GaAs AlAs 超晶格中热导率的各向异性

We combine the transient thermal grating and time-domain thermoreflectance techniques to characterize the anisotropic thermal conductivities of GaAs/AlAs superlattices from the same wafer. The transient grating technique is sensitive only to the in-plane thermal conductivity, while timedomain thermoreflectance is sensitive to the thermal conductivity in the crossplane direction, making them a powerful combination to address the challenges associated with characterizing anisotropic heat conduction in thin films. We compare the experimental results from the GaAs/AlAs superlattices with firstprinciples calculations and previous measurements of Si/Ge SLs. The measured anisotropy is smaller than that of Si/Ge SLs, consistent with both the massmismatch picture of interface scattering and with the results of calculations from density-functional perturbation theory with interface mixing included.

We used two noninvasive optical techniques in tandem to measure the anisotropic thermal properties of two GaAs/AlAs SLs grown with metal organic chemical vapor deposition (MOCVD). The TDTR technique was used to measure the cross-plane thermal conductivities, and TTG23−29 measured the in-plane thermal diffusivities. These two techniques have not been combined before to characterize the anisotropy of thermal transport in thin films and SLs. Both techniques use lasers to induce a temperature rise at the sample surface and to monitor the temperature decay via heat conduction. The main difference between the two techniques lies in the spatial distribution of the excitation laser light at the sample surface, as illustrated in Figure 1A and B. In TDTR (Figure 1A), we used a large (60 μm) laser spot; hence on short time scales the heat transfer occurs primarily in the cross-plane direction. In TTG (Figure 1B), two interfering beams created a sinusoidal intensity pattern with a small period (on the order of a few micrometers). The heat transport in TTG occurs in both the in-plane and the cross-plane directions; however, as will be explained below, the temporal signature of the thermal grating decay is only sensitive to the in-plane thermal conductivity. This unique characteristic of the measurement was recognized only recently. By combining the TDTR and TTG techniques, we measured the room temperature anisotropic thermal conductivity of GaAs/AlAs SLs and compared the results with previously reported measurements of cross- and in-plane conductivities on different GaAs/AlAs SLs. The results are also compared to first-principles calculations of SL thermal conductivity, accounting for both intrinsic and extrinsic phonon scattering channels, and contrasted with experimental data on Si/Ge superlattices, which have a larger acoustic mismatch.

Two 3.5 μm thick GaAs/AlAs SLs, an 8 nm × 8 nm (with a total period of 16 nm) SL and a 2 nm × 2 nm (with a total period of 4 nm) SL, were epitaxially grown by MOCVD on the] direction of a GaAs wafer with a 500 nm GaAs buffer. X-ray diffraction performed on the samples confirmed both the planarity and the period thicknesses of 4 and 16 nm for the two SL samples. For the TDTR measurements, a 90 nm Al opticalthermal transducer layer was deposited with e-beam evaporation.

Fig1

In the TTG method, in contrast with the TDTR method, we do not use a metal coating. This is an important difference that deserves a brief discussion. Measuring an uncoated sample entails a number of advantages: no sample preparation is required and the measurements are affected neither by the metal film thickness nor by the thermal boundary resistance at the metal−semiconductor interface. Thus, there is no need to employ multiparameter fitting. In fact, in the model we use to fit the data, given by eq 4, the only variable parameter is the parameter of interest. On the other hand, photothermal measurements on uncoated semiconductors are often complicated by the presence of photoexcited carriers. Indeed, laser excitation creates both a temperature grating and a concentration grating of the excited carriers with the latter also influencing the refractive index and contributing to diffraction of the probe beam. However, the ambipolar diffusion coefficient of photoexcited carriers in both bulk GaAs33 and GaAs quantum wells34 exceeds 10 cm² /s, whereas the thermal diffusivity measured on our samples is well below 0.1 cm² /s. Consequently, the carrier grating should decay much faster than the thermal grating. Indeed, we do observe a small initial spike of duration ∼0.5 ns limited by the resolution of the detection electronics, which may be caused by the photoexcited carriers. On the longer time scale of interest the carrier contribution should be entirely negligible, which is confirmed by the fact that the thermal model alone generally fits the data very well.

In summary, we have shown that the TTG technique is sensitive to the in-plane thermal conductivities of thin films and combined it with the TDTR technique to probe the anisotropies in the thermal conductivities of GaAs/AlAs SLs. The anisotropy in the thermal conductivities of GaAs/AlAs SLs is much smaller than that of the Si/Ge SLs reported in the past, consistent with pictures obtained from first-principles simulation. Although DFPT simulations produced similar anisotropy values compared to experimental data for both GaAs/AlAs and Si/Ge SLs, the thermal conductivity values predicted are about a factor of 2 higher than experiments, potentially due to local thickness fluctuations.