1. technology and computing

Thermal Properties of In-Sb-Te Thin Films for Phase Change
Memory Application
Huu Tan Nguyena,b, Andrzej Kusiaka, Jean-Luc Battagliaa, Cecile Gaborieaua, Yanick
Anguya, Roberto Fallicab, Claudia Wiemerb, Alessio lamperti, Massimo Longob.
a
Laboratory I2M, UMR CNRS 8503, University of Bordeaux, 351 cours de la Libération, 33405
Talence, France.
b
Laboratorio MDM, IMM-CNR, Unità di Agrate Brianza, Via C. Olivetti 2, 20864 Agrate Brianza,
(MB), Italy
Abstract: Phase change memories (PCM) are typically based on compounds of the Ge-Sb-Te (GST)
ternary system. Nevertheless, a major drawback of PCM devices is the failure to fulfill automotivelevel or military-grade requirements (125°C continuous operation), due to the low crystallization
temperature of GST. To overcome this limitation, alloys belonging to the In-Sb-Te (IST) system have
been proposed, which have demonstrated high crystallization temperature, and fast switching. Thermal
properties of the chalcogenide alloy and of its interfaces within the PCM cell can influence the
programming current, reliability and optimized scaling of PCM devices. The Modulated Photothermal
Radiometry (MPTR) technique was implemented to measure the thermal conductivity of IST thin
films as well as the thermal boundary resistance at the interface with other surrounding materials (a
metal and a dielectric). The experiment was carried out in situ from room temperature up to 550oC in
order to investigate the intrinsic thermal properties at different temperatures and the significant
structural rearrangement upon the phase transition. Two different stoichiometries for the IST ternary
alloy were deposited by Metal Organic Chemical Vapor Deposition (MOCVD) on a Si substrate
covered with thermal SiO2 and then capped with a Platinum layer that acts as an optical and thermal
transducer. Additional data from Raman and XRD lead to complementary analysis.
Keywords: Phase change memory, In-Sb-Te ternary, Thermal conductivity, Thermal boundary
resistance.
1.
Introduction
Among many emerging non-volatile memories (NVM) technologies, phase change memory (PCM)
has been considered as an interesting alternative to Flash technology in terms of both performance and
scalability perspectives [1, 2]. From a chemical viewpoint, PCM based on semiconducting glasses
belonging to the IV, V, VI of the periodic table, can be switched rapidly from high resistance
(disordered amorphous phase) to a low resistance (ordered crystalline phase) by heating the material
above its crystallization temperature. The switching back to the amorphous phase is achieved by
melting and quenching the material fast enough that it solidifies in the amorphous state. Since it shows
some suitable characteristics, Ge2Sb2Te5 (GST) has emerged as the most common and most studied
phase change material [3]. Up to date, GST composition and its forms modified with nitrogen, oxygen,
silicon, and antimony, among others, have been intensively studied due to its fast phase transformation
and reliable stability of the amorphous phase. However, GST faces some drawbacks: i) a high melting
temperature (~620 °C), causing a high reset current and easily inducing thermal cross-talk problems in
high density memory arrays and ii) a low crystallization temperature (around 160°C) , therefore a
relatively low stability of the amorphous phase. Phase change materials with a high crystallization
temperature enable the fulfillment of automotive level or military-grade requirements; to address this
issue, the In-Sb-Te ternary system has been proposed, showing a broadly changeable characteristic
owing to their composition [4-7]. From previous literature reports, higher indium content has been
demonstrated to increase the crystallization temperature by hindering the crystalline growth. Besides,
these alloys exhibit several combinations of some intermediate binary and ternary phases at different
temperatures; as it is the case for the stoichiometric ternary alloy In3Sb1Te2 (IST), having the ability to
hold multi bits in a single cell [8]. In addition, In-Sb-Te shows the same space group number (S.G.
#225, Fm-3m) as GST as well as nearly the same lattice parameter (IST: a=6.12 Å, GST: a=6.03 Å)
[9].
This paper presents the temperature dependent thermal conductivity measurement of In-Sb-Te thin
films. Due to the low thickness of the chalcogenide material in real memory cells, the value of the
thermal boundary resistance (TBR) at the interface between the In-Sb-Te and neighbor materials
involved in the PCM cell is comparable to the one of the thermal resistance of the In-Sb-Te itself that
is the ratio between its thickness (in the direction of the heat flow) and its thermal conductivity.
Therefore, accurate measurement of the TBRs is also required to address the heat transfer analysis
within the PCM device. Moreover, it must be noted that the TBR improves the heating efficiency of
each PCM cell by increasing its thermal insulation, therefore avoiding the thermal cross-talk between
neighbor cells and allowing the cell scalability.
2.
Experimental details
In-Sb-Te films were grown by Metal Organic Chemical Vapor Deposition (MOCVD) technique at
260oC, 104 Pa, on a Si(100) wafer 550 µm thick, covered with SiO2 ( eSiO = 50 nm ) obtained by
2
thermal oxidation of the substrate (see Fig. 1). Electronic grade solution trimethylindium In(CH3)3,
trisdimethylaminoantimony (N(CH3)2)3Sb and diisopopyltelluride (C3H7)2Te were used as In, Sb and
Te metal-organic precursors, respectively. We obtained two compositions of In-Sb-Te: In3Sb2.7Te0.8
(A) and In3Sb2.5Te1.2 (B), corresponding to a Te content of 12 at.% and 17 at.%, respectively. The
thermal characterization required having different In-Sb-Te film thicknesses eIST . Therefore four
thicknesses were deposited for the two stoichiometries: In3Sb2.7Te0.8 (30, 50, 65, 90 nm) and
In3Sb2.5Te1.2 (30, 50, 70, 105 nm). X-ray diffraction (XRD) analysis demonstrated that as-deposited InSb-Te layers are formed by a mixture of polycrystalline face centered cubic InSb0.85Te0.15 phase and a
Te-rich amorphous component [10, 11].
Fig. 1 - Layout of the sample for the MPTR characterization
Modulated photo thermal radiometry (MPTR) was used to measure the In-Sb-Te film thermal
conductivity kIST . This technique consists in applying a periodic heat flux at the surface of the sample
to mesure the phase lag and the amplitude between this thermal excitation and the temperature
variation on the heated area (front face technique). This relative variation is measured using an
infrared (IR) HgCdTe detector that collects the emitted radiation by the heated surface. A lock-in
amplifier with both signals at the input allows measuring the phase and the amplitude. The
temperature of the sample is controlled by a furnace working in inert atmosphere (Ar or N2). The
sample was heated at a rate of 25 °C/min and annealed for 2 min at the required temperature before
starting the measurement (duration of the experiment characterization is about 4 hours).
A laser of 1064 nm wavelength generates the surface periodic heat source. The laser beam is
modulated by an acousto-optic modulator using the square signal issued from a function generator.
The laser beam has a Gaussian profile, with a diameter of the spot at the sample surface equal to 3
mm. In order to avoid the phase lag due to the acousto-optic modulator driver, a nanosecond time rise
photodiode is used to measure the reference signal relative variation. The IR detector wavelength
detection limits fall between 5 and 12 µm. Since the detector wavelength operating range is higher
than that of the laser, the measurement is not disturbed by the photonic source; moreover, an optical
filter with 1.6 µm cut-off wavelength is used to reject all the VIS and NIR radiation arriving on the IR
detector. The zone viewed by the detector is the image of the infrared sensitive element on the sample
corresponding to a circle of 0.5 mm in diameter. Since the generating heat flux at the surface comes
from a laser source, the MPTR technique requires controlling the heat flux absorption. Therefore, a 50
nm Pt layer is deposited on top of the In-Sb-Te thin films to act as an optical transducer for the laser
absorption (at least 30% in the laser wavelength) at the surface (In-Sb-Te extinction coefficient is low
in the crystalline phase [3-6]). On the other hand, the Pt layer is intended to avoid a possible oxidation
and evaporation of the In-Sb-Te at high temperatures. The frequency range swept during the
experiment is [754-21123] Hz. The mean free path of heat carriers (electrons and phonons) in the InSb-Te is less than the minimal investigated film thickness. Therefore, the heat diffusion model
describing the heat transfer in the sample during the experiment is based on the classical Fourier
relation connecting the heat flux and the temperature gradient within the material. This model enables
the calculation of the theoretical phase and amplitude as a function of frequency. For all the
experienced frequencies f the heat penetration depth, which is a π f (where a is the thermal
diffusivity of the material) was higher than the In-Sb-Te and SiO2 film thickness. On the other hand,
within the whole frequency range, the Si substrate behaves as a semi-infinite medium. Therefore, these
two layers behave as pure thermal resistances as: RIST = eIST kIST and RSiO = eSiO kSiO , respectively.
2
The thermal conductivity of amorphous SiO2 is kSiO = 1.45W.m .K
-1
2
o
-1
2
2
and it does not significantly
o
vary in the measured 100 C to 550 C range. On the other hand, the Pt layer thermal resistance (ePt/kPt)
can be neglected with respect to the two previous ones as Pt is assumed to be thermally thin. We call
i
i
i
+ RIST-SiO
+ RISiO
is the
R = RIST + RSiO + Ri the thermal resistance of the stack, where Ri = RPt-IST
-Si
2
2
2
sum of the thermal boundary resistances at the interfaces between the three layers. Since the thermal
properties as a function of temperature of the Si substrate are well known, only R has to be identified.
This is achieved by minimizing the quadratic gap between the theoretical and the experimental data
(phase and amplitude). The Levenberg-Marquardt algorithm is used to perform the minimization.
The contact resistance at the interface between Si and thermal SiO2 has been measured by [12] as:
i
RSiO
= 4.5 × 10−9 K.m 2 .W -1 .
-Si
2
The thermal resistance at the interface between In-Sb-Te and SiO2 and that between Pt and In-Sb-Te
can be estimated starting from the diffuse mismatch model (DMM) asymptotic formulation at high
(
i
temperature [13-15]: R1−2
(T ) = 4 τ 12 v1 ρ1 (T ) C p (T )
1
)
−1
, where 1 and 2 denote the two materials
forming the interface. In this relation, material 1 is such that Θ D,1 < Θ D,2 , where Θ D denotes the
Debye temperature. In addition, v1 , ρ1 and C p are the sound velocity, the density and the specific
1
heat of material 1, respectively. The phonon transmission coefficient τ 12 is calculated using the Debye
approximation as: τ 12 = v2 −2
(v
−2
1
)
+ v2 −2 .
ρ (T )
c p (T )
⎡ kg.m -3 ⎤
⎣
⎦
⎡ J.K -1.kg -1 ⎤
⎣
⎦
Θ D ⎡⎣ K ⎤⎦
vl ⎡⎣ m.s-1 ⎤⎦
vt ⎡⎣ m.s-1 ⎤⎦
@300K
@300K
21450@300K
130@300K
240
4174
1750
Pt
23820@800K
144@800K
(ref. [13])
(ref. [21])
(ref. [21])
(ref. [13])
(ref. [13])
6310@300K
210@300K
190
3100
1900
In3Sb2.7Te0.8
(ref. [10])
(In3Sb1Te2)
(In3Sb1Te2)
(In3Sb1Te2)
(In3Sb1Te2)
2200@300K
787@300K
552
6633
4038
SiO2
2650@800K
1230@800K
(ref. [20])
(ref. [17])
(ref. [17])
(ref. [19, 20])
(ref. [19, 20])
Tab. 1 - Thermophysical properties for Pt, In-Sb-Te (calculated from phonon DOS of In3Sb1Te2 [10])
and SiO2.
The density In3Sb2.7Te0.8 has been measured and the specific heat, the Debye temperature and the
sound velocity for crystalline In-Sb-Te are calculated starting from the phonon DOS obtained through
DFT simulation on In3Sb1Te2 reported by [22]. Data for Pt and SiO2 are found in the literature and
reported
in
Tab.
1.
Thus
we
i
found RPt-IST
= 2.93× 10−9 K.m 2 .W -1
and
i
RIST-SiO
= 7.29 × 10−9 K.m 2 .W -1 , which led to Ri = 1.47 × 10−8 K.m 2 .W -1 .
2
3.
Results and discussion
Measurements were performed from room temperature up to 550oC. The thermal resistance for the
two stoichiometries at each investigated temperature was identified. A linear regression was obtained
at each temperature, leading to the In-Sb-Te thermal conductivity kIST as well as to the total thermal
boundary resistance R i according to the expression of R. Values for the two In-Sb-Te compositions
are reported in Fig. 2a and 2b for the thermal conductivity and total boundary resistance, respectively.
The values obtained for the as-deposited IST using the 3ω method are also reported in the figure [23].
As expected, Fig. 2a shows that the two compositions present very close values for the thermal
conductivity. In addition, they are also very close to the values found for the Ge2Sb2Te5 alloy [16].
During thermal annealing, the thermal conductivity monotonically increased owing to the on- going
crystallization of the materials. A more pronounced increase occurred at 350 °C for the low- Te film
and at 400 °C for the high-Te film. These two temperatures are lower than that found by [10] where
the structural and electrical characterization jointly proved that these compounds exhibit a metastable
behavior at above 464 °C, when the stoichiometric phase is formed and the InSb0.8Te0.2 phase
temporarily disappears. With respect to the error bars, the data fluctuated rather significantly between
400oC and 450oC. Going back to room temperature (RT) after 550oC led to stable values for the
thermal conductivity. As showed in Fig. 2b, the total thermal boundary resistance decreases when
temperature increases, which agrees with the theoretical prediction related to phonon anharmonic
scattering. In returning to RT a high TBR value was measured at 10−7 K.m 2 .W -1 . This value is ten
times higher than the theoretical value for R i previously found.
a)
Fig. 2 - a) In-Sb-Te thermal conductivity kIST
b)
(comparison with GST [16]) and b) total thermal
resistance R i for the two compounds as a function of temperature (blue filled square: comparison with
theoretical value calculated using the DMM).
a)
b)
Fig. 4 - XRD patterns of In-Sb-Te samples a) as-deposited and b) annealed at 550°C.
XRD measurements on the as-deposited and annealed at 550°C In-Sb-Te samples are reported in
figures 4a) and 4b), respectively. It appears that after annealing the only composition that allows
retrieving the data for the annealed sample is In6TeO12. It thus demonstrates that the In-Sb-Te film is
fully oxidized. On the other hand, Time of Flight Secondary Ion Mass Spectroscopy (Tof-SIMS)
measurements reported in Fig. 5 clearly reveal a very significant species inter-diffusion between InSb-Te and Pt, as well as the presence of oxygen migration at 550 °C.
5
10
3
10
2
10
1
10
0
4 3 4 a d
0
100
200
300
5
10
4
10
3
10
2
10
1
10
0
C
O H
1 8 O
3 0 S i
S iO
S i2
In
S b
T e
InO
S bO
T e O
P t
P tO
4
C
O H
1 8 O
3 0 S i
S iO
S i2
In
S b
T e
InO
S bO
T e O
P t
P tO
Inte ns ity (a rb. units )
10
10
Inte ns ity (a rb. units )
10
4 3 4 a nn
0
100
200
300
T im e (s )
T im e (s )
a)
b)
0 4 1 5a)
_ 1 2 as-deposited
-­‐ C s 1 k e V G a 2 5 k e VIn-Sb-Te and b) annealed In-Sb-Te
Fig. 5 - Tof-SIMS measurements 4on
at V550°C.
40415_ 13 -­‐ C s 1ke
G a 25ke V
4.
Conclusion
The thermal conductivity of In-Sb-Te thin films with a Te content of 12 and 17%, deposited by
MOCVD, was measured in the [RT-550°C] temperature range using the MPTR technique. The total
thermal boundary resistance between the In-Sb-Te film and a metal (Pt) and dielectric (SiO2) layers
was measured also as a function of temperature. Measured thermal conductivity is consistent with the
one of Ge2Sb2Te5 in the same temperature range. However, XRD and Tof-SIMS experiments showed
that the In-Sb-Te annealed at 550°C is fully oxidized due to massive Pt- In-Sb-Te species interdiffusion allowing oxygen migration through the Pt film. Major modification of the layer properties
are also confirmed by the fact that going back to RT after the thermal cycle up to 550°C, the measured
TBR is ten times higher than the theoretical value. Therefore, a more precise control of annealing
atmosphere and thermal budget in the in the 400-550°C temperature range are required.
Acknowledgements
This work is partially supported by the EC (within FP7 SYNAPSE Project, #310339). HTN is
supported by the Italian French University within the Vinci project 2011.
References
[1] M. Wuttig, Nature Mater. 4, 265 (2005)
[2] M. Lankhorst, B. Ketelaars and R. Wolters, Nature Mater. 4, 347 (2005)
[3] A. Kolobov, P. Fons, A. Frenkel, A. Ankudinov, J. Tominaga and T. Uruga, Nature Mater. 3, 703 (2004)
[4] J.-K. Ahn, K.-W. Park, H.-J. Jung, S.-V. Pammi, S.-G. Hur and Soon-Gil Yoon , ECS Transactions, 25 (8),
1129-1133 (2009).
[5] J.-K. Ahn, H.-J. Cho, K.-W. Park and Soon-Gil Yoon, Journal of The Electrochemical Society 157 (6),
D353-D356 (2010)
[6] J.-K. Ahn, H.-J. Jung, K.-W. Park and Soon-Gil Yoon , Nano Lett. 10, 472-477 (2010).
[7] K. Daly-Flynn and D. Strand, “InSbTe phase change materials for high performance multi-level recording,”
in Optical Data Storage Conference, Hawaii, 2002.
[8] Y. Maeda, H. Andoh, I. Ikuta, and H. Minemura, Journal of Applied Physics 64, 1715 (1988).
[9] J. M. Yoon, D. O. Shin, Y. Yin, H. K. Seo, D. Kim, Y. I. Kim, J. H. Jin, Y. T. Kim, B.-S. Bae, S. O. Kim and
J. Y. Lee, Nanotechnology 23, 255301 (2012).
[10] R. Fallica, T. Stoycheva, C. Wiemer, and M. Longo, Phys. Status Solidi RRL, 1–5 (2013).
[11] T. Stoycheva, M. Longo, R. Fallica, F. Volpe, C. Wiemer, Thin Solid Films 533, 66–69 (2013).
[12] R. Kato and I. Hatta, Int. J. of Thermophysics 29, 2062 (2008).
[13] E. T. Swartz and R. O. Pohl, Rev. Mod. Phys. 61, 605 (1989).
[14] J.-L. Battaglia,1, V. Schick, C. Rossignol, A. Kusiak, I. Aubert, A. Lamperti, and C. Wiemer, App. Phys.
Lett. 102, 181907 (2013).
[15] L. De Bellis, P. E. Phelan, and R. S. Prasher, J. Thermophys. Heat Transfer 14, 144–150 (2000).
[16] J.-L. Battaglia, A. Kusiak, V. Schick, A. Cappella, C. Wiemer, M. Longo, and E. Varesi, J. Appl. Phys. 107,
044314 (2010).
[17] J. P. Reifenberg, D. L. Kencke, and K. E.Goodson, IEEE Electron Device Letters 29, 1112 (2008).
[18] D. T. Dekadjevi, C. Wiemer, S. Spiga, S. Ferrari, M. Fanciulli, G. Pavia and A. Gibaud, Appl. Phys. Lett.
83, 2148 (2003).
[19] T. Yamane, N. Nagai, S.- I. Katayama, and M. Todoki, J. Appl. Phys. 91, 9772 (2002).
[20] D. T. Dekadjevi, C. Wiemer, S. Spiga, S. Ferrari, M. Fanciulli, G. Pavia and A. Gibaud, Appl. Phys. Lett.
83, 2148 (2003).
[21] A. P. Caffrey, P. E. Hopkins, J. M. Klopf, and P. M. Norris, Microscale Thermophys. Eng. 9, 365 (2005).
[22] J. H. Los, T. D. Kühne, S. Gabardi and M. Bernasconi, Phys. Rev. B 88, 174203 (2013).
[23] R. Fallica, C. Wiemer, T. Stoycheva, E. Cianci, M. Longo, H. T. Nguyen, A. Kusiak, J.-L. Battaglia,
Microelectronic Engineering 120, 3-8 (2014).