1. Theory

Definition: An ALD reactor grows films by alternating self-limiting surface reactions: a precursor pulse, a purge, a co-reactant pulse and a purge make one cycle (KTH; Stanford).
Self-limitation: TMA reacts with surface OH groups until they are used up, and water then restores OH while releasing methane (Stanford).
Growth per cycle: Typically 0.5 to 1.5 angstrom per cycle, so thickness is set by counting cycles (Stanford; Purdue).
ALD window: Growth is constant over a temperature range bounded by condensation, slow activation, decomposition and desorption (Stanford).
Conformality: Without line of sight, thermal ALD coats aspect ratios above 2000:1 (Stanford).

2. Working Principle

Thermal ALD: Water or ozone as oxidant at 1 to 10 Torr, reactor up to 250 to 300 $^{\circ}$ C (KTH; Purdue; Penn).
Plasma ALD: Ar/O $_{2}$ plasma radicals as oxidant; remote plasma keeps ion energies below about 20 eV (Purdue; KTH; Stanford).
Purge time: Waits lengthen from 3 s at 300 $^{\circ}$ C to 20 to 60 s at 100 to 150 $^{\circ}$ C (Penn).
Precursors: TMA, TDMAH or HFDMA, TDMAZ, TDMAT and aminosilanes (Purdue; Penn).

3. Applications

High-k gate dielectrics: HfO $_{2}$ and ZrO $_{2}$ with k above 14 to 15 (KTH; Purdue).
Conformal liners and barriers: Deep trenches and high-aspect-ratio structures (Stanford).
Low-temperature films: Al $_{2}$ O $_{3}$ from 25 $^{\circ}$ C upward and SiO $_{2}$ from 80 $^{\circ}$ C (KTH; Penn).
Thickness metrology: Ellipsometry and X-ray reflectometry for angstrom-level films (my reading; Spectroscopic Ellipsometer and X-Ray Reflectometer (XRR) pages).

Cycle Chemistry and Thickness

Name	Formula	Related Formulae	Variables/Constants
TMA half-reaction	$A l (C H_{3})_{3} (g) + : S i - O H (s) \to : S i - O - A l (C H_{3})_{2} (s) + C H_{4}$	$H_{2} O step: A l - C H_{3} \to A l - O H and A l - O - A l, + C H_{4}$	$S t an f or d : o n e cy c l e i s T M A p u l se, p u r g e, w a t er p u l se, p u r g e; K T H : s am e f o u r s t e p s, A l_{2} O_{3} a t 200^{\circ} C g r o w in g l in e a r l y w i t h cy c l es$ $(m y r e a d in g : e a c h ha l f - r e a c t i o n s t o p s o n ce t h e s u r f a ce g r o u p s a r e u se d u p, w hi c h i s t h e se l f - l imi t a t i o n)$
Thickness by counting	$t = N_{cy c} \times GP C$	$GP C = 0.5 to 1.5 \overset{˚}{A} /cycle$	$S t an f or d : t y p i c a l r an g e, A l_{2} O_{3} ab o u t 1 \overset{˚}{A} /cycle; P u r d u e : ab o u t 1 \overset{˚}{A} /cycle in g e n er a l; K T H : f i l m s o f 1 t o 500 nm$ $(co m p u t e d : 10 nm o f A l_{2} O_{3} a t 1.0 \overset{˚}{A} /cycle n ee d s 100 cy c l es)$
Throughput	$R = \frac{GP C}{t _{cy c l e}}$	$t_{cy c l e} \approx 1 s to > 1 min$	$S t an f or d : ab o u t 3 s p er cy c l e in t h e e x am pl e; P e nn : w ai t s o f 3 s a t 300^{\circ} C r i s in g t o 20 s a t 150^{\circ} C f or T M A an d w a t er$ $(co m p u t e d : 1 \overset{˚}{A} /cycle in 3 s i s 2 nm/min; 100 cy c l es t ak e ab o u t 10 min a t 300^{\circ} C an d 67 min a t 150^{\circ} C w i t h 40 s o f w ai t s p er cy c l e)$

ALD Window and Precursors

Name	Formula	Related Formulae	Variables/Constants
ALD window	$GP C (T) \approx const inside the window$	$limits: condensation, activation, decomposition, desorption$	$S t an f or d : g r o w t h s a t u r a t e d in s i d e t h e w in d o w, f o u r l imi t in g r e g im es o u t s i d e i t; P e nn : A l_{2} O_{3} 0.91 t o 1.0 \overset{˚}{A} /cycle f r o m 150 t o 300^{\circ} C$ $(m y r e a d in g : l o w t e m p er a t u r e r i s k s co n d e n s a t i o n or s l o w r e a c t i o n, hi g h t e m p er a t u r e d eco m p os i t i o n or d esor pt i o n)$
Precursor-dependent growth	$GP C_{H f O_{2}} = 1.12 \to 0.98 \overset{˚}{A} /cycle (100 \to 250^{\circ} C)$	$GP C_{T i O_{2}} \approx 0.4 \overset{˚}{A} /cycle, GP C_{A l_{2} O_{3}, O_{3}} = 0.6 (100^{\circ} C) to 0.9 (250^{\circ} C)$	$P e nn : H F D M A + H_{2} O, T D M A T + H_{2} O an d T M A + O_{3} r ec i p es; in d e x 2.092 f or H f O_{2} an d 2.1 t o 2.5 f or T i O_{2}$ $(m y r e a d in g : r i s in g GP C a t l o w t e m p er a t u r e hin t s a t l e f t o v er w a t er, n ee d in g t h e l o n g er 60 s w ai t s P e nn u ses)$
Low-temperature SiO2	$GP C_{S i O_{2}} = 0.5 (80^{\circ} C) \to 1.15 \overset{˚}{A} /cycle (350^{\circ} C)$	$n_{633} = 1.52 (80^{\circ} C) \to 1.45 (\geq 150^{\circ} C)$	$P e nn : T D M A S + O_{3}; K T H : minim u m t e m p er a t u r es, S i O_{2} ab o v e 100^{\circ} C, A l_{2} O_{3} ab o v e 25^{\circ} C, H f O_{2}, T i O_{2} an d Z r O_{2} ab o v e 80^{\circ} C$ $(m y r e a d in g : t h e hi g h in d e x a t 80^{\circ} C s i g na l s in co m pl e t e o x i d a t i o n)$

Conformality, Plasma and High-k

Name	Formula	Related Formulae	Variables/Constants
Conformality	$A R_{t h er ma l} > 2000 : 1$	$A R_{pl a s ma} \approx 20 : 1$	$S t an f or d : n o l in e o f s i g h t, s t o p - v a l v e e x p os u r e f or e x t r e m e a s p ec t r a t i os, r a d i c a l r eco mbina t i o n l imi t in g pl a s ma A L D$ $(m y r e a d in g : r a d i c a l s a r e l os t o n t r e n c h w a l l s, w h er e a s t h er ma l p r ec u r sor s a r e n o t)$
Thermal and plasma pressures	$p_{t h er ma l} = 1 to 10 Torr$	$p_{d i r ec t} = 1 to 10 mTorr (E_{i} > 100 eV), p_{r e m o t e} = 1 to 1000 mTorr (E_{i} < 20 eV)$	$K T H; S t an f or d : s a t u r a t i o n w i t h p r ess u r e ab o u t 100 mTorr ab o v e ba se, r e m o t e I C P a v o i d in g i o n d ama g e; P u r d u e : A r / O_{2} pl a s ma$ $(m y r e a d in g : r e m o t e pl a s ma s k ee p t h e r a d i c a l s b u t d r o p t h e d ama g in g i o n s)$
High-k dielectrics	$k_{A l_{2} O_{3}} = 6 to 9, k_{H f O_{2}} > 15, k_{Z r O_{2}} > 14$	$k_{T i O_{2}} > 20, k_{T a_{2} O_{5}} > 22, k_{S i O_{2}} = 3.9$	$K T H : t ab l e w i t h minim u m t e m p er a t u r es; P u r d u e : T D M A H an d T D M A Z p r ec u r sor s$ $(co m p u t e d : H f O_{2} ab o v e 15 i s m or e t han 3.8 t im es S i O_{2})$

Important Questions

Ques: What makes atomic layer deposition self-limiting?

Each half-reaction consumes a finite set of surface sites. Stanford writes the TMA step as $A l (C H_{3})_{3} + : S i - O H \to : S i - O - A l (C H_{3})_{2} + C H_{4}$ . The water step converts the remaining methyl groups to Al-OH and Al-O-Al bridges, again releasing methane, and purges separate the two pulses. KTH gives the same four-step cycle and shows Al $_{2}$ O $_{3}$ from TMA and water at 200 $^{\circ} C$ growing linearly with cycle number. Stanford’s saturation curves show growth per cycle levelling off once the dose saturates the surface, and it recommends a pressure rise of about 100 $mTorr$ above base. Because each step stops by itself, thickness depends on cycle count, not on flux or time (my reading). Reference: J. Provine and M. Rincon, ‘Atomic Layer Deposition: Introduction to the Theory and Cambridge Nanotech Savannah and Fiji’, Stanford University, NNIN ALD Roadshow (1 November 2012): TMA half-reaction Al(CH3)3 + Si-O-H -> Si-O-Al(CH3)2 + CH4 and water step converting methyls to Al-OH and Al-O-Al with methane release, one cycle of TMA pulse, purge, H2O pulse, purge, ALD window with desorption-, condensation-, activation-energy- and decomposition-limited regimes, typical 0.5 to 1.5 angstrom/cycle, Al2O3 about 1 angstrom/cycle with a linear fit of 1.259 angstrom/cycle, saturation curves of growth against dose, cycles of about 3 s in the example and from about 1 s to over 1 min, pressure about 100 mTorr above base for saturation, conformality without line of sight, aspect ratios above 2000:1 with stop-valve exposure, plasma ALD widening the window, avoiding decomposition, remote ICP avoiding ion damage, faster cycles and smaller nucleation delay but radical recombination limiting it to about 20:1, alcohol replacing water (fetch-tool summary); KTH Royal Institute of Technology, Lecture 8 ‘CVD and ALD’: ALD for 1 to 500 nm, cycle of precursor pulse, purge, co-reactant pulse, purge, TMA + H2O Al2O3 at 200 C growing linearly with cycles, minimum temperatures and dielectric constants Al2O3 above 25 C (k 6 to 9), SiO2 above 100 C (3.9), HfO2 above 80 C (above 15), TiO2 above 80 C (above 20), Ta2O5 above 100 C (above 22), ZrO2 above 80 C (above 14), thermal ALD 1 to 10 torr, direct plasma ALD 1 to 10 mTorr with ions above about 100 eV, remote plasma ALD 1 to 1000 mTorr with ions below about 20 eV (fetch-tool summary).

Ques: How is thickness controlled, and how fast is ALD?

By counting cycles, at the cost of speed. Stanford gives typical growth of 0.5 to 1.5 $\overset{˚}{A} /cycle$ and Purdue about 1 $\overset{˚}{A} /cycle$ in general, so 10 $nm$ of Al $_{2}$ O $_{3}$ takes about 100 cycles (computed). Stanford lists cycle times from about 1 s to over 1 $min$ and gives a 3 s example cycle, which at 1 $\overset{˚}{A} /cycle$ is 2 $nm/min$ (computed). Penn’s TMA and water recipes lengthen the waits from 3 s at 300 $^{\circ} C$ to 20 s at 150 $^{\circ} C$ . With two waits per cycle, 100 cycles take about 10 $min$ at 300 $^{\circ} C$ but 67 $min$ at 150 $^{\circ} C$ (computed). KTH places ALD at 1 to 500 $nm$ . The technique suits thin films, not thick ones (my reading). Reference: J. Provine and M. Rincon, ‘Atomic Layer Deposition: Introduction to the Theory and Cambridge Nanotech Savannah and Fiji’, Stanford University, NNIN ALD Roadshow (1 November 2012): TMA half-reaction Al(CH3)3 + Si-O-H -> Si-O-Al(CH3)2 + CH4 and water step converting methyls to Al-OH and Al-O-Al with methane release, one cycle of TMA pulse, purge, H2O pulse, purge, ALD window with desorption-, condensation-, activation-energy- and decomposition-limited regimes, typical 0.5 to 1.5 angstrom/cycle, Al2O3 about 1 angstrom/cycle with a linear fit of 1.259 angstrom/cycle, saturation curves of growth against dose, cycles of about 3 s in the example and from about 1 s to over 1 min, pressure about 100 mTorr above base for saturation, conformality without line of sight, aspect ratios above 2000:1 with stop-valve exposure, plasma ALD widening the window, avoiding decomposition, remote ICP avoiding ion damage, faster cycles and smaller nucleation delay but radical recombination limiting it to about 20:1, alcohol replacing water (fetch-tool summary); Purdue Birck Nanotechnology Center, ‘Veeco Fiji G3 ALD’ wiki page: thermal ALD with water and plasma ALD with Ar/O2 plasma radicals, reactor up to 250 C and chuck to 500 C, valves at 150 C, standard Al2O3 and HfO2 recipes at 110 to 250 C, precursors TMA, TDMAH, TDMAZ and bis(diethylamino)silane, about 1 angstrom per cycle as the general principle, pieces up to 200 mm wafers, self-limiting reactions giving uniformity and conformality (fetch-tool summary); University of Pennsylvania Singh Center Nanofab, ‘ALD Recipe’ page (Cambridge Nanotech Savannah recipes): TMA + H2O at 300, 250, 200 and 150 C with 0.015 s pulses and waits of 3, 5, 8 and 20 s giving 1.0, 0.94, 1.0 and 0.91 angstrom/cycle, 30 s waits at 80 C, TMA + O3 0.9 (250 C) and 0.6 (100 C) angstrom/cycle, TDMAS + O3 SiO2 0.5, 0.7, 0.8, 0.9, 1.1 and 1.15 angstrom/cycle from 80 to 350 C with index 1.52 at 80 C falling to 1.45 to 1.44 above 120 C, TDMAT + H2O TiO2 about 0.4 angstrom/cycle with index 2.1 to 2.5, HfO2 from HFDMA + H2O 0.98, 1.01, 1.04 and 1.12 angstrom/cycle at 250, 200, 150 and 100 C with waits of 5, 10, 20 and 60 s and index 2.092 at 632 nm for 1000 angstrom at 200 C, dose and exposure still to be optimised (fetch-tool summary); KTH Royal Institute of Technology, Lecture 8 ‘CVD and ALD’: ALD for 1 to 500 nm, cycle of precursor pulse, purge, co-reactant pulse, purge, TMA + H2O Al2O3 at 200 C growing linearly with cycles, minimum temperatures and dielectric constants Al2O3 above 25 C (k 6 to 9), SiO2 above 100 C (3.9), HfO2 above 80 C (above 15), TiO2 above 80 C (above 20), Ta2O5 above 100 C (above 22), ZrO2 above 80 C (above 14), thermal ALD 1 to 10 torr, direct plasma ALD 1 to 10 mTorr with ions above about 100 eV, remote plasma ALD 1 to 1000 mTorr with ions below about 20 eV (fetch-tool summary).

Ques: What is the growth per cycle of TMA/water Al2O3, and which value is more likely?

About 1.0 $\overset{˚}{A} /cycle$ is more likely for a saturated process. Penn’s Savannah recipes give 1.0, 0.94, 1.0 and 0.91 $\overset{˚}{A} /cycle$ at 300, 250, 200 and 150 $^{\circ} C$ , and Purdue and Stanford both state about 1 $\overset{˚}{A} /cycle$ . Stanford’s plotted linear fit, however, gives 1.259 $\overset{˚}{A} /cycle$ . Three statements and four measured recipes cluster at 0.9 to 1.0, while the 1.26 figure comes from one fitted plot without stated conditions. A fit can include a nucleation offset or extra growth from incomplete purging (my reading). The 1.0 $\overset{˚}{A} /cycle$ value is therefore judged more likely. Penn’s ozone process differs, 0.6 to 0.9 $\overset{˚}{A} /cycle$ , so the oxidant must always be stated with a GPC (my reading). Reference: University of Pennsylvania Singh Center Nanofab, ‘ALD Recipe’ page (Cambridge Nanotech Savannah recipes): TMA + H2O at 300, 250, 200 and 150 C with 0.015 s pulses and waits of 3, 5, 8 and 20 s giving 1.0, 0.94, 1.0 and 0.91 angstrom/cycle, 30 s waits at 80 C, TMA + O3 0.9 (250 C) and 0.6 (100 C) angstrom/cycle, TDMAS + O3 SiO2 0.5, 0.7, 0.8, 0.9, 1.1 and 1.15 angstrom/cycle from 80 to 350 C with index 1.52 at 80 C falling to 1.45 to 1.44 above 120 C, TDMAT + H2O TiO2 about 0.4 angstrom/cycle with index 2.1 to 2.5, HfO2 from HFDMA + H2O 0.98, 1.01, 1.04 and 1.12 angstrom/cycle at 250, 200, 150 and 100 C with waits of 5, 10, 20 and 60 s and index 2.092 at 632 nm for 1000 angstrom at 200 C, dose and exposure still to be optimised (fetch-tool summary); Purdue Birck Nanotechnology Center, ‘Veeco Fiji G3 ALD’ wiki page: thermal ALD with water and plasma ALD with Ar/O2 plasma radicals, reactor up to 250 C and chuck to 500 C, valves at 150 C, standard Al2O3 and HfO2 recipes at 110 to 250 C, precursors TMA, TDMAH, TDMAZ and bis(diethylamino)silane, about 1 angstrom per cycle as the general principle, pieces up to 200 mm wafers, self-limiting reactions giving uniformity and conformality (fetch-tool summary); J. Provine and M. Rincon, ‘Atomic Layer Deposition: Introduction to the Theory and Cambridge Nanotech Savannah and Fiji’, Stanford University, NNIN ALD Roadshow (1 November 2012): TMA half-reaction Al(CH3)3 + Si-O-H -> Si-O-Al(CH3)2 + CH4 and water step converting methyls to Al-OH and Al-O-Al with methane release, one cycle of TMA pulse, purge, H2O pulse, purge, ALD window with desorption-, condensation-, activation-energy- and decomposition-limited regimes, typical 0.5 to 1.5 angstrom/cycle, Al2O3 about 1 angstrom/cycle with a linear fit of 1.259 angstrom/cycle, saturation curves of growth against dose, cycles of about 3 s in the example and from about 1 s to over 1 min, pressure about 100 mTorr above base for saturation, conformality without line of sight, aspect ratios above 2000:1 with stop-valve exposure, plasma ALD widening the window, avoiding decomposition, remote ICP avoiding ion damage, faster cycles and smaller nucleation delay but radical recombination limiting it to about 20:1, alcohol replacing water (fetch-tool summary).

Ques: What is the ALD window, and what happens outside it?

It is the temperature range where growth per cycle stays constant. Stanford describes each process as having an ideal window where growth saturates at a monolayer, bounded by condensation-limited, activation-energy-limited, decomposition-limited and desorption-limited regimes. In my reading, the first two lie below the window and the last two above it. Penn’s data show the window in practice: TMA and water stay at 0.91 to 1.0 $\overset{˚}{A} /cycle$ from 150 to 300 $^{\circ} C$ . HfO $_{2}$ from HFDMA rises from 0.98 to 1.12 $\overset{˚}{A} /cycle$ as the temperature falls from 250 to 100 $^{\circ} C$ , which suggests leftover water at low temperature (my reading). Penn’s TDMAS SiO $_{2}$ index of 1.52 at 80 $^{\circ} C$ , against 1.45 above 120 $^{\circ} C$ , signals incomplete oxidation (my reading). Reference: J. Provine and M. Rincon, ‘Atomic Layer Deposition: Introduction to the Theory and Cambridge Nanotech Savannah and Fiji’, Stanford University, NNIN ALD Roadshow (1 November 2012): TMA half-reaction Al(CH3)3 + Si-O-H -> Si-O-Al(CH3)2 + CH4 and water step converting methyls to Al-OH and Al-O-Al with methane release, one cycle of TMA pulse, purge, H2O pulse, purge, ALD window with desorption-, condensation-, activation-energy- and decomposition-limited regimes, typical 0.5 to 1.5 angstrom/cycle, Al2O3 about 1 angstrom/cycle with a linear fit of 1.259 angstrom/cycle, saturation curves of growth against dose, cycles of about 3 s in the example and from about 1 s to over 1 min, pressure about 100 mTorr above base for saturation, conformality without line of sight, aspect ratios above 2000:1 with stop-valve exposure, plasma ALD widening the window, avoiding decomposition, remote ICP avoiding ion damage, faster cycles and smaller nucleation delay but radical recombination limiting it to about 20:1, alcohol replacing water (fetch-tool summary); University of Pennsylvania Singh Center Nanofab, ‘ALD Recipe’ page (Cambridge Nanotech Savannah recipes): TMA + H2O at 300, 250, 200 and 150 C with 0.015 s pulses and waits of 3, 5, 8 and 20 s giving 1.0, 0.94, 1.0 and 0.91 angstrom/cycle, 30 s waits at 80 C, TMA + O3 0.9 (250 C) and 0.6 (100 C) angstrom/cycle, TDMAS + O3 SiO2 0.5, 0.7, 0.8, 0.9, 1.1 and 1.15 angstrom/cycle from 80 to 350 C with index 1.52 at 80 C falling to 1.45 to 1.44 above 120 C, TDMAT + H2O TiO2 about 0.4 angstrom/cycle with index 2.1 to 2.5, HfO2 from HFDMA + H2O 0.98, 1.01, 1.04 and 1.12 angstrom/cycle at 250, 200, 150 and 100 C with waits of 5, 10, 20 and 60 s and index 2.092 at 632 nm for 1000 angstrom at 200 C, dose and exposure still to be optimised (fetch-tool summary).

Ques: When is plasma ALD used instead of thermal ALD?

When lower temperature or faster nucleation matters more than conformality. Stanford lists the advantages of plasma-enhanced ALD: a wider window from lower activation energy, no precursor decomposition on thermally limited substrates, faster cycles, fewer contaminants and a smaller nucleation delay, with O $_{2}$ plasma removing the nucleation delay for Pt. Its limitation is that radical recombination prevents aspect ratios above about 20:1, against more than 2000:1 for thermal ALD. KTH gives the pressures: thermal ALD at 1 to 10 $Torr$ , direct plasma at 1 to 10 $mTorr$ with ions above about 100 $eV$ , and remote plasma at 1 to 1000 $mTorr$ with ions below about 20 $eV$ . Purdue’s Fiji uses Ar/O $_{2}$ plasma radicals, the same radical-driven idea as the PECVD Reactor page but in a self-limiting cycle (my reading). Reference: J. Provine and M. Rincon, ‘Atomic Layer Deposition: Introduction to the Theory and Cambridge Nanotech Savannah and Fiji’, Stanford University, NNIN ALD Roadshow (1 November 2012): TMA half-reaction Al(CH3)3 + Si-O-H -> Si-O-Al(CH3)2 + CH4 and water step converting methyls to Al-OH and Al-O-Al with methane release, one cycle of TMA pulse, purge, H2O pulse, purge, ALD window with desorption-, condensation-, activation-energy- and decomposition-limited regimes, typical 0.5 to 1.5 angstrom/cycle, Al2O3 about 1 angstrom/cycle with a linear fit of 1.259 angstrom/cycle, saturation curves of growth against dose, cycles of about 3 s in the example and from about 1 s to over 1 min, pressure about 100 mTorr above base for saturation, conformality without line of sight, aspect ratios above 2000:1 with stop-valve exposure, plasma ALD widening the window, avoiding decomposition, remote ICP avoiding ion damage, faster cycles and smaller nucleation delay but radical recombination limiting it to about 20:1, alcohol replacing water (fetch-tool summary); KTH Royal Institute of Technology, Lecture 8 ‘CVD and ALD’: ALD for 1 to 500 nm, cycle of precursor pulse, purge, co-reactant pulse, purge, TMA + H2O Al2O3 at 200 C growing linearly with cycles, minimum temperatures and dielectric constants Al2O3 above 25 C (k 6 to 9), SiO2 above 100 C (3.9), HfO2 above 80 C (above 15), TiO2 above 80 C (above 20), Ta2O5 above 100 C (above 22), ZrO2 above 80 C (above 14), thermal ALD 1 to 10 torr, direct plasma ALD 1 to 10 mTorr with ions above about 100 eV, remote plasma ALD 1 to 1000 mTorr with ions below about 20 eV (fetch-tool summary); Purdue Birck Nanotechnology Center, ‘Veeco Fiji G3 ALD’ wiki page: thermal ALD with water and plasma ALD with Ar/O2 plasma radicals, reactor up to 250 C and chuck to 500 C, valves at 150 C, standard Al2O3 and HfO2 recipes at 110 to 250 C, precursors TMA, TDMAH, TDMAZ and bis(diethylamino)silane, about 1 angstrom per cycle as the general principle, pieces up to 200 mm wafers, self-limiting reactions giving uniformity and conformality (fetch-tool summary); PECVD Reactor page of this vault.

Ques: Which films does ALD provide for advanced devices?

Mainly high-k oxides and conformal layers. KTH lists minimum deposition temperatures and dielectric constants: Al $_{2}$ O $_{3}$ above 25 $^{\circ} C$ with k of 6 to 9, SiO $_{2}$ above 100 $^{\circ} C$ with 3.9, HfO $_{2}$ above 80 $^{\circ} C$ with k above 15, ZrO $_{2}$ above 80 $^{\circ} C$ with k above 14, TiO $_{2}$ above 80 $^{\circ} C$ with k above 20, and Ta $_{2}$ O $_{5}$ above 100 $^{\circ} C$ with k above 22. HfO $_{2}$ thus gives more than 3.8 times the permittivity of SiO $_{2}$ (computed). Purdue lists TDMAH and TDMAZ for HfO $_{2}$ and ZrO $_{2}$ , and Penn measures HfO $_{2}$ at an index of 2.092. Films this thin are measured on the Spectroscopic Ellipsometer and X-Ray Reflectometer (XRR) pages (my reading). Reference: KTH Royal Institute of Technology, Lecture 8 ‘CVD and ALD’: ALD for 1 to 500 nm, cycle of precursor pulse, purge, co-reactant pulse, purge, TMA + H2O Al2O3 at 200 C growing linearly with cycles, minimum temperatures and dielectric constants Al2O3 above 25 C (k 6 to 9), SiO2 above 100 C (3.9), HfO2 above 80 C (above 15), TiO2 above 80 C (above 20), Ta2O5 above 100 C (above 22), ZrO2 above 80 C (above 14), thermal ALD 1 to 10 torr, direct plasma ALD 1 to 10 mTorr with ions above about 100 eV, remote plasma ALD 1 to 1000 mTorr with ions below about 20 eV (fetch-tool summary); Purdue Birck Nanotechnology Center, ‘Veeco Fiji G3 ALD’ wiki page: thermal ALD with water and plasma ALD with Ar/O2 plasma radicals, reactor up to 250 C and chuck to 500 C, valves at 150 C, standard Al2O3 and HfO2 recipes at 110 to 250 C, precursors TMA, TDMAH, TDMAZ and bis(diethylamino)silane, about 1 angstrom per cycle as the general principle, pieces up to 200 mm wafers, self-limiting reactions giving uniformity and conformality (fetch-tool summary); University of Pennsylvania Singh Center Nanofab, ‘ALD Recipe’ page (Cambridge Nanotech Savannah recipes): TMA + H2O at 300, 250, 200 and 150 C with 0.015 s pulses and waits of 3, 5, 8 and 20 s giving 1.0, 0.94, 1.0 and 0.91 angstrom/cycle, 30 s waits at 80 C, TMA + O3 0.9 (250 C) and 0.6 (100 C) angstrom/cycle, TDMAS + O3 SiO2 0.5, 0.7, 0.8, 0.9, 1.1 and 1.15 angstrom/cycle from 80 to 350 C with index 1.52 at 80 C falling to 1.45 to 1.44 above 120 C, TDMAT + H2O TiO2 about 0.4 angstrom/cycle with index 2.1 to 2.5, HfO2 from HFDMA + H2O 0.98, 1.01, 1.04 and 1.12 angstrom/cycle at 250, 200, 150 and 100 C with waits of 5, 10, 20 and 60 s and index 2.092 at 632 nm for 1000 angstrom at 200 C, dose and exposure still to be optimised (fetch-tool summary); Spectroscopic Ellipsometer page of this vault; X-Ray Reflectometer (XRR) page of this vault.

References

J. Provine and M. Rincon, “Atomic layer deposition: introduction to the theory and Cambridge Nanotech Savannah and Fiji,” Stanford University, NNIN ALD Roadshow, 1 November 2012. Available: https://snfguide.stanford.edu/files/sections/diplayfiles/ald_tutorial_2012-11-01_su_intro.pdf
University of Pennsylvania Singh Center Nanofab, “ALD recipe” (Cambridge Nanotech Savannah). Available: https://www.seas.upenn.edu/~nanosop/ALD_Recipes.htm
Purdue Birck Nanotechnology Center, “Veeco Fiji G3 ALD,” BNC wiki. Available: https://purdue.atlassian.net/wiki/spaces/BNCWiki/pages/534183944/Veeco+Fiji+G3+ALD
KTH Royal Institute of Technology, “Lecture 8: CVD and ALD,” course slides. Available: https://www.kth.se/social/upload/510bc6e8f276547a4fe3c222/Lecture%208.pdf
Spectroscopic Ellipsometer page of this vault.
X-Ray Reflectometer (XRR) page of this vault.
PECVD Reactor page of this vault.

Quartz 5

Explorer

Atomic Layer Deposition (ALD) Reactor