Masami Inoue

Mechanism of Radical Control in Capacitive RF Plasma for ULSI Processing

The radicals of capacitive plasmas actually used in mass production were analyzed using various measurement systems. The composition of radicals in bulk plasma depends on the gas chemistry, the dissociation process, and interaction with the wall. It is revealed that parent gas (C4F8) is dissociated by multiple collision with electrons according to τne , where τ is the residence time, ne is the electron density, σ is the dissociation collision cross section and v is the electron velocity. A high-performance etching process, which can realize 0.09 µm contact holes with aspect ratio of 11, was achieved using a short residence time to suppress the excess dissociation and the control of deposition species through the addition of O2 to C4F8/Ar plasma as well as the reduction of the density of F radicals through the reaction with the Si wall.

Mechanism of C4F8 dissociation in parallel-plate-type plasma

To investigate the mechanism of C4F8 dissociation in parallel-plate-type plasma, we used several of the latest diagnostic tools and made extensive measurements of electrons, radicals, and ions under conditions that greatly suppressed the effects of plasma-surface interaction. These measurements showed that the amount of light fluorocarbon radicals and ions increased with increasing electron density. The dissociation of C4F8 was analyzed by using rate equations, after confirming the stability and uniformity of the plasma. The total dissociation rate coefficient of C4F8 was 1×10−8 cm3/s, and CF2 radicals were mainly generated from products of C4F8 dissociation. F was mainly generated from CF2 by electron-impact dissociation and lost by pumping. We could estimate that the C2F4 density was roughly comparable to the densities of CF and CF3, and that the surface loss probability of C2F4 increased with increasing electron density. C2F4 might play an important role in the etching because of its rich polymerizatio...

Plasma-Wall Interactions in Dual Frequency Narrow-Gap Reactive Ion Etching System

The effects of plasma-wall interactions on the plasma chemistry in a dual frequency narrow-gap reactive ion etching (RIE) system is investigated as a function of electron density, residence time, and partial pressure of additive O2 gas. It is found that there is a critical point in the residence time, where the dissociation dominant region and the wall-interaction dominant region are separated. The net flux of chemically reactive species is estimated. The net flux of carbon-inclusive ions is of the order of 1016 cm-2s-1, and is larger than that of CFx radicals. Carbon-inclusive and silicon-inclusive ions dominate the total flux of chemically reactive species. The deposition rate of the fluorocarbon film is strongly dependent on the O2 partial pressure, and is controlled by the chemical etching with oxygen as well as the reduction of fluorocarbon radical density in the gas phase. Based on the estimation of the net flux of chemically reactive species and the effect of added oxygen, dominant chemical species that control the etching reaction in the RIE system are discussed.

Characterization of 100 MHz inductively coupled plasma (ICP) by comparison with 13.56 MHz ICP

The effect of the excitation frequency on the dissociative process of the C4F8 gas was investigated by comparing a 100 MHz [very high-frequency (VHF)] inductively coupled plasma (ICP) with a 13.56 MHz [radio frequency (rf)] ICP. The same apparatus except for the wave generator and matching network was used for both ICPs in order to investigate the frequency effect as precisely as possible. The electron density and electron temperature in an Ar plasma were measured by using a Langmuir probe. From the dependence of the radial distribution of the Ne on the ICP source power, it was found that the rf ICP was produced in a cylindrical space under the coil area, while the VHF ICP was generated throughout the reactor. In C4F8/Ar plasma, the CFx (x=1, 2, 3) radical densities near the reactor wall were measured by using appearance mass spectrometry, and the F radical density was measured by using actinometry through the optical emission spectroscopy of Ar (750.4 nm) and F (703.7 nm). The degree of dissociation of t...

CF and CF2 Radical Densities in 13.56-MHz CHF3/Ar Inductively Coupled Plasma

Radial and axial distributions of CF and CF2 radical densities in CHF3/Ar inductively coupled plasma (ICP) generated by a radio frequency of 13.56 MHz were evaluated using laser-induced fluorescence (LIF). The radical densities in the bulk plasma were estimated by combining the density distributions measured by LIF with the absolute radical density measured by appearance mass spectrometry (AMS) near the reactor wall. Axial and radial distributions of CF and CF2 radical densities were extremely hollow. The radial distribution was strongly correlated with the electron density distribution and was significantly influenced by the dissociation process resulting from the electron excitation in the reactor space. The axial distribution, on the other hand, was mainly determined by the surface reaction of radicals on the wafer.

Replicating characteristics by SR lithography

Process optimization and pattern replication on various substrates by synchrotron radiation lithography was carried out to evaluate the problems for 0.15 micrometers level resists patters for 1-Gbit dynamic random access memory. It was found that the exposure latitude was rather restricted by the resist residue remaining between lines (scum) and the pattern collapse than the normally used +/- 10-percent critical dimensions. A simple Fresnel diffraction calculation including the phase-shifting effect and mask contrast showed that the occurrence of the scum was mainly determined by the optical images of x-rays, and could not be significantly improved by the resist process condition. We used the mask/wafer proximity gap of 20 micrometers to get a good optical image and 5000-angstrom resist thickness to suppress the pattern collapse. On the other hand, the replicating characteristics on the light element substrates were similar to that on the Si substrate, especially good on the SiN substrate, but the residues caused by secondary electrons ont he metal substrates and the catalytic reaction on the Pt substrate were observed. It was shown that protection layers could suppress those residues and serve a good pattern profile.

Evaluation of Acid Diffusibility in a Chemical Amplification Resist Using Acidic Water-Soluble Film

A new method is proposed to evaluate the acid diffusibility in a chemical amplification resist using acidic overcoat film. This method is easy and effective for both positive- and negative-tone resists and is not influenced by the uncertainty in optical images. The acid diffusion coefficient was estimated for various process conditions. The acid loss reaction was introduced in Ficks law, and the calculation showed a good agreement with experimental results. In order to obtain good profile patterns with sufficient controllability, the distribution of acids in the resist must be controlled as well as diffusion length. The resist patterns replicated using X-ray lithography corresponded to those expected from experimental results.

Plasma research activities in the association of super-advanced electronics technologies

Association of super-advanced electronics technologies (ASET) is a Japanese electronics research and development consortium that was founded on 29 February 1996. The target of the plasma research group in ASET is to make breakthroughs for future dry etching technology by investigating the mechanisms of dry etching scientifically. The plasma research group is investigating plasma diagnostics, plasma generation and its transportation, plasma surface reaction and vapor phase reaction, plasma modeling/simulation directed toward plasma control, and the mechanism of silicon oxide etching in high aspect ratio, narrow contact holes, and will develop a new chemistry, a new plasma source, and a new monitoring method.

MONITORING OF ELECTRON ENERGY DISTRIBUTION CHANGE FROM OPTICAL EMISSION FOR NONMAGNETIC ULTRAHIGH-FREQUENCY PLASMA

Fractional electron density, which is partial electron density in energy distribution, has been measured in a high-density non-magnetic ultrahigh-frequency (UHF) plasma from optical emission of rare gases (Xe, Ar, He). The technique was applied to rare gas mixed plasma, as well as fluorocarbon gas containing plasma. In the calculation procedure, the fractional electron density was assumed to be constant between the two threshold energies of the different emissions. In the experiment, total electron density was changed by changing the UHF source power without changing electron temperature measured by a single probe. However, in UHF plasma, the fractional electron density between the threshold energy of ArI and HeI emissions increased more than total electron density increase. This result was obtained both for Ar/He plasma and Ar/C4F8/He/Xe plasma. On the other hand, fractional electron density over the threshold energy of He increased at about the same rate of or less than the total electron density increase. In addition, the fractional electron density between the threshold energy of XeI and ArI in Ar/C4F8/He/Xe plasma increased less than total electron density increase. These results indicate that the electron temperature, which is commonly used as a typical index value of electron energy of the plasma, does not reflect the fine structure of EEDF. The optical technique can supplement this point, especially at the high energy tail of EEDF.

Analysis of Overlay Accuracy in 0.14 µm Device Fabrication using Synchrotron Radiation Lithography

We applied synchrotron radiation lithography to the fabrication of a 0.14 µ m rule dynamic random access memory structure using high dielectric film capacitor and trench isolation techniques. An overlay accuracy of 80 nm was obtained. In order to clarify the contributions of the error causes to the overlay accuracy, we mathematically classified the overlay error into four components: rotation, translation, magnification and in-plane deformation (IPD). A rotation and translation error was estimated to be 38 nm. A wafer process-induced error such as a rotation and translation due to a degradation of the alignment mark worsens the overlay error by 33 nm. We have also found that the maximum magnification error including the mask magnification was 4 ppm and the magnification correction was necessary to achieve an overlay accuracy less than 50 nm. The IPD was estimated to be 60 nm, and the major cause was a mask overlay error. The analyzed results suggest that an IPD less than 30 nm and a rotation/translation component less than 30 nm are required to achieve an overlay error less than 50 nm.