1. MAGNETIC MATERIALS

Permalloy (NiFe), other Fe-based materials, Co and Cu are widely used in the data storage industry and in magnetic sensors and actuators. A long-standing problem is that there has been no high rate dry etching processes for magnetic materials, seriously limiting the ability to batch-fabricate submicron magnetic devices. The situation is even more problematic for inherently nanoscale magnetic devices, such as spin-dependent electron scattering multilayer structures, spin-valves or spin-dependent tunneling devices, all of which are based on controlling the layer structure at the nanometer scale, while retaining the bulk microstructure essential for the ferromagnetism and/or spin-polarized band structures. Magnetic materials, whose main constituents are transition elements, are virtually inert in conventional dry etch processes like reactive ion etching. Patterning of magnetic devices such as magnetic read/write heads has been accomplished predominantly by Ar+ ion milling or additive deposition processes such as electroplating or lift-off. Typical recording heads currently have a track width of ~2mm, but to achieve disk drives with areal recording density of _10 Gbit-in-2 sub-micron trackwidth magnetic write heads will be necessary.

In this first year of the MURI we have successfully developed high-rate (_1,000Å-min-1) dry etch processes for Cu, Co, NiFe, NiFeCo, TaN and other magnetic materials using an Electron Cyclotron Resonance (ECR) high ion density plasma source. Six different plasma chemistries have been investigated, namely Cl2/Ar, BCl3/Ar, CH4/H2/Ar, IBr/Ar, ICl/Ar and SF6/Ar. The etching was characterized in terms of rate, anisotropy, surface roughness and presence of any etch residues. Typical etch rate results for different materials are shown in Figure 1 for Cl2/Ar, as a function of applied microwave power. The rates increase with microwave power as the ion density in the plasma increases from ~1010cm-3 at 400W to ~5X1011cm-3 at 1000W. We have found that Cl2/Ar produces the fastest rates of the plasma densities investigated. Similar results are shown in Figure 2 for Cu etching and for Ag, Co and Fe in two other plasma chemistries in Figure 3. The remarkable result is that even though the etch products such as CuClX, NiClX, etc. are normally quite involatile at room temperature, and require sample heating for desorption, the presence of the high ion flux under ECR etching conditions produces efficient sputter-assisted desorption of these species before a thick selvedge layer forms. Normally there is a build-up of the involatile etch product on the surface, which prevents further reaction with the substrate and thus prevents any etching occurring. We have found that by balancing the chlorine radical-to-argon ion ratio, we can efficiently sputter away the chlorinated surface layer and achieve high rates. If the chlorine concentration in the plasma is too high then a thick chlorinated selvedge layer forms, and if the Ar+ ion density is too high then the etch rates are low and the surfaces rough as one reverts to the physical sputtering regime. Figure 4 shows AES surface scans of an as-etched Cu surface, showing the presence of chlorine-residues (top) and after a subsequent in-situ H2 plasma clean (bottom), which reduces this residue.

The significance of this discovery is that it is applicable to virtually any hard-to-etch material. The use of Cu as interconnects in the Si semiconductor industry has been complicated by the lack of a room temperature etch process, necessitating the use of CVD processes.

We will continue to develop the etch process for NiFe, NiFeCo, Cu etc. in the coming contract period using another high ion density plasma source, the Inductively Coupled Plasma, which is easier to scale-up to large substrate sizes than ECR, and since it operated at radio frequencies does not require expensive waveguiding or magnets and is easier to tune.

2. COMPOUND SEMICONDUCTORS

One of the goals of the MURI is to establish processes for integration of magnetic materials with semiconductors to provide increased functionality through on-chip electronics connected to the magnetic devices. We have investigated the dry etching of InP, GaAs, InGaP, AlInP, AlGaP and related alloys in the same type of ECR plasma that were developed for patterning of the magnetic materials.

Figure 5 shows the ternary etch rates (x = 0.5 in all cases) in 1.5mTorr discharges of 2Cl2/13Ar, as a function of applied rf power for both RIE (0W microwave power) and ECR (1000W microwave power) conditions. At low rf powers the etch rates are typically 10-30 times higher for ECR conditions as a result of the higher Ar+ ion flux and dissociation of the chlorine. Optical emission spectroscopy showed significantly higher atomic (and ionic Cl+ and Cl2+) chlorine densities in the ECR discharges. The etch rates also increase with rf power as the ion energy increases and leads to more efficient removal of the etch products. The etched surface morphologies were smoother for all these materials at high microwave powers and high Cl2 concentrations in the Cl2/Ar mixtures. We ascribe this to the need to optimize the neutral/ion ratio in order to efficiently sputter away the group III chlorides without forming a thick selvedge, or reaction layer. If the ion density is too high the preferential sputtering of P occurs, leading to surface roughing. If the ion density is too low, then the chlorinated, group III-rich surface is produced. In both cases the resultant surface is non-stoichiometric and unsuitable for further processing.

Auger Electron Spectroscopy analysis of the surfaces showed there is no detectable chlorine residue after etching, and only the presence of oxygen from the native oxide and adventitious carbon from exposure to ambient when transferring the sample between the etch reactor and the AES system. Aluminum oxide appears to comprise a major component of the immediate surface. In a sample etched in the same type of plasma but under RIE conditions there is a clear In-enrichment relative to Al suggesting that equi-rate removal of the group III chlorides is not occurring. In both samples there is a P deficiency to a depth of Û 100è, based on a typical Ar sputter rate of ~60è min-1 during the AES depth profiling.

By performing all the studies in the same reactor, it can be shown conclusively that the major reason for the smooth, high rate etching of In-containing III-V materials in chlorine-based plasma chemistries is the presence of the high ion current under ECR conditions. The absence of this high ion flux in RIE discharges leads to formation of a thick selvedge layer consisting mainly of InCl3, which produces rough morphologies, non-stoichiometric surfaces and low etch rates. The advantages of operating at high ion densities in that simple BCl3 or Cl2 plasma chemistries can be employed for etching deep features in InGaAlP multilayer structures. The absence of polymer-forming gases is advantageous from the viewpoint of chamber cleanliness and wafer throughput.

While InGaP and AlInP show fast etching in both Cl2 and BCl3 chemistries, the rates of Al0.5Ga0.5P are much lower in any discharges. Since the volatility of the products for AlGaP is high it is logical to conclude that the high bond strengths reduce formation of the chlorides and hence results in the slow removal rates. The ion/neutral ratio controls both the etch rates and surface morphologies for InGaP and AlInP, with the best results when formation of a thick InCl3 reaction layer is avoided. If the ion flux is too high, preferential sputtering at the surface can occur particularly for InGaP and AlGaP, whereas if the chlorine is too high the surface of the two materials becomes "poisoned" by the InCl3 selvedge layer. Ar appears to be the better candidate for the additive than N2 for fast etch and smooth surface.

As a whole, Cl2/Ar plasma chemistry is likely to be the best choice for the etching of InGaAlP system as long as a load lock chamber is used. Although BCl3/Ar discharge can provide fast etch comparable with the former discharge it is more difficult to obtain good surface morphologies and stoichiometry, since it produces B-residues as well as Cl-residues on the surface, in contrast to Cl2/Ar discharges.

Figure 2. Etch rate of Cu (top) as a function of microwave power in Cl2/Ar, and features formed using SiNX masks.

Figure 3. Etch rates of Fe, Co and Ag in SF6/Ar in SF6/Ar or Ch4/H2/Ar ECR discharges.

Figure 4. AES surface scans of Cu after Cl2/Ar etching (top) and after subsequent in-situ H2 plasma clean.