AVS2011 Session PS-WeA: Plasma Sources

This paper describes a new remote source technology employing a high efficiency VHF electrostatic coupling method to produce a versatile and robust remote plasma generator. The design approach is shown capable of producing high density discharges across a broad range while operating at relatively low voltages. The concept is scalable and adaptable to most any chemistry used for cleaning, etching and even deposition. This paper describes some of the important design elements incorporated into these new source devices along with early results illustrating a broad performance range exceeding many capabilities of the alternative technologies.

Conventional bias systems use sine wave voltage systems to achieve ion energy distribution functions for the creation of thin films. By combining dual frequency sine waves, the mean energy and its spread can be independently controlled¹. Arbitrary wave-shaping has been suggested to create single energy near-delta function distributions². We investigate a system where the waveform is defined a priori where two elements are run in a feed forward system to control the instantaneous IEDF. We evaluate the effectiveness of this system in an argon/oxygen plasma at typical operating pressures inside the 10-150 mT range at plasma densities in the low 10¹⁰ cm^-3 range.

1 S. Shannon et al.; J. Appl. Phys. 97, 103304 (2005)

2 Wendt A. et. al., “Method and apparatus for plasma processing with control of ion energy distribution at the substrates”, US Patent 6201208, March 13, 2001

Inductively coupled plasma (ICP) sources, commonly used for semiconductor and conductor etching, embody the concept of functional separation between plasma production and ion energy control, wherein the inductive coupling through the coils is only responsible for the plasma generation while the bias determines the ion energies. Plasma etching of microelectronics structures at advanced technological nodes (< 3x nm), especially complicated structures such as multi-gate MOSFETs and 3D memory stacks, are placing great emphasis on control of ion energy distributions (IEDs) to finely discriminate etching thresholds.[1] Sinusoidal biases typically provide broad IEDs, making such control difficult to achieve. One promising alternative is non-sinusoidal bias waveforms, which have been demonstrated to provide such control.[2] However, there are issues associated with passing non-sinusoidal signals through the finite impedance of the match, transmission line, and the substrate.

Multi-frequency bias is compatible with current manufacturing hardware, and has been successfully used for controlling IEDs in capacitively coupled plasmas.[3] This approach may be utilized to enable IED control in ICP etchers as well. We investigate the impact of multiple bias frequencies in this paper. A high frequency applied in addition to a relatively low frequency bias causes the sheath potential to vary in a complicated manner due to the non-linear nature of the sheath. As a result, IED exhibits a complex dependence on relative bias voltages and frequencies.

In this work, the effect of applying bias at multiple frequencies will be discussed using results from a computational investigation. The 2-dimensional plasma equipment model, HPEM[4], has been modified to enable power deposition at multiple frequencies on the same electrode. Results will be discussed for Ar/Cl₂ plasma utilizing a 13.56 MHz bias in addition to a bias at a different frequency in an ICP chamber. The additional frequency is varied over a wide range and its consequences assessed on the ion and radical flux and IEDs incident on the wafer. We found that, in addition to modulating the IEDs, the flux composition is different depending on the frequency due to the secondary plasma generation by the alternate frequency.

[1] A. Nitayama and H. Aochi, ECS Trans. 18, 89 (2009).

[2] A. Agarwal and M.J. Kushner, J. Vac. Sci. Technol A 27, 37 (2009).

[3] S. Shannon, D. Hoffman, J.-G. Yang, A. Paterson, and J. Holland, J. Appl. Phys. 97, 103304 (2005).

Low pressure (sub-20 mTorr) capacitively coupled plasmas (CCP) are playing an increasingly important role in technological applications. As the mean free path becomes commensurate with the discharge dimensions, the fluid assumptions inherent in plasma and sheath models start to break down and ought to be reexamined. We focus on one aspect of the CCP operation in this paper, namely the electron energy distribution (EED) at electrodes and surfaces, and use kinetic particle-in-cell (PIC) models to understand the temporal behavior of the EED. Kinetic results are compared to fluid representation of the EED at electrodes to identify deficiencies in the fluid model at low pressures and propose solutions.

The sheath at the plasma-surface interface ensures that the electrons remain confined in the bulk plasma. However, during certain phases of the radio-frequency (RF) cycle in a CCP, the sheath collapses and the electrons exit at the surface. Energy distribution of these electrons contains useful information about the bulk plasma and the sheath. One can probe into the energy characteristics of these electrons using dc probes embedded in the electrode. Analysis of the resulting probe data can be used to determine the electron temperature, the electron density, and the EED in the bulk plasma. If a fluid model is used for this analysis, the electrons are assumed to be governed by the Boltzmann relation where their density and flux depend exponentially on the sheath voltage. Electrons are however highly non-equilibrium near the sheaths in CCPs and the Maxwellian distribution assumption (implicit in the Boltzmann relation) is questionable. Furthermore, most probe analysis models are dc-based. Low pressure situations demand further scrutiny as even the bulk plasma EED tends to become non-Maxwellian.

1 and 2-dimensional PIC model of CCPs are used for this investigation. These models consider plasma chemistry using the Monte Carlo technique. Simulations are done for Ar and N₂ plasmas under a variety of conditions (13.56 – 60 MHz RF frequency, RF voltage of 100 – 500 V, 5 – 100 mTorr gas pressure). The 1-dimensional PIC model is used to examine the EED at the electrodes where the sheath undergoes substantial variation during the RF cycle. The 2-dimensional model is used to investigate the EED at small metal surfaces (e.g., a probe) away from the primary electrodes. Dc voltage is also applied to the probe electrode in the 2-dimensional simulations. It is found that, in addition to a non-Maxwellian contribution from electrons adjacent to the sheath, the EED also contains high energy electrons which are the remnant of electrons that were accelerated at the opposite sheath.

There are several principles in the consideration of plasma etcher design. This paper addresses two important areas: (1) the ability of tailoring the electron energy distribution function (EEDf), (2) the ability of adjusting the charging and neutralization of surface features (the electron shading effect). Stochastic heating by high frequency RF (VHF) energizes the Maxwellian bulk into the energetic tail population for efficient ionization. Such energetically bottom-up heating also indiscriminately populates the below-ionization energetic group that drives chemistry such as molecular dissociation and VUV production. A generic DC/RF system has a RF biased wafer-electrode with a high-negative DC superimposed opposing electrode. The DC/RF system dominates its electron heating with an energetically top-down process. The secondary electrons emitted from the high-negative DC surface disseminate the beam-energy into a distribution of energetic-electrons through collisions and more importantly, various beam-wave instabilities. These energetic electrons are trapped between the sheaths of the two parallel plates when the RF sheath field is sufficiently strong, dissipating their energies mainly into ionization. The energetic part of the EEDf reveals (in descending order) a group of ballistic-electron associated with the applied –DC voltage, an energy continuum, and finally stop at a lower middle-energy peak in the range of ~ 40eV to 300eV depending on the process. The data show extremely efficient ionization by these energetic electrons and as a result, the Maxwellian bulk remains relatively unchanged at T_e~1.8eV regardless the bias RF power and process pressure. Such energetically decoupled EEDf enables increased ionization without increasing molecular dissociation. In one RF period, the trapping of these energetic electrons diminishes as the RF sheath collapses. By pulse-synchronizing the –DC voltage and the bias RF voltage, ion-bombardment excitation of the wafer surface can alternate with high-energy electron-bombardment neutralization of the wafer surface. Ion-bombardment of the wafer surface occurs when the bias RF pulse is high and the –DC pulse is at a medium level. As the RF and the DC pulses are synchronously altered to a low RF voltage and a high-negative DC voltage, ion-bombardment excitation of the wafer diminishes and the wafer surface sees an increased current of energetic (>25eV) electrons. High aspect ratio SEM data show that the signatures of electron shading (e.g., bending and twisting of the features) are eliminated when synchronous DC/RF pulsing is implemented allowing energetic electrons to reach the bottom of high aspect ratio features.

Negative plasma potentials were obtained in DC hot filament unmagnetized electropositive argon plasma in two configurations. For sufficiently low plasma density (<10⁶ cm^-3) bounded by conducting walls, double layers provide ion and electron confinement near the walls. Similar results were observed in higher density plasmas (~10⁹ cm^-3) when a thin dielectric coating of oil covered the surface of the conducting walls. The potential profiles, measured using emissive probes in the limit of zero emission, from the center of the plasma to the potential minima are quite similar in shape to those observed when the plasma has positive plasma potentials. The primary electrons emitted from the filaments are important for charge conservation and for modification of the Bohm criteria but are not important for current balance.

This work was supported by US Department of Energy grants No. DE-AC02-09CH11466 and No. DE-FG02-97ER54437, and the Fusion Energy Sciences Fellowship Program administered by Oak Ridge Institute for Science and Education under a contract between the U.S. Department of Energy and the Oak Ridge Associated Universities.