Vapor HF and Plasma Dry Clean Tool

1. Theory

• Definition: Dry cleans remove oxide without liquid, either with anhydrous HF vapour and an alcohol catalyst or with a remote NH${}_3$/NF${}_3$ plasma that converts oxide to a salt sublimed by heating (EPFL; Polifab; ASM patent; Applied Materials patent).
• Vapour HF chemistry: SiO${}_2$ + 4HF gives SiF${}_4$ and water through silanol intermediates; an alcohol such as ethanol ionises HF and acts as a catalyst, and the water produced is also a catalyst that must be controlled (Polifab; EPFL; KLA).
• Why vapour: Liquid HF can destroy tiny structures or make them stick by cohesion, and wet HF corrodes exposed aluminium, so release etches use vapour (Polifab; KLA).
• Aluminium attack: HF etches Al only when dissociated by water, so keeping water away protects the metal (Gablech et al.).
• Plasma dry clean chemistry: NH${}_4$F and NH${}_4$HF${}_2$ react with oxide to form ammonium hexafluorosilicate, which sublimes near 100 ${}^{\circ }$C into SiF${}_4$, NH${}_3$ and HF (ASM patent; Applied Materials patent).
• Selectivity: The plasma clean etches oxide faster than nitride, by more than about 7:1, and barely attacks silicon (ASM patent; Applied Materials patent).

2. Working Principle

• Vapour HF release tool: HF, ethanol and N${}_2$ at a fixed 125 Torr, with HF partial pressure setting rates of 12 to 175 nm/min (EPFL).
• Production vapour etchers: HF and alcohol at 75 to 150 Torr and elevated temperature, oxide rates of 0.01 to 1 $\mu$m/min, single-wafer to 25-wafer batch modules (KLA).
• Heated-substrate HF etcher: Wafer heated to 35 to 60 ${}^{\circ }$C over an HF reservoir, giving 4 to 10 $\mu$m/h (Polifab).
• Remote-plasma pre-clean: NH${}_3$ and NF${}_3$ excited away from the wafer, salt formed at about 22 to 40 ${}^{\circ }$C, then a short anneal sublimes it (ASM patent; Applied Materials patent).
• Cycling: Thick oxide is removed in two to five steps of about 25 to 200 angstrom each (Applied Materials patent).

3. Applications

• MEMS release: Removing sacrificial oxide under polysilicon or SOI structures without stiction, with yields typically near 100 per cent, in microphones, CMOS MEMS and micromirror arrays (EPFL; Polifab; KLA).
• Aluminium-bearing devices: Releasing structures with unprotected Al mirrors and bond pads (KLA; Gablech et al.).
• Pre-epitaxy and pre-contact clean: Native-oxide removal before deposition, an alternative to the HF dip of the RCA Wet Bench page (ASM patent; my reading).
• Damage-free drying: Avoiding the capillary forces covered on the Spin Rinse Dryer and IPA Vapor Dryer page (my reading).

Vapour HF Etch

Name	Formula	Related Formulae	Variables/Constants
Vapour HF etch of oxide	$Si{O}_2+4HF\to Si{F}_4\uparrow +2H_{2}O$	$\text{catalyst: }{C}_2{H}_{5}OH\text{(not consumed)}$	$\begin{array}{r}Polifab:simplifiedreactionthrough\\ silanolintermediates,wateraffectingthe\\ rate;EPFL:ethanolacatalystwithout\\ whichSi{O}_{2}isnotetched,productsSi{F}_{4}gas\\ andwatervapour\end{array}$ $\begin{array}{r}(myreading:thecatalystionizesHFon\\ thesurface,sotheetchneedsanadsorbed\\ polarfilm)\end{array}$
uEtch rate range	$R_{ox}=12\text{to}175nm/min$	$p_{tot}=125\mathrm{Torr}$	$\begin{array}{r}EPFL:fiverecipesatfixedtotalpressure\\ differinginHFpartialpressure,minimum\\ ratesonthermaloxideatmaximumloading,\\ depositedoxidesfaster\end{array}$ $\begin{array}{r}(computed:Recipe5isabout15times\\ fasterthanRecipe1)\end{array}$
Heated-substrate vapour etcher	$R_{ox}=4\text{to}10\mu m/h$	$T_{sub}=35\text{to}60{}^{\circ }C$	$\begin{array}{r}Polifab:Idonusetcher,etching\\ spontaneousatroomtemperature,\\ controlledheatingsettinghowmuchvapour\\ reachesthesubstrate\end{array}$ $\begin{array}{r}(computed:4to10\mu m/his67to167\\ nm/min,overlappingtheEPFLrange)\end{array}$
Alcohol-catalysed mechanism	$Si{O}_2+2H{F}_2^{-}+2A{H}^{+}\to Si{F}_4+2H_{2}O+2A$	$R_{ox}=0.01\text{to}1\mu m/min,p=75\text{to}150\mathrm{Torr}$	$\begin{array}{r}A:alcohol;KLA:alcoholionisesHFandis\\ acatalyst,waterby-productalsoa\\ catalystthatmustbecontrolledand\\ removed,reducedpressureandelevated\\ temperaturekeepby-productsinthegas\\ phase\end{array}$ $\begin{array}{r}(computed:0.01to1\mu m/minis10to1000\\ nm/min)\end{array}$

NH3/NF3 Plasma Dry Clean

Name	Formula	Related Formulae	Variables/Constants
Etchant formation	$N{H}_3+N{F}_3\stackrel{\text{plasma}}{\to }N{H}_{4}F+N{H}_{4}F\cdot HF$	$N{H}_{4}H{F}_2\equiv N{H}_{4}F\cdot HF$	$\begin{array}{r}AppliedMaterials:effluentsthoughtto\\ includeN{H}_{4}FandN{H}_{4}F.HF;ASM:N{H}_{4}FandN{H}_{4}H{F}_2\\ fromammoniawithN{F}_3,HFor{F}_2\end{array}$ $\begin{array}{r}(myreading:thetwopatentsnamethesame\\ bifluoridesaltintwonotations)\end{array}$
Salt formation on oxide	$Si{O}_2\to (N{H}_4{)}_{2}Si{F}_6(s)+H_{2}O\uparrow$	$T_{sub}\approx 22\text{to}40{}^{\circ }C$	$\begin{array}{r}ASM:Si{F}_{4}combineswithN{H}_{4}Ftoform\\ ammoniumhexafluorosilicate,substrate22\\ to28{}^{\circ }C;AppliedMaterials:substrate\\ belowabout100{}^{\circ }C,30{}^{\circ }Cinexamples\end{array}$ $\begin{array}{r}(myreading:reactionstoichiometryisnot\\ balancedineitherpatent,soonlythe\\ productsaregiven)\end{array}$
Sublimation anneal	$(N{H}_4{)}_{2}Si{F}_6(s)\stackrel{\approx 100{}^{\circ }C}{\to }Si{F}_4+N{H}_3+HF$	$t_{anneal}<30\mathrm{s}$	$\begin{array}{r}AppliedMaterials:sublimationproducts\\ andper-stepannealtime;ASM:about100\\ {}^{\circ }C,broadrange80to500{}^{\circ }C\end{array}$ $\begin{array}{r}(myreading:thesaltlayerself-limits\\ eachstep,sothickoxidesareremovedin\\ cycles)\end{array}$

Ratios, Selectivity and Cycling

Name	Formula	Related Formulae	Variables/Constants
Gas ratio	${\left(\frac{H}{F}\right)}_{atomic}=\frac{3n_{N{H}_3}}{3n_{N{F}_3}}=\frac{n_{N{H}_3}}{n_{N{F}_3}}$	$2:1\text{(conventional)},4.5:1\text{(ASM)},10:1\text{(smooth)}$	$\begin{array}{r}AppliedMaterials:conventionalSiConi\\ about2:1H:F,smoothprocess350\mathrm{sccm}N{H}_3\\ with35\mathrm{sccm}N{F}_3;ASM:N{H}_{3}tofluorinegas\\ molarratioabout4.5:1\end{array}$ $\begin{array}{r}(computed:threeHperN{H}_{3}andthreeFper\\ N{F}_3,sotheatomicratioequalstheflow\\ ratio;350/35=10)\end{array}$
Oxide to nitride selectivity	$S_{ox:SiN}>7:1\text{(one cycle)}$	$S_{ox:SiN}\approx 14:1\text{to}150:1\text{(second cycle)}$	$\begin{array}{r}ASM:greaterthanabout7:1,about6:1to\\ 150:1,second-cycleexamples;Applied\\ Materials:oxide-selective,littleattack\\ ofsilicon,nonumericvalue\end{array}$ $\begin{array}{r}(myreading:thehighersecond-cycle\\ valuesreflectathinnerremainingoxide,\\ notadifferentchemistry)\end{array}$
Cycled removal	$\Delta d_{step}\approx 25\text{to}200\mathring{\mathrm{A}}$	$N_{steps}=2\text{to}5$	$\begin{array}{r}AppliedMaterials:per-steplimitandthe\\ shareoffilmremovedperstepfortwoto\\ fivesteps,plasma20to100\mathrm{W}at50to90\\ \mathrm{kHz},1to10\mathrm{Torr}\end{array}$ $\begin{array}{r}(computed:2nmofnativeoxidefitsina\\ singlestep)\end{array}$

Metal Compatibility and Wet Alternative

Name	Formula	Related Formulae	Variables/Constants
Why water attacks aluminium	$HF+H_{2}O\to H_3{O}^{+}+F^{-}\text{(etches Al)}$	$H_{2}O+S{O}_3\to H_{2}S{O}_4\text{(oleum removes water)}$	$\begin{array}{r}Gablechetal.:HFetchesAlonlywhen\\ dissociatedbywater,theetchitself\\ produceswateratthesurface,oleumbinds\\ it;KLA:wetHFcorrodesexposedAl,\\ vapourprocesscompatiblewithunprotected\\ Al\end{array}$ $\begin{array}{r}(myreading:adry,low-pressurevapour\\ processkeepsthewatertoosparseto\\ dissociateHFattheAlsurface)\end{array}$
Wet anhydrous alternative	$R_{ox}=0.93\pm 0.05\mu m/min(70{}^{\circ }C)$	$\frac{\Delta R_{\square ,Al}}{R_{\square ,Al}}\approx 7.6\%(\approx 9\min )$	$\begin{array}{r}Gablechetal.:48percentHFwith20per\\ centoleum,about14timesfasterthan6:1\\ BOEatabout70nm/min,butneeding\\ critical-pointdrying\end{array}$ $\begin{array}{r}(computed:930/70=13.3,consistentwith\\ their14times)\end{array}$
Material compatibility	$\text{sacrificial: thermal, TEOS, SOI, quartz (PECVD, spin-on conditional)}$	$\text{metals: Al, Au, TiW, Ni (Cu conditional)}$	$\begin{array}{r}KLA:compatibilitytable,greenandyellow\\ marks,Si-richLPCVDnitrideyellowas\\ protectivelayer;EPFL:siliconnitride,\\ Ti,TiN,Ta,TaN,Ge,polymersandglasses\\ incompatibleintheuEtch(nitridewith\\ exceptions)\end{array}$ $\begin{array}{r}(myreading:yellowmarksmeanconditional\\ use;nolegendappearsinthebrief)\end{array}$

Important Questions

Ques: How does vapour HF etch silicon dioxide?

Through a surface reaction that needs a catalyst. Polifab gives the simplified reaction $Si{O}_2+4HF\to Si{F}_4+2H_{2}O$, proceeding through silanol intermediates, with the water produced affecting the etch rate. EPFL describes its SPTS uEtch as using anhydrous HF and ethanol vapour carried by nitrogen; ethanol acts as a catalyst that is not consumed, and without it the oxide is not etched. KLA writes the catalysed step as $Si{O}_2+2H{F}_2^{-}+2A{H}^{+}\to Si{F}_4+2H_{2}O+2A$, with the alcohol A ionising HF, and notes that the water produced is also a catalyst that must be controlled and removed. Polifab likewise notes that the water produced affects the rate. In my reading, controlling water is therefore controlling the rate. Reference: Polifab, Politecnico di Milano, ‘HF vapor release (Idonus)’ equipment page: only vapour contacts the substrate because liquid HF would destroy tiny structures or make them stick by cohesion forces, simplified reaction SiO2 + 4HF -> SiF4 + 2H2O through silanol intermediates with water affecting the rate, etching spontaneous at room temperature, controlled substrate heating 35 to 60 C setting the rate, typical rate 4 to 10 um/h, MEMS release on SOI, up to 100 mm wafers, 100 to 160 mL etchant (fetch-tool summary); EPFL Center of MicroNanoTechnology, ‘SPTS uEtch’ equipment page: isotropic vapour-phase release of sacrificial SiO2 without liquids or supercritical drying, anhydrous HF with ethanol vapour carried by N2, products SiF4 gas and water vapour, ethanol a catalyst not consumed and SiO2 not etched without it, ethanol keeping the surface dry, water vapour able to cause stiction, total pressure fixed at 125 Torr with five recipes of different HF partial pressure, minimum thermal-oxide rates 12, 38, 125, 153 and 175 nm/min, deposited oxides faster, rate falling with exposed oxide area (macro-loading), 200 to 250 C pre-bake, incompatible materials including polymers, glasses, Ge, Ti, TiN, Ta, TaN and silicon nitride with exceptions, minimum step 15 min (fetch-tool summary); SPTS Technologies (KLA), ‘Introduction to HF Vapor Etch’, application brief HF-Intro-Q1/23 (2023), read from the PDF supplied by the user: SiO2 + 4HF -> SiF4 + 2H2O; wet HF and rinse liquids cause stiction on drying and corrode exposed metals, most notably aluminium; vapour penetrates small features and allows longer undercuts; alcohol A ionises HF and acts as catalyst, SiO2 + 2HF2- + 2AH+ -> SiF4 + 2H2O + 2A; water by-product also a catalyst and must be controlled and removed; Primaxx HF/alcohol process at reduced pressure and elevated temperature, 75 to 150 torr, residue-free, vertical and lateral oxide rates 0.01 to 1 um/min; device yields typically about 100 per cent; compatible with unprotected Al; reduced pressure keeps by-products in the gas phase for high selectivity to metals and avoids localized loading; material table: thermal oxide, TEOS, SOI bonded oxide and quartz green as sacrificial oxide, PECVD and spin-on oxide yellow; alumina, ALD alumina, aluminium and silicon carbide green as protective layer, Si-rich LPCVD nitride yellow; aluminium, gold, TiW and nickel green and copper yellow as metal/electrode/adhesion; Monarch300 13-wafer batch, Monarch25 25-wafer, Monarch3 3-wafer and uEtch single-wafer systems.

Ques: Why is vapour HF preferred over liquid HF for MEMS release?

Because liquid causes stiction. Polifab states that only vapour contacts the substrate, since liquid HF would destroy tiny structures or make them stick together through cohesion forces. KLA adds that wet HF and rinse liquids pull structures together as they dry, and quotes vapour-release yields of typically about 100 per cent. EPFL describes the uEtch as releasing devices with no liquids and no supercritical drying, but warns that the water vapour the reaction produces can itself cause stiction, so etch conditions must be matched to each device; HF partial pressure and total pressure set the rate and hence the water formed per cycle. The capillary physics is that of the Spin Rinse Dryer and IPA Vapor Dryer page. Reference: Polifab, Politecnico di Milano, ‘HF vapor release (Idonus)’ equipment page: only vapour contacts the substrate because liquid HF would destroy tiny structures or make them stick by cohesion forces, simplified reaction SiO2 + 4HF -> SiF4 + 2H2O through silanol intermediates with water affecting the rate, etching spontaneous at room temperature, controlled substrate heating 35 to 60 C setting the rate, typical rate 4 to 10 um/h, MEMS release on SOI, up to 100 mm wafers, 100 to 160 mL etchant (fetch-tool summary); SPTS Technologies (KLA), ‘Introduction to HF Vapor Etch’, application brief HF-Intro-Q1/23 (2023), read from the PDF supplied by the user: SiO2 + 4HF -> SiF4 + 2H2O; wet HF and rinse liquids cause stiction on drying and corrode exposed metals, most notably aluminium; vapour penetrates small features and allows longer undercuts; alcohol A ionises HF and acts as catalyst, SiO2 + 2HF2- + 2AH+ -> SiF4 + 2H2O + 2A; water by-product also a catalyst and must be controlled and removed; Primaxx HF/alcohol process at reduced pressure and elevated temperature, 75 to 150 torr, residue-free, vertical and lateral oxide rates 0.01 to 1 um/min; device yields typically about 100 per cent; compatible with unprotected Al; reduced pressure keeps by-products in the gas phase for high selectivity to metals and avoids localized loading; material table: thermal oxide, TEOS, SOI bonded oxide and quartz green as sacrificial oxide, PECVD and spin-on oxide yellow; alumina, ALD alumina, aluminium and silicon carbide green as protective layer, Si-rich LPCVD nitride yellow; aluminium, gold, TiW and nickel green and copper yellow as metal/electrode/adhesion; Monarch300 13-wafer batch, Monarch25 25-wafer, Monarch3 3-wafer and uEtch single-wafer systems; EPFL Center of MicroNanoTechnology, ‘SPTS uEtch’ equipment page: isotropic vapour-phase release of sacrificial SiO2 without liquids or supercritical drying, anhydrous HF with ethanol vapour carried by N2, products SiF4 gas and water vapour, ethanol a catalyst not consumed and SiO2 not etched without it, ethanol keeping the surface dry, water vapour able to cause stiction, total pressure fixed at 125 Torr with five recipes of different HF partial pressure, minimum thermal-oxide rates 12, 38, 125, 153 and 175 nm/min, deposited oxides faster, rate falling with exposed oxide area (macro-loading), 200 to 250 C pre-bake, incompatible materials including polymers, glasses, Ge, Ti, TiN, Ta, TaN and silicon nitride with exceptions, minimum step 15 min (fetch-tool summary); Spin Rinse Dryer and IPA Vapor Dryer page of this vault.

Ques: What etch rates does a vapour HF tool give, and which figure is more likely?

Tens to hundreds of nm/min; the 15 $nm/min$ figure is a low-end value. KLA gives 0.01 to 1 $\mu$m/min, that is 10 to 1000 $nm/min$ (computed), at 75 to 150 $\mathrm{Torr}$. EPFL lists minimum thermal-oxide rates of 12 to 175 $nm/min$ at 125 $\mathrm{Torr}$, measured at maximum loading, with deposited oxides faster. Polifab quotes 4 to 10 $\mu m/h$, or 67 to 167 $nm/min$ (computed). Gablech et al., citing an earlier study, call anhydrous HF with alcohol slow at about 15 $nm/min$. That value sits at EPFL’s lowest recipe and KLA’s lower end, so the 15 $nm/min$ figure is less likely to be typical. Three tool sources agree on roughly 70 to 175 $nm/min$ for productive recipes (my reading). Reference: SPTS Technologies (KLA), ‘Introduction to HF Vapor Etch’, application brief HF-Intro-Q1/23 (2023), read from the PDF supplied by the user: SiO2 + 4HF -> SiF4 + 2H2O; wet HF and rinse liquids cause stiction on drying and corrode exposed metals, most notably aluminium; vapour penetrates small features and allows longer undercuts; alcohol A ionises HF and acts as catalyst, SiO2 + 2HF2- + 2AH+ -> SiF4 + 2H2O + 2A; water by-product also a catalyst and must be controlled and removed; Primaxx HF/alcohol process at reduced pressure and elevated temperature, 75 to 150 torr, residue-free, vertical and lateral oxide rates 0.01 to 1 um/min; device yields typically about 100 per cent; compatible with unprotected Al; reduced pressure keeps by-products in the gas phase for high selectivity to metals and avoids localized loading; material table: thermal oxide, TEOS, SOI bonded oxide and quartz green as sacrificial oxide, PECVD and spin-on oxide yellow; alumina, ALD alumina, aluminium and silicon carbide green as protective layer, Si-rich LPCVD nitride yellow; aluminium, gold, TiW and nickel green and copper yellow as metal/electrode/adhesion; Monarch300 13-wafer batch, Monarch25 25-wafer, Monarch3 3-wafer and uEtch single-wafer systems; EPFL Center of MicroNanoTechnology, ‘SPTS uEtch’ equipment page: isotropic vapour-phase release of sacrificial SiO2 without liquids or supercritical drying, anhydrous HF with ethanol vapour carried by N2, products SiF4 gas and water vapour, ethanol a catalyst not consumed and SiO2 not etched without it, ethanol keeping the surface dry, water vapour able to cause stiction, total pressure fixed at 125 Torr with five recipes of different HF partial pressure, minimum thermal-oxide rates 12, 38, 125, 153 and 175 nm/min, deposited oxides faster, rate falling with exposed oxide area (macro-loading), 200 to 250 C pre-bake, incompatible materials including polymers, glasses, Ge, Ti, TiN, Ta, TaN and silicon nitride with exceptions, minimum step 15 min (fetch-tool summary); Polifab, Politecnico di Milano, ‘HF vapor release (Idonus)’ equipment page: only vapour contacts the substrate because liquid HF would destroy tiny structures or make them stick by cohesion forces, simplified reaction SiO2 + 4HF -> SiF4 + 2H2O through silanol intermediates with water affecting the rate, etching spontaneous at room temperature, controlled substrate heating 35 to 60 C setting the rate, typical rate 4 to 10 um/h, MEMS release on SOI, up to 100 mm wafers, 100 to 160 mL etchant (fetch-tool summary); I. Gablech, J. Brodsky, J. Pekarek and P. Neuzil, ‘Infinite selectivity of wet SiO2 etching in respect to Al’, Micromachines 11, 365 (2020), read from the PDF supplied by the user: HF etches Al only when dissociated by water into H3O+ and F-, so water must be absent; SiO2 + 4HF -> SiF4 + 2H2O produces that water at the surface; anhydrous HF gas with alcohol vapour gives a slow rate of about 15 nm/min on thermal oxide in complex, expensive equipment (their ref. 20); 73 per cent HF gives 1.6 um/min without attacking Al but is a controlled chemical; 48 per cent HF with 20 per cent oleum binds water by H2O + SO3 -> H2SO4, etching thermal oxide at (0.93 plus or minus 0.05) um/min at about 70 C, about 14 times faster than 6:1 BOE at about 70 nm/min; Al sheet resistance up by about 7.6 per cent after about 9 min; dry vapour etch needs no critical-point drying while the wet method does; BOE etches Al and PECVD protection layers have pinholes.

Ques: How does an NH3/NF3 remote-plasma pre-clean remove native oxide?

By converting the oxide into a salt and subliming it. The Applied Materials SiConi patent states that the NF${}_3$ and NH${}_3$ plasma effluents are thought to include NH${}_4$F and NH${}_4$F.HF, which react with silicon oxide to form solid ammonium hexafluorosilicate while ammonia and water leave as vapour; heating then sublimes the solid into SiF${}_4$, NH${}_3$ and HF. The ASM patent describes the same route with NH${}_4$F and NH${}_4$HF${}_2$, the same bifluoride salt written differently (my reading). Salt formation runs with the wafer at about 22 to 28 ${}^{\circ }\mathrm{C}$ (ASM) or 30 ${}^{\circ }\mathrm{C}$ (Applied Materials examples), and sublimation near 100 ${}^{\circ }\mathrm{C}$ in both. Because the salt caps the surface, removal is cycled in steps of about 25 to 200 angstrom (Applied Materials). Reference: J. Tang, N. Ingle and D. Yang (Applied Materials), ‘Smooth SiConi etch for silicon-containing films’, US Patent 8,501,629 (2013): NF3 and NH3 plasma effluents thought to include NH4F and NH4F.HF, reacting with silicon oxide to form solid ammonium hexafluorosilicate with ammonia and water vapour, sublimation into SiF4, NH3 and HF, substrate below about 100 C during etch with 30 C in examples, sublimation near 100 C (ranges above 75 C to 115 to 150 C), conventional SiConi about 2:1 hydrogen-to-fluorine atomic ratio, smooth process below 1:1 or above 5:1 with 350 sccm NH3 and 35 sccm NF3 (about 10:1), 1 to 10 Torr, 20 to 100 W at 50 to 90 kHz, conformal and oxide-selective with little etching of amorphous, crystalline or polycrystalline silicon, multi-step etch with about 25 to 200 angstrom per step and anneals under about 30 s, no numeric selectivity given (fetch-tool summary); J. Tolle, M. G. Goodman, R. M. Vyne and E. R. Hill (ASM IP Holding), ‘Plasma pre-clean module and process’, US Patent 9,299,557 (2016): pre-clean material of ammonium fluoride NH4F and ammonium hydrogen fluoride NH4HF2 formed from ammonia and NF3, HF or F2, reacting with silicon oxide to give water vapour and SiF4 that combines with NH4F to form ammonium hexafluorosilicate, substrate at about 22 to 28 C during formation, heating to about 80 to 500 C and about 100 C for sublimation, oxide-to-nitride selectivity greater than about 7:1, about 6:1 to 150:1, about 8:1 to 12:1, second-cycle examples about 14:1 and about 30:1 to 150:1, ammonia to NF3, HF or F2 molar ratio about 4.5:1, a stated halogen-to-hydrogen gas ratio of about 3:1 to 10:1, remote-plasma ammonia with unactivated NF3 about 3:1 (fetch-tool summary).

Ques: What selectivity does the plasma pre-clean have to nitride and silicon?

High to both, with nitride quantified only by ASM. The ASM patent claims oxide-to-nitride selectivity greater than about 7:1, ranges of about 6:1 to 150:1 and 8:1 to 12:1, and second-cycle examples of about 14:1 and about 30:1 to 150:1. The Applied Materials patent gives no number but calls the process largely conformal and selective to silicon oxide, not readily etching amorphous, crystalline or polycrystalline silicon. In my reading the higher second-cycle figures arise because less oxide remains, so the reported ratio reflects nitride loss over a shorter effective exposure. For a pre-epitaxy clean, low silicon attack matters more than nitride selectivity, while for a pre-contact clean beside nitride spacers the 7:1 floor is the figure to check (my reading). Reference: J. Tolle, M. G. Goodman, R. M. Vyne and E. R. Hill (ASM IP Holding), ‘Plasma pre-clean module and process’, US Patent 9,299,557 (2016): pre-clean material of ammonium fluoride NH4F and ammonium hydrogen fluoride NH4HF2 formed from ammonia and NF3, HF or F2, reacting with silicon oxide to give water vapour and SiF4 that combines with NH4F to form ammonium hexafluorosilicate, substrate at about 22 to 28 C during formation, heating to about 80 to 500 C and about 100 C for sublimation, oxide-to-nitride selectivity greater than about 7:1, about 6:1 to 150:1, about 8:1 to 12:1, second-cycle examples about 14:1 and about 30:1 to 150:1, ammonia to NF3, HF or F2 molar ratio about 4.5:1, a stated halogen-to-hydrogen gas ratio of about 3:1 to 10:1, remote-plasma ammonia with unactivated NF3 about 3:1 (fetch-tool summary); J. Tang, N. Ingle and D. Yang (Applied Materials), ‘Smooth SiConi etch for silicon-containing films’, US Patent 8,501,629 (2013): NF3 and NH3 plasma effluents thought to include NH4F and NH4F.HF, reacting with silicon oxide to form solid ammonium hexafluorosilicate with ammonia and water vapour, sublimation into SiF4, NH3 and HF, substrate below about 100 C during etch with 30 C in examples, sublimation near 100 C (ranges above 75 C to 115 to 150 C), conventional SiConi about 2:1 hydrogen-to-fluorine atomic ratio, smooth process below 1:1 or above 5:1 with 350 sccm NH3 and 35 sccm NF3 (about 10:1), 1 to 10 Torr, 20 to 100 W at 50 to 90 kHz, conformal and oxide-selective with little etching of amorphous, crystalline or polycrystalline silicon, multi-step etch with about 25 to 200 angstrom per step and anneals under about 30 s, no numeric selectivity given (fetch-tool summary).

Ques: What NH3 to NF3 ratio is used, and which value is more likely correct?

An ammonia-rich ratio, between about 2:1 and 10:1. Applied Materials gives about 2:1 hydrogen-to-fluorine for conventional SiConi and uses 350 $\mathrm{sccm}$ NH${}_3$ with 35 $\mathrm{sccm}$ NF${}_3$, or 10:1, for its smooth process. With three H per NH${}_3$ and three F per NF${}_3$, the atomic ratio equals the flow ratio (computed). ASM gives an ammonia-to-fluorine-gas molar ratio of about 4.5:1, but elsewhere states a halogen-to-hydrogen gas ratio of 3:1 to 10:1. These disagree in direction. The ammonia-rich reading is more likely correct: it matches ASM’s own 4.5:1 example and both Applied Materials values, and NH${}_4$F formation needs excess ammonia (my reading). The reversed phrase is probably a drafting slip. Reference: J. Tang, N. Ingle and D. Yang (Applied Materials), ‘Smooth SiConi etch for silicon-containing films’, US Patent 8,501,629 (2013): NF3 and NH3 plasma effluents thought to include NH4F and NH4F.HF, reacting with silicon oxide to form solid ammonium hexafluorosilicate with ammonia and water vapour, sublimation into SiF4, NH3 and HF, substrate below about 100 C during etch with 30 C in examples, sublimation near 100 C (ranges above 75 C to 115 to 150 C), conventional SiConi about 2:1 hydrogen-to-fluorine atomic ratio, smooth process below 1:1 or above 5:1 with 350 sccm NH3 and 35 sccm NF3 (about 10:1), 1 to 10 Torr, 20 to 100 W at 50 to 90 kHz, conformal and oxide-selective with little etching of amorphous, crystalline or polycrystalline silicon, multi-step etch with about 25 to 200 angstrom per step and anneals under about 30 s, no numeric selectivity given (fetch-tool summary); J. Tolle, M. G. Goodman, R. M. Vyne and E. R. Hill (ASM IP Holding), ‘Plasma pre-clean module and process’, US Patent 9,299,557 (2016): pre-clean material of ammonium fluoride NH4F and ammonium hydrogen fluoride NH4HF2 formed from ammonia and NF3, HF or F2, reacting with silicon oxide to give water vapour and SiF4 that combines with NH4F to form ammonium hexafluorosilicate, substrate at about 22 to 28 C during formation, heating to about 80 to 500 C and about 100 C for sublimation, oxide-to-nitride selectivity greater than about 7:1, about 6:1 to 150:1, about 8:1 to 12:1, second-cycle examples about 14:1 and about 30:1 to 150:1, ammonia to NF3, HF or F2 molar ratio about 4.5:1, a stated halogen-to-hydrogen gas ratio of about 3:1 to 10:1, remote-plasma ammonia with unactivated NF3 about 3:1 (fetch-tool summary).

Ques: Why does vapour HF spare aluminium while wet HF attacks it?

Because Al is etched only by water-dissociated HF. Gablech et al. explain that HF attacks Al only when water splits it into $H_3{O}^{+}$ and $F^{-}$, and the etch reaction $Si{O}_2+4HF\to Si{F}_4+2H_{2}O$ produces that water at the surface. KLA notes that wet HF corrodes exposed metals, most notably aluminium, while its reduced-pressure HF and alcohol process keeps by-products in the gas phase and is compatible with unprotected Al mirrors and bond pads. Gablech et al. reach the same result in liquid by adding oleum, $H_{2}O+S{O}_3\to H_{2}S{O}_4$, to 48 per cent HF: thermal oxide etched at 0.93 $\mu$m/min at about 70 ${}^{\circ }\mathrm{C}$ while Al sheet resistance rose only about 7.6 per cent. Their wet route still needs critical-point drying. Reference: I. Gablech, J. Brodsky, J. Pekarek and P. Neuzil, ‘Infinite selectivity of wet SiO2 etching in respect to Al’, Micromachines 11, 365 (2020), read from the PDF supplied by the user: HF etches Al only when dissociated by water into H3O+ and F-, so water must be absent; SiO2 + 4HF -> SiF4 + 2H2O produces that water at the surface; anhydrous HF gas with alcohol vapour gives a slow rate of about 15 nm/min on thermal oxide in complex, expensive equipment (their ref. 20); 73 per cent HF gives 1.6 um/min without attacking Al but is a controlled chemical; 48 per cent HF with 20 per cent oleum binds water by H2O + SO3 -> H2SO4, etching thermal oxide at (0.93 plus or minus 0.05) um/min at about 70 C, about 14 times faster than 6:1 BOE at about 70 nm/min; Al sheet resistance up by about 7.6 per cent after about 9 min; dry vapour etch needs no critical-point drying while the wet method does; BOE etches Al and PECVD protection layers have pinholes; SPTS Technologies (KLA), ‘Introduction to HF Vapor Etch’, application brief HF-Intro-Q1/23 (2023), read from the PDF supplied by the user: SiO2 + 4HF -> SiF4 + 2H2O; wet HF and rinse liquids cause stiction on drying and corrode exposed metals, most notably aluminium; vapour penetrates small features and allows longer undercuts; alcohol A ionises HF and acts as catalyst, SiO2 + 2HF2- + 2AH+ -> SiF4 + 2H2O + 2A; water by-product also a catalyst and must be controlled and removed; Primaxx HF/alcohol process at reduced pressure and elevated temperature, 75 to 150 torr, residue-free, vertical and lateral oxide rates 0.01 to 1 um/min; device yields typically about 100 per cent; compatible with unprotected Al; reduced pressure keeps by-products in the gas phase for high selectivity to metals and avoids localized loading; material table: thermal oxide, TEOS, SOI bonded oxide and quartz green as sacrificial oxide, PECVD and spin-on oxide yellow; alumina, ALD alumina, aluminium and silicon carbide green as protective layer, Si-rich LPCVD nitride yellow; aluminium, gold, TiW and nickel green and copper yellow as metal/electrode/adhesion; Monarch300 13-wafer batch, Monarch25 25-wafer, Monarch3 3-wafer and uEtch single-wafer systems.

Ques: Is silicon nitride compatible with vapour HF?

Only conditionally, and the sources differ in emphasis. KLA’s compatibility table marks Si-rich LPCVD nitride yellow as a protective layer, while alumina, ALD alumina, aluminium and silicon carbide are green; among metals, Al, gold, TiW and nickel are green and copper yellow. EPFL lists silicon nitride, with exceptions, as incompatible with its uEtch, alongside Ti, TiN, Ta, TaN, Ge, polymers and several glasses. The brief gives no legend, so reading yellow as conditional is my reading. Both sources therefore agree that stoichiometric nitride is not a safe etch stop, and the more likely position is that only silicon-rich nitride survives, which explains the exceptions EPFL mentions (my reading). For a robust etch stop, alumina or silicon carbide is the safer choice (KLA). Reference: SPTS Technologies (KLA), ‘Introduction to HF Vapor Etch’, application brief HF-Intro-Q1/23 (2023), read from the PDF supplied by the user: SiO2 + 4HF -> SiF4 + 2H2O; wet HF and rinse liquids cause stiction on drying and corrode exposed metals, most notably aluminium; vapour penetrates small features and allows longer undercuts; alcohol A ionises HF and acts as catalyst, SiO2 + 2HF2- + 2AH+ -> SiF4 + 2H2O + 2A; water by-product also a catalyst and must be controlled and removed; Primaxx HF/alcohol process at reduced pressure and elevated temperature, 75 to 150 torr, residue-free, vertical and lateral oxide rates 0.01 to 1 um/min; device yields typically about 100 per cent; compatible with unprotected Al; reduced pressure keeps by-products in the gas phase for high selectivity to metals and avoids localized loading; material table: thermal oxide, TEOS, SOI bonded oxide and quartz green as sacrificial oxide, PECVD and spin-on oxide yellow; alumina, ALD alumina, aluminium and silicon carbide green as protective layer, Si-rich LPCVD nitride yellow; aluminium, gold, TiW and nickel green and copper yellow as metal/electrode/adhesion; Monarch300 13-wafer batch, Monarch25 25-wafer, Monarch3 3-wafer and uEtch single-wafer systems; EPFL Center of MicroNanoTechnology, ‘SPTS uEtch’ equipment page: isotropic vapour-phase release of sacrificial SiO2 without liquids or supercritical drying, anhydrous HF with ethanol vapour carried by N2, products SiF4 gas and water vapour, ethanol a catalyst not consumed and SiO2 not etched without it, ethanol keeping the surface dry, water vapour able to cause stiction, total pressure fixed at 125 Torr with five recipes of different HF partial pressure, minimum thermal-oxide rates 12, 38, 125, 153 and 175 nm/min, deposited oxides faster, rate falling with exposed oxide area (macro-loading), 200 to 250 C pre-bake, incompatible materials including polymers, glasses, Ge, Ti, TiN, Ta, TaN and silicon nitride with exceptions, minimum step 15 min (fetch-tool summary).

Ques: When is a dry clean chosen instead of the wet HF dip?

When liquid or drying would cause harm. The RCA Wet Bench page covers the HF dip that strips native oxide in a wet sequence, and the Spin Rinse Dryer and IPA Vapor Dryer page shows that drying can collapse tall structures and leave watermarks. Vapour HF avoids liquids entirely for MEMS release (EPFL; Polifab), and spares unprotected aluminium (KLA), though EPFL lists polymers, several glasses, germanium, titanium, tantalum compounds and silicon nitride as incompatible with its tool. The NH${}_3$/NF${}_3$ plasma clean removes oxide selectively to nitride and silicon at near-room temperature and finishes with a short anneal (ASM patent; Applied Materials patent). In my reading it suits pre-epitaxy and pre-contact cleans, since it can run in a vacuum module next to deposition. Reference: RCA Wet Bench page of this vault; Spin Rinse Dryer and IPA Vapor Dryer page of this vault; EPFL Center of MicroNanoTechnology, ‘SPTS uEtch’ equipment page: isotropic vapour-phase release of sacrificial SiO2 without liquids or supercritical drying, anhydrous HF with ethanol vapour carried by N2, products SiF4 gas and water vapour, ethanol a catalyst not consumed and SiO2 not etched without it, ethanol keeping the surface dry, water vapour able to cause stiction, total pressure fixed at 125 Torr with five recipes of different HF partial pressure, minimum thermal-oxide rates 12, 38, 125, 153 and 175 nm/min, deposited oxides faster, rate falling with exposed oxide area (macro-loading), 200 to 250 C pre-bake, incompatible materials including polymers, glasses, Ge, Ti, TiN, Ta, TaN and silicon nitride with exceptions, minimum step 15 min (fetch-tool summary); Polifab, Politecnico di Milano, ‘HF vapor release (Idonus)’ equipment page: only vapour contacts the substrate because liquid HF would destroy tiny structures or make them stick by cohesion forces, simplified reaction SiO2 + 4HF -> SiF4 + 2H2O through silanol intermediates with water affecting the rate, etching spontaneous at room temperature, controlled substrate heating 35 to 60 C setting the rate, typical rate 4 to 10 um/h, MEMS release on SOI, up to 100 mm wafers, 100 to 160 mL etchant (fetch-tool summary); SPTS Technologies (KLA), ‘Introduction to HF Vapor Etch’, application brief HF-Intro-Q1/23 (2023), read from the PDF supplied by the user: SiO2 + 4HF -> SiF4 + 2H2O; wet HF and rinse liquids cause stiction on drying and corrode exposed metals, most notably aluminium; vapour penetrates small features and allows longer undercuts; alcohol A ionises HF and acts as catalyst, SiO2 + 2HF2- + 2AH+ -> SiF4 + 2H2O + 2A; water by-product also a catalyst and must be controlled and removed; Primaxx HF/alcohol process at reduced pressure and elevated temperature, 75 to 150 torr, residue-free, vertical and lateral oxide rates 0.01 to 1 um/min; device yields typically about 100 per cent; compatible with unprotected Al; reduced pressure keeps by-products in the gas phase for high selectivity to metals and avoids localized loading; material table: thermal oxide, TEOS, SOI bonded oxide and quartz green as sacrificial oxide, PECVD and spin-on oxide yellow; alumina, ALD alumina, aluminium and silicon carbide green as protective layer, Si-rich LPCVD nitride yellow; aluminium, gold, TiW and nickel green and copper yellow as metal/electrode/adhesion; Monarch300 13-wafer batch, Monarch25 25-wafer, Monarch3 3-wafer and uEtch single-wafer systems; J. Tolle, M. G. Goodman, R. M. Vyne and E. R. Hill (ASM IP Holding), ‘Plasma pre-clean module and process’, US Patent 9,299,557 (2016): pre-clean material of ammonium fluoride NH4F and ammonium hydrogen fluoride NH4HF2 formed from ammonia and NF3, HF or F2, reacting with silicon oxide to give water vapour and SiF4 that combines with NH4F to form ammonium hexafluorosilicate, substrate at about 22 to 28 C during formation, heating to about 80 to 500 C and about 100 C for sublimation, oxide-to-nitride selectivity greater than about 7:1, about 6:1 to 150:1, about 8:1 to 12:1, second-cycle examples about 14:1 and about 30:1 to 150:1, ammonia to NF3, HF or F2 molar ratio about 4.5:1, a stated halogen-to-hydrogen gas ratio of about 3:1 to 10:1, remote-plasma ammonia with unactivated NF3 about 3:1 (fetch-tool summary); J. Tang, N. Ingle and D. Yang (Applied Materials), ‘Smooth SiConi etch for silicon-containing films’, US Patent 8,501,629 (2013): NF3 and NH3 plasma effluents thought to include NH4F and NH4F.HF, reacting with silicon oxide to form solid ammonium hexafluorosilicate with ammonia and water vapour, sublimation into SiF4, NH3 and HF, substrate below about 100 C during etch with 30 C in examples, sublimation near 100 C (ranges above 75 C to 115 to 150 C), conventional SiConi about 2:1 hydrogen-to-fluorine atomic ratio, smooth process below 1:1 or above 5:1 with 350 sccm NH3 and 35 sccm NF3 (about 10:1), 1 to 10 Torr, 20 to 100 W at 50 to 90 kHz, conformal and oxide-selective with little etching of amorphous, crystalline or polycrystalline silicon, multi-step etch with about 25 to 200 angstrom per step and anneals under about 30 s, no numeric selectivity given (fetch-tool summary).

References

1. SPTS Technologies (KLA), “Introduction to HF vapor etch,” application brief HF-Intro-Q1/23, 2023. Available: https://www.kla.com/documents/HF-Intro-US-Feb2023-KLA.pdf
2. I. Gablech, J. Brodsky, J. Pekarek and P. Neuzil, “Infinite selectivity of wet SiO2 etching in respect to Al,” Micromachines, vol. 11, p. 365, 2020, doi:10.3390/mi11040365. Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC7230285
3. EPFL Center of MicroNanoTechnology, “SPTS uEtch” equipment page. Available: https://www.epfl.ch/research/facilities/cmi/equipment/etching/spts-uetch/
4. Polifab, Politecnico di Milano, “HF vapor release (Idonus)” equipment page. Available: https://www.polifab.polimi.it/equipment/hf-vapor-release-idonus/
5. J. Tolle, M. G. Goodman, R. M. Vyne and E. R. Hill, “Plasma pre-clean module and process,” US Patent 9,299,557, ASM IP Holding, 2016. Available: https://uspto.report/patent/grant/9299557
6. J. Tang, N. Ingle and D. Yang, “Smooth SiConi etch for silicon-containing films,” US Patent 8,501,629, Applied Materials, 2013. Available: https://patents.google.com/patent/US8501629
7. RCA Wet Bench page of this vault.
8. Spin Rinse Dryer and IPA Vapor Dryer page of this vault.

Related pages in this vault: RCA Wet Bench, Spin Rinse Dryer and IPA Vapor Dryer, Megasonic Single-Wafer Cleaner, Defect Sources and Classification, Wet Etch Bench.