Supporting Online Material for Phonon and Electron Transport through Ge 2 Sb 2 Te 5 Films and Interfaces Bounded by Metals Jaeho Lee 1, Elah Bozorg-Grayeli 1, SangBum Kim 2, Mehdi Asheghi 1, H.-S. Philip Wong 3, and Kenneth E. Goodson 1 1 Department of Mechanical Engineering, Stanford University, Stanford, California 94305 2 IBM T. J. Watson Research Center, Yorktown Heights, New York 10598 3 Department of Electrical Engineering, Stanford University, Stanford, California 94305 A. Electrical Resistivity and Specific Contact Resistance Measurements The electrical resistivity and specific contact resistance of Ge 2 Sb 2 Te 5 films are measured by three independent methods respectively, and the measurement results validate each other. As one of the most widely used test structures for metal-semiconductor interfaces, Cross-Bridge Kelvin Resistor (CBKR) allows direct measurement of specific contact resistance [1, 2]. The CBKR structure (Fig. 1) forces the current (I) through Ge2Sb2Te5 and TiN legs and measures the contact voltage (Vc) using the other legs. The contact resistance (R c = V c /I) and the overlap area (A c ) provide the specific contact resistance. However, the contact area misalignment due to process variations and lateral current crowding around the contact area can lead to significant errors [2, 3]. Several techniques to improve the measurement accuracy had been proposed such as using novel patterns and numerical simulations [3, 4]. Fig. 1. A schematic for Cross-Bridge Kelvin Resistor (CBKR) that measures the specific contact resistance of Ge 2Sb 2Te 5 and TiN. The four-point configuration eliminates the probe and pad contributions and allows direct measurement of the specific contact resistance. The linear transfer length method (TLM) structure (Fig. 2) measures both the electrical resistivity and the specific contact resistance (ρ c ) by utilizing the transfer length (l t ), which is defined as the distance the current flows through Ge 2 Sb 2 Te 5 under the TiN contact. Assuming the sheet resistance (R sh ) of Ge 2 Sb 2 Te 5 is much larger than that of TiN, the transfer length can be estimated by l t = (ρ c /R sh ) 1/2. A linear TLM structure is prepared by a 100-µm-wide Ge 2 Sb 2 Te 5 leg with intermittent TiN contacts of gap spacing that varies from 6 µm to 60 µm (Fig. 2b). The specific contact resistance is extracted from the measured resistances as a function of the gap spacing, assuming all contacts are identical.
Fig. 2. Experimental data and a schematic for the linear transfer length method (L-TLM) for measuring the sheet resistance of Ge 2Sb 2Te 5 and the specific contact resistance of Ge 2Sb 2Te 5 and TiN contacts. The resistance increases linearly with the gap spacing. The slope provides the sheet resistance and the y-axis intercept provides the transfer length, which are used to calculate the specific contact resistance. Since the linear-tlm structures can be affected by parasitic current in non-isolated regions, we also used the circular-tlm (Fig. 3). The circular-tlm structure is prepared by patterning an inner circular TiN of diameter (D) 100 µm on a Ge 2 Sb 2 Te 5 film with a ring shaped spacing (d) that varies from 2 µm to 23 µm with respect to an outer TiN (Fig. 3). When the inner TiN diameter much larger than the gap spacing, the ring geometry can be reduced to a linear-tlm model with a correction factor (C = (D/2d) ln(l t +2d/D)). Similar to the linear-tlm measurements, the specific contact resistance can be extracted from the measured resistances as a function of the gap spacing. Compared to the CBKR, the TLM is more suitable for non-uniform contact resistances and less sensitive to the misalignment. However, the specific contact resistance measurements with TLM are indirect and depend on the sheet resistance of Ge 2 Sb 2 Te 5, which can increase the measurement uncertainty. Fig. 3. Experimental data and a schematic for the circular transfer length method (C-TLM) for measuring the sheet resistance of Ge 2Sb 2Te 5 and the specific contact resistance of Ge 2Sb 2Te 5 and TiN contacts. The corrected resistance increases linearly with the gap spacing. The slope provides the sheet resistance and the y-axis intercept provides the transfer length, which are used to calculate the specific contact resistance. The three independent measurement techniques used in this work validate the results for each other and offer consistent information. The electrical resistivity can be obtained from the sheet resistance data from the TLM measurements with the film thickness information. Alternatively, a commercial sheet resistance measurement system is directly used on blanket Ge 2 Sb 2 Te 5 films.
B. Thermal Conductivity and Boundary Resistance Measurements Using Picosecond Thermoreflectance A picosecond time-domain thermoreflectance confines the temporally heated region to the TiN-Ge 2 Sb 2 Te 5 -TiN sandwich structure. The beam from a passively-modelocked 1064 nm Nd:YVO 4 laser with a 9.2 ps pulse width and 82 MHz repetition is split into a pump and probe component. The pump pulse travels along a fixed path and generates a temperature excursion in an Al transducer deposited on the sandwich structure [5]. The probe pulse, travelling along a variably-delayed path, interrogates the normalized temperature of the transducer as a function of time after the arrival of the pump pulse [6]. This results in a temperature decay curve spanning from 0.1 ns before the pump pulse arrives to 3.5 ns after. We fit for this thermal decay trace using a 3-D axially-symmetric model [7] of heat diffusion through a multilayer stack [8]. The thermal properties at different points within the stack affect different time domains of the thermal decay trace (Fig. 4). These domains correspond to different thermal penetration depths, which are determined by the measurement time scale and the thermal time constant of different regions within the sample. The thermal time constant of the Al transducer controls the behavior of the first 0.5 ns after the arrival of the pump beam (Region I). This time constant includes the total capacitance of the transducer, the intrinsic Al thermal resistance, and the Al- TiN thermal boundary resistance. Since the intrinsic resistance of the 50 nm Al layer is small, and since the heat capacity is well known, we extract the Al-TiN thermal boundary resistance (3 m 2 K/GW). From ~0.5 ns to ~2.5 ns after the pump pulse arrives, the thermal resistance of the top TiN layer and underlying Ge 2 Sb 2 Te 5 dictate the decay behavior (Region II). Since the thermal conductivity of TiN is well-characterized, we extract a combination of the intrinsic Ge 2 Sb 2 Te 5 thermal resistance and the Ge 2 Sb 2 Te 5 -TiN thermal boundary resistance. Lastly, the thermal decay behavior from ~2.5 ns to before the arrival of the next pump pulse is sensitive to the total resistance of the TiN- Ge 2 Sb 2 Te 5 -TiN stack (Region III). Our knowledge of the TiN thermal conductivity allows us to obtain the combination of the intrinsic Ge 2 Sb 2 Te 5 thermal resistance and both Ge 2 Sb 2 Te 5 -TiN thermal boundary resistances. Under the assumption that thermal boundary resistance does not differ significantly for the two Ge 2 Sb 2 Te 5 -TiN interfaces, and using the results from Region II, we separate out the intrinsic Ge 2 Sb 2 Te 5 thermal conductivity. Fig. 4. Representative thermal decay trace for a sample consisting of Al, TiN, Ge 2Sb 2Te 5, and TiN layers on a Si substrate. The thermal decay in each temporal domain (I, II, and III) is more sensitive to a different set of thermal properties due to the time dependent heat diffusion. The first time domain (I) is most sensitive to the Al-TiN thermal boundary resistance, and the second time domain (II) is more sensitive to the Ge 2Sb 2Te 5 thermal conductivity and the top TiN-Ge 2Sb 2Te 5 thermal boundary resistance. The third time domain (III) is sensitive to the entire film of Ge 2Sb 2Te 5 including the bottom Ge 2Sb 2Te 5-TiN thermal boundary resistance.
C. X-ray Diffraction Data Fig. 5. Impact of annealing on the phase distributions for the 30-nm-thick Ge 2 Sb 2 Te 5 film, observed using X-ray diffraction with using Cu Ka radiation with a temperature increment of 10 C in 5 minutes. The peaks are normalized with respect to the highest peak in each sample. The phase transition to fcc phase begins at 150 C and to the hcp phase begins at 360 C. The phase transition temperatures may be lower with longer annealing time.
Fig. 6. Impact of annealing on the phase distributions for the 150-nm-thick Ge 2 Sb 2 Te 5 film, observed using X-ray diffraction with using Cu Ka radiation with a temperature increment of 10 C in 5 minutes. The peaks are normalized with respect to the highest peak in each sample. The phase transition to fcc phase begins at 110 C and to the hcp phase begins at 200 C. The phase transition temperatures may be lower with longer annealing time. Reference [1] W. M. Loh, S. E. Swirhun, E. Crabbe, K. Saraswat, and R. M. Swanson, An accurate method to extract specific contact resistivity using cross-bridge kelvin resistors, IEEE Electron Device Lett., vol. 6, pp. 441 443, Sept. 1985. [2] M. Finetti, A. Scorzoni, and G. Soncini, Lateral current crowding effects on contact resistance measurement in four terminal resistor test patterns, IEEE Electron Devices Lett., vol. 5, no. 12, pp. 524 526, Dec. 1984 [3] M. Ono, A. Nishiyama, A. Toriumi, A simple approach to understanding measurement errors in the cross-bridge Kelvin resistor and a new pattern for measurements of specific contact resistivity, Solid- State Electronics, Vol. 46, (2002), pp. 1325-1331 [4] W. M. Loh, S. E. Swirhun, T. A. Schreyer, R. M. Swanson, and K. C. Saraswat, Modeling and measurement of contact resistances, IEEE Electron Device Lett., vol. 6, pp. 441 443, Sept. 1985. [5] J. A. Rowlette, R. D. Kekatpure, M. A. Panzer, M. L. Brongersma,and K. E. Goodson "Nonradiative recombination in strongly interacting silicon nanocrystals embedded in amorphous silicon-oxide films," Physical Review B 80 (4), 045314 (2009). [6] K. Ujihara, "Reflectivity of Metals at High Temperatures," Journal of Applied Physics 43 (5), 2376-2383 (1972). [7] D. G. Cahill, "Analysis of Heat Flow in Layered Structures for Time-Domain Thermoreflectance," Review of Scientific Instruments 75 (12), 5119-5122 (2004). [8] A. Feldman, "Algorithm for Solutions of the Thermal Diffusion Equation in a Stratified Medium with a Modulated Heat Source," High Temperatures - High Pressures 31, 293-298 (1999).