PRINCIPALES CUENTAS QUE SE EMPLEAN EN LA CONTABILIDAD COMERCIAL

The etching of AlN was studied for two reasons, namely:

• AlN is a piezoelectric material that can be used to drive nitride or III-N/NCD MEMS structures.

• AlN can act as sacrificial layer, as discussed in chapter 4. Therefore it may act as stop layer for the dry etching of other III-N materials.

3.2.1

Preliminary results

AlN grown epitaxially on different substrates by different techniques was used for the preliminary study:

• AlN grown by MOCVD on sapphire • AlN grown by MOCVD on SiC • AlN grown by MBE on Si(111)

The first dry etching attempts of these materials were extremely slow, with mea- sured etch rates under 1 nm/min. Nevertheless, if the processes were maintained for long periods the etch rate increased drastically, as shown in figure 3.7. This is because the native oxide of AlN forms a compact layer on the sample surface that is very dif- ficult to etch. Actually, the etching is only performed by the physical component of the process because the reactivity of AlO3 is virtually inexistent. The thickness of the

oxide layer is variable and depends on the defect density of the material and sample processing prior to the etching experiments.

In order to study of the etching of AlN without the effect of the native oxide layer, a special sample of AlN was grown by MOCVD on SiC, using a 30 nm cap of GaN . The GaN cap layer acts as a diffusion barrier in order to prevent the oxidation of the AlN.

3.2.2

Etch rate and selectivity to GaN

For the etching experiments of AlN with the GaN cap, the chamber pressure was fixed to 40 mtorr and only VDC and the SF6 content in the SiCl4 : Ar : SF6 mixture

were varied. For all the studied etching conditions, it was observed that after etching the first 25 nm to 30 nm, corresponding to the GaN cap, the etch rate was stable. Thus, it may be confirmed that the GaN cap acted as a successful diffusion layer, preventing the formation of the difficult-to-etch native oxide on the AlN layer.

Observing the etch rate plot versus SF6concentration in figure 3.8, it is noticeable

that for concentrations of this gas lower than 20% the etch rate does not significantly depend on VDC. This is due to the hardness of AlN , that makes this material ex-

tremely difficult to etch. For SF6 proportion of 20% the influence of VDC is much

stronger. This is explained by the physical-chemical interaction that activates the for- mation of N F3 allowing a more efficient removal of the N planes. This effect was

already observed for the GaN samples but with a less drastic influence.

The measured etch rates are high for what is expected for AlN , which is usually very difficult to etch, and all the studied conditions led to reproducible and uniform results. Comparing the etch rates to those obtained for the GaN etches, it is observed that the selectivity is maximized for SF6 proportions that enhance the etch rate of

GaN but are not sufficiently high to trigger the formation of N F3 during the etching

of AlN . At these conditions, the VDCthat produces highest selectivities is 400 V as for

this plasma voltage the GaN etch rate is highly incremented and the AlN sputtering component is still small. On the whole, an etch rate selectivity of GaN to AlN of 13 is achievable while preserving large GaN etch rates, 55 nm/min. These conditions allow the etching of large GaN structures precisely an relying on an AlN etch stop layer.

If higher selectivity is required, small proportions of O2 may be added to the gas

mixture. Smith et al. [109] claim to have obtained selectivities as high as 50. This is because the oxidation of GaN does not has a significant impact on the etch rate, as opposed to the large influence showed in AlN . These observations have been con- firmed in the experiments carried out for this dissertation. Nevertheless, the addition of O2 might difficult the masking technology, as photoresist masks will be rapidly

etched in plasmas containing active oxygen species. Probably, this technique would require the use of metal masks, which could difficult the production of functional MEMS structures.

3.2.3

Surface morphology

The AlN samples could not be studied by means of scanning electron microscopy. This is because the material exhibits strong insulating characteristics, due to the large band gap. Thus, the sample gets rapidly charged and SEM imaging is impossible because the impinging electron get deflected by the accumulated electrostatic charge on the AlN surface. Consequently, only AFM survey was used for characterizing the etched surface.

The roughness of the unetched surface has been considered to be similar to the surface roughness of the AlN . The presence of a thin GaN cap has a negligible influence on the surface roughness, as the growth of GaN by MOCVD has been opti- mized. The measurements of the roughness of the etched surface was very similar to the values measured on the original surface. No defect was induced during etching. Conversely, the chemical polishing effect described for the GaN etching study was not observed in the case of the RIE experiments with AlN .

Lastly, no residues were found to accumulate neither on the etched surface nor on the pattern sidewall. This indicates that either the reaction by-products are removed efficiently or that they form a continuous layer on the etched surface. In any case, no influence on the resulting surface morphology was found for any of the studied conditions.