb2 Te5 . Apart from the application in
Q value and from the pair-distribution function (PDF)
data storage, GeTe may have potential use as a ther-
analysis [14]. Such investigations has been performed
moelectric material [2]. These have motivated us to re-
on the X-ray diffraction data by Matsunaga et al. [13]
visit the high temperature ferroelectric phase transition
and lead to the conclusion that in the local scale the
in GeTe.
high temperature phase of GeTe still exhibits distinct
GeTe is a narrow band-gap semiconductor [3] and is short and long Ge–Te bonds, contrary to the conven-
ferroelectric at room temperature with a Curie temper- tional structure refinement with only the Bragg intensi-
ature of about 705 K. The low temperature ferroelectric ties considered which suggests that the high temperature
phase has a rhombohedraly distorted NaCl type crystal cubic phase reveals solely one type of the Ge–Te bond.
structure with the space group R3m [4–8]. Structural Additionally, the two distinct bond distances observed in
distortions involve relative displacement of the Ge and the local scale hardly change across the rhombohedral-to-
Te sublattices along the body cell diagonal and subse- cubic phase transition. Results of Matsunaga et al. [13]
quent rhombohedral shear deformation along [111] direc- were also supported by the EXAFS studies of Fons et al.
tion which changes the rhombohedral angle from its fcc [12]. These findings suggested that the phase transition
value of 60◦ to α. The Ge and Te atoms are six fold in GeTe is not displacive but order-disorder type. On
coordinated by each other with three shorter (2.83 Å) the other hand, the most recent lattice dynamical cal-
and three longer (3.15 Å) bonds. This is often described culations [11] based on density functional theory (DFT)
as Peierls distortion [9] due to reduced coupling between show that the rhombohedral-to-cubic phase transition in
the p-type orbitals that constitute basis of the bonding GeTe is indeed displacive in its origin and becomes driven
in GeTe. by the condensation of exactly three components of the
GeTe undergoes a ferroelectric phase transition in triply degenerate optical transverse soft phonon mode at
which the low temperature rhombohedral R3m structure the Brillouin zone center. Moreover, the displacive char-
transform to the cubic F m3̄m structure at high temper- acter of the phase transition in GeTe has been further
ature of about 600–700 K [7, 8]. The transition tem-
 2

supported by the recent electron and X-ray diffraction

three-fold rotation axis
studies as well as the Raman scattering experiments of
Polking et al. [15].
We have revisited the ferroelectric phase transition in
GeTe by the high resolution neutron powder diffraction
on a modern high power spallation neutron source and
determined the temperature variation of the lattice pa-
rameters, unit cell volume, positional parameters, and
bond distances more accurately and in much finer tem-
perature steps across the rhombohedral-to-cubic phase
transition than were done for this compound before [8].
Also, the DFT phonon calculations were performed to
better understand the phase transition and the associ-
ated anomalous volume contraction in GeTe.

FIG. 1: (Color online) Structure of the low temperature

II. EXPERIMENTAL AND CALCULATION phase of GeTe (space group R3m) shown in the pseudo-cubic
METHODS (thick black lines), hexagonal (dotted lines), and rhombohe-
dral (thick blue/gray lines) representations.
GeTe powder samples were obtained from Alfa Aesar.
The samples are claimed to be 99.999% pure and are
200 Mesh. We checked the samples by the X-ray pow-
der diffraction and found the presence of small amount
of GeO2 impurity. Neutron powder diffraction (NPD)
measurements were performed on the time-of-flight pow- units cells used in the two investigations as well as the
der diffractometer, POWGEN, located at the Spallation relationship between them. The low temperature struc-
Neutron Source at Oak Ridge National Laboratory. The ture of GeTe with the space group R3m can be expressed
data were collected with neutrons of central wavelength in the pseudo-cubic, rhombohedral or hexagonal crystal-
1.333 Å, covering d-spacing range from 0.42 to 5.4 Å. lographic representations as illustrated in Fig. 1. The
Approximately 5 g of GeTe sample was loaded in a vana- distorted rocksalt structure with the lattice parameter ac
dium container of 10 mm diameter and measured in a and angle α is related to hexagonal unit cell with the lat-
traditional ILL furnace within the temperature range of tice constants
√ a and c via the relation a = 2ac sin(α/2)
309–973 K. Structure refinement was carried out using and c = ac 3 + 6 cos α [22]. In the hexagonal repre-
the FullProf suite [16]. sentation the Ge and Te atoms occupy 3a(0, 0, x) and
The present theoretical studies use the DFT method 3a(0, 0, 1 − x) Wyckoff positions, respectively. In addi-
implemented in the vasp code [17] together with the tion, the distortion parameter ∆x = 0.25 − xGe = xTe −
direct method approach [18] to provide necessary in- 0.75 describes the relative shift of the Ge and Te sublat-
put data to calculate the temperature dependence of the tices from the values xGe = 0.25 and xTe = 0.75 which
mean-squared vibrational amplitudes (Uij ) of the Ge and are characteristic for the cubic GeTe structure. The de-
Te atoms in the low and high temperature phases of gree of distortion from the cubic NaCl-type structure is
GeTe within the harmonic theory. The Uij tensor is ob- also reflected in the deviation ∆α of the rhombohedral
tained from the calculated diagonal and off-diagonal par- angle α from the cubic value of 90◦ (∆α = 90◦ − α).
tial phonon densities of states [19]. The volume thermal
expansions for the low and high temperature structures We recall that Chattopadhyay et al. [8] performed
of GeTe are evaluated according the quasiharmonic ap- structural refinement of their single crystal data using
proximation (QHA) [20]. Details of the lattice dynamics the pseudo-cubic cell with the F 13m setting whereas we
calculations on GeTe system can be found in Ref. [11]. have done structure refinement of our NPD data in the
The pair-distribution functions of GeTe were obtained hexagonal setting of a rhombohedral cell. At 309 K,
using the PDFgui program [21]. the hexagonal cell is described by the lattice parame-
ters ah = 4.1651 Å and ch = 10.6704 Å, correspond-
ing to the pseudo-cubic cell with the lattice parameters
III. EXPERIMENTAL RESULTS
ac = 5.9818 Å and αc = 88.2615◦ . Although we re-
fined the powder diffraction data in the hexagonal cell
A. Crystal structure and unit cell transformation we choose the pseudo-cubic representation as the most
convenient to make comparison between the present re-
Before we discuss the present NPD results and com- sults and those of Chattopadhyay et al. [8]. The usage
pare them with the results of neutron diffraction exper- of the pseudo-cubic representations is also advantageous
iments on single crystals [8], we describe the different because α = 90◦ in the high temperature phase.
 3

FIG. 3: (Color online) Results of the Rietveld profile refine-

ment for the low temperature rhombohedral phase of GeTe
(a) and the high temperature cubic phase of GeTe (b). The
refinements were performed by taking into account the impu-
rity phase GeO2 . The orange and green small vertical lines
FIG. 2: (Color online) Temperature dependence of diffraction indicate the positions of the Bragg peaks corresponding to
diagrams for GeTe measured at wide d-range (a) and d-ranges the GeTe and GeO2 phases, respectively. The blue curves
corresponding respectively to the cubic (222)c (b), (220)c (c), at the bottom denote differences between the measured and
and (200)c (d) reflections. Note the phase transition at Tc = calculated intensities. The conventional Rietveld discrepancy
600 K, development of the rhombohedral distortions below Tc , parameters for the refinement (a) were Rp = 20.9%, Rwp =
and vanishing splitting of the diffraction peaks at the onset 21.3%, Re = 8.48% and χ2 = 6.323 and those for the refine-
of the rhombohedral-to-cubic phase transition. ment (b) were Rp = 26.6%, Rwp = 18.9%, Re = 11.9% and
χ2 = 2.509.

B. Neutron Powder Diffraction Results

these parameters discrepancy rather than agreement pa-
rameters because these parameters are larger when the
Figures 2 (a)–(d) show the temperature variation of the discrepancy (and not the agreement) is larger. It occurs
diffraction diagram represented by contour plots. The that the impurity phase does not affect the refined pa-
large d-range is presented in Fig. 2(a), while the d-ranges rameters neither of the low nor high temperature GeTe
corresponding to the cubic (222)c , (220)c , and (200)c phases. We performed such refinements for the data mea-
reflections are depicted in Fig. 2(c)-(d). These reflec- sured at several temperatures from ∼300 K to ∼1000 K.
tions are split due to distortions of the Peierls type be- The parameters obtained from these refinements will be
low Tc = 600 K. The splitting progressively diminishes described and discussed in the following section.
with increasing temperature and finally the double-peak
structure disappears at the onset of structural transfor-
mation. The single peak observed above 600 K is a direct
evidence of the cubic symmetry of GeTe. IV. DISCUSSION
Diffraction intensities of the GeTe samples have been
refined together with the impurity GeO2 phase. Re- The rhombohedral-to-cubic phase transition in GeTe
sults of the Rietveld refinements performed at 310 K is characterized by pronounced changes in the positional
(low temperature structure) and 923 K (high tempera- parameters xGe and xTe . The xGe (xTe ) increases (de-
ture structure) are shown in Fig. 3. The conventional Ri- creases) with increasing temperature. The initial increase
etveld discrepancy parameters for the refinement (a) were (decrease) in xGe (xTe ) is almost linear up to ∼ 500 K and
Rp = 20.9%, Rwp = 21.3%, Re = 8.48% and χ2 = 6.323 becomes non-linear while approaching the transition tem-
and those for the refinement (b) were Rp = 26.6%, Rwp = perature Tc = 600 K. One observes sudden raise (drop)
18.9%, Re = 11.9% and χ2 = 2.509. The suffixes p, wp of xGe (xTe ) at Tc . Above Tc , the xGe and xTe take on
and e mean profile, weighted profile and expected from the values characteristic for the cubic GeTe phase. These
the counting statistics, respectively. We preferred to call changes are also revealed by ∆x and ∆α, see Fig. 4, as the
 4

0.015 rameters of the phase transition in GeTe [8]. The contin-

(a) uous shift of the Ge and Te atomic positions ∆x which
is directly related to the amplitude of the soft-phonon
mode Γ− 4 (Q) [11] breaks the symmetry elements cor-
0.010
responding to the four-fold rotation axis and a mirror
plane perpendicular to this axis. It also destroys the
∆x

symmetry inversion center of the high temperature cubic

phase and induces polarization along the three-fold ro-
0.005
tation axis. Thus, the primary order parameter ∆x (Q)
is the driving force of ferroelectricity appearing in GeTe
below Tc . On the other hand, the ∆α is actually the
lattice strain  which breaks only the symmetry elements
0.000
corresponding to the four-fold rotation axis and a mirror
300 400 500 600 700 800 900 plane perpendicular to it, but it does not affect the sym-
T (K)
metry inversion center of the GeTe system. The neutron
2.00 diffraction experiments on both powder and single crystal
(b) samples show the linear coupling between ∆α and ∆x2 ,
which conforms to the Landau theory [24]. The single
1.50 crystal data [8] display, however, slightly smaller slope
(0.89 × 10−4 deg−1 ) than the present NPD data for pow-
der samples (1.0 × 10−4 deg−1 ). On the other hand, our
Δα (deg.)

1.00 theoretical calculations give the linear coupling between

∆α and ∆x2 of 1.2 × 10−4 deg−1 .
The short (s) and long (l) bond lengths determined
0.50 from our neutron measurements on powder samples
amount to s = 2.82 Å and l = 3.18 Å at room tem-
perature. They are very close to those obtained from the
0.00 neutron diffraction studies on single crystal samples [8].
300 400 500 600 700 800 900
The s and l bonds in the low temperature rhombohe-
T (K) dral phase follow the course of xGe and xTe . They vary
smoothly with temperature to reach the average value of
∼ 3 Å above Tc = 600 K, as shown in Fig. 5. This unique
FIG. 4: (Color online) Temperature variations of (a) distor-
distance is certainly the Ge–Te bond length in the cubic
tion parameter ∆x and (b) deviation ∆α of the distortion
angle from 90◦ . The present NPD data (circles) are com-
phase which subsequently slightly grows at still higher
pared to the neutron diffraction experiments performed on temperatures due to the thermal expansion of the cubic
single crystals by Chattopadhyay et al. [8] (squares). GeTe lattice. Nevertheless the visible shift between the
The continuous lines are just guides to the eye. single crystal neutron diffraction data and those mea-
sured on powder samples which arises from a difference
in the respective transition temperatures, there is a good
distortion parameters are directly related to xGe and xTe . qualitative agreement between these two sets of data.
Both ∆x and ∆α decrease continuously from ∆x = 0.14 Such a behavior as revealed by the Ge–Te bond lenghts
and ∆α = 1.98◦ at 300 K to zero at Tc = 600 K. The in both single crystal and powder diffraction studies is
∆x and ∆α obtained for single crystals of GeTe [8] show however claimed [12] to be the case only in the so-called
very similar behavior as that observed for powder sam- average structure. The PDF analysis of the total diffrac-
ples. They also progressively diminish with temperature, tion data on GeTe shows that the high temperature phase
however with smaller slopes than those determined for exhibits two distinct bond lengths which hardly change
our powder samples, and finally they approach zero val- across the phase transition [13]. The local scale structural
ues at 700 K. The ∆x = 0 and ∆α = 0 indicate that the distortions, evidenced by unequal Ge–Te bond distances
R3m structure undergoes transformation into the F m3̄m above Tc , suggested that the transition in GeTe could be
structure at Tc . Here we note the difference in the transi- of the order-disorder type [12, 13]. To obtain additional
tion temperature Tc between our powder samples and the information about the local structure of GeTe above Tc ,
single crystal GeTe samples investigated by Chattopad- we have simulated the PDF spectra of its high tempera-
hyay et al. [8]. Such a significant spread in transition ture phase with the static structural lattice distortions
temperature of GeTe compound has been known and is generated according to the triply degenerate unstable
usually attributed to the sample stoichiometry as well as soft-phonon mode of Γ− 4 symmetry [11]. This mode,
the free charge carrier concentration [23]. while frozen, leads to relative displacements of the Ge
Distortions ∆x and ∆α has been considered respec- and Te sublattices along the cubic cell diagonal. Results
tively as the primary (Q) and secondary () order pa- of these calculations for perfectly ordered (∆x0 = 0) and
 5

3.20 5

4 Δx0
3.15
Δx1
3 Δx2
3.10

2
3.05

G(r) (Å-2)
s , l (Å)

1
3.00
0
2.95
−1
2.90
−2
2.85
−3
2.80
−4
300 400 500 600 700 800 900
5 10 15 20 25 30
T (K)
r (Å)
FIG. 5: (Color online) Temperature dependences of short (s)
and long (l) Ge–Te bonds determined in the present NPD FIG. 6: (Color online) Pair distribution function G(r) cal-
measurements (circles) and those provided by the single crys- culated for the cubic GeTe phase with the static distortions
tal neutron diffraction [8] (squares). The continuous lines are ∆x0 = 0, ∆x1 = 7.94 × 10−3 , and ∆x2 = 14.56 × 10−3 .
just guides to the eye. Simulation are performed at the experimental r-range and
Qmax = 15 Å−1 [13].

disordered (∆x1 = 7.94 × 10−3 and ∆x2 = 14.56 × 10−3 ) 221

2
high temperature cubic phases of GeTe are shown 2 in
220
Fig. 6. The distortions ∆x1 and ∆x2 correspond 2 re-
2
spectively to the Ge–Te bond lengths of (s1 = 2.91 2 Å, 219
l1 = 3.10 Å) and (s2 = 2.84 Å, l2 = 3.19 Å). Here we
218
note that ∆x2 also reflects the Ge–Te bond lengths deter-
V (Å3)

mined above Tc by using the experimental PDF data [13]. 217

We observe that the simulated spectra of cubic GeTe with
216
and without static distortions are essentially identical.
Moreover, they are closely related to the experimental 215 ∆V = 0.6%
PDF data reported by Matsunaga et al. [13]. This find-
ing enables us to suggest that the PDF results can hardly 214
Tc = 600 K
provide a definite answer about the displacive or order- 213
disorder type of the phase transition in GeTe compound 300 400 500 600 700 800 900
since they probe the average static lattice distortions but T (K)
not a dynamical nature of this transition connected with
the phonon dynamics [11]. FIG. 7: (Color online) Temperature dependence of the GeTe
Figure 7 indicates that both powder and single crys- pseudo-cubic unit cell volume (V ). The present neutron pow-
tal samples of GeTe exhibit temperature-induced vol- der diffraction results (circles) are compared to the single crys-
ume reduction ∆V at the ferroelectric rhombohedral-to- tal neutron diffraction data obtained by Chattopadhyay et al.
cubic phase transition. The unit cell volume decreases [8]. Solid lines denote results of the DFT-QHA calculations
by ∆V ≈ 0.6% at Tc = 600 K, as indicated by our NPD for the low and high temperature phases of GeTe.
data. The volume anomaly at the ferroelectric structural
transition in GeTe can be considered as an electrostric-
tion effect in analogy with magnetostriction effects at close vicinity of Tc , where the QHA approach is unable to
the magnetic ordering. Also, the volume contraction at reproduce the volume reduction at the onset of the phase
Tc has been interpreted by Chattopadhyay et al. [8] as transformation. Our calculations indicate that the vol-
the extra volume due to the presence of lone pairs in ume of the rhombohedral phase is higher than the volume
the rhombohedral phase and the absence of this excess of the cubic phase. Indeed, in the rhombohedral struc-
volume in the high temperature cubic phase. Results ture the Ge and Te atoms are displaced from each other
of the DFT and QHA calculations performed for both with respect to their position in the cubic structure. The
the low temperature rhombohedral and high tempera- relative shift of the Ge and Te sublattices along the body
ture cubic structures of GeTe support to some extent the cell diagonal (∆x) is accompanied by angular distortion
present experimental data. The agreement between cal- of the GeTe lattice (∆α) as well. Both distortions re-
culations and experiment remains reasonable, except the sult in the larger volume of the rhombohedral structure
 6

in comparison with the rocksalt one. These distortions 6

are dynamical in their origin as they are driven by the Δx = 0
2
soft phonon mode Γ− 4 , as shown and discussed 4 in our 5 Δx > 0 Ge
recent paper [11]. The DFT calculations show 22that the
rhombohedrally distorted GeTe is more energetically4 sta- 4
R3 ge
ble than the undistorted cubic one. At the ground R3 te
state

B (Å2)
the difference in their Helmholtz free energies 2amounts 3
to 26 meV per formula unit, i.e., it lies in the 22range of
Ge
thermal excitations. The present experiments also 2 show
2 2
that each of GeTe phases expands upon heating2 outside
the temperature range where the phase transition 2 occurs. Te

The volume thermal expansion coefficient of the rhombo- 1

hedral GeTe equals 4.59 × 10−5 K−1 at 300 K, whereas Te

it amounts to 7.67 × 10−5 K−1 at 650 K for the cubic 0
GeTe. 300 400 500 600 700 800 900
T (K)
The phase-change material GeTe is not a unique sys-
tem undergoing temperature-induced volume collapse at
the phase transition. The well-known example is the FIG. 8: (Color online) Experimental (symbols) and theoreti-
volume contraction of ice at its melting temperature. cal (lines) isotropic temperature factors B determined for Ge
This phenomenon remains, however, unexplained quanti- (circles) and Te (squares) atoms in the rhombohedral and cu-
bic phases of GeTe. The solid lines denote calculated B fac-
tatively so far. The temperature-induced volume reduc-
tors for the rhombohedral and ideal cubic GeTe structures.
tion is also found to exist in solid-to-solid phase transi- Dashed lines correspond to calculated B factors for the cu-
tions in diverse condensed matter systems. Even earlier- bic GeTe structure with static distortion ∆x = 14.56 × 10−3
known industrially important canonical ferroelectric ma- (s = 2.84 Å, l = 3.19 Å). Calculations are performed within
terial like BaTiO3 shows such volume decrease at its the harmonic approximation. Theoretical isotropic tempera-
ferroelectric phase transition at 393 K [25]. One no- ture factors are evaluated as B = 8π 2 hU i, where hU i is the
table example is the orbital order-disorder transition or trace of Uij tensor [20].
orbital melting in the strongly correlated electron sys-
tem LaMnO3 [26] – the recognized parent compound of
colossal magnetoresistive manganites that lure a host of gle parameter. Our calculations show that the thermal
condensed matter scientists for spintronics and other de- motions of Ge and Te atoms in the rhombohedral GeTe
vice applications. Reduction of the LaMnO3 volume at are almost isotropic as Uxx Ge
≈ UzzGe Te
and Uxx ≈ UzzTe
.
the orbital order-disorder transition is assisted by the Similar dependences were also observed in the neutron
considerable change in the atomic mean-squared vibra- diffraction study on single crystals of GeTe [8]. The
tional amplitudes [26]. Very similar effect is observed in nearly isotropic Uij tensor arises from very small differ-
the present studies on GeTe too. The isotropic temper- ence between the atomic on-site force constants along
ature factors BGe and BTe depicted in Fig. 8 increase and perpendicular to the three-fold rotational axis in the
with increasing temperature and show λ-type behavior rhombohedral GeTe. This negligible difference remains
in the close vicinity of the rhombohedral-to-cubic phase in accordance with the structural features of the rhom-
transition. The BGe and BTe measured for single crys- bohedral phase for which the structural distortions of the
tals of GeTe [8] are much more scattered compared to Peierls-type are indeed small. The mean-squared vibra-
the present data. They are also limited to temperatures tional amplitudes are obviously different for the Ge and
not exceeding Tc and therefore they do not reveal such a Te atoms due to the difference in their masses. Also, they
characteristic change at Tc as BGe and BTe in our exper- exhibit typical increase with increasing temperature ac-
iments. cording to the applied harmonic approximation with the
It is well know that the atomic temperature factors Uij growing their values over two times between 300 and
remain sensitive to the immediate atomic environment. 600 K. The harmonic approach is, however, unable to de-
Experimental atomic thermal displacements contain both scribe properly the anomalous behavior of Uij in the close
static and dynamic effects. The dynamic contribution vicinity of Tc . Nevertheless, it occurred to be helpful in
to the atomic thermal displacements is gained from the explaining the behavior of atomic thermal vibrations in
present DFT calculations. The Uij tensor for both Ge the cubic phase of GeTe. We notice that the slope of BTe
and Te atoms in the R3m structure contains two indepen- practically does not change it course with increasing tem-
dent components, namely Uxx = Uyy and Uzz which rep- perature, except the tiny range of temperatures around
resent the atomic thermal vibrations perpendicular and Tc = 600 K when it shows small discontinuity. There is,
parallel to the three-fold rotational axis, respectively (cf. however, considerable jump in the thermal vibrations of
Fig. 1). The site symmetries of the Ge and Te atoms in the Ge atoms in the cubic GeTe phase as well as visi-
the F m3̄m structure constrain their mean-squared dis- ble change in the slope of BGe above Tc . The results of
placements to be isotropic and hence described by a sin- our calculations show the the on-site force constants of
 7

Ge atoms in the cubic GeTe are almost two times lower more accurate determination of the temperature depen-
compared to the respective on-site force constants in the dences of the lattice and structural parameters of GeTe
rhombohedral GeTe structure, i.e., the Ge atoms in the compared to the previous studies [8]. The measured vari-
cubic phase are more loosely bound inside the lattice ation of the GeTe volume over a wide temperature range
which in turn allows for the larger values of the ampli- is however a robust result of Bragg diffraction and the
tudes of their mean-squared displacements. The mean- volume anomalous behavior at the phase transition is the
ingful decrease in the on-site force constants of the Ge most important result. The temperature variation of the
atoms while going from the rhombohedral to cubic GeTe structural parameters across the rhombohedral-to-cubic
accounts for the increase in BGe slope above Tc = 600 K. phase transition provides us microscopic mechanism be-
It is interesting to note that the cubic phase of GeTe with hind the volume discontinuity and enables us to remove
incorporated static lattice distortions of Peierls-type ∆x the veil of mystery around this transition.
shows higher B-factors in comparison with the respec- It is interesting to note that the interpretation of the
tive factors for the ideal cubic structure (∆x = 0). In local probes results [12, 13] viz. the PDF analysis of the
addition, the effect of ∆x > 0 is much more pronounced total diffraction intensities or EXAFS, give an impres-
for BGe than for BTe . Our calculations indicate that sion that apart from a small linear thermal expansion
progressive increase of ∆x results in the gradual growth nothing happens during the phase transition, namely the
and small changes in slopes of B-factors due to modi- short and long Ge–Te bond lengths do not change within
fied force constants which decrease with decreasing ∆x. experimental errors in the whole wide temperature range
In principle, this observation allows to distinguish the investigated (300–800 K). One also wonders whether a
high temperature GeTe structure with persisting static hypothetical static model compatible with the bond dis-
lattice distortions (retained distinct short and long Ge– tances obtained by the local probes would reproduce the
Te bonds) above the phase transition from that with the robust result of volume contraction at the rhombohedral-
local distortions vanished (unique Ge–Te bond length). to-cubic phase transition in GeTe. Perhaps some kind
of a dynamical model, though definitely difficult to con-
struct, might reproduce the volume decrease. Hence, an
V. SUMMARY AND CONCLUSIONS answer to the question whether transition is of displacive
or order-disorder type lies in the dynamics of phase tran-
Chattopadhyay et al. [8] carried out high tempera- sition and the role of soft modes, as has been recently
ture single crystal neutron diffraction investigations on shown by the DFT and phonon calculations [11]. The
GeTe almost three decades ago when the powder neutron conventional PDF analysis of the total scattering data
diffraction technique was not very mature. The present does not analyze the energy and therefore it lacks an in-
NPD investigations on GeTe have been carried out on a formation about dynamics of the phase transition. The
modern high resolution powder diffractometer at a spal- high temperature phase contains the dynamics of broken
lation neutron source and have benefited from the enor- symmetry phase and the PDF local probe just sees the
mous progress made during recent years in the neutron snap-shot of low frequency soft phonon mode that hap-
powder diffraction technique. The high temperature sin- pened to enter into the window of local probe. The two
gle crystal neutron diffraction is very time consuming and distinct Ge–Te bond lengths seen by the local probes at
requires temperature stability for long periods whereas the high temperature phase are just dynamical signature
neutron powder diffraction is relatively free from these of the low temperature phase still persisting at temper-
constraints. The extinction effects can be very large in atures exceeding Tc . Since the PDF measures instan-
single crystal diffraction whereas it is often practically taneous structure and cannot distinguish between static
negligible in powder diffraction. Also, the NPD technique and dynamic correlations, interpretation of the PDF re-
appears much attractive in describing the phase transi- sults based solely on the static bond distances could be
tions in relatively simple structures like GeTe due to its verified.
less sensitivity to the crystal domain structure which sig-
nificantly complicates both the data collection and the
data treatment. VI. ACKNOWLEDGMENTS
The results of our NPD measurements on GeTe are
based on Rietveld refinement involving only Bragg in- The research conducted at SNS was sponsored by
tensities. An information contained in the background the Scientific User Facilities Division, Office of Basic
diffuse scattering from the sample is not taken into ac- Energy Sciences, US Department of Energy. Interdis-
count, and hence our results are related to the average ciplinary Center for Mathematical and Computational
structure, but not to the local structure or dynamics. Modeling (ICM), Warsaw University, Poland and the
Although the current experimental results could not un- IT4Innovations National Supercomputing Center, VSB-
ambiguously resolve controversy about the nature of the Technical University, Ostrava, Czech Republic are ac-
phase transition in GeTe, viz. whether this phase transi- knowledged for providing the computer facilities under
tion is of displace or order-disorder type, they allowed for Grants No. G28-12 and Reg. No. CZ.1.05/1.1.00/02.0070.
 8

[1] For a review see M. Wuttig and N. Yamada, Nature tern matching and analysis program version 2010, ILLB,
Mater 6, 824 (2007) and the references therein. CEA-CNRS, France [[Link]
[2] E. M. Levin, M. F. Besser and R. Hanus, J. Appl. Phys. [17] G. Kresse and D. Joubert, Phys. Rev. B 59, 1758 (1999);
114, 083713 (2013). G. Kresse and J. Furthmüller, Phys. Rev. B 54, 11169
[3] R. Tsu, W. E. Howard, and L. Esaki, Phys. Rev. 172, (1996); Comput. Mater. Sci. 6, 15 (1996).
779 (1968); Solid State Commun. 5, 167 (1967). [18] K. Parlinski, Z. Q. Li, and Y. Kawazoe, Phys. Rev. Lett.
[4] K. Schubert and H. Fricke, Z. Naturforsch. 6a, 781 78, 4063 (1997); K. Parlinski, software PHONON ver.
(1951). 6.14 (Krakow, Poland, 2013).
[5] K. Schubert and H. Fricke, Z. Metallk. 44, 457 (1951). [19] U. D. Wdowik, K. Parlinski, T. Chatterji, S. Rols, and
[6] J. Goldak, C. S. Barrett, D. Innes, and W. Youdelis, J. H. Schober, Phys. Rev. B 82, 104301 (2010).
Chem. Phys. 44, 3323 (1966). [20] D. C. Wallace, Thermodynamics of Crystals (Wiley, New
[7] E. Steigmeier and G. Harbecke, Solid State Comm. 8, York, 1972).
1275 (1970). [21] P. Juhás, T. Davis, C. L. Farrow, and S. J. L. Billinge,
[8] T. Chattopadhyay, J. X. Boucherle and H. G. von Schner- J. Appl. Crystallogr. 46, 560 (2013); C. L. Farrow, P.
ing, J. Phys. C 20, 1431 (1987). Juhás, J. W. Liu, D. Bryndin, E. S. Božin, J. Bloch, Th.
[9] R. E. Peierls, Quantum theory of solids (Oxford Univer- Proffen, and S. J. L. Billinge, J. Phys.: Condens. Matter
sity Press, Oxford, 1955). 19, 335219 (2007); Th. Proffen and S. J. L. Billinge, J.
[10] K. M. Rabe and J. D. Joannopoulos, Phys. Rev. Lett. Appl. Crystallogr. 32, 572 (1999).
59, 570 (1987); Phys. Rev. B 36, 6631 (1987). [22] M. O’Keefee and B. G. Hyde, Crystal structures, Miner-
[11] U. D. Wdowik, K. Parlinski, S. Rols, and T. Chatterji, alogical Society of America, Washington (1996).
Phys. Rev. B 89, 224306 (2014). [23] F. Tong, X. S. Miao, Y. Wu, Z. P. Chen, H. Tong, and
[12] P. Fons, A. V. Kolobov, M. Krbal, J. Tominaga, K. S. X. M. Cheng, Appl. Phys. Lett. 97, 261904 (2010); F.
Andrikopoulos, S. N. Yannopoulos, G. A. Voyiatzis and J. Espinoza-Beltrán, G. Trapage, J. App. Phys. 112,
T. Uruga, Phys. Rev. B 82, 155209 (2010). 052018 (2012); E. M. Levin, M. F. Besser, and R. Manus,
[13] T. Matsunaga, P. Fons, V. Kolobov, J. Tominaga and N. J. Appl. Phys. 114, 083713 (2013).
Yamada, Appl. Phys. Lett. 99, 231907 (2011). [24] R. A. Cowley, Adv. Phys. 29, 1 (1980); R. Clarke, Phys.
[14] For example see T. Egami and S. J. L. Billinge, Under- Rev. B 18, 4920 (1978).
neath the Bragg peaks: structural analysis of complex ma- [25] G. Shirane and A. Takeda, J. Phys. Soc. Japan 7, 1
terials (Pargamon, 2003). (1952).
[15] M. J. Polking, J. J. Urban, D. J. Milliron, H. Zheng, E. [26] T. Chatterji, F. Fauth, B. Ouladdiaf, P. Mandal and B.
Chan, M. A. Caldwell, S. Raoux, C. F. Kisielowski, J. Ghosh, Phys. Rev. B 68, 052406 (2003); U. D. Wdowik,
W. Ager III, R. Ramesh and A. P. Alivisatos, Nano Lett. B. Ouladdiaf, T. Chatterji, J. Phys.: Condens. Matter.
11, 1147 (2010). 23, 245402 (2011).
[16] J. Rodriguez-Carvajal, FULLPROF, a Rietveld and pat-