1. Theory

Definition: An ICP etcher creates a dense plasma with an RF coil and applies a separate RF bias to the wafer chuck, so ion density and ion energy are set independently (Purdue; Lee et al.).
High density at low pressure: $1 0^{11}$ to $1 0^{12}$ ions per cm $^{3}$ at 1 to 10 mTorr, against 10 to 100 mTorr for parallel-plate RIE (KTH; MIT).
Trade-off: Lower pressure gives more anisotropy but more substrate damage (MIT).
Rate at low bias: More ions at lower energy give faster etching with less physical damage than RIE (my reading, from the GaN comparison below).

2. Working Principle

Source: A planar spiral coil at 13.56 MHz, for example 600 W, couples power into the plasma (Lee et al.).
Bias: A separate 13.56 MHz supply on the chuck sets the DC bias, below 1 kV (Lee et al.; Purdue).
GaN process: Cl $_{2}$ with Ar or BCl $_{3}$ at 10 to 30 mTorr and 70 $^{\circ}$ C (Lee et al.).
Diagnostics: Langmuir probe for ion current (Lee et al.; Langmuir Probe page).

3. Applications

Compound semiconductors: GaN mesas and ridges at 0.4 to 0.85 $μ$ m/min (Lee et al.; GaN HEMT page).
Deep silicon etching: ICP sources underlie the Bosch process (Purdue).
Low-damage pattern transfer: High rate with modest bias (Purdue; MIT).

Source, Density and Bias

Name	Formula	Related Formulae	Variables/Constants
Decoupled sources	$P_{I C P} \to n_{i} (ion flux)$	$P_{bia s} \to V_{d c} (ion energy)$	$P u r d u e : i o n e n er g y a d j u s t ab l e in d e p e n d e n tl y o f d e n s i t y, bia s b e l o w 1 k V; L ee e t a l . : 600 W co i l an d a se p a r a t e 13.56 MHz bia s, b o t h a t 13.56 MHz$ $(m y r e a d in g : in a p a r a l l e l - pl a t e R I E o n e g e n er a t or se t s b o t h, w hi c h i s t h e l imi t a t i o n I C P r e m o v es)$
Density and pressure	$n_{i} = 1 0^{11} to 1 0^{12} c m^{- 3}$	$p = 1 to 10 mTorr$	$K T H : a g ain s t R I E a t 10 t o 100 mTorr; M I T : d e n s i t y n o l o n g er d e p e n d s o n p r ess u r e, l o w er p r ess u r e g i v es m or e ani so t r o p y b u t m or e d ama g e$ $(co m p u t e d : 10 t im es l o w er p r ess u r e m e an s a 10 t im es l o n g er m e an f r ee p a t h)$
Bias comparison	$V_{d c, R I E} = 520 to 561 V$	$V_{d c, I C P} \approx - 120 V$	$U C S B : R I E 5 a t 5 mTorr an d 200 W; L ee e t a l . : v a l u e p r in t e d a s 2120 V, p r o bab l y - 120 V$ $(co m p u t e d : t h e I C P bia s ma g ni t u d e i s 3.1 t o 4.7 t im es l o w er t han t h e R I E^{'} s 375 t o 561 V)$

GaN Rate, Chemistry and Selectivity

Name	Formula	Related Formulae	Variables/Constants
GaN rate: RIE against ICP	$r_{R I E} = 0.146 μ m/min (C l_{2})$	$r_{I C P} = 0.40 to 0.85 μ m/min$	$U C S B : C l_{2} 10 sccm a t 561 V; L ee e t a l . : C l_{2} / A r 4000 t o 6500 an d C l_{2} / B C l_{3} ab o u t 8500 \overset{˚}{A} /min$ $(co m p u t e d : I C P 2.7 t o 5.8 t im es f a s t er a t r o ug h l y a q u a r t er o f t h e bia s)$
Chemistry	$r (B C l_{3}) = 0.009 \to r (C l_{2}) = 0.146 μ m/min$	$θ_{w a l l} : 4 5^{\circ} (B C l_{3}) \to 7 6^{\circ} (C l_{2} / B C l_{3})$	$U C S B : c h l or in e s u ppl i es t h e e t c han t, B C l_{3} a l o n e e t c h es s l o w l y an d n e a r l y i so t r o p i c a l l y, m or e C l_{2} d ee p e n s mi cr o - t r e n c h es; L ee e t a l . : 10 p er ce n t B C l_{3} r ai s in g t h e r a t e in C l_{2}$ $(m y r e a d in g : B C l_{3} i s u se f u l in s ma l l am o u n t s, f or e x am pl e t o r e m o v e na t i v e o x i d e, a s t h e U C S B p r e - c l e an s ug g es t s)$
Mask selectivity	$S_{G a N / S i O_{2}} = 3.7 to 5.2 (ICP)$	$S_{G a N / S i O_{2}} = 5.4 to 10.5 (RIE, Cl_{2} rich)$	$L ee e t a l . : 30 mTorr, 10 t o 25 p er ce n t B C l_{3} or A r; U C S B : hi g h es t 10.5 a t 15 mTorr, 150 W, 375 V$ $(m y r e a d in g : t h e se l ec t i v i t y r e f l ec t s c h e mi s t r y an d bia s, n o t t h e so u r ce t y p e a l o n e)$

Process Dependences

Name	Formula	Related Formulae	Variables/Constants
Pressure and rate	$r_{10 m T or r} = 400 \to r_{30 m T or r} = 650 nm/min$	$T_{s u b} = 70^{\circ} C$	$L ee e t a l . : C l_{2} w i t h 10 p er ce n t A r; in d u c t i v e p o w er, bia s an d t e m p er a t u r e g e n er a l l y r ai se t h e r a t e$ $(co m p u t e d : a 1.6 t im es r i se f or t h r ee t im es t h e p r ess u r e)$
Ion flux diagnostics	$J_{i} from a Langmuir probe$	$relative ion current density$	$L ee e t a l . : p r o b e m e a s u r e m e n t s; L an g m u i r P r o b e p a g e$ $(m y r e a d in g : t h e p r o b e co n f i r m s t ha t co i l p o w er, n o t bia s, c han g es t h e i o n s u ppl y)$
Etch time	$t = \frac{d}{r}$	$t_{2 μ m} : 14 min (RIE), 2.4 min (ICP)$	$U C S B : 2 μ m G a N l a y er s; L ee e t a l . : 8500 \overset{˚}{A} /min$ $(co m p u t e d : 2 μ m a t 0.146 an d 0.85 μ m/min)$

Important Questions

Ques: What does an ICP etcher do that a parallel-plate RIE cannot?

It separates ion supply from ion energy. Purdue states that in high-density sources such as ICP the plasma potential, and hence ion energy, can be adjusted independently of plasma density, with bias below 1 kV. Lee et al. describe the hardware: a 600 $W$ planar spiral coil at 13.56 $MHz$ to make the plasma and a separate 13.56 $MHz$ supply on the bottom electrode for bias. KTH gives high-density plasmas of $1 0^{11}$ to $1 0^{12}$ $c m^{- 3}$ at 1 to 10 $mTorr$ , against 10 to 100 $mTorr$ for RIE. MIT adds that density no longer depends on pressure. In a parallel-plate RIE one generator sets both density and energy (Parallel-Plate Reactive Ion Etcher (RIE) page; my reading). Reference: Purdue ECE 695Q, lecture 47 (nanoHUB, April 2016): remote high-density plasma reactors at 1 to 10 mTorr with bias below 1 kV, ICP sustaining plasma at much lower pressure, plasma potential (ion energy) adjustable independently of plasma density, Bosch deep RIE alternating passivation and etch at low pressure to reduce ion scattering and microloading (fetch-tool summary); Y. H. Lee, H. S. Kim, G. Y. Yeom, J. W. Lee, M. C. Yoo and T. I. Kim, ‘Etch characteristics of GaN using inductively coupled Cl2/Ar and Cl2/BCl3 plasmas’, Journal of Vacuum Science and Technology A 16(3) (1998): 600 W 13.56 MHz planar spiral Cu coil, separate 13.56 MHz bias on the bottom electrode, dc bias printed as 2120 V (probably -120 V with the sign lost), 10 to 30 mTorr, 70 C substrate, relative ion current by Langmuir probe, Cl2/Ar about 4000 angstrom/min at 10 mTorr and 6500 at 30 mTorr with 10 per cent Ar, Cl2/BCl3 close to 8500 angstrom/min at 30 mTorr with 10 per cent BCl3, GaN/SiO2 selectivity 3.7 at 10 per cent BCl3 and up to 5.2 at 25 per cent BCl3 and 4.8 at 25 per cent Ar, inductive power, bias and temperature generally raising rate (fetch-tool summary); KTH Royal Institute of Technology, Lecture 7 ‘Etching’: high-density plasmas of 1e11 to 1e12 ions per cm3 at 1 to 10 mTorr against 1e8 to 1e12 in general plasmas, RIE at 10 to 100 mTorr, sheath 100 to 700 eV in RIE mode (fetch-tool summary); MIT 6.152J lecture 17 ‘Etching’ (2005): inductively coupled high-density plasma whose density no longer depends on pressure, reaching high density at 1 to 10 mTorr, lower pressure giving more anisotropy but more substrate damage, RIE at 10 to 100 mTorr (fetch-tool summary); Parallel-Plate Reactive Ion Etcher (RIE) page of this vault.

Ques: How much faster is ICP than RIE for GaN?

About three to six times, at about a quarter of the bias. UCSB’s RIE#5, with no ICP power, etched GaN at up to 0.146 $μ m/min$ in Cl $_{2}$ at 5 $mTorr$ and 200 $W$ , with a DC bias of 561 $V$ . Lee et al.’s ICP reached about 400 $nm/min$ in Cl $_{2}$ /Ar at 10 $mTorr$ , 650 $nm/min$ at 30 $mTorr$ , and close to 850 $nm/min$ in Cl $_{2}$ /BCl $_{3}$ , with a bias printed as 2120 V that is probably -120 $V$ . That is 2.7 to 5.8 times faster (computed). A 2 $μ m$ layer takes about 14 $min$ in the RIE but 2.4 $min$ in the ICP (computed). Higher ion flux at lower energy explains the gain (my reading). Reference: UCSB Nanofabrication Laboratory, ‘Plasma-etching of GaN using (Cl2, BCl3) based chemistry and RIE#5 etcher’ (November 2004): RIE without ICP power, 0.4 um SiO2 mask on 2 um GaN, room temperature, 5 mTorr and 200 W bias at 520 to 561 V dc bias giving 0.009 (BCl3), 0.012 (BCl3/Ar), 0.088 (Cl2/BCl3 2/8), 0.133 (5/5) and 0.146 um/min (Cl2 10 sccm), 0.131 um/min at 15 mTorr, 150 W and 375 V with Cl2 20 sccm, GaN/SiO2 selectivity 0.91, 1.1, 5.4, 8.3, 5.9 and 10.5, about 45 degree sidewalls for BCl3 and BCl3/Ar and 76 degrees for Cl2/BCl3, micro-trenches deepening with Cl2, BCl3 pre-clean reducing pillars from sputtered mask oxide (fetch-tool summary); Y. H. Lee, H. S. Kim, G. Y. Yeom, J. W. Lee, M. C. Yoo and T. I. Kim, ‘Etch characteristics of GaN using inductively coupled Cl2/Ar and Cl2/BCl3 plasmas’, Journal of Vacuum Science and Technology A 16(3) (1998): 600 W 13.56 MHz planar spiral Cu coil, separate 13.56 MHz bias on the bottom electrode, dc bias printed as 2120 V (probably -120 V with the sign lost), 10 to 30 mTorr, 70 C substrate, relative ion current by Langmuir probe, Cl2/Ar about 4000 angstrom/min at 10 mTorr and 6500 at 30 mTorr with 10 per cent Ar, Cl2/BCl3 close to 8500 angstrom/min at 30 mTorr with 10 per cent BCl3, GaN/SiO2 selectivity 3.7 at 10 per cent BCl3 and up to 5.2 at 25 per cent BCl3 and 4.8 at 25 per cent Ar, inductive power, bias and temperature generally raising rate (fetch-tool summary).

Ques: What DC bias did Lee et al. use, and which value is more likely?

About -120 $V$ is more likely than the printed 2120 V. The text of Lee et al. gives the DC bias as 2120 V, which reads as a minus sign lost in the scanned text. Three points support -120 $V$ . Purdue states that high-density reactors run with bias below 1 kV. UCSB’s parallel-plate RIE, which has no separate source and therefore needs high bias, ran at only 375 to 561 $V$ . And a 2120 V bias would make the GaN etch strongly physical, with selectivity to SiO $_{2}$ near 1, as UCSB’s high-bias BCl $_{3}$ runs showed, whereas Lee et al. report 3.7 to 5.2 (my reading). The ICP bias is therefore 3.1 to 4.7 times lower than the RIE bias (computed). Reference: Y. H. Lee, H. S. Kim, G. Y. Yeom, J. W. Lee, M. C. Yoo and T. I. Kim, ‘Etch characteristics of GaN using inductively coupled Cl2/Ar and Cl2/BCl3 plasmas’, Journal of Vacuum Science and Technology A 16(3) (1998): 600 W 13.56 MHz planar spiral Cu coil, separate 13.56 MHz bias on the bottom electrode, dc bias printed as 2120 V (probably -120 V with the sign lost), 10 to 30 mTorr, 70 C substrate, relative ion current by Langmuir probe, Cl2/Ar about 4000 angstrom/min at 10 mTorr and 6500 at 30 mTorr with 10 per cent Ar, Cl2/BCl3 close to 8500 angstrom/min at 30 mTorr with 10 per cent BCl3, GaN/SiO2 selectivity 3.7 at 10 per cent BCl3 and up to 5.2 at 25 per cent BCl3 and 4.8 at 25 per cent Ar, inductive power, bias and temperature generally raising rate (fetch-tool summary); Purdue ECE 695Q, lecture 47 (nanoHUB, April 2016): remote high-density plasma reactors at 1 to 10 mTorr with bias below 1 kV, ICP sustaining plasma at much lower pressure, plasma potential (ion energy) adjustable independently of plasma density, Bosch deep RIE alternating passivation and etch at low pressure to reduce ion scattering and microloading (fetch-tool summary); UCSB Nanofabrication Laboratory, ‘Plasma-etching of GaN using (Cl2, BCl3) based chemistry and RIE#5 etcher’ (November 2004): RIE without ICP power, 0.4 um SiO2 mask on 2 um GaN, room temperature, 5 mTorr and 200 W bias at 520 to 561 V dc bias giving 0.009 (BCl3), 0.012 (BCl3/Ar), 0.088 (Cl2/BCl3 2/8), 0.133 (5/5) and 0.146 um/min (Cl2 10 sccm), 0.131 um/min at 15 mTorr, 150 W and 375 V with Cl2 20 sccm, GaN/SiO2 selectivity 0.91, 1.1, 5.4, 8.3, 5.9 and 10.5, about 45 degree sidewalls for BCl3 and BCl3/Ar and 76 degrees for Cl2/BCl3, micro-trenches deepening with Cl2, BCl3 pre-clean reducing pillars from sputtered mask oxide (fetch-tool summary).

Ques: How does gas chemistry affect a chlorine-based GaN etch?

Chlorine supplies the etchant; BCl $_{3}$ and Ar modify it. UCSB found BCl $_{3}$ alone etched GaN at only 0.009 $μ m/min$ with about 45 degree sidewalls, close to isotropic, while Cl $_{2}$ /BCl $_{3}$ at 2/8 gave 0.088 $μ m/min$ and 76 degree sidewalls. Pure Cl $_{2}$ reached 0.146 $μ m/min$ , but more Cl $_{2}$ deepened micro-trenches at the ridge corners. A BCl $_{3}$ pre-clean reduced surface pillars that UCSB attributes to oxide sputtered from the mask. In the ICP, Lee et al. found that adding 10 per cent BCl $_{3}$ to Cl $_{2}$ raised the rate to about 850 $nm/min$ . In my reading, a small BCl $_{3}$ fraction helps by removing native oxide, while too much starves the plasma of chlorine. Reference: UCSB Nanofabrication Laboratory, ‘Plasma-etching of GaN using (Cl2, BCl3) based chemistry and RIE#5 etcher’ (November 2004): RIE without ICP power, 0.4 um SiO2 mask on 2 um GaN, room temperature, 5 mTorr and 200 W bias at 520 to 561 V dc bias giving 0.009 (BCl3), 0.012 (BCl3/Ar), 0.088 (Cl2/BCl3 2/8), 0.133 (5/5) and 0.146 um/min (Cl2 10 sccm), 0.131 um/min at 15 mTorr, 150 W and 375 V with Cl2 20 sccm, GaN/SiO2 selectivity 0.91, 1.1, 5.4, 8.3, 5.9 and 10.5, about 45 degree sidewalls for BCl3 and BCl3/Ar and 76 degrees for Cl2/BCl3, micro-trenches deepening with Cl2, BCl3 pre-clean reducing pillars from sputtered mask oxide (fetch-tool summary); Y. H. Lee, H. S. Kim, G. Y. Yeom, J. W. Lee, M. C. Yoo and T. I. Kim, ‘Etch characteristics of GaN using inductively coupled Cl2/Ar and Cl2/BCl3 plasmas’, Journal of Vacuum Science and Technology A 16(3) (1998): 600 W 13.56 MHz planar spiral Cu coil, separate 13.56 MHz bias on the bottom electrode, dc bias printed as 2120 V (probably -120 V with the sign lost), 10 to 30 mTorr, 70 C substrate, relative ion current by Langmuir probe, Cl2/Ar about 4000 angstrom/min at 10 mTorr and 6500 at 30 mTorr with 10 per cent Ar, Cl2/BCl3 close to 8500 angstrom/min at 30 mTorr with 10 per cent BCl3, GaN/SiO2 selectivity 3.7 at 10 per cent BCl3 and up to 5.2 at 25 per cent BCl3 and 4.8 at 25 per cent Ar, inductive power, bias and temperature generally raising rate (fetch-tool summary).

Ques: Is GaN selectivity to an oxide mask better in ICP or RIE?

It depends more on chemistry and bias than on the source. Lee et al. report GaN/SiO $_{2}$ selectivity of 3.7 with 10 per cent BCl $_{3}$ at 30 $mTorr$ , rising to 5.2 with 25 per cent BCl $_{3}$ and 4.8 with 25 per cent Ar. UCSB’s RIE gave 0.91 to 1.1 in BCl $_{3}$ -rich mixtures, 5.4 to 8.3 with Cl $_{2}$ /BCl $_{3}$ , and 10.5 with pure Cl $_{2}$ at 15 $mTorr$ , 150 $W$ and 375 $V$ , the lowest bias it used. In my reading, the best selectivity comes from chlorine-rich gas at low ion energy, which an ICP can combine with a high rate. Selectivity sets mask thickness: 2 $μ m$ of GaN at 5:1 consumes 0.4 $μ m$ of oxide, the whole UCSB mask (computed). Reference: Y. H. Lee, H. S. Kim, G. Y. Yeom, J. W. Lee, M. C. Yoo and T. I. Kim, ‘Etch characteristics of GaN using inductively coupled Cl2/Ar and Cl2/BCl3 plasmas’, Journal of Vacuum Science and Technology A 16(3) (1998): 600 W 13.56 MHz planar spiral Cu coil, separate 13.56 MHz bias on the bottom electrode, dc bias printed as 2120 V (probably -120 V with the sign lost), 10 to 30 mTorr, 70 C substrate, relative ion current by Langmuir probe, Cl2/Ar about 4000 angstrom/min at 10 mTorr and 6500 at 30 mTorr with 10 per cent Ar, Cl2/BCl3 close to 8500 angstrom/min at 30 mTorr with 10 per cent BCl3, GaN/SiO2 selectivity 3.7 at 10 per cent BCl3 and up to 5.2 at 25 per cent BCl3 and 4.8 at 25 per cent Ar, inductive power, bias and temperature generally raising rate (fetch-tool summary); UCSB Nanofabrication Laboratory, ‘Plasma-etching of GaN using (Cl2, BCl3) based chemistry and RIE#5 etcher’ (November 2004): RIE without ICP power, 0.4 um SiO2 mask on 2 um GaN, room temperature, 5 mTorr and 200 W bias at 520 to 561 V dc bias giving 0.009 (BCl3), 0.012 (BCl3/Ar), 0.088 (Cl2/BCl3 2/8), 0.133 (5/5) and 0.146 um/min (Cl2 10 sccm), 0.131 um/min at 15 mTorr, 150 W and 375 V with Cl2 20 sccm, GaN/SiO2 selectivity 0.91, 1.1, 5.4, 8.3, 5.9 and 10.5, about 45 degree sidewalls for BCl3 and BCl3/Ar and 76 degrees for Cl2/BCl3, micro-trenches deepening with Cl2, BCl3 pre-clean reducing pillars from sputtered mask oxide (fetch-tool summary).

Ques: How do pressure and temperature affect ICP etch rate?

Both raise it in the reported range. Lee et al. found the Cl $_{2}$ /Ar rate rising from about 400 $nm/min$ at 10 $mTorr$ to 650 $nm/min$ at 30 $mTorr$ , a 1.6 times rise for three times the pressure (computed). They state that inductive power, bias power and substrate temperature, 70 $^{\circ} C$ in their work, generally increased the GaN rate. They tracked relative ion current density with a Langmuir probe, the method of the Langmuir Probe page. MIT warns that lowering pressure gives more anisotropy but more substrate damage. In my reading, an ICP recipe for device mesas therefore balances pressure for rate and profile against low bias for low damage, which matters for the GaN HEMT page structures. Reference: Y. H. Lee, H. S. Kim, G. Y. Yeom, J. W. Lee, M. C. Yoo and T. I. Kim, ‘Etch characteristics of GaN using inductively coupled Cl2/Ar and Cl2/BCl3 plasmas’, Journal of Vacuum Science and Technology A 16(3) (1998): 600 W 13.56 MHz planar spiral Cu coil, separate 13.56 MHz bias on the bottom electrode, dc bias printed as 2120 V (probably -120 V with the sign lost), 10 to 30 mTorr, 70 C substrate, relative ion current by Langmuir probe, Cl2/Ar about 4000 angstrom/min at 10 mTorr and 6500 at 30 mTorr with 10 per cent Ar, Cl2/BCl3 close to 8500 angstrom/min at 30 mTorr with 10 per cent BCl3, GaN/SiO2 selectivity 3.7 at 10 per cent BCl3 and up to 5.2 at 25 per cent BCl3 and 4.8 at 25 per cent Ar, inductive power, bias and temperature generally raising rate (fetch-tool summary); MIT 6.152J lecture 17 ‘Etching’ (2005): inductively coupled high-density plasma whose density no longer depends on pressure, reaching high density at 1 to 10 mTorr, lower pressure giving more anisotropy but more substrate damage, RIE at 10 to 100 mTorr (fetch-tool summary); Langmuir Probe page of this vault; GaN HEMT page of this vault.

References

Y. H. Lee, H. S. Kim, G. Y. Yeom, J. W. Lee, M. C. Yoo and T. I. Kim, “Etch characteristics of GaN using inductively coupled Cl2/Ar and Cl2/BCl3 plasmas,” Journal of Vacuum Science and Technology A, vol. 16, no. 3, 1998. Available: https://swb.skku.edu/_res/pnpl/etc/1998-05.pdf
UCSB Nanofabrication Laboratory, “Plasma-etching of GaN using (Cl2, BCl3) based chemistry and RIE#5 etcher,” November 2004. Available: https://wiki.nanofab.ucsb.edu/w/images/c/cb/08-Plasma_Etching_of_GaN-RIE5.pdf
Purdue ECE 695Q, lecture 47, April 2016, nanoHUB. Available: https://nanohub.org/resources/25351/download/2016.04.22-ECE695Q-L47.pdf
KTH Royal Institute of Technology, “Lecture 7: etching,” course slides. Available: https://www.kth.se/social/upload/510f795cf276544e1ddda13f/Lecture%207%20Etching.pdf
MIT 6.152J Micro/Nano Processing Technology, “Lecture 17: etching,” MIT OpenCourseWare, 2005. Available: https://ocw.mit.edu/courses/6-152j-micro-nano-processing-technology-fall-2005/59b9f31d836355a3380aa156925508bc_lecture17.pdf
Parallel-Plate Reactive Ion Etcher (RIE) page of this vault.
Langmuir Probe page of this vault.
GaN HEMT page of this vault.

Quartz 5

Explorer

Inductively Coupled Plasma Etcher (ICP-RIE)