Atomic Layer Etch (ALE) Tool

1. Theory

• Definition: Atomic layer etching removes material in cycles of a self-limited surface modification and a separate removal step, ideally one monolayer per cycle (Huard et al.; KLA).
• Plasma ALE: Chlorine saturates silicon as SiCl${}_x$, then argon ions with energy between the passivated and bare sputter thresholds, about 10 and 50 eV, remove only the modified layer (Huard et al.).
• Synergy: How much of the removal is self-limited; overlapping steps or stray chlorine turn ALE back into continuous etching (Huard et al.; KLA).
• Thermal ALE: Spontaneous, self-limiting reactions such as fluorination with HF followed by ligand exchange with TMA, needing volatile products (George).

2. Working Principle

• Plasma ALE in an ICP: Ar/Cl${}_2$ 70/30 at 0 V bias, then Ar with at most ppm Cl${}_2$ at about 30 V bias, 20 mTorr, 300 W (KLA; Inductively Coupled Plasma Etcher (ICP-RIE) page).
• Purges: Seconds between steps, so ALE is about 7.5 to 10 times slower than continuous etching (Huard et al.).
• Thermal ALE reactor: Sequential vapour exposures like ALD in reverse, for example HF and TMA for Al${}_2$O${}_3$ (George; Atomic Layer Deposition (ALD) Reactor page).

3. Applications

• FinFET gates and 3D corners: Over-etch cut from more than 100 to 175 per cent down to about 25 per cent (Huard et al.; KLA).
• Self-aligned contacts: SiO${}_2$ over Si${}_3$N${}_4$ with low shoulder loss (KLA).
• High-k and metal trimming: Al${}_2$O${}_3$, HfO${}_2$, W and TiN by thermal ALE (George).

Plasma ALE Cycle

Name	Formula	Related Formulae	Variables/Constants
Plasma ALE cycle	$Si+Cl\to SiC{l}_x\text{(self-limited)}$	$SiC{l}_x+A{r}^{+}\to \text{volatile products}$	$\begin{array}{r}Huardetal.(2017):0.5spassivation,\\ 3.0sionstep,2.77Clpersiteafter\\ about50ms;KLA:chlorinationlowersthe\\ sputterthreshold,Clflux7.0\times 10^{17}then\\ 7.9\times 10^{14}\mathrm{c}{\mathrm{m}}^2/s\end{array}$ $\begin{array}{r}(computed:thechlorinefluxdropsabout\\ 890timesbetweensteps)\end{array}$
Ion energy window	$E_{th,SiC{l}_x}\approx 10\mathrm{eV}<E_i<E_{th,Si}\approx 50\mathrm{eV}$	$E_i=24\mathrm{eV}\text{(model)}$	$\begin{array}{r}Huardetal.:physicalsputteringabove50\\ \mathrm{eV},synergy71percentat65\mathrm{eV}forAR\\ 10;KLA:low-biasRIEat30Vstill\\ reaching75\mathrm{eV}\end{array}$ $\begin{array}{r}(myreading:thewindowissetbythe\\ passivatedandbaresputterthresholds)\end{array}$
Etch per cycle	$EP{C}_{ideal}=1\text{ML}(0.9\text{to}1.1)$	$EP{C}_{exp}\approx 1.3\text{to}14\text{angstrom}$	$\begin{array}{r}Huardetal.:themodelMLisa3angstrom\\ cell,aSi(100)layeris1.36angstrom;\\ KLA:0.9to1.1MLwindow\end{array}$ $\begin{array}{r}(computed:themodelMLis2.2Si(100)\\ layers)\end{array}$

Synergy and Non-Ideality

Name	Formula	Related Formulae	Variables/Constants
Synergy	$Sy=\frac{\text{self-limited removal}}{\text{total removal}}\times 100\%$	$S{y}_{KLA}=\frac{EPC}{EP{C}_{ideal}}$	$\begin{array}{r}Huardetal.(2017):100percent\\ self-limited,0percentcontinuous;KLA:\\ Si{O}_{2}quasi-ALEsynergy0.80at1.5srising\\ to0.9\end{array}$ $\begin{array}{r}(myreading:thetwodefinitionsmeasure\\ differentthings;seetheQ/A)\end{array}$
Non-ideality	$C_i/C_n=0.02,0.01,10^{-3}\Rightarrow Sy=63,76,96\%$	$C{l}_2:1,10,100\mathrm{ppm}\Rightarrow Sy=98,94,71\%$	$\begin{array}{r}{C}_i/C_n:iontoneutralfluxratioduring\\ passivation;Huardetal.:AR2,roughness\\ 1.8,1.6,1.4and0.9againstCi/Cn\end{array}$ $\begin{array}{r}(myreading:anyoverlapofthetwo\\ half-stepsturnsALEbackintoRIE)\end{array}$
Over-etch and time	$\text{over-etch}:>100\%\to \approx 25\%\text{(2017)},175\%\to 25\%\text{(KLA)}$	$t_{ALE}=15\text{to}20\min \text{against}\approx 2\min$	$\begin{array}{r}Huardetal.:FinFETgatecase,200pulses\\ with5spurges;KLA:3Dcorners,\\ aspect-ratio-independentidealALE\end{array}$ $(computed:ALE7.5to10timesslower)$

Oxide, Thermal and Conversion ALE

Name	Formula	Related Formulae	Variables/Constants
SiO2 quasi-ALE	$Ar/C_4{F}_8=95/5,300\mathrm{W},0\mathrm{V}\to Ar,600\mathrm{W},45\mathrm{V}$	$EPC=f(t_p)$	$\begin{array}{r}KLA:40\mathrm{mTorr},polymerthicknessgrows\\ withpassivationtime{t}_p,etchstopatlong\\ t_p,S{i}_3{N}_{4}neverreachingsteadyetch,\\ selectiveSi{O}_{2}overS{i}_3{N}_4\end{array}$ $\begin{array}{r}(myreading:fluorocarbondepositionis\\ notself-limiting,hencequasi-ALE)\end{array}$
Thermal ALE	$A{l}_2{O}_3:HF\text{then TMA},0.51\mathring{\mathrm{A}}/cycle$	$\Delta m=-15.9ng/c{\mathrm{m}}^2\text{per cycle}$	$\begin{array}{r}George:fluorinationthenligandexchange,\\ spontaneousself-limitingreactionswith\\ volatileproducts,TMAselectiveforA{l}_2{O}_3\\ andHf{O}_2\end{array}$ $(computed:impliedfilmdensity3.1g/c{\mathrm{m}}^3)$
Conversion ALE	$W:O_3,BC{l}_3,HF(207{}^{\circ }C),2.56\mathring{\mathrm{A}}/cycle$	$TiN:O_3,HF,0.2\mathring{\mathrm{A}}/cycle(\ge 250{}^{\circ }C)$	$\begin{array}{r}George:Wagainsta3.19angstrombcc\\ cell,W{O}_{3}at4.19\mathring{\mathrm{A}}/cyclewithWasetch\\ stop,Sn(acac{)}_2/HFforHf{O}_2,GaAs,GaN,InP\\ andPbSe\end{array}$ $\begin{array}{r}(computed:Wremoves0.80unitcellsper\\ cycle)\end{array}$

Important Questions

Ques: How does plasma atomic layer etching work?

By separating surface modification from removal. Huard et al. describe a cycle in which chlorine passivates silicon until the surface is fully covered with SiCl${}_x$, reaching 2.77 Cl per site after about 50 ms. Low-energy argon ions then remove only that layer. Their ideal pulses were 0.5 s of passivation and 3.0 s of ion bombardment. KLA’s modelled reactor ran Ar/Cl${}_2$ 70/30 at 0 $\mathrm{V}$ bias, then argon with 100 ppm Cl${}_2$ at 30 $\mathrm{V}$, in a 300 $\mathrm{W}$ ICP at 20 $\mathrm{mTorr}$. The chlorine flux fell from $7.0\times 10^{17}$ to $7.9\times 10^{14}$ $\mathrm{c}{\mathrm{m}}^2/s$, about 890 times (computed). KLA notes that chlorination lowers the sputter threshold of the top layer, which makes the removal step selective to it. Reference: C. M. Huard, Y. Zhang, S. Sriraman, A. Paterson, K. J. Kanarik and M. J. Kushner, ‘Atomic layer etching of 3D structures in silicon: self-limited and nonideal reactions’, Journal of Vacuum Science and Technology A 35(3), 031306 (2017): ALE as a self-limited passivation step and a removal step, Cl passivating Si to SiClx then low-energy Ar+ removing it, synergy Sy as material removed by self-limited reactions over total per cycle with 100 per cent fully self-limited and 0 per cent continuous, ideal EPC of 1 ML where the model ML is a 3 angstrom cell (Si(100) layer 1.36 angstrom), experimental EPC about 1.3 to 14 angstrom per cycle, ion window of perhaps a few tens of eV with thresholds 10 eV for SiClx and 50 eV for bare Si and 24 eV used, physical sputtering above 50 eV and Sy 71 per cent at 65 eV for AR 10, ion flux during passivation Ci/Cn of 0.02, 0.01 and 1e-3 giving Sy 63, 76 and 96 per cent, Cl2 contamination of 1, 10 and 100 ppm in the ion step giving 98, 94 and 71 per cent, 0.5 s passivation and 3.0 s ion steps, 2.77 Cl per site after about 50 ms, about 4e15 cm-2 ion fluence to steady roughness, FinFET case with Ar/Cl2 70/30, 42 ms and 0.55 s, 200 pulses taking 15 to 20 min with 5 s purges against about 2 min continuous and over-etch falling from over 100 per cent to about 25 per cent (fetch-tool summary); C. Huard, P. Panneerchelvam and A. Agarwal (KLA-Tencor), ‘Computational modeling as a tool to optimize and understand the atomic layer etching process’, AVS NCCAVS Plasma Applications Group (about 2018): extreme selectivity, high rate and ARDE-free etching hard to achieve together, chlorination lowering the sputter threshold, synergy defined as EPC over ideal EPC, ideal Si EPC window 0.9 to 1.1 ML, 24 eV ideal Ar+, low-bias RIE at 30 V reaching 75 eV maximum ion energy, reactor at 20 mTorr, 200 sccm, 300 W ICP at 10 MHz with Ar/Cl2 70/30 at 0 V and Ar + 100 ppm Cl2 at 30 V, Cl flux 7.0e17 and 7.9e14 cm-2 s-1, SiO2 quasi-ALE with Ar/C4F8 95/5 at 40 mTorr and 300 W then Ar at 600 W and 45 V, EPC depending on passivation time, synergy 0.80 at 1.5 s rising to about 0.9 before etch stop, Si3N4 not reaching steady etching, ALE needing about 25 per cent over-etch against 175 per cent for low-bias RIE, aspect-ratio-independent ideal ALE at AR 2 and 10, continuous etch at 40 V stopping on polymer and at 100 V eroding the nitride shoulder (fetch-tool summary).

Ques: What sets the ion energy window?

The sputter thresholds of the modified and unmodified surface. Huard et al. use 10 $\mathrm{eV}$ for chemically enhanced sputtering of SiClx and 50 $\mathrm{eV}$ for physical sputtering of bare silicon, describe the real window as perhaps a few tens of $\mathrm{eV}$, and use 24 $\mathrm{eV}$ ions in the ideal case. Above 50 $\mathrm{eV}$ bare silicon sputters: at 65 $\mathrm{eV}$ synergy fell to 71 per cent for aspect ratio 10. KLA shows why ordinary low-bias RIE cannot stay in the window: at 30 V bias its maximum ion energy reached 75 $\mathrm{eV}$. In my reading, ALE therefore needs a tight ion energy distribution, not just a low average bias, which is easier in an ICP with separate bias control. Reference: C. M. Huard, Y. Zhang, S. Sriraman, A. Paterson, K. J. Kanarik and M. J. Kushner, ‘Atomic layer etching of 3D structures in silicon: self-limited and nonideal reactions’, Journal of Vacuum Science and Technology A 35(3), 031306 (2017): ALE as a self-limited passivation step and a removal step, Cl passivating Si to SiClx then low-energy Ar+ removing it, synergy Sy as material removed by self-limited reactions over total per cycle with 100 per cent fully self-limited and 0 per cent continuous, ideal EPC of 1 ML where the model ML is a 3 angstrom cell (Si(100) layer 1.36 angstrom), experimental EPC about 1.3 to 14 angstrom per cycle, ion window of perhaps a few tens of eV with thresholds 10 eV for SiClx and 50 eV for bare Si and 24 eV used, physical sputtering above 50 eV and Sy 71 per cent at 65 eV for AR 10, ion flux during passivation Ci/Cn of 0.02, 0.01 and 1e-3 giving Sy 63, 76 and 96 per cent, Cl2 contamination of 1, 10 and 100 ppm in the ion step giving 98, 94 and 71 per cent, 0.5 s passivation and 3.0 s ion steps, 2.77 Cl per site after about 50 ms, about 4e15 cm-2 ion fluence to steady roughness, FinFET case with Ar/Cl2 70/30, 42 ms and 0.55 s, 200 pulses taking 15 to 20 min with 5 s purges against about 2 min continuous and over-etch falling from over 100 per cent to about 25 per cent (fetch-tool summary); C. Huard, P. Panneerchelvam and A. Agarwal (KLA-Tencor), ‘Computational modeling as a tool to optimize and understand the atomic layer etching process’, AVS NCCAVS Plasma Applications Group (about 2018): extreme selectivity, high rate and ARDE-free etching hard to achieve together, chlorination lowering the sputter threshold, synergy defined as EPC over ideal EPC, ideal Si EPC window 0.9 to 1.1 ML, 24 eV ideal Ar+, low-bias RIE at 30 V reaching 75 eV maximum ion energy, reactor at 20 mTorr, 200 sccm, 300 W ICP at 10 MHz with Ar/Cl2 70/30 at 0 V and Ar + 100 ppm Cl2 at 30 V, Cl flux 7.0e17 and 7.9e14 cm-2 s-1, SiO2 quasi-ALE with Ar/C4F8 95/5 at 40 mTorr and 300 W then Ar at 600 W and 45 V, EPC depending on passivation time, synergy 0.80 at 1.5 s rising to about 0.9 before etch stop, Si3N4 not reaching steady etching, ALE needing about 25 per cent over-etch against 175 per cent for low-bias RIE, aspect-ratio-independent ideal ALE at AR 2 and 10, continuous etch at 40 V stopping on polymer and at 100 V eroding the nitride shoulder (fetch-tool summary); Inductively Coupled Plasma Etcher (ICP-RIE) page of this vault.

Ques: How is synergy defined, and which definition is more useful?

The two sources define it differently. Huard et al. (2017) define synergy as the material removed by self-limited reactions divided by the total removed per cycle, from 100 per cent for ideal ALE to 0 per cent for continuous etching. The KLA slides define it as EPC divided by the ideal EPC, reporting 0.80 at 1.5 s of passivation for SiO${}_2$. The first measures purity of mechanism, the second completeness of removal (my reading). A process can etch exactly one monolayer per cycle partly by continuous sputtering and score well on the KLA measure but poorly on the 2017 one. The 2017 definition is more useful for judging self-limitation, the defining property of ALE (my reading). Reference: C. M. Huard, Y. Zhang, S. Sriraman, A. Paterson, K. J. Kanarik and M. J. Kushner, ‘Atomic layer etching of 3D structures in silicon: self-limited and nonideal reactions’, Journal of Vacuum Science and Technology A 35(3), 031306 (2017): ALE as a self-limited passivation step and a removal step, Cl passivating Si to SiClx then low-energy Ar+ removing it, synergy Sy as material removed by self-limited reactions over total per cycle with 100 per cent fully self-limited and 0 per cent continuous, ideal EPC of 1 ML where the model ML is a 3 angstrom cell (Si(100) layer 1.36 angstrom), experimental EPC about 1.3 to 14 angstrom per cycle, ion window of perhaps a few tens of eV with thresholds 10 eV for SiClx and 50 eV for bare Si and 24 eV used, physical sputtering above 50 eV and Sy 71 per cent at 65 eV for AR 10, ion flux during passivation Ci/Cn of 0.02, 0.01 and 1e-3 giving Sy 63, 76 and 96 per cent, Cl2 contamination of 1, 10 and 100 ppm in the ion step giving 98, 94 and 71 per cent, 0.5 s passivation and 3.0 s ion steps, 2.77 Cl per site after about 50 ms, about 4e15 cm-2 ion fluence to steady roughness, FinFET case with Ar/Cl2 70/30, 42 ms and 0.55 s, 200 pulses taking 15 to 20 min with 5 s purges against about 2 min continuous and over-etch falling from over 100 per cent to about 25 per cent (fetch-tool summary); C. Huard, P. Panneerchelvam and A. Agarwal (KLA-Tencor), ‘Computational modeling as a tool to optimize and understand the atomic layer etching process’, AVS NCCAVS Plasma Applications Group (about 2018): extreme selectivity, high rate and ARDE-free etching hard to achieve together, chlorination lowering the sputter threshold, synergy defined as EPC over ideal EPC, ideal Si EPC window 0.9 to 1.1 ML, 24 eV ideal Ar+, low-bias RIE at 30 V reaching 75 eV maximum ion energy, reactor at 20 mTorr, 200 sccm, 300 W ICP at 10 MHz with Ar/Cl2 70/30 at 0 V and Ar + 100 ppm Cl2 at 30 V, Cl flux 7.0e17 and 7.9e14 cm-2 s-1, SiO2 quasi-ALE with Ar/C4F8 95/5 at 40 mTorr and 300 W then Ar at 600 W and 45 V, EPC depending on passivation time, synergy 0.80 at 1.5 s rising to about 0.9 before etch stop, Si3N4 not reaching steady etching, ALE needing about 25 per cent over-etch against 175 per cent for low-bias RIE, aspect-ratio-independent ideal ALE at AR 2 and 10, continuous etch at 40 V stopping on polymer and at 100 V eroding the nitride shoulder (fetch-tool summary).

Ques: What makes real ALE non-ideal?

Overlap between the two half-cycles. Huard et al. model ions arriving during passivation: an ion-to-neutral flux ratio of 0.02, 0.01 and $10^{-3}$ gave synergies of 63, 76 and 96 per cent, with roughness falling from 1.8 to 1.4. Chlorine left in the ion step did the same: 1, 10 and 100 ppm gave 98, 94 and 71 per cent. Experimental etch per cycle spans about 1.3 to 14 angstrom, against an ideal of one monolayer. Their model monolayer is a 3 angstrom cell, against 1.36 angstrom for a Si(100) layer, or 2.2 layers (computed). In my reading, any overlap turns ALE back into RIE, so purges and gas switching matter as much as the plasma. Reference: C. M. Huard, Y. Zhang, S. Sriraman, A. Paterson, K. J. Kanarik and M. J. Kushner, ‘Atomic layer etching of 3D structures in silicon: self-limited and nonideal reactions’, Journal of Vacuum Science and Technology A 35(3), 031306 (2017): ALE as a self-limited passivation step and a removal step, Cl passivating Si to SiClx then low-energy Ar+ removing it, synergy Sy as material removed by self-limited reactions over total per cycle with 100 per cent fully self-limited and 0 per cent continuous, ideal EPC of 1 ML where the model ML is a 3 angstrom cell (Si(100) layer 1.36 angstrom), experimental EPC about 1.3 to 14 angstrom per cycle, ion window of perhaps a few tens of eV with thresholds 10 eV for SiClx and 50 eV for bare Si and 24 eV used, physical sputtering above 50 eV and Sy 71 per cent at 65 eV for AR 10, ion flux during passivation Ci/Cn of 0.02, 0.01 and 1e-3 giving Sy 63, 76 and 96 per cent, Cl2 contamination of 1, 10 and 100 ppm in the ion step giving 98, 94 and 71 per cent, 0.5 s passivation and 3.0 s ion steps, 2.77 Cl per site after about 50 ms, about 4e15 cm-2 ion fluence to steady roughness, FinFET case with Ar/Cl2 70/30, 42 ms and 0.55 s, 200 pulses taking 15 to 20 min with 5 s purges against about 2 min continuous and over-etch falling from over 100 per cent to about 25 per cent (fetch-tool summary).

Ques: What does ALE gain over continuous etching, and at what cost?

Less over-etch and no aspect-ratio dependence, at the cost of time. In Huard et al.’s FinFET gate case, ALE with Ar/Cl${}_2$ 70/30, 42 ms passivation and 0.55 s ion steps needed 200 pulses. With 5 s purges that took 15 to 20 $\min$, against about 2 $\min$ for continuous etching, 7.5 to 10 times slower (computed). In return, over-etch fell from over 100 per cent to about 25 per cent. KLA reports 25 per cent over-etch for 3D corners against 175 per cent for low-bias RIE, and aspect-ratio-independent rates at AR 2 and 10. For SiO${}_2$, its C${}_4$F${}_8$/Ar quasi-ALE depends on passivation time, with etch stop when polymer accumulates. Reference: C. M. Huard, Y. Zhang, S. Sriraman, A. Paterson, K. J. Kanarik and M. J. Kushner, ‘Atomic layer etching of 3D structures in silicon: self-limited and nonideal reactions’, Journal of Vacuum Science and Technology A 35(3), 031306 (2017): ALE as a self-limited passivation step and a removal step, Cl passivating Si to SiClx then low-energy Ar+ removing it, synergy Sy as material removed by self-limited reactions over total per cycle with 100 per cent fully self-limited and 0 per cent continuous, ideal EPC of 1 ML where the model ML is a 3 angstrom cell (Si(100) layer 1.36 angstrom), experimental EPC about 1.3 to 14 angstrom per cycle, ion window of perhaps a few tens of eV with thresholds 10 eV for SiClx and 50 eV for bare Si and 24 eV used, physical sputtering above 50 eV and Sy 71 per cent at 65 eV for AR 10, ion flux during passivation Ci/Cn of 0.02, 0.01 and 1e-3 giving Sy 63, 76 and 96 per cent, Cl2 contamination of 1, 10 and 100 ppm in the ion step giving 98, 94 and 71 per cent, 0.5 s passivation and 3.0 s ion steps, 2.77 Cl per site after about 50 ms, about 4e15 cm-2 ion fluence to steady roughness, FinFET case with Ar/Cl2 70/30, 42 ms and 0.55 s, 200 pulses taking 15 to 20 min with 5 s purges against about 2 min continuous and over-etch falling from over 100 per cent to about 25 per cent (fetch-tool summary); C. Huard, P. Panneerchelvam and A. Agarwal (KLA-Tencor), ‘Computational modeling as a tool to optimize and understand the atomic layer etching process’, AVS NCCAVS Plasma Applications Group (about 2018): extreme selectivity, high rate and ARDE-free etching hard to achieve together, chlorination lowering the sputter threshold, synergy defined as EPC over ideal EPC, ideal Si EPC window 0.9 to 1.1 ML, 24 eV ideal Ar+, low-bias RIE at 30 V reaching 75 eV maximum ion energy, reactor at 20 mTorr, 200 sccm, 300 W ICP at 10 MHz with Ar/Cl2 70/30 at 0 V and Ar + 100 ppm Cl2 at 30 V, Cl flux 7.0e17 and 7.9e14 cm-2 s-1, SiO2 quasi-ALE with Ar/C4F8 95/5 at 40 mTorr and 300 W then Ar at 600 W and 45 V, EPC depending on passivation time, synergy 0.80 at 1.5 s rising to about 0.9 before etch stop, Si3N4 not reaching steady etching, ALE needing about 25 per cent over-etch against 175 per cent for low-bias RIE, aspect-ratio-independent ideal ALE at AR 2 and 10, continuous etch at 40 V stopping on polymer and at 100 V eroding the nitride shoulder (fetch-tool summary).

Ques: How does thermal ALE differ from plasma ALE?

It uses spontaneous chemistry instead of ions. George lists its requirements: spontaneous, sequential, self-limiting reactions with favourable thermochemistry and volatile products. Al${}_2$O${}_3$ is fluorinated by HF and then removed by ligand exchange with TMA, losing 15.9 $ng/c{\mathrm{m}}^2$, or 0.51 $\mathring{\mathrm{A}}/cycle$, per cycle. That implies a film density of 3.1 $g/c{\mathrm{m}}^3$ (computed). Tungsten is oxidised by O${}_3$, converted with BCl${}_3$ and fluorinated by HF at 207 ${}^{\circ }\mathrm{C}$, removing 2.56 $\mathring{\mathrm{A}}/cycle$, 0.80 of its bcc cell (computed). TiN etches at 0.2 $\mathring{\mathrm{A}}/cycle$ at 250 ${}^{\circ }\mathrm{C}$ or above. Selectivity comes from which metals form volatile complexes (George). In my reading, the sequence mirrors the ALD Reactor page in reverse. Reference: S. M. George (University of Colorado Boulder), ‘Mechanisms for thermal atomic layer etching’, AVS NCCAVS Plasma Applications Group slides: spontaneous, sequential, self-limiting reactions needing favourable thermochemistry and volatile products, Al2O3 with HF and TMA at -15.9 ng/cm2 and 0.51 angstrom per cycle, ligand exchange, HfO2 + 4Sn(acac)2 + 4HF -> Hf(acac)4 + 4SnF(acac) + 2H2O, Sn(acac)2/HF for GaAs, GaN, InP and PbSe, TiN with O3 and HF at 0.2 angstrom per cycle at 250 C and above, W with O3, BCl3 and HF at 207 C at 2.56 angstrom per cycle against a 3.19 angstrom bcc cell, WO3 4.19 angstrom per cycle with W as etch stop, TMA selective for Al2O3 and HfO2 and Sn(acac)2 for Al2O3, HfO2 and ZrO2 through stable volatile complexes (fetch-tool summary); Atomic Layer Deposition (ALD) Reactor page of this vault.

References

1. C. M. Huard, Y. Zhang, S. Sriraman, A. Paterson, K. J. Kanarik and M. J. Kushner, “Atomic layer etching of 3D structures in silicon: self-limited and nonideal reactions,” Journal of Vacuum Science and Technology A, vol. 35, p. 031306, 2017. Available: https://cpseg.eecs.umich.edu/pub/articles/JVSTA_35_031306_2017.pdf
2. C. Huard, P. Panneerchelvam and A. Agarwal, “Computational modeling as a tool to optimize and understand the atomic layer etching process,” KLA-Tencor, AVS NCCAVS Plasma Applications Group. Available: https://nccavs-usergroups.avs.org/wp-content/uploads/PAG2018/PAG818-2-Huard1-KLA.pdf
3. S. M. George, “Mechanisms for thermal atomic layer etching,” University of Colorado Boulder, AVS NCCAVS Plasma Applications Group. Available: https://nccavs-usergroups.avs.org/wp-content/uploads/PAG2018/PAG818-3-George-UnivCol.pdf
4. Inductively Coupled Plasma Etcher (ICP-RIE) page of this vault.
5. Atomic Layer Deposition (ALD) Reactor page of this vault.

Related pages in this vault: Inductively Coupled Plasma Etcher (ICP-RIE), Atomic Layer Deposition (ALD) Reactor, Parallel-Plate Reactive Ion Etcher (RIE), Deep Reactive Ion Etcher (Bosch DRIE).