Thermal transport across metal silicide-silicon interfaces: An experimental comparison between epitaxial and nonepitaxial interfaces

Silicides are used extensively in nano- and microdevices due to their low electrical resistivity, low contact resistance to silicon, and their process compatibility. In this work, the thermal interface conductance of ${\mathrm{TiSi}}_2, {\mathrm{CoSi}}_2$, NiSi, and PtSi are studied using time-domain thermoreflectance. Exploiting the fact that most silicides formed on Si(111) substrates grow epitaxially, while most silicides on Si(100) do not, we study the effect of epitaxy, and show that for a wide variety of interfaces there is no dependence of interface conductance on the detailed structure of the interface. In particular, there is no difference in the thermal interface conductance between epitaxial and nonepitaxial silicide/silicon interfaces, nor between epitaxial interfaces with different interface orientations. While these silicide-based interfaces yield the highest reported interface conductances of any known interface with silicon, none of the interfaces studied are found to operate close to the phonon radiation limit, indicating that phonon transmission coefficients are nonunity in all cases and yet remain insensitive to interfacial structure. In the case of ${\mathrm{CoSi}}_2$, a comparison is made with detailed computational models using (1) full-dispersion diffuse mismatch modeling (DMM) including the effect of near-interfacial strain, and (2) an atomistic Green' function (AGF) approach that integrates near-interface changes in the interatomic force constants obtained through density functional perturbation theory. Above 100 K, the AGF approach significantly underpredicts interface conductance suggesting that energy transport does not occur purely by coherent transmission of phonons, even for epitaxial interfaces. The full-dispersion DMM closely predicts the experimentally observed interface conductances for ${\mathrm{CoSi}}_2$, NiSi, and ${\mathrm{TiSi}}_2$ interfaces, while it remains an open question whether inelastic scattering, cross-interfacial electron-phonon coupling, or other mechanisms could also account for the high-temperature behavior. The effect of degenerate semiconductor dopant concentration on metal-semiconductor thermal interface conductance was also investigated with the result that we have found no dependencies of the thermal interface conductances up to ($n$ or $p$ type) $\approx 1\times {10}^{19} {\mathrm{cm}}^{-3}$, indicating that there is no significant direct electronic transport and no transport effects that depend on long-range metal-semiconductor band alignment.

Figure 1

(a) XRD results of ${\mathrm{CoSi}}_2$ on intrinsic Si(100) and Si(111) wafer; (b) XRD results of NiSi on intrinsic Si(100) and Si(111) wafer; (c) XRD results of PtSi on intrinsic Si(100) and Si(111) wafer; (d) XRD results of ${\mathrm{TiSi}}_2$ on intrinsic Si(100) and Si(111) wafer; (e) XRD phi scan of the in-plane diffraction for PiSi(020)/Si(111), NiSi(200)/Si(111), and ${\mathrm{CoSi}}_2$(111)/Si(111) samples.

Figure 2

HRTEM of an epitaxial ${\mathrm{CoSi}}_2$-Si interface.

Figure 3

Schematic of the Si-${\mathrm{CoSi}}_2$ interface supercell with a 8B interfacial atomic configuration. The dashed rectangular boxes indicate the unit cells of bulk Si and bulk strained ${\mathrm{CoSi}}_2$.

Figure 4

The interfacial thermal conductance of ${\mathrm{CoSi}}_2, {\mathrm{TiSi}}_2$, NiSi, and PtSi on intrinsic Si(100) and Si(111) wafer. ${\mathrm{CoSi}}_2$/Si-1, ${\mathrm{CoSi}}_2$/Si-2, and ${\mathrm{CoSi}}_2$/Si-3 represent samples made under different conditions. ${\mathrm{CoSi}}_2$/Si-1: HF treated wafer + films deposited by co-sputtering, ${\mathrm{CoSi}}_2$/Si-2: HF treated wafer + films made by reactive growth method, ${\mathrm{CoSi}}_2$/Si-3: RF bias treated wafer + films made by reactive growth method. For ${\mathrm{TiSi}}_2$, NiSi, and PtSi films are made by reactive growth method on RF-bias cleaned substrates. The interfacial thermal conductance of Al/Si(100) is also attached as a reference.

Figure 5

(a) Modeling results for ${\mathrm{CoSi}}_2$(111)-Si(111) interfaces using various models: the full-dispersion diffuse mismatch model (green), the atomistic Green's function method for interface of 8B (red) and 8A (orange) and the radiation limit (black). Experimental data at room temperature are shown for comparison (blue squares). (b) Comparison between experimental thermal interface conductance of ${\mathrm{TiSi}}_2$-Si(111) interface (blue squares), the full-dispersion DMM calculation of ${\mathrm{TiSi}}_2$(001), ${\mathrm{TiSi}}_2$(010), ${\mathrm{TiSi}}_2$(100), ${\mathrm{TiSi}}_2$(111)-Si(111), and the radiation limit (black). (c) Comparison between experimental thermal interface conductance of an epitaxial NiSi(200)-Si(111) interface (blue squares), the full-dispersion DMM calculation (green), and the radiation limit (black). (d) Comparison between experimental thermal interface conductance of an epitaxial PtSi(020)-Si(111) interface (blue squares), the full-dispersion DMM calculation (green), and the radiation limit (black).

Figure 6

The substrate doping effects on the interfacial thermal conductance. The thermal conductance values are normalized with their correspondent undoped values. This includes the interfacial thermal conductance of ${\mathrm{CoSi}}_2$/Si(111) and ${\mathrm{CoSi}}_2$/Si(100) made by co-deposition, ${\mathrm{CoSi}}_2$/Si(111), ${\mathrm{CoSi}}_2$/Si(100), ${\mathrm{TiSi}}_2$/Si(111), NiSi/Si(111), and PtSi/Si(111) made by reactive growth. “*” indicates that the interface is not epitaxial. (For original data please refer to Fig. 5 of Ref. [38].)