A Brief Tour Around the World of RF Plasma Processing
-Neil Benjamin, Lam Research Corporation
Today it is virtually universal to use RF excited plasmas to process semiconductor materials. This may extend to hundreds of process steps as Deposition, Etching, Stripping, Cleaning and Surface treatments are all in the RF-plasma repertoire. In order to do so, multiple factors and timelines have had to converge, including Electronics development, specifically RF technology and devices, and Plasma Technology A.K.A. Gaseous Electronics, specifically Dry Processing as applied to: Semiconductors, Flat panel displays, P-V solar panels, MEMs devices etc.
It is less than 70 years since the invention of active solid state electronics in 1947, but the semiconductor industry is now mature and consolidated while continuing to advance according to Moore’s law. In the same period RF delivery systems have also progressed from high power vacuum tubes/valves to solid state devices in the 1980s. Most such RF systems use 50Ω transmission lines (for I.S.M.) so that matching networks are used to optimize power transfer to the antenna load impedance. Plasma technology use for semiconductor production did not start until the late 1960s / early 70s. In particular, despite the engineering complications of dealing with RF excitation, RF plasmas became popular because of their suitability for use with dielectric materials, and ameliorating the potential for damage caused by passing DC currents through delicate devices during manufacture.
The technology involved in plasma processing has remained basically the same for nigh on 50 years, but the demands on RF performance, control and consistency have escalated beyond all recognition. I will address some typical RF-plasma issues that continue to challenge us with examples taken from the current state of the art.
Feature Scale Modeling of Semiconductor Processes
-Philip Stout, Applied Materials
An overview of Monte Carlo feature scale modeling work will be presented. The two major areas of discussion will be etching and metallization processes. In high aspect ratio (HAR) oxide etch processes the mask gates the amount of etchants and passivants entering the feature and has a large influence on the resulting etched profile. Mask sidewall slopes alter the path of ions entering the feature thereby modifying the ion strike map inside the feature. Mask geometry also influences polymer deposition within mask and bow formation in oxide. Mechanisms for off-axis profiles and profile distortion include: off-axis ion incidence to wafer, non-uniform polymer deposition at opening, re-deposition of etch byproducts, feature geometry (mask), mask reflow, charging in feature, and off-angle yield curve peaks. Two cases illustrate the interplay of these profile distortion mechanisms: pattern distortion dependence on etch stop layer charging properties, and the influence of a tilted hard mask on HAR trench oxide etch profile. Feature scale models can be used to study integration issues in multi-step processes. A thirteen step spacer double patterning integration has been studied showing the importance of the spacer etch step. An STT-MRAM (Spin Transfer Torque - Magnetoresistive Random Access Memory) etch process will be discussed. Removal of metal sidewall deposits resulting from re-deposition of sputtered MTJ metal layers is a major issue. The study looks at ion beam etching. The metallization topics reviewed will include copper physical vapor deposition (PVD) in dual-damascene (DD) features, predicting across wafer coverage in feature, and copper reflow studies. In DD features a sloped inner via sidewall can have faster yields than the trench bottom. With reactor models supplying across wafer flux and AEDFs it is possible to predict feature coverage properties as a function of wafer position. With smaller feature sizes copper reflow is being explored as a means to fill via and trench structures for back end of line interconnects. Using a simple hopping surface diffusion model, reflow behavior is shown. The model predicts the initial reflow causes rounding of the Cu surfaces and a shrinking of the opening as the surfaces round to a more minimal surface configuration.
Optimization of Linear Scanning Magnetron Array Performance
-Vladimir Kudriavtsev, Alexandru Riposan, Robert Norris, David W. Brown, Chris Smith, Terry Bluck, Intevac, Inc.
In this presentation we discuss Linear Scanning Magnetic Array (LSMA) technology for magnetron sputtering in conjunction with in-line substrate processing. In this approach, the magnet array (pole) scans over planar target spreading the erosion pattern in a controlled fashion. Thus, high quality, dense films with good uniformity can be produced at significant advantages over static magnetrons, such as significantly higher target utilization, longer uptime, and prevention/removal of target defects related to re-deposition. We review the influence of magnet motion acceleration/deceleration, the influence of endpoint motion offset (stagger), and the influence of magnet - to - substrate velocity ratio, on target utilization and lead-to-trail edge film uniformity. Trade-offs between uniformity and target utilization were established and characterized. The optimization method we use employs a combination of theoretical simulations and experimental measurements. Theoretical analysis utilizes ANSYS static magnetic field simulations, erosion profile calculations including motion integration effects, and ray tracing method for sputtering film thickness calculations (MATLAB). The structure and uniformity of LSMA-deposited thin films was characterized experimentally by XRF and SEM, and the target erosion measured by weight and erosion profiles of spent targets. We have demonstrated that, with a judicious design approach, an optimal range of operating parameters can be defined and target utilizations above 70% can be reached, while maintaining deposition uniformity below 2% with excellent film properties. This makes the LSMA plasma source (using planar targets) more economically competitive than static and rotatable magnetrons.
Molecular Dynamics Simulations of Atomic Layer Etching by Low Energy Ions
-Jun-Chieh Wang, Shahid Rauf, Jason Kenney, Leonid Dorf, and Ken Collins, Applied Materials, Inc.
In the semiconductor industry, the use of atomic layer etching (ALE) makes it feasible to accurately control the critical dimensions to nanometer level or smaller. In ALE, the target substrate is first exposed to a reactive gas that passivates the surface, which is then followed by ion bombardment with energy below the sputtering threshold. It is critical to precisely control the ion energy and flux during the etching process to remove the topmost layer of the passivated surface without damaging the underlying substrate. Once the passivation layer is removed, the etch process stops. The passivation and etching steps are repeated until one has etched to the desired thickness. In contrast to conventional plasma etch processes, microfabrication using ALE promises high selectivity and low damage to the substrate. In this presentation, we discuss the properties of ALE using results from molecular dynamics (MD) simulations. The simulation procedure is conceptually similar to those described in previous publications [1,2]. In this study, a crystalline Si(100)-(2x1) or amorphous surface (made by low energy Ar+ ion bombardment) was generated and equilibrated at room temperature. The bottom layers were fixed in space, and the periodic boundary conditions were applied laterally to remove the boundary effect. The ions are modeled as energetic neutrals. The surface was passivated by repeated bombardment with low energy Cl atoms at normal incident, which was followed by Ar+ or Cl+ ion bombardment to remove the passivation topmost layers. The Berendsen scheme is used between ion/neutral impacts to remove the energy from the surface region and cool the surface layer to room temperature. The Stillinger Weber (SW) type potentials are used for Si-Si, Si-Cl and Cl-Cl interactions. The Ar-Si and Ar-Cl interactions were modeled using Moliere potentials. The leap-frog form of Verlet algorithm was used to numerically integrate the Newton’s equation of motion. The MD is applied to study several variants of the ALE process. The fundamental properties of Si etching are also investigated for both bare and Cl-passivated Si surfaces with several ions including Ar+, Cl+ and Cl2+. These fundamental studies are used to interpret our layer-by-layer ALE experiments in our laboratory.