AVS2010 Session PS-ThM: Plasma Surface Interactions (Fundamentals & Applications) III

III-V semiconductors remain unequalled materials to fabricate high efficient emitters. Anisotropic etching of III-V heterostructures is a key building-block for the processing of such photonic devices (e.g. low-optical loss waveguides or optical cavities with high quality factors, ...) and the ICP etching technique is now widely used for this purpose. Several Cl₂-containing chemistries have been proposed to date for the smooth and high-aspect-ratio etching of InP-based or GaAs-based heterostructures required to reach the NIR region. We have evidenced that in many cases anisotropic etching is due to a passivation mechanism involving SiOx deposition on the III-V sidewalls [JVSTB 26, 666 (2008)]. SiOx passivation simply occurs when a Si wafer is used as the sample tray; this configuration corresponds actually to most commercial ICP etch systems having an electrode diameter of 4-in or more, used to etch III-V samples of not more than 2-in size. However this will not be the case in future large surface processing of III-Vs, when the III-V wafer becomes of the same size of the electrode, or when III-V dies bonded onto a 200/300mm wafer have to be etched, with the wafer surface covered by protecting layer that is not silicon as it may be the case in III-V/Si photonic technologies. Other passivation mechanisms have therefore to be found. For example a Si-containing gas could be added in order to maintain SiOx passivation. In this work we have investigated the Cl₂-SiH₄ chemistry for this purpose. It is found that highly anisotropic etching (aspect ratio > 30 for micropillars) of III-V patterns with a high etch rate (> 0.6 µm/min) can be obtained by optimizing the SiH₄/Cl₂ ratio, independently from the nature of the sample tray. A high selectivity > 1:30 is also obtained with the process using a metallic/ dielectric mask. Ex-situ EDX-TEM analysis of the thin (10-50 nm thick for a 3-µm etch depth) passivation layer deposited on the sidewalls of etched sub-micrometer pillars shows that in optimized conditions this layer consists of micro-crystalline silicon We also confirm that the deposition of the passivation layer is enhanced by H addition, as previously proposed in Cl₂-H₂ chemistry [JVSTA 27, 262, 2009]. We will further discuss the possibility to use HBr/SiH₄ plasma for the anisotropic etching of InP-based or GaAs-based heterostructures, and the effect of oxygen or nitrogen addition in the gas phase on the composition of the passivation layer. We also will compare the respective effects of SiCl₄ and SiH₄ addition on the etched surface passivation process.

III-Nitride semiconductors such as gallium nitride are widely used in light emitter device manufacturing¹. GaN physical properties also open new prospects in microelectronics developments². Combining a wide bandgap (3.4eV), strong chemical bonds and high electron mobility, GaN based devices should operate under higher temperature, higher power and higher frequency than typical silicon devices. Due to inert chemical nature of GaN, wet etching is limited³. As a consequence, it is necessary to use dry etching method⁴ to obtain a reliable MESA structure. Chlorine based plasmas are commonly used because GaCl₃ is the most volatile etching product. Etch rate is also found to strongly depend on physical sputtering. GaN etching requirements for power device applications are different from those concerning photonic devices. Due to the power density supplied to the next generation of power devices, deep structures as high as 10µm should be build. As a comparison, the etched depth needed for light emitter are of the order of a few hundreds of nanometers. Deep GaN etching implies etch rate issues as well as surface roughness defects. We showed that these etching defects are linked with dislocations and nanopipes inherent to the substrates and revealed during etching⁵. Experiments were mainly performed in two Inductively Coupled Plasma (ICP) reactors: an industrial Alcatel 601E, composed of an ICP source and a diffusion chamber and an ICP-RIE Corial 200IL without diffusion chamber. For a better understanding of the etching mechanisms, different diagnostics are used to characterize the plasma. Optical emission spectroscopy, Langmuir probe and mass spectrometry are performed as a function of process parameters. We observe that etching behaviour depends on cover plate material. An optimum etch rate as a function of source power is measured by using a silicon cover plate. OES and Langmuir probe measurements suggest that silicon cover plate, which is etched by chlorine radicals, can be a limitation in the etching performance of the process. Different chemistries are studied as source of active species, sputtering ions or molecule scavenging impurities. We have shown that oxygen impurities are responsible for the columnar defects. We also show that adding a small amount of nitrogen in the chemistry could increase the selectivity with SiO₂ mask.

¹S. Nakamura & al., Appl. Phys. Lett., 67, 1868 (1995)

²G.T. Dang & al., IEEE Trans. On Elec. Dev., 47, 692-696 (2000)

³D. Zhuang and J.H. Edgar, Mat. Sci. and Eng., 48, 1–46 (2005)

⁴S.J. Pearton & al., J. Appl. Phys., 86 (1999)

⁵J. Ladroue, A. Meritan, M. Boufnichel, P. Lefaucheux, P. Ranson and R. Dussart, JVST A submitted (2010)

Recombination and other possible reactions of Cl and O on chamber walls in chlorine and oxygen plasmas were studied by the “spinning wall” technique. With this method, a small cylinder (1”dia. x 1”high) in the chamber wall was rapidly rotated, periodically exposing its surface to the plasma and then to the differentially pumped diagnostic chamber housing an Auger electron spectrometer for in-situ surface analysis. The plasma chamber also contained a silicon electrode that can be rf-biased and sputtered in an inert (Ar) plasma, or etched in a chlorine plasma. Using this technique, we previously measured Cl and O atom recombination on plasma-conditioned anodized Al and stainless steel surfaces by monitoring desorption of Cl₂ and O₂ with a mass spectrometer or through a pressure rise. We also previously found a substantial increase in O atom recombination probabilities due to trace amounts of Cu deposited in-situ by thermal evaporation. In the current study, a smooth Ni-coated Al substrate was used. This substrate was exposed to Si sputtered from the rf-biased electrode and was then oxidized and conditioned by long exposure to oxygen plasma. Traces of Ti were deposited on this surface with the evaporator, followed by the oxygen plasma conditioning (Ti:Si:O::5:47:48 atomic percents, averaged over the ~10 nm depth probed by Auger). The O atom recombination probability on this Ti-contaminated substrate was found to be ~30% lower than on the Ti-free substrate, i.e. the opposite to that observed for Cu contamination. Ti was then etched away in the chlorine plasma, leaving an oxy-silicon-chloride surface (Si:O:Cl:: 47:39:14). The Si electrode was then etched in the chlorine plasma while rotating the substrate, coated the rotating substrate with etching products changing the substrate surface composition to Si:O:Cl :: 54:19:27. Immediately after extinguishing the rf bias to the Si electrode, the Cl atom recombination probability was found to be lower than on the more oxygen-rich Si surface. During rf bias, however, the total product yield is higher, indicating that other products (i.e. SiCl_x) could also be desorbing from the substrate as a result of the etching of the Si electrode.

The interaction of H₂ gas discharges with carbon-based materials has been of sustained interest in many technological fields. In this work we have used a well characterized, inductively coupled plasma system (ICP) to study the interaction of H₂/Ar and D₂/Ar discharges with hard a-C:H films. The erosion of a-C:H is monitored in real time by ellipsometry, optical emission spectroscopy and plasma properties are characterized by a Langmuir probe. Our experimental setup allows for varying the reactor geometry over a wide range by changing the plasma generation substrate distance. H₂/Ar and D₂/Ar plasma interaction with a-C:H were performed using low pressure (30 mTorr) 600 W 13.56 MHz RF inductive power plasma with different substrate bias voltages. Employing real-time ellipsometry, we were able to monitor the detailed kinetics of the formation of a 1-5 nm thick hydrogenated layer of lower density than the a-C:H substrate, followed by steady-state erosion. The influence of various plasma parameters on modified surface properties and erosion will be reported. We also will present real-time data which gives insight into the dynamic flux of carbon atoms into the plasma produced by erosion a-C:H for different conditions . The measurements of the modified surface layers are compared with “Stopping and Range of Ions in Matter (SRIM)” simulations for different conditions. The atomistic details of surface processes will also be be compared with molecular dynamics simulations of the UCB group.

We gratefully acknowledge support of this work by DOE’s Plasma Science Center for Predictive Control of Plasma Kinetics: Multi-phase and Bounded Systems (University of Michigan ).

Porous low-dielectric-constant (low-k) materials such as porous SiOCH film are essential for interlayer dielectric film in high performance ULSI devices. To establish extremely precise etching processes of the low-k film for the next generation devices, it is required to understand the surface reaction and damage formation mechanism during plasma processing, while developing a sophisticated methodology to control the etching and ashing processes. We developed an integrated monitoring system equipped with in-situ spectroscopic ellipsometry, Fourier transform infrared reflection absorption spectroscopy (FT-IR RAS), a substrate temperature monitor using an optical fiber-type low-coherence interferometer [1] and an absolute density monitor for H and N radicals [2,3]. The integrated monitoring system was installed in a dual frequency capacitively coupled plasma etch reactor and we investigated H₂/N₂ plasma interactions on the low-k film. The in-situ monitoring during the plasma etching or ashing is crucial for the clarification of damage mechanism because the damaged films are easily modified during air exposure. Furthermore, the effect of each particle, i.e. ions, photons and radicals, was investigated individually by ‘PAPE’ method [4] that uses small plates, such as Si, SiO₂ and MgF₂, on or above the film substrate during the plasma exposures. So far, we considered that damages on the p-SiOCH are determined by chemical reactions of H radicals that reduce the Si-CH₃ bonds and N radicals that have an effect of inhibition of the damages. It was also confirmed that a portion of Si-O-Si linear structure in the SiOCH film changed to network and cage structures with decrease in Si-CH₃ bond during the plasma exposure. The effects of the temperature during etching on the etch profile were also examined for a variety of H₂/N₂ gas mixture ratio. The higher the H radical density and the temperature, the lager the undercut in the low-k pattern profile. Especially, the temperature increase after plasma ignition was found to be a cause of the profile deformation. Based on the above results, we proposed an autonomously-controlled etch system that realized a real-time feedback control for the fine pattern etching while monitoring the wafer temperature, radical densities and so on. It was demonstrated that real-time radical-density control upon the temperature was effective for obtaining precise pattern profiles.

[1] K. Takeda, et al., J. Appl. Phys., 43, 7737 (2004).

[2] S. Takashima, et al., Appl. Phys. Lett., 75, (25), 3929 (1999).

[3] S. Takashima, et al., J. Vac. Sci. Technol., A 19, 599 (2001).

[4] S. Uchida, et al., J. Appl. Phys. 103, 073303 (2008).