AVS2012 Session PS2-ThM: Low Damage Processing

The Si recess, which is caused after post-wet treatment in the Si gate etching process, can have a big effect on transistor properties such as threshold voltage (V_th) and off-state leakage current (I_off). We previously studied what caused the Si recess [1] and found through our analysis of a molecular dynamics (MD) calculation and a beam experiment that it could be the oxidation enhanced diffusion induced by incident hydrogen during HBr/O₂ gate etching [2]. The fluctuation in the Si recess depth (Δd_R) as well as in the critical dimension (ΔCD) is one of the key factors causing fluctuations in V_th (ΔV_th) and I_off (ΔI_off). The pattern dependence of the incident particle flux is related to Δd_R, therefore a prediction technology that considers this dependence is necessary in order to understand the effects on ΔV_th and ΔI_off.

To model the dependence, we assumed that three factors­–mask open area ratio at the wafer level (global), chip level (semi-local), and local level (local)–affect ΔCD and Δd_R. We performed experiments using a dual frequency capacitively coupled plasma system. We used wafers ranging from 60 to 91 % (the global range (R_G) and the semi-local (R_S)) with various patterns of the photo-resist mask on the Poly-Si film. These samples were treated by the HBr/O₂ process under a pressure of 30 mTorr.

We found that ΔCD had positive and linear correlations with the global, semi-local, and local levels, which was consistent with the trend of the by-product (SiBr_x) intensity and with that of the taper angles of the etched profiles. We also clarified that ΔCD was affected by the amount of SiBr_x generated within several times of the mean free path area for the semi-local dependence, and that it was the solid angle (S) viewing from a pattern, not the pattern space, that had a good correlation with variations in ΔCD as a control indicator. We used this experimental knowledge to model the SiBr_x flux and created a Si gate etching simulation that demonstrated the ΔCD value and etched profile trends. We also found that Δd_R depended on (R_G+R_S)S as well as on the dosage of incident particles, considering the relationship between d_R and the ion energy reduced by the SiBr_x deposition depth [3]. Furthermore, when we used the analytical transistor model [3], we could predict ΔV_th and ΔI_off enhanced by the Si recess with (R_G+R_S)S.

[1] T. Ohchi et al., JJAP 47, 5324 (2008).

[2] T. Ito et al., JJAP 50, 08KD02 (2011).

Experimental - Samples were Si-doped GaN epitaxial layer with thickness of 1 μm grown by metal organic chemical vapor deposition (MOCVD) on sapphire substrates. After removed native oxides off chemically, the samples of GaN were partially etched off by using capacitively coupled plasma (CCP) reactor. Pure argon gas of 50 sccm was introduced and a pressure of 10 Pa was maintained. Plasmas were sustained by applying radio frequency (13.56 MHz or 100 MHz) power to a sample stage which temperature was maintained at 600°C. After the plasma processes for 10 min., morphology of the sample was measured by the atomic force microscopy (AFM), chemical composition was evaluated by the x-ray photoelectron spectroscopy (XPS).

AFM images of initial sample was shown smooth surface and also surface morphology was not significantly changed Just after annealed at 600°C for 10 min. However, after exposed Ar plasma at elevated temperature of 600°C for 10min, surface roughness was observed in particular. This is considered that the observed roughness caused by forming metallic-Ga precipitates on the surface, which was supported by results increase of metallic Ga bonding by XPS analysis. Namely, this indicates that favorable volatilization of nitrogen created Ga-droplets and promoted to roughen the surface at 600°C.

On contrary, surface morphology after N₂ plasma was relatively smooth although the high temperature. The results in smoothness were possibly resulted by difference in Ga droplet formation because of suppression of metallic Ga formation by reacting of reactive N species.

In this study, we report a method to control surface properties on plasma etched GaN through experiments carried out with samples at elevated temperatures. Further detailed discussion will be conducted.

1) R. Kawakami et al., Thin Solid Film. 516, 3478 (2008).

2) R. Kometani et al., Jpn. Soc. Appl. Phys. Spring Meeting (2012), 30a-M-11.

The III-V compound quantum dot (QD) has recently become extremely attractive when researching the quantum effect and developing novel opt/electronic devices such as high-efficiency intermediate-band solar cells. In the latter one, it is very important to precisely control the geometry size and alignment of QDs to form an intermediate-band. A high in-plane density is also necessary for photovoltaic devices in order to obtain a large optical gain. Molecular beam epitaxy have widely used and have been greatly researched for forming QDs. However, it is still a great challenge for realizing ideal nanostructures using self-organized bottom-up processes.

Our proposed top-down process of using bio-template [1] and neutral beam etching (NBE) [2] has great potential to fabricate defect-free, high-density (more than 7×10¹¹ cm^-2), sub-20-nm-in-diameter GaAs QDs structures [3]. This process uses 7-nm-in-diameter iron cores of ferritin (protein included iron core) as the etching mask and damage-free NB with eliminating UV photons and high-energy ions to etch GaAs without defects.

In this study, we successfully fabricated high aspect ratio nano-pillars etched the GaAs/AlGaAs stack-layered structure with 95 nm height and 15 nm the diameter by using a NBE. The diameter of the GaAs nanodisk could be precisely controlled from 12 to 18 nm by a combination of Hydrogen-radical treatment and Cl₂-NBE. This is because a taper profile of GaAs-NBO film could be controlled by Hydrogen-radical treatment time due to the isotropic and slow etching rate. In addition, not only iron cores but also this tapered GaAs-NBO film can work as etching masks because of the high NBE selectivity of GaAs-NBO. Thus, the diameter of the GaAs nanodisk could be controlled using the surface GaAs-NBO film taper profile produced by the prior Hydrogen-radical treatment time before NBE. This result means that our fabricated nanodisk array structures have great potential for high performance III-V compound optical QDs devices.

[1] I. Yamashita. Thin Solid Films. 393. 12. (2001).

[2] S. Samukawa. Jpn. J. Appl. Phys. 45. 2395. (2006).

[3] X. Y. Wang, et al., Nanotechnology, 22. 365301. (2011).

To fabricate MEMS devices, it is necessary to etch large-scale three-dimensional (3D) structures. Plasma etching is widely used for this purpose, but plasma irradiates charged particles and high energy UV photons and they may cause problems.

Ion sheath exists between bulk plasma and the etched sample. It accelerates and collimates ions in plasma and enables vertical etching. However, when the sample has large-scale 3D structure whose height is similar to or larger than the sheath thickness, the sheath is distorted along the surface. This causes distorted acceleration of ions and results in distorted etching shape. Also, high energy UV photons from plasma generates defects and deteriorates the mechanical property. It is reported that quality factor (q factor) and resonance frequency (f) were decreased by plasma irradiation [1]. In this study, we investigated 3D, damage-free, and high aspect ratio silicon etching for MEMS application using neutral beam (NB) to solve these problems. NB source developed by Samukawa et al. [2] can achieve high neutralization efficiency and elimination of UV irradiation damage.

First, large-scale 3D structure etching was investigated by using a sample with a vertical step of up to 725 µm high. In cases of plasma etchings (Cl₂ ICP etching and Bosch process), distortion of etching shape was observed. The etching shape distortion is concluded to be due to ion sheath distortion, based on the dependence on the step height and plasma parameters. On the other hands, no distortion was observed in the case of NB etching. This is because accelerated and collimated neutral particles without electric charge are used in NB etching.

Next, etching damage on mechanical property of silicon was investigated using microcantilever structure. Ar plasma and neutral beam were irradiated at the surface of the samples. Plasma irradiation degraded q factor and f of the cantilever, but neutral beam did not. This is due to crystalline defects generated by the UV irradiation from plasma. It is regarded that phonons are scattered by these defects, which causes loss of vibration energy and decrease of q factor and f.

In conclusion, MEMS process suffers both of sheath distortion and UV irradiation problems and neutral beam can solve these problems.

A part of this work was supported by the New Energy and Industrial Technology Development Organization (NEDO). This work is partly supported by Creation of Innovation Centers for Advanced Interdisciplinary Research Areas Program.

[1] M. Tomura et al., Jpn. J. Appl. Phys. 40, 04DL20 (2010).