Hot-wire assisted atomic layer deposition of tungsten films
Mengdi Yang is PhD Student in the Reseach group of Integrated Devices and Systems. Her supervisor is Jurriaan Schmitz from the Faculty of Electrical Engineering, Mathematics and Computer Science.
This thesis aims to establish a novel technique of atomic layer deposition (ALD) for the future ultra-large-scale integration (ULSI) of microelectronics. In last decades, chemical vapor deposition (CVD) is a dominant means for film deposition. However, the downscaling of modern ULSI manufacturing demands ALD to achieve conformal and uniform thin film with very precise control of thicknesses on structures of increasing complexity. Recently, plasma-enhanced ALD (PEALD) is largely adopted in industries to enable deposition of layers at lower substrate temperatures compared to thermal ALD. Moreover, PEALD can also provide deposition of single-elements such as Si and selected metals. However, plasma has some drawbacks. To provide a possible alternative, in this thesis we develop hot-wire assisted ALD (HWALD), where a heated tungsten (W) filament is utilized instead of a plasma to generate radicals. HWALD is expected to be another candidate for deposition in future ULSI technology. Particularly, this thesis focuses on the application of HWALD for W deposition by providing sequential pulses of atomic hydrogen (at-H) and WF6.
Chapter 2 describes two reactors which were used for HWALD W. The cold-wall reactor has a much larger volume than the hot-wall one. The hot wire in the cold-wall reactor is situated much further away from the substrate compared to the hot-wall reactor. In both reactors, there is no direct line-of-sight between the hot wire and the substrate. A spectroscopic ellipsometer (SE) is installed to in-situ monitor the film growth. Measured film thicknesses have been verified by other techniques and the optical models have been established and validated. Furthermore, tellurium (Te) etching experiments were conducted to confirm the existence of atomic hydrogen and its delivery to the substrate. The at-H was generated by cracking H2 on the hot wire and could be transferred to the substrate surface to provide a reasonable etch rate of Te. The total process pressure, Ar flow rates and other parameters affected the etching. Although WF6 gas was introduced not via the hot wire, it was found to diffuse upwards to the hot wire resulting in back-stream diffusion. This effect has an influence on the subsequent W deposition.
Chapter 3 presents results of tungsten films deposition in the cold-wall reactor. Besides ALD, CVD and etching modes of the W film were observed. This can be explained by the back-stream diffusion: WF6 could diffuse to the hot-wire, resulting in WF6 decomposition and generation of a flux of fluorine-containing species, such as fluorine (F) and tungsten subfluorides (WFx, x<6). The fluorine could cause etching of the grown W film, whereas WFx could mix with at-H, leading to CVD. And the fluorine containing species would adsorb on the cold walls and evaporate into gas phase during experiments. It is found that a higher gas pressure strengthened etching whereas a lower pressure enhanced CVD. By selecting the proper process pressure and limiting the dose of WF6, optimal conditions have been found to maintain the ALD mode. Under these chosen conditions, HWALD W films were deposited with a W purity approaching 99 at%. Further, we compared HWALD W with CVD W in terms of growth kinetics and properties. For CVD, the samples were made in a mixture of WF6 and either molecular or atomic hydrogen. Resistivity of the CVD W was around 20 µΩ·cm, whereas it was as high as 100 µΩ·cm for the HWALD films. X-ray diffraction (XRD) revealed that the HWALD W crystallized as β-W, whereas both CVD films were in the α-W phase.
Chapter 4 demonstrates results of HWALD W in the hot-wall reactor. The X-ray photoelectron spectroscopy (XPS) analysis revealed high-purity films, reaching 99 at.% of W. Remarkably, XRD proved the high-purity α-phase W, compared to β-phase W obtained in the cold-wall reactor. The α-phase was further verified by the d-spacing values of W obtained from high-resolution transmission electron microscopy (HR-TEM) images. The resistivity measurements by means of four point probe, transfer length method test structures and the Drude-Lorentz SE model all revealed a low resistivity of 15 µΩ·cm for the HWALD W. The HR-TEM analysis of the films showed a uniform and conformal coverage on high aspect ratio structures (up to an aspect ratio of 36), confirming the effective ALD process and the sufficient diffusion of both WF6 and at-H into deep trenches. Finally, it is found that W layers start to become electrically continuous in a thickness range of 2-3 nm.
As described in the last two chapters, W obtained in two different reactors possessed different crystalline structure. Thus, Chapter 5 aimes to find the factors which are decisive for the formed. Impurites, i.e. N2O, O2, NH3 and H2O, were added upon the standard HWALD process to investigate their effects. O2 and water have a retarding effect on W growth but the HWALD process can be re-initiated after stopping their supply. In contrast, nitridizing species (N2O and NH3) have a permanent terminating effect. However, W deposited with O2 impurites still resulted in α-phase. Furthermore, the effects of WF6 overdose were studied. The surplus of WF6 appeared to lead to the formation of β-phase W. Extra fluorine-containing species were thus identified as the likely root cause of β-phase formation.
Chapter 6 proposed an inherent area-selective HWALD of W. The nucleation and growth of HWALD W on various substrates were studied. No nucleation was found on a thermally-grown SiO2 surfaces nor on (ALD-grown) TiN and Al2O3 surfaces. On the contrary, HWALD W could be successfully deposited on W and Co surfaces. Moreover, the native oxides of these metals could be reduced by at-H, having no influence on the subsequent deposition of W by HWALD. Due to the nucleation delays on different surfaces, an area-selective HWALD W process was achieved on W/SiO2 and Co/SiO2 patterned surfaces. Furthermore, it is found that applying an a-Si seed layer of thickness even below 1 nm was sufficient to enable the effective nucleation on surfaces which are inert to HWALD process.
To sum up, this thesis presents results of HWALD W. The deposited W has a supreme property in terms of low resistivity and high purity. However this process can be further developed. At this stage, the growth rate per cycle (varying between 0.01 up to 0.02 nm/cycle) is comparable with that of other metals deposited by ALD, having however a long cycle time of 21.5 s due to the extended purge requirements and leading to a long deposition time. More efforts can be made to shorten the cycle time. For example, the purge times can be shortened by further limiting the WF6 dose and adsorption on cold surfaces. Alternatively, a spatial ALD[1] can be adopted to avoid the long purge times.
