Relative merits of Cl2 and CO/NH3 plasma chemistries for dry etching of magnetic random access memory device elements

Abstract

Long term stability of dry etched magnetoresistive random access memory elementsHigh density submicron magnetoresistive random access memory (invited)A typical magnetic random access memory stack consists of NiFe/Cu/NiFeCo multilayers, sandwiched by contact and antioxidation layers. For patterning of submicron features without redeposition on the sidewalls, it is desirable to develop plasma etch processes with a significant chlorinated etch component in addition to simple physical sputtering. Under conventional reactive ion etch conditions with Cl 2 -based plasmas, the magnetic layers do not etch because of the relatively involatile nature of the chlorinated reaction products. However, in high ion density plasmas, such as inductively coupled plasma, etch rates for NiFe and NiFeCo up to ϳ700 Å min Ϫ1 are achievable. The main disadvantage of the process is residual chlorine on the feature sidewalls, which can lead to corrosion. We have explored several options for avoiding this problem, including use of in situ and ex situ cleaning processes after the Cl 2 -etching, or by use of a noncorrosive plasma chemistry, namely CO/NH 3 . In the former case, removal of the chlorine residues with in situ H 2 plasma cleaning ͑to form volatile HCl that is pumped away͒, followed by ex situ solvent rinsing, appears effective in preventing corrosion. In the latter case, the CO/NH 3 plasma chemistry produces metal carbonyl etch products, that are desorbed in the simultaneous presence of an ion flux. The etch rates with CO/NH 3 are much lower than with Cl 2 over a broad range of source powers ͑0-1500 W͒, radio frequency chuck powers ͑50-450 W͒, pressures ͑1-30 mTorr͒ and plasma compositions. We have tried substitution of CO 2 for CO, and addition of Ar to produce faster etch rates, without success. Maximum rates of ϳ300 Å min Ϫ1 for NiFe and NiFeCo were obtained with CO/NH 3 under optimum conditions. The etched sidewalls tend to be sloped because of mask erosion during plasma exposure, in contrast to the case of Cl 2 -based chemistries where the sidewalls are vertical.

Suboxide/subnitride formation on Ta masks during magnetic material etching by reactive plasmas

“…Also, for magnetic metals, there have been various attempts to devise efficient RIE processes for their smallscale patterning. For example, etching of magnetic materials by Cl 2 based plasmas has been proposed, [5][6][7][8] but under typical conditions of chlorine etching of magnetic materials, metal chlorides are not volatile and corrosion occurs on etched metal surfaces.…”

Correlation between dry etching resistance of Ta masks and the oxidation states of the surface oxide layers

“…Plasma etching of magnetic materials such as CoFeB, NiFe, and PtMn has been extensively studied with plasmas of CO/NH 3 or CH 3 OH gases. [9][10][11][12][13][14][15][16][17][18] The advantages of these processes include no corrosion of magnetic materials during the process and high etching selectivity of hard mask materials such as Ta, Ti, and TiN over magnetic materials. Matsui et al 12 pointed out that the high mask selectivity was due to the formation of a hardening surface layer on the mask by nitridation, carbonization, or oxidation.…”

Negative electron-beam resist hard mask ion beam etching process for the fabrication of nanoscale magnetic tunnel junctions

“…8,9 For such device applications, pattering, and therefore etching, the component magnetic materials becomes a critical process for fabricating stable devices. Reactive ion etching (RIE) [10][11][12][13]15 based on chlorine and methanol gases has been used for nanoscale magnetic materials etching. But it shows after-corrosion and oxidation problems.…”