CCl 4 -based reactive ion etching of semi-insulating GaAs and InP (original) (raw)

CCl4\mathrm{CCl}_{4}-based reactive ion etching of semi-insulating GaAs and InP

Š. Haščík, P. Eliáš, J. Šoltýs, J. Martaus
Institute of Electrical Engineering, Slovak Academy of Sciences
Dúbravská cesta 9, 84104 Bratislava, Slovak Republic
e-mail: elektase@savba.sk

1. Hotový

Department of Microelectronics, Slovak University of Technology Ilkovicova 3, 81219 Bratislava, Slovak Republic

Received 30 April 2006
Reactive ion etching of semi-insulating (100) GaAs and (100) InP substrates in CCl4/He\mathrm{CCl}_{4} / \mathrm{He}-based plasma at 25∘C25^{\circ} \mathrm{C} was performed at rf power between 43 and 153 W and reactor pressures of 0.8 and 1.5 Pa . The etching rates of GaAs and InP and surface quality were evaluated by AFM. With the pressure, the etching rates of GaAs and InP increased and decreased, respectively. InP was removed at 0.8 Pa at rates (∼8÷54 nm min−1)\left(\sim 8 \div 54 \mathrm{~nm} \mathrm{~min}^{-1}\right) comparable with those of GaAs (∼3÷71 nm min−1)\left(\sim 3 \div 71 \mathrm{~nm} \mathrm{~min}^{-1}\right). At 1.5 Pa , InP was etched at negligible rates (∼1÷4 nm min−1)\left(\sim 1 \div 4 \mathrm{~nm} \mathrm{~min}^{-1}\right). GaAs exhibited none or little trenching at edges of resist patterns at 0.8 and 1.5 Pa . InP was slightly trenched at low bias at 0.8 Pa , but it was heavily trenched at 1.5 Pa over the entire interval of rf power used. GaAs surfaces were smoother compared with those of InP after etching. The average root-mean-square roughness σ\sigma of 10×10μ m210 \times 10 \mu \mathrm{~m}^{2}-sized areas of GaAs and InP surfaces after RIE at 0.8 Pa , 100 W , and -150 V was 0.9 nm and ∼1.8 nm\sim 1.8 \mathrm{~nm}, respectively.

PACS: 52.77.Bn, 81.65.Cf
Key words: plasma, etching, RIE, InP, CCl4\mathrm{CCl}_{4}

1 Introduction

GaAs and InP semiconductor materials are used in a vast number of microelectronic and optoelectronic devices, such as heterojunction bipolar transistors, high electron mobility transistors, lasers, and microwave and optoelectronic integrated circuits. In their technology, dry etching plays a crucial role. As the devices are designed ever smaller, the etching is expected to meet stringent requirements that include minimized crystal damage, room temperature etching, and high selectivity, high directionality, and high dimensional control. The reactive ion etching (RIE) of InP and GaAs is usually performed in chlorine-based (Cl2,SiCl4,BCl3\left(\mathrm{Cl}_{2}, \mathrm{SiCl}_{4}, \mathrm{BCl}_{3}\right., CCl2 F2,CCl4\mathrm{CCl}_{2} \mathrm{~F}_{2}, \mathrm{CCl}_{4} mixed with Ar,H2,He)\left.\mathrm{Ar}, \mathrm{H}_{2}, \mathrm{He}\right) or in hydrocarbon-based plasmas (CH4\left(\mathrm{CH}_{4}\right. and C2H6\mathrm{C}_{2} \mathrm{H}_{6} mixed with Ar,H2,He)\left.\mathrm{Ar}, \mathrm{H}_{2}, \mathrm{He}\right) [1]. Semura et al. [2] showed that GaAs can be etched smoothly at relatively high rates in CCl4/H2\mathrm{CCl}_{4} / \mathrm{H}_{2}-based plasma. Rawal et al. formed via holes in GaAs by means of a RIE process based in CCl2 F2/CCl4\mathrm{CCl}_{2} \mathrm{~F}_{2} / \mathrm{CCl}_{4} plasma [3] This paper reports on initial experiments with the RIE of semi-insulating (100) GaAs and (100) InP substrates in CCl4/He\mathrm{CCl}_{4} / \mathrm{He}-based plasma. Atomic force microscopy (AFM)

was used to evaluate the etching rates into GaAs and InP and surface quality as functions of electrode rf power and reactor pressure.

2 Experiment

Standard (100)-oriented semi-insulating InP and GaAs wafers were used for the experiment. The wafers were provided with etching mask patterns defined in a ∼1.4μ m\sim 1.4 \mu \mathrm{~m}-thick positive-tone AZ5214-E resist layer using standard photolithography. The wafers were etched in RIE mode in CCl4/He\mathrm{CCl}_{4} / \mathrm{He}-based plasma in a ROTH & RAU MICROSYS 350 machine. The chlorine-based plasma was generated by a radio-frequency (rf) field at 13.56 MHz supplied via a stainless steel electrode ( ∅200 mm\emptyset 200 \mathrm{~mm} ). The temperature of the electrode was stabilized at 25∘C25^{\circ} \mathrm{C} by He flown into the chamber at 4 sccm . Before the introduction of CCl4/He\mathrm{CCl}_{4} / \mathrm{He}, the chamber was evacuated to a background pressure <5×10−4 Pa<5 \times 10^{-4} \mathrm{~Pa}. During etching the flow of CCl4/He\mathrm{CCl}_{4} / \mathrm{He} was 13.6 sccm . The wafers, put on the electrode by a loadlock system, were etched under conditions summarized in Tables 1 and 2. The RIE process was evaluated by means of atomic force microscopy (AFM), scanning electron microscopy (SEM), and optical microscopy. The AFM probed the surfaces in non-contact mode with a tip of a typical tip curvature radius of 10 nm .

Table 1. RIE of (100) InP and (100) GaAs during 10 min at 25∘C25^{\circ} \mathrm{C} at pressure p=0.8 Pap=0.8 \mathrm{~Pa}

3 Results and discussion

The first round of experiments, summarized in Table 1, was performed at 0.8 Pa and rf powers ranged between 43 and 153 W with the sample electrode self-biased between -100 and -200 V . The samples were etched for 10 min in each run. The etching rates of GaAs and InP were comparable and increased with bias voltage (figure 1). GaAs exhibited no trenching at edges of resist patterns. InP was trenched only at low bias (figure 2). After RIE GaAs surfaces were smoother compared with those of InP. The average root-mean-square roughness σ\sigma was evaluated by AFM for samples etched under the run 4 conditions: GaAs and InP exhibited a σ∼0.9 nm\sigma \sim 0.9 \mathrm{~nm}

Table 2. RIE of (100) InP and (100) GaAs during 10 min at 25∘C25^{\circ} \mathrm{C} at pressure p=1.5 Pap=1.5 \mathrm{~Pa}

and ∼1.8 nm\sim 1.8 \mathrm{~nm}, respectively over 10×10μ m210 \times 10 \mu \mathrm{~m}^{2}-sized areas. The AFM scans of GaAs and InP are exemplified in figure 3 and 4 , respectively.

Fig. 1. Etching rates of semi-insulating (100) GaAs and (100) InP substrates for CCl4\mathrm{CCl}_{4}-based reactive ion etching (RIE) at 0.8 Pa and rf power ranged between 43 and 153 W with the sample electrode self - biased between -100 and 200 V .

Fig. 2. Trenching of InP at the resist edge of an etching mask pattern for the RIE at 0.8 Pa,43 W0.8 \mathrm{~Pa}, 43 \mathrm{~W}, and −100 V;d11-100 \mathrm{~V} ; d_{11} and d1pd_{1 p} denote the top-to-trench-bottom distance and the top-to-plateau distance, respectively.

The second round of experiments, summarized in Table 2 and Fig. 5, was performed at 1.5 Pa and rf powers ranged between 50 and 100 W with the sample electrode self-biased between -75 and -147 V . The samples were etched for 10 min in each run. Under the conditions, the etching of GaAs significantly differed from that of InP: while the etching rate of GaAs increased with bias voltage, the etching rate of InP remained negligible. The GaAs surface exhibited none or very low trenching at edges of resist patterns, but InP was increasingly heavily trenched as bias increased.

The experiments showed that while GaAs was etched away faster with increased

Fig. 3. Typical surface of GaAs after 10 min long RIE at 0.8 Pa,100 W0.8 \mathrm{~Pa}, 100 \mathrm{~W} and 150 V . The average root-mean-square roughness σ\sigma of such a surface was ∼0.9 nm\sim 0.9 \mathrm{~nm} for 10×10μ m210 \times 10 \mu \mathrm{~m}^{2}-sized areas.

Fig. 4. Typical surface of InP after 10 min long RIE at 0.8 Pa,100 W0.8 \mathrm{~Pa}, 100 \mathrm{~W} and 150 V . The average root-mean-square roughness σ\sigma of such a surface was ∼1.8 nm\sim 1.8 \mathrm{~nm} for 10×10μ m210 \times 10 \mu \mathrm{~m}^{2}-sized areas.

Fig. 5. Etching rates of semi-insulating (100) GaAs and (100) InP substrates for CCl4−\mathrm{CCl}_{4}- based RIE at 1.5 Pa and rf power ranged between 50 and 100 W with the sample electrode self-biased between -75 and -147 V .
chamber pressure, InP followed an opposite trend: its etching rate decreased as the chamber pressure was increased. While at 1.5 Pa InP was removed at negligible rates, at 0.8 Pa it was removed at rates close those of GaAs. This suggests that the etching of InP via the removal of its surface chlorides was possible in a plasma in which the contribution of ion bombardment was more intensive. The results are in agreement with what is known about the behaviour of chlorinated GaAs and InP surfaces under thermal and plasma excitation. They can be explained considering that GaAs and InP differ in the formation of surface chlorides [4-5], in the volatility of the chlorides [6] and in their ion-bombardment-stimulated desorption [7-9]. Of GaClx,AsClx,InClx\mathrm{GaCl}_{x}, \mathrm{AsCl}_{x}, \mathrm{InCl}_{x}, and PClx(x=1,2,3)\mathrm{PCl}_{x}(x=1,2,3) chloride species that form on GaAs and InP, respectively, the removal of GaCl3,AsCl3,InCl3\mathrm{GaCl}_{3}, \mathrm{AsCl}_{3}, \mathrm{InCl}_{3}, and PCl3\mathrm{PCl}_{3} is believed to be the most important for their dry etching. The species are removed via an interrelated action of thermal and chemical activation combined with ion bombardment during etching. It is accepted that it is harder to remove InCl3\mathrm{InCl}_{3} from the surface of InP

than GaCl3\mathrm{GaCl}_{3} from GaAs because the volatility of InCl3\mathrm{InCl}_{3} is lower than that of GaCl3\mathrm{GaCl}_{3}. If the surface of InP is not thermally activated, InCl3\mathrm{InCl}_{3} can form an etch-limiting (stopping or at least slowing) layer over the substrate during etching [10]. Also, the differences between the desorption of InCl3\mathrm{InCl}_{3} and PCl3\mathrm{PCl}_{3} and the tendency of InP to cover itself with InCl3\mathrm{InCl}_{3} can lead during dry etching to the formation of island-like structures on InP, and consequently rougher surfaces compared with those of GaAs [11]. The surface quality difference is evident from figures 3 and 4.

4 Conclusion

Reactive ion etching of semi-insulating (100) GaAs and (100) InP substrates in CCl4/He\mathrm{CCl}_{4} / \mathrm{He}-based plasma at 25∘C25^{\circ} \mathrm{C} was performed at rf power between 43 and 153 W and reactor pressures of 0.8 and 1.5 Pa . The etching rates of GaAs and InP and surface quality were evaluated by AFM. With the pressure, the etching rates of GaAs and InP increased and decreased, respectively. InP was removed at 0.8 Pa at rates (∼8÷54 nm min−1)\left(\sim 8 \div 54 \mathrm{~nm} \mathrm{~min}^{-1}\right) comparable with those of GaAs (∼3÷71 nm min−1)\left(\sim 3 \div 71 \mathrm{~nm} \mathrm{~min}^{-1}\right). At 1.5 Pa , InP was etched at negligible rates (∼1÷4 nm min−1)\left(\sim 1 \div 4 \mathrm{~nm} \mathrm{~min}^{-1}\right). GaAs exhibited none or little trenching at edges of resist patterns at 0.8 and 1.5 Pa . InP was slightly trenched at low bias at 0.8 Pa , but it was heavily trenched at 1.5 Pa over the entire interval of rf power used. GaAs surfaces were smoother compared with those of InP after etching. The average root-mean-square roughness σ\sigma of 10×10μ m210 \times 10 \mu \mathrm{~m}^{2}-sized areas of GaAs and InP surfaces after RIE at 0.8 Pa,100 W0.8 \mathrm{~Pa}, 100 \mathrm{~W}, and -150 V was 0.9 nm and ∼1.8 nm\sim 1.8 \mathrm{~nm}, respectively.

This research has been sponsored under projects APVV-51-045705, APVT-20-021004, VEGA Agency 2/6096/26, 2/6097/26, and Slovak-Czech bilateral project SK-95/CZ-80. The financial support of the Center of Excellence CENG, SAS is acknowledged.

References

[1] D. E. Ibbotson: Pure and Appl. Chem 60 (1988) 703.
[2] S. Semura, H. Saitoh: J. Vac. Sci. Technol. A 2(1984) 474.
[3] D. S. Rawal, V. R. Agarwal, H. S. Sharma, B. K. Sehgal, R. Gulati, H. P. Vyas: J. Electrochem. Soc. 150 (2003) G395.
[4] W. C. Simpson, J. A. Yarmoff: Annu. Rev. Phys. Chem. 47 (1996) 527.
[5] W. H. Hung, J. T. Hsieh, H. L. Hwang, H. Y. Hwang, C. C. Chang: Surf. Sci. 418 (1998) 46.
[6] S. C. McNevin: J. Vac. Sci. Technol. B 4 (1986) 1216.
[7] M. Balooch, D. R. Olander, W. J. Siekhaus: J. Vac. Sci. Technol. B 4 (1986) 794.
[8] D. G. Lishan, E. L. Hu: J. Vac. Sci. Technol. B 8 (1990) 1951.
[9] S. J. Pearson: Mat.Sci.Eng. B 40 (1996) 101.
[10] V. M. Donnelly, D. L. Flamm, C. W. Tu, D. E. Ibbotson: J. Electrochem. Soc. 129 (1982) 2533.
[11] M. Vernon, T. R. Hayes, V. M. Donnelly: J. Vac. Sci. Technol. A 10 (1992) 3499.