Full characterization of ultrathin 5-nm low-$k$ dielectric bilayers: Influence of dopants and surfaces on the mechanical properties

Ultrathin films and multilayers, with controlled thickness down to single atomic layers, are critical for advanced technologies ranging from nanoelectronics to spintronics to quantum devices. However, for thicknesses less than 10 nm, surfaces and dopants contribute significantly to the film properties, which can differ dramatically from that of bulk materials. For amorphous films being developed as low dielectric constant interfaces for nanoelectronics, the presence of surfaces or dopants can soften films and degrade their mechanical performance. Here we use coherent short-wavelength light to fully and nondestructively characterize the mechanical properties of individual films as thin as 5 nm within a bilayer. In general, we find that the mechanical properties depend both on the amount of doping and the presence of surfaces. In very thin (5-nm) silicon carbide bilayers with low hydrogen doping, surface effects induce a substantial softening—by almost an order of magnitude—compared with the same doping in thicker (46-nm) bilayers. These findings are important for informed design of ultrathin films for a host of nano- and quantum technologies, and for improving the switching speed and efficiency of next-generation electronics.

Figure 1

Dynamic EUV diffraction from transverse and longitudinal acoustic waves. (a) After ultrafast laser excitation, the hot Ni nanolines impulsively expand, launching acoustic waves in the sample. After a controlled time delay, an EUV probe pulse diffracts from the sample surface, and the scattered light is collected by a charge-coupled device camera. The acoustic waves dynamically change the EUV diffraction efficiency, as shown in (b) and (c). (b) A longitudinal acoustic wave is launched downward into the film (inset). Reflections from the film-substrate interface imprint a discrete series of echoes in the data (see arrows in inset). (c) At early times, we observe the longitudinal breathing mode (left) of the nanolines. At longer times, we observe a surface acoustic wave (right), whose penetration depth is confined to a fraction of the grating period. The surface acoustic wave and longitudinal acoustic wave velocities provide the two independent components of the isotropic film's elastic tensor.

Figure 2

Compositional characterization of the 5-nm SiC:H bilayer. (a) STEM image obtained using a HAADF detector. Strain from the deposited films blurs the atomic contrast peaks in the Si substrate, as expected [41]. HAADF intensity also drops at the interface of the two films in the bilayer. We attribute this to a reduction in the density of the bottom SiC:H layer, as it is strained by the layers above it. (b) EDS image of the sample showing Si (blue) and N (pink). Nitrogen exists at the bottom interfaces of the two SiC:H layers due to the nitrogen plasma clean applied to improve the film adhesion. (c) Horizontally binned lineout of the HAADF contrast for the full cross section taken for STEM characterization. This lineout extends upwards into the focused ion beam (FIB)-deposited Pt layer used for STEM, and downwards into the strained region of the Si substrate.

Figure 3

Surface-induced softening compared to hydrogenation-induced softening. (a) Elastic constant ranges for the 5-nm SiC:H top layer (red circles), the 46-nm SiC:H top layer (green triangles), and the highly hydrogenated 44-nm SiOC:H film (blue squares). Each point represents a configuration simulated in the FEA model that agrees with the data, within uncertainty. The 46-nm film maintains a bulklike rigid bond network, due to its low hydrogenation and large thickness. The 5-nm film of the same material is significantly softened due to the terminated bonds at its surface (orange). This is distinct from the softening observed in highly hydrogenated SiOC:H, where hydrogenation breaks up the rigid bond network in the volume of the film (blue). Note the top schematics are to illustrate the differences between samples, and are not exact. The auxetic boundary, defined by $c_{11}=2c_{44}$, is the limit where Poisson's ratio becomes negative. (b) The same data as in (a), expressed in terms of Poisson's ratio and Young's modulus.

Figure 4

STEM cross section of 46-nm SiC:H bilayer. (a) Full HAADF image. SiC:H films are known to strain Si substrates, and the full depth of the strained volume is visible here in the blurring of the atomic peaks imaged in the substrate region. We attribute the low-intensity region between the two SiC:H layers to a reduction in density as the top layer and N-rich layer similarly strain the bottom layer. Note the thickness of this strained region is similar to the layer thickness for the 5-nm bilayer in Fig. 2, indicating the same effect is lowering the HAADF intensity in the bottom layer of both SiC:H samples. (b) Lineout by horizontally binning (a). The faint intensity striations within each SiC:H layer are a result of the four-step deposition process. Each step is identical, with no change in film composition.