Photo-Enhanced Etching of n-Si

The etch rate of n-type Si in diluted HF solutions was investigated as a function of the bias voltage applied to the Si/electrolyte interface in the dark and under illumination. It was observed that the etch rate depends very sensitively on the minority carrier flow through this interface. For an illumination intensity of 5.3x 1016 photons/cm2s (2 = 550 nm) and the depleted Si/electrolyte interface biased slightly (less than 1 V) in reverse, the etch rate is increased by a factor of more than 1000 as compared to the etch rate under open-circuit condition. This effect can be used to create etch patterns during device processing without prior masking the semiconductors. Using the same effect it should be possible to trim the thickness of Si layers on (semi-) insulating substrates for the fabrication of enhancement-mode FETs. PACS: 73.40.Mr, 82.30.Lp, 85.30.De, 85.30.Tv Etching solutions for semiconductors consist usually of an oxident and a complexing agent which causes the dissolution of the oxidized semiconductor surface (see e.g. [1]). For Si the oxident is very often nitric acid (HNO3) and the complexing agent is usually fluoric acid (HF). Without oxident the etch rate is considerably reduced. The etch rate of the (! 11) surface of n-Si, e.g., is in the range between 15-1700 gm/min depending on the HF : HNO3 ratio [2], whereas it is as low as 0.3 A/min at 25 ~ in 48% HF solution alone I-3, 4]. Hu and Kerr [4] reported slightly larger values than 0.3 A/mi.n after diluting the HF solution and adding NaC2H30 2 or NaF. But even then, the etch rate does not exceed 0.78 A/min. These authors observed also that these low etch rates were independent of the Oz content of the etching solution. In contrast to this, however, we report in the following on a series of experiments that show how one can increase the etch rate of HF solutions considerably by using light in lieu of an oxident. We tested how and under what conditions the etch rate of Si can be controlled by the illumination, since the availability of a light-controlled etch rate has several promising consequences for device processing applications as we will discuss subsequently. * Present address: Schott Glaswerke, Hattenbergstrasse 10, D-6500 Mainz, Fed. Rep. Germany For the etching experiments we used (111) surfaces of prepolished n-type Si wafers with 0.3-0.8 x 1015 cm-3 charge carriers. After providing the back sides of the samples with ohmic In contacts, we connected them with silver paste to the electrode of a sampleholder. Thereafter the samples were partly covered with black wax in order that only the uncovered surface could be attacked by the etching solution. The area of the uncovered Si surface varied between 4 and 25 mm 2. This surface was the working electrode in a threeelectrode electrochemical cell. A saturated calomel electrode (SCE) was the reference electrode and a 4 cm 2 platinum sheet the auxiliary electrode. The electrolyte consisted of 1 part HF (48%) diluted in 9 to 19 parts deionized water. The temperature was 24 + 3 ~ The n-Si/electrolyte interface was either kept in the dark or illuminated through an interference filter with a transmission maximum at a wavelength of 550 nm and a full width at half maximum of 25 nm. The penetration depth into Si for light with this range of wavelength is about 1.5 gm. Figure 1 shows the I-V characteristics of the n-Si/electrolyte interface without (Fig. la) and with illumination (Fig. lb) for a small interval of voltages. For zero current there is a voltage of -0.37 V between the sample and SCE in the dark and 0 . 5 5 V under illumination with 5.3 x 1016 photons/cm 2 s. The open244 H.J. Hoffmann and J. M. Woodall