Handbook of Thin Film Process Technology

98/1 Reactive Sputtering

David A Glocker

• Publisher: CRC Press
• Year: 2018
• ISBN:
  9781351089685
• Edition: 1
• Topic: Naturwissenschaften
• Subtopic: Physik

The Handbook of Thin Film Process Technology is a practical handbook for the thin film scientist, engineer and technician. This handbook is regularly updated with new material, and this volume is a special issue on reactive sputtering which will be of interest to a wide range of industrial and academic researchers in addition to owners of the main Handbook. Some recent developments in the reactive sputtering field are covered, including unbalanced magnetron sputtering and pulsed reactive sputtering. The articles contain a wealth of practical information relating to applications, practice and manufacturing techniques.

A5

---

Reactive Sputtering

A5.0 Reactive Sputtering: Introduction and General Discussion

William D Westwood

A5.0.0 INTRODUCTION

Reactive sputtering is a method to deposit films which have a different composition from the target by adding a gas to the sputtering system to produce a reaction of the gas with the target material. The reaction may be controlled to dope the film (e.g. nitrogen doped Ta films) or enough gas may be supplied to ensure complete reaction of the sputtered metal. In the former case, the main sputtering gas is Ar or another inert gas but the fully reacted film may be sputtered either in a similar gas mixture (e.g. Ar/N₂) or in the reactive gas alone. Reactive sputtering was discussed briefly at the end of section A3.0.

Reactive sputtering was probably first observed by Grove [¹] in the initial discovery of the sputtering phenomenon; he sputtered Ag in residual air due to the poor vacuum available at the time. Reactive sputtering has often been carried out inadvertantly because of inadequate understanding of the reactivity in the discharges used for sputtering. However, the first real application of reactive sputtering was to deposit Ta₂N films for resistors in thin film hybrid circuits [¹]. By controlling the flow rate of N₂, the electrical properties of the film was adjusted to the required value by doping the Ta. If the N₂ flow is too high, the completely reacted TaN is deposited. The sputtering of these resistor films was thoroughly covered in a book [¹] but this predated the use of magnetron sputtering systems. The development of reactive sputtering methods and the range of applications have been reviewed [⁴, ⁵].

Reactive sputtering for film deposition was carried out initially by diode sputtering with simple DC power supplies. Both metallic (e.g. Ta) and non-metallic (e.g. NiFe₂O₄) targets were sputtered; the oxide ceramic targets were made sufficiently conducting by firing in a reducing atmosphere [¹]. High quality dielectric films were deposited; e.g. Ta₂Oi optical waveguides [¹]. Although dielectric layers formed on the targets, they were sufficiently conducting for the low current densities of diode sputtering. However, when high current density magnetrons were used, serious arcing caused defects in the films. The arcing is caused by the charging of the insulating layers by the ion current in the plasma and the subsequent breakdown of the dielectric when the breakdown field of the dielectric is exceeded. When the current density is low, this charge is dissipated by the leakage current through the layer but the commercial requirement is for high deposition rates.

The solution to the problem of insulating targets was solved in the 1960s by the introduction of RF (typically 13.56 MHz) diode sputtering [¹]. RF sputtering is discussed in A3.0.1. Due to the much higher mobility of electrons than ions, the target potential is established at – V₀ relative to the plasma, where V₀ is the amplitude of the RF voltage; the target can be positive relative to the plasma for only a small part of each RF cycle so that the positive (ions) and negative (electrons) charges reaching it over the complete RF cycle are equal, as required by the RF circuit. In fact, RF sputtering is ideally suited to two identical targets. In this ‘symmetric’ case, the processes at one electrode are identical to those at the other except for a phase difference; both targets are sputtered at the same rate. However, this ‘symmetric’ case was not well suited to the diode system geometries in use at the time. To make the RF diode system similar to the DC diode, the symmetry was broken by increasing the capacitance of one electrode relative to the other; this reduces the potential difference between the ‘substrate electrode’ and the plasma so that it approximates the anode in the DC diode.

The symmetric RF diode system was used, with the targets facing (or opposed), in cases where negative ions (e.g. O⁻) are sputtered from the target (e.g. ZnO, highT_c oxides); the substrates are located off to the side to avoid the flux of highly energetic species [¹].

RF power supplies were applied to magnetron targets and solved the problem of insulating targets becoming charged. However, it did not solve the other problem with many insulating targets; i.e. the intrinsically low sputter yield of some materials. For example, the yield of A1 from A1₂0₃ is about 5% of the yield from Al.

Another problem with RF magnetron sputtering was explored by Este and Westwood [¹]; the sputtering rate was only about 50% of that for DC or low frequency (<30 kHz) sputtering because only 50% of the power registered by the RF supply was dissipated in the targets. Although a full explanation for this effect is not yet available, the loss of sputtering power at high frequency is real.

Thus, there are two basic problems with sputtering insulating targets or insulating regions or layers on metal targets; first, charging of the target layers; second, the low sputter yield for these materials. The latter may still occur if the compound layer on the target is a conductor; e.g. TiN is a good conductor but the yield of Ti is still only 25% of that from Ti.

Different approaches have been taken to try to solve these problems. In the following parts of this section, several techniques are described in detail; section A5.1 addresses the control issues involved with maintaining the target free of the compound using basically a DC power supply, although it is actually modulated in a step-wise function to minimize the effect of any arcing which may occur. This is essentially an extreme case of ‘doping’ the film; i.e. the reactive gas must be controlled at the correct level. In doping, such as for the TaN_x resistor films, the N₂ is controlled to give the correct value ofx. In depositing a fully reacted compound like Ta₂O₅, the O₂ has to be controlled to prevent oxide formation on the target. The specification of ‘fully reacted’ is important. The Ta₂O₅ film can be visually transparent but still be oxygen deficient and thus have poor dielectric properties. However, a slight oxygen deficiency is advantageous in oxide films (e.g. SnO₂) used as transparent conductors.

Section A5.2 deals with the use of full wave modulation of the power to prevent target charging even when the conducting target is covered by an insulating layer. It is usually preferable to use a metal target; fabrication is simpler, and less expensive, than for a ceramic target which involves calcining, pressing and firing. The metal target is also likely to have a lower impurity content and is easier to cool because of the higher thermal conductance. However, there are some materials (e.g. LiNbO₃) for which a metallic target is not feasible; the sputter yields are typically low in these cases. Full wave modulation may be either AC (∼ 40 kHz) or RF (13.56 MHz) frequencies.

The most appropriate method will depend on the application. A major consideration is often the deposition rate which can be obtained. Table A5.0.1 gives the maximum possible deposition rates for some examples, assuming that the erosion and deposition areas are identical. These calculations are based upon the sputter yields for the elements and the measured ratios of deposition rates of metal and compound from a target covered by the compound layer. The rate per W for the element (row 5) is calculated from the elemental yields for 500 eV Ar⁺ sputtering (row 1); the current density is 2 mA cm⁻². As an example, for a typical circular planar magnetron with an erosion track area of 50 cm² (i.e. uniform sputtering from a track with inner and outer radii of 3 and 5 cm), the erosion rate per kW of sputtering power is given in row 6. If the reactive gas simply added to the element and the deposition and erosion areas were equal, the compound deposition rates in row 7 of the table would be obtained; i.e. the element is sputtered from the target by Ar⁺ and the atoms react at the substrate. The deposition rate therefore increases.

These deposition rates cannot be achieved in a real system; first, the elemental sputtering rate decreases since some reactive gas atoms reach the target and are sputtered instead of the element;...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Table of Contents
• A5 Reactive Sputtering Coordinating Editor: W D Westwood
• Appendix A: List of Contributors
• Subject Index