Knoops Handbook of Crystal Growth 2014 Pre Print

Atomic Layer Deposition
Harm C.M. Knoops, Stephen E. Potts,

Ageeth A. Bol, and W.M.M. (Erwin) Kessels
Eindhoven University of Technology
[email protected]
Photo: Area-selective atomic layer deposition research in

Eindhoven, the Netherlands. (Photo by Bart van Overbeeke)
Please cite this chapter as:
H.C.M. Knoops, S.E. Potts, A.A. Bol, and W.M.M. Kessels, Ch. 27 - Atomic Layer Deposition
(pp. 1101–1134) in Handbook of Crystal Growth, edited by T. Kuech (Elsevier, 2015).
doi: 10.1016/B978-0-444-63304-0.00027-5
FOR TEACHING AND SCHOLARLY USE ONLY - DO NOT DISTRIBUTE

1
Abstract
Atomic layer deposition (ALD), also referred to historically as atomic layer epitaxy (ALE), is a
vapor-phase deposition technique for preparing ultrathin films with precise growth control. ALD
is currently rapidly evolving, mostly driven by the continuous trend to miniaturize electronic
devices. In addition many other innovative technologies are increasingly benefitting from the
high-quality thin films. This chapter describes – on an elementary level – the key features of
ALD. A standard ALD process scheme is used to discuss the relevant concepts of the technique.
Materials that can be deposited by ALD are discussed, including typical precursors and co-
reactants that can be used. Several example chemistries, specifically for ALD of Al2O3, HfO2, TiN,
and Pt, are presented to illustrate the variety in surface chemistry. ALD reactor types are de-
scribed and, finally, some cases are addressed to illustrate the virtues and practicalities of ALD
that are important to advancing present-day and emerging thin film applications.
Contents
Abstract ........................................................................................................................................... 1
1. Thin films and the need for precise growth control................................................................ 2
2. From atomic layer epitaxy to atomic layer deposition ........................................................... 4
3. Basics of ALD ............................................................................................................................ 5
4. Materials, Precursors, and Co-reactants ................................................................................. 8
5. ALD Chemistries ..................................................................................................................... 13
6. ALD reactors .......................................................................................................................... 19
7. ALD virtues and practicalities ................................................................................................ 21
8. Conclusion ............................................................................................................................. 26
References .................................................................................................................................... 27
2
on a workpiece surface, the substrate. This vaporized

1. Thin films and the need for form of the material is obtained from a target material by
purely physical processes, such as thermal or laser-
precise growth control induced evaporation or sputtering by energetic ion bom-
Thin films are ubiquitous in present-day technologies, bardment. The method of CVD, on the other hand, in-
with applications spanning from surface coatings to the volves chemical reactions. These chemical reactions take
most-advanced nanoelectronics. In the field of electron- place by volatile precursor molecules that decompose at
ics especially, thin films are very pervasive with an ever the surface leaving a thin film and volatile by-products.
increasing number of material systems, functionalities The chemical reactions are thermally driven, most fre-
and applications. They are key in the continuing the roll- quently by heating the substrate. The chemical reactions
out of increasingly powerful devices for computation, can also be enhanced by reactive species created in the
data storage, communication, energy scavenging, energy gas phase, e.g., in plasma-enhanced CVD.
storage, and sensing. Many of these applications require
films that are of high (electronic) quality, that are rela- To date, the techniques of PVD and CVD continue to
tively thin (< 1000 nm) and that can be deposited by dominate the field of electronic thin film deposition.
vapor-phase deposition techniques compatible with However the continuous trend in the minimization of the
electronic device manufacturing. Often there is no strict critical device dimensions and the processing of devices
requirement on the microstructure of the films as long as on increasingly larger substrate materials sets stricter
they fulfill the demands on the quality, although in some and new demands on the film deposition techniques,
cases (poly)crystalline or amorphous films really are vital. which become difficult to meet with PVD and CVD. There
Obviously, there are also various requirements in terms is an increasing need for growth control in terms of:
of processing temperature and throughput of the tech-
nique used to deposit the thin films.  Thickness control: The deposition of high-quality
ultrathin films with a thickness control at the subna-
When considering polycrystalline and amorphous films, nometer level;
the use of vapor phase deposition techniques such as
 Uniformity and Conformality: The uniformity of the
physical vapor deposition (PVD) and chemical vapor dep-
films on large substrates, such large wafers, sheet
osition (CVD) has skyrocketed in the past few decades.
materials and foils; excellent conformality for surface
PVD describes a variety of techniques that are based on
features including trenches, pores, surface rough-
the condensation of a vaporized form of the film material
ness, etc.;
Fig. 1. The coverage metrics of a film on a substrate with 3D features. Coverage of the planar surface is evaluat-
ed using the uniformity while the coverage of 3D features is evaluated using the conformality. The growth con-
trol over the film thickness itself is another important metric. The ability to achieve these metrics at low temper-
atures is an additional important aspect.
Published as: H.C.M. Knoops, S.E. Potts, A.A. Bol, and W.M.M. Kessels, Ch. 27 - Atomic Layer Deposition
3
Fig. 2. (a) Cross-section of a field-effect transistor at the 45-nm technology node. The transistor contains a hafni-
um-based high-k dielectric, with a thickness of ~3 nm, as the gate oxide (intel.com). (b) Illustration of film
growth in which the thickness is controlled at the subnanometer level. (c) Intel’s “Tri-gate” FinFET transistors at
the 22-nm technology node. The gates are the taller ridges (going up from left to right), and the fins are the
lower ridges that link the gates (going down from left to right) (intel.com). (d) GaP nanowire coated by a thin
film of Al2O3 deposited by atomic layer deposition. (e) Electron microscope cross-section of a CIGS solar cell,
where the buffer layer (in yellow) has to cover the rough CIGS absorber layer (www.zsw-bw.de). (f) 5-inch Plastic
OLED from LG Display (lgdnewsroom.com).
 Low temperature: The ability to deposit high-quality ter level (Fig. 2b), uniformly over 300 mm silicon wafers,
materials with a high purity and a high density (no is necessary. With the advent of 3D transistors in the 22
voids or pinholes) at low substrate temperatures. nm node and beyond, also an excellent conformality over
These coverage metrics, as well as the need for low sub- the semiconductor fins is required (Fig. 2c). A high con-
strate temperatures, are illustrated in Fig. 1. To exemplify formality is also vital for many other emerging nanoscale
the increasing need for precise growth control, a few devices, e.g., those involving thin films deposited on
very demanding, emerging applications of thin films are nanomaterials such as nanowires (Fig. 2d), nanotubes,
shown in Fig. 2. The first example concerns state-of-the- and nanoparticles. These nanoscale devices have applica-
art field-effect transistors, which are the fundamental tions in electronics, sensing and energy technologies.
building blocks of modern electronic devices. Since the Dense conformal films without pinholes are also required
45-nm technology node of transistors, thermally grown in large area applications such as in photovoltaics. For
SiO2 has become inadequate and high-quality nanome- example, in thin film chalcogenide solar cells based on
ter-thick films of materials such as HfO2 need to be de- copper-indium-gallium-selenide (CIGS), thin pinhole-free
posited (Fig. 2a). A thickness control at the subnanome- buffer layers need to be deposited on the polycrystalline
4
absorber layer, which has a randomly corrugated surface polycrystalline materials one atomic layer at a time.
topology (Fig. 2e). The deposition process needs to take However the method was also used to deposit amor-
place at reasonably low substrate temperatures. The phous dielectric thin films, such as Al2O3, for the same
latter is even more important for flexible electronics application. During the 1980s, the research efforts on ALE
involving polymeric or otherwise organic materials (Fig. strongly increased when there was interest in fabricating
2f). Many of such materials require temperatures lower elemental semiconductors, single crystal III-V and II-VI
than 200 or even 100 °C. Clearly these demands are in- semiconductors, and ordered heterostructures by ALE.
creasingly harder to meet with techniques that are not Organometallic compounds such as metal alkyls were
intrinsically self-limiting (PVD and CVD), are flux- used as precursors. Although good results were reported
controlled (PVD and CVD),* or that take place at (highly) for epitaxial growth, the overall success of ALD for crys-
elevated substrate temperatures (CVD). talline materials was limited due to the complicated
surface chemistry and that ALE did not bring obvious
In this chapter, the method of atomic layer deposition benefits over the traditional molecular beam epitaxy
(ALD) will be introduced with the aim of describing the (MBE) and metal organic vapor phase epitaxy (MOVPE)
basics of this technique, showing that ALD is able to meet methods. As a consequence, the activities in this area of
the abovementioned requirements for precise growth ALE diminished in the mid 1990s. However, in the late
control, as well as describing typical ALD chemistries and 1990s into the early 2000s, interest in the method by the
ALD reactors. Also, some particular cases will be consid- silicon-based microelectronics industry began to in-
ered to show how ALD manifests itself in practice. It is crease.4,6 At that time, the method was also coined atom-
not within the scope of this chapter to give an in-depth ic layer deposition (ALD), a name which was more appro-
and comprehensive overview of the method, the materi- priate considering the fact that the interest in amorphous
als deposited by ALD and the results obtained with re- films strongly increased. This increase was most promi-
spect to its rapidly growing number of applications. Such nent in highly-insulating oxide films to be used as low-
information can be found elsewhere.1–4 In the next sec- leakage high-k dielectrics in complementary metal-oxide-
tion, the relationship between ALD and epitaxy will be semiconductor (CMOS) transistors and dynamic random
briefly treated addressing the historical development of access memory (DRAM) devices. Ruled by the trend of
the technique from atomic layer epitaxy (ALE) to atomic miniaturization in electronic devices, these applications
layer deposition (ALD). have been the main driving force for the development of
ALD processes and equipment. These applications were
furthermore the first main commercial applications of
2. From atomic layer epitaxy to ALD besides the production of electroluminescent film
atomic layer deposition display devices and thin film magnetic heads in hard
The origin of ALD dates back to the 1970s, when the drives. In the last decade, many more potential applica-
technique was developed to deposit polycrystalline II-VI tions of ALD have been identified as the need for precise
thin films, in particular ZnS, for electroluminescent film growth control has become widespread (as has been
display devices.4,5 The first experiments were carried out addressed in the previous section). The library of materi-
with elemental source materials such as zinc and sulfur. als possible by ALD, including metal oxides, nitrides, car-
The method was called atomic layer epitaxy (ALE), as the bides, chalcogenides, metals, etc., has significantly ex-
prime interest was to deposit films of crystalline and panded over the last years, as addressed later in this
chapter (Sec. 4). The interest lies currently mostly in
amorphous and polycrystalline films.1
*
For flux-controlled processes, the growth rate on a
surface is directly related to the flux of source species. Now that the relevance and historical background of ALD
PVD and CVD are examples of this. The disadvantage of have been discussed, the following sections will show the
flux-controlled deposition is that, in many cases, the flux basics and important aspects of ALD.
has to be precisely regulated to achieve controlled, uni-
form and conformal growth.
5
3. Basics of ALD As mentioned before, the various steps in a typical, ideal

In this section, the basics of ALD, as well as several con- ALD cycle are represented in Fig. 3. This figure shows a
cepts that are important in the understanding of ALD, are schematic representation of the self-limiting surface
explained on the basis of an ideal ALD process. Such a reactions during the two half-cycles (in the first half-cycle
process and its inherent properties provide a means for the precursor is dosed; in the second half-cycle the co-
understanding the behavior of real, potentially less-ideal, reactant) as well as the saturation of the surface cover-
ALD processes that are discussed in the later sections. age in every step of the cycle. The saturation of both
half-cycles leads to a characteristic amount of growth per
Standard ALD process scheme cycle, often abbreviated to GPC. It is also illustrative to
In ALD, thin films are built up in cycles in which the sur- express the surface reactions during the half-cycles in
face is exposed to various vapor- or gas-phase species in equations. Considering a binary material AB and precur-
alternating, separated doses. In each cycle, a submono- sors and reactants with both two ligands, the reaction
layer of material is deposited.* As illustrated in Fig. 3, a during the first half-cycle is of the form:
typical cycle consists of four steps: (i) a precursor dosing
s-Y (ads) + AX2 (g)  s-AX (ads) + XY (g) (3.1)
step, where a precursor is typically an inorganic coordina-
tion compound, i.e., a metal center surrounded by chem- where s indicates the surface with surface groups Y, AX2
ical functional groups called ligands; (ii) a purge and/or is the formula for the precursor, with A the first element
pump step; (iii) a co-reactant step, typically involving a to be deposited and X the ligand, and XY is the reaction
small molecule;† and (iv) a purge and/or pump step. The product. The reaction during the second half-cycle is of
precursor, and in many cases also the co-reactant, bring the form:
elements to the surface that lead to film growth. For the
precursor, the element to be deposited is in many cases s-AX (ads) + BY2 (g)  s-ABY (ads) + XY (g) (3.2)
the metal center, while for the reactant, it is typically a
where BY2 is the formula for the co-reactant, with B the
non-metal such as O, N, S, etc. For ALD, it is vital that the
second element to be deposited and Y the ligand, and XY
precursor and co-reactants react with the surface in a
is again the reaction product. Note that in other ALD
self-limiting way. This means that the precursor mole-
processes the number of ligands and surface groups can
cules and co-reactant species react with surface sites
differ, resulting in more diverse reaction products. To
and/or surface chemical groups as long as these are pre-
reach a certain film thickness the two half-cycles (3.1)
sent or accessible; hence the surface reactions eventually
and (3.2) are repeated in an ABAB… fashion.
saturate and stop. The precursor molecules and co-
reactants react neither with themselves nor with the
Temperature window
surface groups that they create. In the purge and/or
The chemical and physical conditions necessary to obtain
pump steps, the gaseous reaction products that may be
self-limiting growth differ for each ALD process and lead
generated during the surface reactions, as well as any
to a characteristic GPC. Furthermore, each process is
excess precursor or co-reactant molecules, are removed
deemed to have a specific temperature window in which
from the ALD reactor. This is necessary to avoid reactions
ALD behavior is obtained. An idealized temperature win-
between precursor and co-reactant molecules directly in
dow is indicated in Fig. 4, where the GPC is plotted as a
the gas phase or on the surface, as this could lead to an
function of temperature. For ideal ALD growth, the tem-
undesired CVD component.
perature window represents the temperature range over
which the GPC shows a weak or no temperature depend-
*
ence, as indicated by the horizontal line in Fig. 4. Outside
Due to steric hindrance and limited density of reactive groups
the temperature window, several chemical and physical
on the surface, the growth per cycle is typically less than a
monolayer. Note that this does not affect the ability to grow a processes can disrupt the ALD behavior. At low tempera-
closed, dense film. tures, some precursors and co-reactants can condense
†
The co-reactant can also be referred to as a precursor, typical- on the surface, leading to an increase in GPC. Alterna-
ly the second precursor.
6
Fig. 3. A schematic representation of the various steps in an ALD cycle consisting of two half-reactions. The exposures in the
first half-cycle (precursor) and second half-cycle (co-reactant) are self-limiting such that the process stops when all available
surface sites are occupied. The two half-cycles are separated by purge steps. The half-cycles are repeated in an ABAB fash-
ion to build the film up to the target thickness. The lower panels show the resulting coverage, or growth per cycle, as a func-
tion of exposure or time for that particular step. For sufficient exposure, saturated growth is obtained, while insufficient
exposure results in incomplete saturation. For insufficient purging, a CVD component from mixing of the precursor and co-
reactant is obtained.
tively, the reactivity of the molecules with the surface ture, leading to a decrease in growth.
sites can be too low because of limited thermal energy at
low temperatures, which prevents saturation of the reac- Uniformity and Conformality
tion and leads to a decrease in growth. At high tempera- Because of the self-limiting nature of ALD, a difference in
tures, the precursors or co-reactants can decompose, the flux of source species, either at different areas on a
leading to a CVD component and an increase in growth. substrate or in a 3D structure, still affords the same film
Furthermore the film itself or the reactive surface groups thickness. The only requirement that needs to be fulfilled
involved may desorb or etch, above a certain tempera- is that a sufficient flux reaches all areas. The coverage for
surface-controlled growth is compared to the coverage
Fig. 4. The idealized temperature window of the ALD process, indicated by the growth per cycle as a function of tempera-
ture. In the case of ALD behavior, self-limiting growth is obtained and ideally a weak or no dependence on temperature is
observed, while outside the window the ALD behavior is lost due to one of four effects. Condensation: precursors can con-
dense on the surface preventing effective purging; low reactivity: low temperatures can prevent complete reactions from
taking place; decomposition: when precursors decompose, a (parasitic) CVD component can be added to the growth; de-
sorption: the film itself or the precursor can desorb, thereby reducing growth.
7
Fig. 5. The uniformity and conformality for (a) surface-controlled and (b) flux-controlled deposition techniques. The magni-
fied areas show the film coverage of a 3D feature in the substrate. Surface-controlled processes afford a coverage with
constant thickness throughout the structure (a), while flux-controlled processes result in a thickness that depends on the
line-of-sight access to the source (b). Uniformity has to be carefully engineered for flux-controlled processes to achieve an
equal flux of source species over the substrate surface. For surface-controlled growth, the requirement is simply that suffi-
cient flux should reach each area.
Fig. 6. A schematic representation of the various steps in (a) a regular ALD process, (b) a multistep process and (c) a super-
cycle. In a multistep process one or more additional steps are added in the ALD cycle to form for instance an ABC process. In
a supercycle, the steps of two normal ALD processes are combined where m cycles of the first process are followed by n
cycles of the second process. The variables m and n can be chosen so as to obtain the desired composition and structure of
the film.
for flux-controlled growth in Fig. 5. The top half of Fig. 5

8
shows the coverage of relatively large planar areas (uni- porated into ALD films is ever increasing, and those re-
formity). Perfect uniformity would be obtained for ideal ported to date are summarized in Fig. 7. The periodic le
ALD growth, while a flux-controlled process usually needs shows that a significant number of main-group elements,
careful engineering to ensure a constant flux of source transition metals and lanthanides have been reported in
species over the whole surface area, which can be diffi- films deposited by ALD, either as the principal metal
cult to achieve. The magnified areas in the bottom half of (blue squares) or as the non-metal component of binary
Fig. 5 indicate the coverage of a 3D structure (conformal- compounds (gray squares). In principle, ALD is best-
ity). When the flux of source species is sufficient, a per- suited for depositing multi-component materials (such as
fect coverage is obtained for ideal ALD growth (Fig. 5a). A binary, tertiary and quaternary) because of its stepwise
high conformality indicates an equal thickness through- approach. Indeed, binary compounds with a metal and
out a 3D structure, while a low conformality indicates a non-metal component do make up the majority of these
varying thickness, which is typically seen with flux- materials, such as metal oxides, nitrides, carbides, chal-
controlled growth (Fig. 5b). Reactor engineering can cogenides and pnictides. In general, ALD materials are
improve the uniformity for flux-controlled growth, but amporphous and, in some cases, polycrystalline. The
cannot typically improve the conformality. most common materials deposited using ALD are metal
oxides, mainly as a result of the thermodynamic stability
Advanced ALD process schemes of the M–O bond. However, with advanced ALD process
Besides the process described in Fig. 3 with two alternat- schemes such those based on supercycles, many tertiary
ing half-cycles (a so-called AB ALD cycle), there are many and quaternary compounds or alloys are also well-
additional methods of exploiting ALD-type processes by known, such as SrTiO3 (STO), HfxSiyOz or InxSnyOz. In addi-
using more steps. Fig. 6 shows a comparison between tion, single-element films have also been reported, in
various process schemes. The process can consist of particular noble metals such as Ru, Pd and Pt but also
more than two half-cycles in what is called a multistep metals such as W and Mo.
process (sometimes referred to as an ABC ALD cycle); and
In general, for binary compounds, an ALD cycle of two
ALD cycles for a certain material can be alternated with
half-cycles as discussed in Sec. 3 is applied. The choice of
cycles for other materials in what is called a supercycle. In
compound used in each half-cycle can have a significant
a multistep process (Fig. 6b) additional steps can be used
effect on the ALD process and the final film.
to change the process to widen the temperature window
or achieve different material properties. Supercycles (Fig.
Metal Precursors
6c) can be used to grow alloy, doped or multilayer films
Very early ALD processes used elemental (metal) sources;
of specific elemental mixtures. An ALD cycle for the first
however, current metal sources are almost exclusively
process is run m times, which is followed by n cycles of a
inorganic coordination complexes, i.e., a metal center
second ALD process. This supercycle of m+n cycles is then
surrounded by ligands. Examples of typical inorganic
repeated x times until the desired thickness is reached.
complexes used as precursors are shown in Fig. 8. The
The possible materials and typical precursors and co- ligands are essentially what make the metal center vola-
reactants will be discussed in the next section. tile, and they play a key role in determining the charac-
teristics of the precursor. In general, such metal sources
should ideally have the following attributes specifically
4. Materials, Precursors, and Co- for ALD:
reactants  The precursor must be reactive towards the surface

groups and leave reactive surface groups. Without
Materials this feature, precursors and co-reactions would not
To date, a wide variety of technologically important ma- undergo surface chemical reactions and ALD would
terials has been deposited using ALD 1. Moreover, the not occur.
selection of elements that have been successfully incor-
9
Fig. 7. Periodic table denoting the materials deposited by ALD (December 2013), adapted from the literature.1 Metallic
(and metalloid) elements with a blue background are those which have been incorporated into ALD films of compounds
such as oxides, nitrides, carbides etc. The gray background indicates elements forming the non-metallic component of
the films. Underlined symbols indicate that ALD thin films of the pure element have been reported.
Fig. 8. Examples of compounds used as the metal source in ALD. The names in bold refer to the general class of precursor
compound. The names below each image are the names of the specific example compound, followed, where applicable,
by the common name in italic.
 The precursors should be sufficiently volatile.
10
 The precursor must not react with itself or with its examples shown in Fig. 8 show a selection of common
surface-adsorbed species. ligand types used in ALD precursors, although there are
many other ligands possible. Only examples of homolep-
 The precursors should have a sufficiently high de-
tic precursors are covered here, but heteroleptic precur-
composition temperature. If decomposition occurs,
sors with these ligands are also widely reported.
the self-limiting growth control may be lost.
 The reaction products should be volatile and neither Halide-based (F, Cl, Br or I) precursors, such as HfCl4, are
easily adsorb to nor etch the growing film. Excessive highly desirable for industrial processes as they are highly
adsorption of reaction products can affect the self- reactive, stable compounds and are relatively inexpen-
limiting growth, resulting in CVD-like growth. Etching sive. However, halide ligands have a high tendency to
can lead to a lower GPC and reduce the film’s uni- contaminate films especially at low deposition tempera-
formity (see Sec. 7). tures. Additionally, a common reaction product is HX (X =
halide), which can be corrosive towards both the film and
the reactor. Organometallic compounds, such as metal
Although not essential to the ALD mechanism, there are
alkyls (e.g. ZnEt2 or Al2Me6) or metal carbonyls (e.g.,
many other desirable properties for an ALD precursor, in
Fe(CO)5 or Co2(CO)8), do not have these problems and are
that the compound should:
still highly reactive with high vapor pressures. However,
the low energy of the M–C bond affects their thermal
 Have a low toxicity.
stability and shelf-life, potentially making handling diffi-
 Be easy to handle. Generally, liquids are ideal as they
cult; for example, metal alkyls tend to be pyrophoric and
can be stored and transferred relatively easily. Solid
metal carbonyls readily decompose, even at room tem-
compounds can sometimes be difficult to volatilize.
perature. A stronger M–C bond is obtained for metal
 Be easy to synthesize and scale up. cyclopentadienyl (Cp) complexes (e.g., FeCp2 or CoCp2),
 Be low cost. as the Cp ligand helps to saturate the metal center both
electronically and coordinatively through π-bonding,
Some of the aforementioned precursor considerations although the bonding can be too strong with some met-
with respect to volatility are outlined in Fig. 9. A com- als, leaving the Cp difficult to hydrolyze under ALD condi-
pound’s volatility is affected by intermolecular forces, tions.
such as hydrogen-bonding or electrostatic interactions
(mainly van der Waals forces), which in turn are influ- Other popular categories of precursor include alkoxides
enced in varying degrees by the molecular weight of the (OR), such as Ti(OiPr)4, and alkylamides (NR2), such as
precursor and the shape of the molecule (i.e., how easy it Ta(NMe2)5. Alkoxides are generally liquids and exhibit a
is to pack together). Although the vapor pressure of a lower reactivity than alkyl compounds, for example, and
compound can be measured directly, it is often easier to they readily form oligomers, which can potentially affect
obtain a TG curve (thermogravimetric analysis), which volatility, as is apparent for Ti(OMe)4, which exists as a
shows a compound’s weight loss as a function of tem- tetramer with a low melting point.7 However, for higher
perature. As a rule-of-thumb, heavier, symmetrical mole- alkoxides, volatility is not usually a major issue, and they
cules tend to exhibit low volatilities, whereas lighter, are more thermally stable than alkyl precursors and have
highly asymmetric compounds are very volatile (although significantly better shelf-lives. Alkyl amides show an in-
there are many exceptions). A standard method of im- termediate reactivity to that of halides and alkoxides, but
proving volatility is to adopt a heteroleptic precursor (i.e., thermal stability can be very problematic, particularly
a metal center with two or more different ligands), which over long periods of time held in a heated bubbler.
introduces asymmetry, as opposed to a homoleptic mol-
ecule (where all the ligands are the same). Furthermore, Both alkoxides and alkylamides can oligomerise. In order
the intermolecular forces and molecular weight are also to overcome this, ligands that bind to the metal center
influenced by the ligands themselves. The precursor via two or more atoms can be adopted to block vacant
coordination sites on the metal. This can either be to the
11
Fig. 9. Thermogravimetric analysis (a compound’s mass loss as a function of temperature at a set heating rate, usually
10 °C/min) of an ideal and non-ideal ALD precursor. Ideal ALD precursors should be volatile at low temperatures and not
decompose easily, which is denoted by volatilization at a relatively low temperature and characterized by a swift mass
drop to 0%. However, there are many compounds that do not fit these requirements, potentially giving a TG curve as
exemplified above, where the volatilization takes place slowly over a wide temperature range, which is higher than that
required by most ALD processes. In this case the mass loss is likely a sign of decomposition, observable by a large residu-
al mass and loss of ligands, denoted by multiple plateaus and mass drops in the TG curve.
Table I. Typical ALD materials and the common (non-metallic) co-reactants used. R = H or CxHy (any alkyl or aryl group).1
Material Common Co-reactants

H2O, H2O2, ROH, R(CO)H, R(CO)OH, R2CO, O2 gas,
Metal Oxides O3, O2 plasma, O radicals, M(OR)x, CO2, NxOy, air
NH3, NR3, R2NNR2, NH3 plasma, H2 plasma,
Metal Nitrides N2 plasma, H2-N2 plasma, NHx radicals
Metal Carbides CxHy, BR3
Metal Phosphides PR3, POCl3, P(OR)3, PO(OR)3, P(NR2)3
Metal Arsenides AsH3, AsR3, As(NR2)3
Metal Sulfides H2S, S2R2
Metal Selenides H2Se, R2Se, Se(SiR3)2
Metal Tellurides H2Te, Te(SiR3)2
Metal Fluorides HF, MFx
H2 gas, H2 plasma, NH3, NH3 plasma, H2 plasma, N2 plasma, H2-N2 plasma,
Pure element (metal)
NHx radicals, O2 gas, O2 plasma, O radicals, SixHy, formalin
12
same metal center (a chelating ligand), as exhibited by β- amines and hydrazines have also been reported. Analo-
diketonates (and their analogues: β-diketiminates and β- gous compounds of phosphorus and arsenic (phosphines
ketoiminates), amidinates and guanidinates, or by a lig- and arsines) have been used for the deposition of metal
and bridging across two metal centers. All three ligand phosphides and arsenides, respectively. This is a similar
types are noteworthy as they have produced volatile case to the deposition of metal sulfides, selenides and
compounds of late transition metal and lanthanide com- tellurides, where either hydrides or alkyl compounds are
pounds. In general, β-diketonates form compounds with typical. The choice is more restricted for metal fluorides,
high thermal stability, which gives them a long shelf-life, in that only the hydride HF and other metal fluoride
such as [Ir(acac)3] (Fig. 8). The caveat is that their reactiv- compounds have been reported as fluorine sources. The
ity under ALD conditions is not very high. Amidinates, deposition of pure metals has also been made possible
such as [Cu2(iPr-amd)2] or [Dy(iPr-amd)3], offer a higher by a wide range of co-reactants, including H2 and even
reactivity as the M–N bond is weaker than the M–O O2. In general for metals, the co-reactant should be a
bond; however, they still afford good coordinative satu- reducing agent such that the pure ligand-free metal re-
ration. Guanidinates, such as [Gd{(iPrN)2CNMe2}3], are mains. The use of O2 is restricted to the ALD of catalytic
related to amidinates, although the alkylamino moiety on metals that are more thermodynamically stable than
the carbon-backbone gives the ligand more electronic their respective oxide.
flexibility, equivalent to that of a Cp ligand without com-
The main purpose of a co-reactant molecule is to react
promising reactivity.
cleanly with surface ligands, thereby adding a second
Of course, knowledge of metal precursors is only half the component to the film (where required), and reform the
story. The choice of co-reactant is of equal importance original surface groups. To do that, they must be suffi-
when designing a new ALD process, as the interplay be- ciently reactive at the substrate temperature employed.
tween precursor and co-reactant can be quite complex. With this in mind, co-reactant molecules can be placed in
Below, we discuss the myriad of co-reactants possible for an order of reactivity as is done for metal oxides in the
ALD processes. next paragraphs.
H2O is by far the most popular, and reported, oxygen

Co-Reactants source for metal oxide ALD.1 However, H2O adsorbs
A list of the most common ALD materials is given in Table strongly to surfaces, which means that it is difficult to
I, alongside common non-metal sources (or co-reactants) purge from the reactor, particularly at very low deposi-
for those materials. It is clear from the table that metal tion temperatures. H2O is also not sufficiently reactive at
oxides, metal nitrides and elemental metals are the low temperatures. More reactivity can be obtained using
most-studied ALD materials with the widest variety of species with less thermodynamic stability. H2O2, for ex-
possible co-reactants. In general, co-reactants are all ample, contains an O–O bond, which has a low bond
volatile, small molecules, such as elements (H2 or O2), energy, thereby increasing the molecule’s reactivity.
hydrides (H2O, NH3 etc.) or alkyl compounds (BEt3 or Despite this, the use of H2O2 has not been as widely re-
AsMe3). There are other possible compounds, but the ported as H2O, although examples of its use have often
aforementioned categories serve as a good rule-of- been associated with low-temperature depositions.
thumb. In addition, species that are activated by the
input of extra energy can be used, but these will be dis- Energy-enhanced ALD
cussed in the next section. In order to obtain more reactivity, species with relatively
short lifetimes can be used. O3 gas and plasma species
From Table I, it is clear that the greatest range of com-
are examples of co-reactants that both have short life-
pounds is associated with metal oxide depositions. H2O is
times and must therefore be produced in situ, rather
the most commonly-used oxygen source, although H2O2
than being obtained from a chemical supplier, for exam-
and many oxygen-containing organic molecules have also
ple. Their production requires an application of energy to
been successfully employed. For metal nitrides, NH3 is
a gaseous species. Such energy is usually in the form of
the most popular co-reactant, although substituted
13
an electrical discharge, thermal cracking or photodissoci-  Association: bonding of an intact precursor or co-
ation.8 Processes involving such extra energy as this are reactant with the surface without release of ligands
referred to as energy-enhanced ALD. Specifically for O3 (e.g., hydrogen bonding).
gas, O3 molecules are typically stable on the order of  Combustion: an oxidizing co-reactant can combust
minutes in an ALD reactor setting, and so are present surface groups and replace them with an oxidized
long enough to travel from the O3 generator to the reac- surface.
tion chamber. However, the O3 transport line must be  Abstraction: a co-reactant can remove ligands and
kept clean of certain materials as the O3 molecules are release reaction products, without leaving behind
prone to decompose on specific surfaces.9 fragments from the co-reactant. For instance in con-
In the case of plasma, the species are generated closer to junction with reduction of the metal center.
or in the reaction chamber. When employing a plasma,  Reduction: the metal center can be reduced to a
various highly reactive species can contribute to the lower oxidation state by the co-reactant.
surface chemistry, such as ions, electrons, radicals, meta-  Nitridation: the co-reactant can nitridate the materi-
stable species and (UV) photons 10. These plasma species al or surface. This reaction is essentially the nitrogen
are typically produced in an electric discharge by the version of combustion.
application of radio-frequency or microwave energy. The
ions, electrons and photons are basically only present Here, selected examples of specific ALD chemistries are
when the power is on, but the radicals still exist when the presented to illustrate the variety in ALD reaction mech-
power is no longer applied, although their lifetimes are anisms.
also short. The benefit of this situation is that a high
reactivity is obtained and long purge times are often not Aluminum Oxide
necessary. Perhaps the most well-known ALD chemistry is that of
Al2O3, a material that probably also has the widest range
It is clear that each precursor or co-reactant in an ALD of applications of all ALD materials. There are various
cycle has a profound effect on the chemistry of the pro- ALD processes for Al2O3 available, in which [Al(CH3)3]2
cess. Until now, we have been focusing on very general (TMA),* Al2Cl6, and [Al(OiPr)3]4 have been used as the Al-
concepts. In Sec. 5, the chemistries of specific ALD pro- source, alongside water, ozone or an oxygen plasma as
cesses will be covered. the co-reactants. The TMA/water process in particular
has long been considered a model ALD process,2 as it
provides a good demonstration of metal alkyl precursor
5. ALD Chemistries chemistry. Furthermore, the TMA/O2 plasma process also
In the previous section, a general overview of materials demonstrates nicely how an O2 plasma undergoes com-
prepared by ALD is given as well of precursors and co- bustion-like mechanisms with an organically-terminated
reactants used. A variety of reaction chemistries can surface.
apply for the ALD processes:
TMA with Water
 Ligand-exchange: a reaction between surface groups Many ALD processes involve a Brønsted acid-base ligand-
and ligands on precursors, or surface groups and co- exchange reactions between the surface groups and the
reactants, where groups are exchanged leading to incoming precursor(s), releasing volatile reaction prod-
volatile reaction products. These are also referred to
as condensation and hydrolysis reactions. *
Many three-coordinate aluminum compounds exist in a
 Dissociation: reaction in which the precursor or co- dimeric form (i.e., two individual molecules combined to
reactant dissociate into several adsorbed species on form one) in order to gain stability. In particular, TMA is
the surface without releasing reaction products into dimeric at ambient temperatures <70 °C. Above 70 °C,
the vapor phase. the dimer exists in equilibrium with the monomer. For
clarity, it is typically shown in the monomeric form in
many publications.
14
ucts. These ligand-exchange reactions are typically alter- TMA half-cycle (n = 1, 2):
nating condensation and hydrolysis reactions. The two
ALD half-cycles of the TMA/water process are good ex- n s-OH (ads) + ½ [Al(CH3)3]2 (vap) → s-OnAl(CH3)3-n (ads) .
amples of this (Eq. 5.1 and 5.2): + n CH4 (g) (5.3)
TMA half-cycle (condensation, n = 1, 2): O2 plasma half-cycle:
n s-OH (ads) + ½ [Al(CH3)3]2 (vap) → s-OnAl(CH3)3-n (ads) s-Al(CH3) (ads) + 4 O (plasma) → s-AlOH (ads) + CO2 (g) .
.
+ n CH4(g) (5.1) + H2O (vap) (5.4)
Water half-cycle (hydrolysis): The water formed as a reaction product can then react
with any (remaining) s-CH3 groups to form methane and
s-Al(CH3) (ads) + H2O (vap) → s-AlOH (ads) + CH4 (g) (5.2) s-OH, as in the thermal process (Eq. 5.2). As plasmas are
highly reactive, Al2O3 films synthesized by plasma-
In this example, the surface hydroxyl group (s-OH) acts as enhanced ALD tend to have a higher GPC (Fig. 10), higher
a Brønsted acid by donating a proton to a methyl (CH3) density and lower carbon, hydrogen and (excess) oxygen
ligand on the TMA, which is the Brønsted base. This contents than films deposited using water or ozone as
transfer results in the formation of a surface O–Al bond the co-reactant. This is particularly the case at low sub-
and the release of methane as a reaction product. Addi- strate temperatures.13
tionally, TMA may bind either via one methyl group
(monofunctional binding) or two methyl groups (bifunc- Hafnium Oxide
tional binding). Trifunctional binding is energetically very Hafnium oxide is another common ALD material, which is
unlikely due to the geometry of the TMA molecule. a high-k dielectric. In industry, halide-based precursors,
in particular HfCl4, are favored because of the reactivity
In terms of growth per cycle (GPC) as a function of tem-
(with surface OH) and stability of the compound. Other
perature (Fig. 10), it becomes apparent that the process
common precursors are alkoxide- and alkylamide-based,
is well described by the term thermal ALD. At deposition
such as [Hf(OiPr)4] and [Hf(NEtMe)4], which are widely
temperatures below 150 °C, the GPC drops with decreas-
reported in R&D.ALD reactions involving HfCl4 give insight
ing temperature, in essence, there is less thermal energy
into how halide-based ALD processes behave.
available to drive the reaction. At the higher end of the
temperature scale, a drop in GPC is also observed with
HfCl4 with Water
increasing temperature, which stems from a reduction in
Initially, the half-cycles (Eq. 5.5 and 5.6) appear to be
the density of reactive s-OH groups. In this sense, the
similar to the TMA/water process, i.e., alternating con-
TMA/water ALD process illustrates that even a model
densation and hydrolysis reactions.14
ALD process, such as the TMA/water process, does not
necessarily show a temperature window where the GPC HfCl4 half-cycle (n = 1, 2):
is constant with temperature.
n s-OH (ads) + HfCl4 (vap) → s-OnHfCl4-n (ads) + n HCl (g) (5.5)
TMA with an O2 Plasma
For the TMA/O2 plasma ALD process, the first half cycle Water half-cycle:
can essentially be considered the same as in the thermal
s-HfCl (ads) + H2O (vap) → s-Hf(OH) (ads) + HCl (g) (5.6)
process (Eq. 5.3), whereby incoming TMA molecules
react with s-OH groups. In the second half-cycle (O2 The stepwise nature of the ALD cycles can be easily seen
plasma), the reaction products are carbon dioxide, water in the variation of mass with time as measured by a
and methane (Eq. 5.4).11,12 At the most basic level, the quartz crystal microbalance. Fig. 11 shows a large mass
first two reaction products (carbon dioxide and water) increase on adsorption of HfCl4 and a mass loss as water
can be explained by a combustion-like mechanism: displaces the surface chloride ligands as HCl. In addition
to the ligand-exchange reactions, HfCl4 has an affinity to
15
Fig. 10. Variation in growth per cycle (GPC) as a function of substrate temperature for the ALD of Al2O3 from trimethyl-
aluminum, [Al(CH3)3]2 (TMA). The decrease in growth at lower temperatures for the thermal ALD process is a result of
insufficient thermal energy, required to fully drive the surface reactions (see Fig. 4). Above 200 °C there is a decrease in
GPC for thermal ALD and a similar decrease is seen for plasma ALD over the whole temperature range (attributed to the
desorption of reactive surface groups with increasing temperature). This illustrates that there is no ideal temperature
window, as true ALD is occurring at all deposition temperatures yet the GPC is not constant (as in the ideal situation
described in Fig. 4). For plasma ALD at lower temperatures, there is a slightly superlinear increase in the GPC as a result
of reduced film density. However, the Al atoms deposited per cycle as a function of temperature shows a consistent line-
ar plot.13
Fig. 11. Change in mass as a function of time for the HfCl4/water ALD process, measured using a quartz crystal microbal-
ance (QCM). The mass changes are in the order of ng. (a) When only HfCl4 is introduced into the reactor (i.e., no water),
the mass does not change after the first pulse, confirming saturation of the surface sites. (b) Over full ALD cycles, the
mass increases on adsorption/reaction of HfCl4 molecules before reaching a plateau. The mass loss observed on addition
of water is a result of replacement of the heavier chlorine atoms (35.5 g mol-1) with lighter hydroxyl groups (17 g mol-1).
Reprinted from Thin Solid Films, 340, J. Aarik et al., Influence of substrate temperature on atomic layer growth and prop-
erties of HfO2 thin films, pp. 110-116 , copyright (1999), with permission from Elsevier.15
16
oxygen bridges, i.e., Hf–O–Hf.15 In essence, the oxygen TiCl4 with Ammonia
bridges can also serve as reactive surface sites. The HfCl4 In the first half-cycle, a surface covered in NH and NH2
can add across an oxygen bridge (Eq. 5.7) and then the groups is exposed to the TiCl4 precursor, where HCl is
remaining chlorides can undergo hydrolysis with water formed as a reaction product. In the second half-cycle,
(Eq. 5.8). NH3 as a co-reactant removes Cl from the surface and
regenerates the NH and NH2 groups. Again, HCl is formed
HfCl4 half-cycle:
as a reaction product. For simplicity the TiCl4 is assumed
s-Hf(–O–)2Hf (ads) + HfCl4 (vap) → to release one ligand during adsorption.
.
s-Hf(Cl)–O–Hf(OHfCl3) (ads) (5.7)
TiCl4 half-cycle (n = 1, 2):
Water half-cycle:
s-NHn (ads) + TiCl4 (vap) → s-NHn-1TiCl3 (ads) + HCl (g) (5.11)
s-Hf(Cl)–O–Hf(OHfCl3) (ads) + 2 H2O (vap) → .
NH3 half-cycle:
s-Hf(–O–)2Hf(–O–)2Hf (ads) + 4 HCl (g) (5.8)
n s-Cl (ads) + NH3 (vap) → s-NH3-n (ads) + n HCl (g) (5.12)
HfCl4 with Ozone
The deposition of HfO2 using HfCl4 as precursor can also Furthermore, note that N2 and Cl2 also have been sug-
be performed with ozone as an alternative to water as gested as reaction products. In TiCl4, titanium is in the 4+
the co-reactant.16,17 The following mechanism has been oxidation state, while Ti is in the 3+ state in TiN, which
proposed based on known interactions of ozone with means the titanium must therefore be reduced during
surfaces.18 Assuming only oxygen bridges or surface ox- the reaction. Interestingly, XPS analysis has shown that
ide were present, the HfCl4 could either react by adding the reduction takes place during the TiCl4 exposure step
across a Hf–O–Hf bridge or simply by binding to surface instead of the NH3 exposure step.20
oxygen species (Eq. 5.9). However, it is the reaction be-
tween incoming ozone and the surface chloride species TiCl4 with H2-N2 plasma
that remains unconfirmed. It has been suggested that the TiN can also be deposited using an H2-N2 plasma as co-
ozone oxidizes the surface chloride species to chlorine reactant. The first half-cycle is assumed to be similar as
gas (Eq. 5.10) while forming a Hf–O bond. for the thermal process:
HfCl4 half-cycle (n = 1, 2): TiCl4 half-cycle (n = 1, 2):
s-Hf(–O–)2Hf (ads) + HfCl4 (vap) → .

s-NHn (ads) + TiCl4 (vap) → s-NHn-1TiCl3 (ads) + HCl (g) (5.13)
s-Hf(Cl)–O–HfOHfCl3 (ads) (5.9)
while it has been suggested that H and N radicals are the
Ozone half-cycle: most important reactive species in the co-reactant half-
cycle:
s-HfCl2 (ads) + 2 O3 (g) → s-HfO2 (ads) + 2 O2 (g) + Cl2 (g) (5.10)
s-Cl (ads) + (n+1)H (plasma) + N (plasma)→ s-NHn (ads) .
Furthermore, it has been suggested that oxychloride + HCl (g) (5.14)
species, such as Cl2O or ClO, could also be produced.
Here, H radicals also serve as a reducing agent for the Ti
Titanium Nitride center. The energy provided to the surface by the ions in
Titanium nitride is a common ALD metal nitride, which the H2-N2 plasma is furthermore reported to be beneficial
has good barrier properties and a high conductivity.19 to achieving low resistivities. As can be seen in Fig. 12,
Similar to the halide HfCl4, TiCl4 is often used in industry low resistivities have been obtained down to low sub-
because of the stability of the compound. Another com- strate temperatures. Note that in the H2-N2 plasma itself,
mon precursor is alkylamide-based, [Ti(NMe2)4], which is NH3 is also formed, which can contribute to the reaction
reported in R&D.
17
in a secondary thermal ALD reaction (Eq. 5.12). If a pure [Pt(CpMe)Me3] with an O2 Plasma
H2 plasma is used with TiCl4, Ti metal can also be grown.21 Since the thermal process relies on the catalytic nature of
the Pt, nucleation on non-catalytic surfaces is inherently
Platinum Metal difficult. By using an O2 plasma, this dependence is re-
Noble metals such as platinum can be used as electrode moved and nucleation is strongly improved.26 The pre-
materials, because of their low resistivity and resistance cursor step is basically the same as the thermal process,
to oxidation. These materials are also used for their cata- while the co-reactant step differs.
lytic properties.
[Pt(CpMe)Me3] half-cycle:
[Pt(CpMe)Me3] with O2 gas
For thermal ALD processes to noble metals, the catalytic s-O (ads) + 2 [Pt(CpMe)Me3] (vap) → s-PtCxHy (ads) + CO2 (g)
activity and the instability of the respective metal oxides + H2O (vap) + CH4 (g) + CO (g) + H2 (g) (5.18)
can be utilized in the growth chemistry by using O2 as a
O2 plasma half-cycle:
co-reactant.22 Typically for platinum, metal-organic pre-
cursors are used with Cp ligands such as [Pt(CpMe)Me3]. s-PtCxHy (ads) + O (plasma) → s-Pt (ads) + s-O (ads) + CO2 (g) .
The following reaction mechanism was proposed after + H2O (vap) (5.19)
studying the reaction products (Fig. 13).23–25 The first
essential step in the reaction is the dissociative chemi- Note that the half-cycles are again not balanced. Natural-
sorption of O2 on Pt: ly, as soon as Pt is grown, the thermal process can also
occur as the Pt can dissociate O2. When applying low
s-Pt (ads) + O2 (g) → s-Pt (ads) + 2 s-O (ads) (5.15) temperatures and/or when large doses of O radicals are
used while not significantly overdosing the precursor,
Note that this reaction is unbalanced in terms of surface
also PtO2 can be grown (i.e., the noble metal oxide).
sites. The surface covered by a submonolayer of O is the
surface that is present during the precursor step. The
Other ALD Chemistries
following overall reactions are expected to take place
The aforementioned chemistries for the Al2O3, HfO2, TiN
during the half-cycles.
and Pt ALD processes demonstrate how ALD manifests
[Pt(CpMe)Me3] half-cycle: itself in real reaction mechanisms. However, there are
many alternative ALD chemistries, which can seem in-
s-O (ads) + [Pt(CpMe)Me3] (vap) → s-PtCxHy (ads) + CO2 (g) . creasingly complex. For example, the following reaction
+ H2O (vap) + CH4 (g) + CO (g) + H2 (g) (5.16) is known to occur between aluminium chloride and alu-
minium isopropoxide (Eq. 5.20):
O2 gas half-cycle:
AlCl3 (ads/vap) + Al(OiPr)3 (ads/vap)  Al2O3 (s) + 3 iPrCl (g) (5.20)
s-PtCxHy (ads) + O2 (g) → s-Pt (ads) + s-O (ads) + CO2 (g) .
+ H2O (vap) (5.17) Such a chemistry has been exploited to deposit (mixed)
metal oxides (metals = Al, Si, Ti, Zr, Hf) using alternating
Note that these half-cycles are not balanced. As on the cycles of a metal chloride and metal alkoxide precursor,27
bare Pt surface, the O2 in this O2 gas half cycle dissocia- although elevated substrate temperatures (>300 °C)
tively chemisorbs on the Pt to form a self-limiting ad- were necessary. Interestingly, a gaseous co-reactant was
sorbed layer of O atoms. This adsorbed O not only com- not necessary, as the alkoxide ligand already contains
busts ligands during the co-reactant step, but also during oxygen. Other chemistries that have been employed for
the subsequent precursor half-cycle. When the adsorbed ALD include silyl halide elimination (for example, Eq.
O is consumed during the precursor half-cycle, dehydro- 5.21), which relies on the formation of a strong Si–X bond
genation reactions then lead to the formation of CH4 and and release of a volatile reaction product, exemplified by
H2. the ALD of Ge3Sb2.28
18
Fig. 12. Resistivity (at room temperature) of TiN films deposited at 100–400 °C. The films were obtained by plasma ALD
using TiCl4 in combination with an H2-N2 plasma. The resistivity was determined by in situ spectroscopic ellipsometry
(thin films, ~10 nm) and four-point probe measurements (thicker films, >45 nm).45
Fig. 13. Difference Fourier transform infrared (FTIR) spectra during the [Pt(CpMe)Me3] dosing and the O2 dosing. During
the precursor dosing, peaks from the excess precursor (2909 cm-1), CO2 (2360 cm-1), and CH4 (3016 cm-1) can be seen.
During the O2 dosing, peaks from the CO2 are visible. The signal from H2O (present during both steps) is hardly distin-
guishable from the noise in the data.23
19
(e.g., 1 Torr) can be obtained to have both efficient pre-

cursor and co-reactant usage, and effective purging of
3 GeCl2(dioxane) (ads/vap) + 2 Sb(SiEt3)3 (ads/vap)  Ge3Sb2 (s) the reactor volume.
+ 6 Et3SiCl (vap) + dioxane (vap) (5.21)
Showerhead ALD reactor
In essence, these examples demonstrate that there is an
For gases entering the reactor, a showerhead can be
almost limitless variety of reaction mechanisms applica-
used to evenly distribute the precursor and co-reactant
ble to ALD, which essentially mimic reactions carried out
flux. This design is beneficial for processes that are likely
in solution.
to exhibit a small CVD component or suffer from chemi-
cal side-effects (see Sec. 7 for discussion on these side-
effects). When high substrate temperatures are needed,
6. ALD reactors
a heated substrate stage can be employed with the wall
In this section, the basics of ALD reactors are explained
temperature set lower for technical reasons (e.g., avoid-
and various types of ALD reactors are briefly discussed.
ing the use of expensive high-temperature-resistant
First a basic flow-type reactor is discussed, after which
components). Reactors where the wall temperature is
alternative designs are treated.
lower than the substrate temperature are often called
“warm wall” or even “cold wall” reactors. Note that care
Reactor basics
has to be taken to avoid precursor condensation or side
The following requirements are important for an ALD
reactions at the wall influencing the growth at the sub-
reactor. Firstly, the reactor heats the samples to the
strate.
desired deposition temperature. Pulses of precursors and
co-reactants need to be injected into the reactor volume,
Batch ALD reactor
and an inert gas is often used to purge the reactor vol-
To increase the throughput of ALD reactors, batch reac-
ume between these steps. Generally, the reactor volume
tors can be employed. As shown in Fig. 14, such a reactor
is continuously pumped to achieve a certain pressure.
is typically a furnace that can contain many substrates
This pressure is typically between 0.1 and 10 Torr, but
(typically 50-500) that are coated in one run. Because of
pressures at the mTorr level and at atmospheric pressure
the larger reactor volume and slower gas diffusion be-
are also viable.
tween the substrates, the exposure and purging times
The reactor requirements dictate the design. Most ALD need to be longer to maintain ALD behavior. Fortunately,
reactor designs are derived from CVD reactors. However, the advantage of processing multiple wafers at the same
contrary to CVD, ALD is not a continuous process and time outweighs the disadvantage of longer processing
therefore the reactor volume has to be filled and purged times in most cases. Because of the longer cycle times,
efficiently to achieve reasonable cycle times. Further- small CVD components or chemical side-effects will have
more, the pulsing of the precursor and co-reactant has to a larger influence on uniformity than in a single-wafer
be automated with accurate timing. Note that in the tool.
reactor, growth can occur on all reactor surfaces that are
exposed to both precursor and co-reactant. Fig. 14 shows Energy-enhanced ALD reactor
an overview of the reactor types discussed. For ALD processes that require more than thermal ener-
gy to obtain a good film or feasible process, energy-
Flow-type ALD reactor enhanced ALD can be employed. In this case, the co-
A flow-type reactor is essentially a furnace through which reactants are species with a limited lifetime, such as
the gases flow. The substrate rests on the bottom edge, ozone or plasmas. This requires the integration of
adopting the temperature of the walls. Typically, a carrier equipment that generates the species relatively close to
gas is used to transport the precursor and co-reactant. By the substrate.8
choosing the inlet gas flow and pumping speed, an opti-
mum flow (e.g., laminar or streamline flow) and pressure
20
Fig. 14. A schematic of the various types of ALD reactor. The top half shows two single-wafer, temporal reactors for
thermal ALD with a flow-type reactor on the left and a showerhead reactor on the right. Alternative types are shown in
the bottom half, starting from the bottom left: batch, energy-enhanced and spatial.
In the case of ozone, the ozone generator is positioned cals recombine (as shown in Fig. 14). A wide variety of
close to the reactor and connected via an inlet. In princi- plasma sources can be used that can either be directly in
ple, the rest of the reactor design can be similar to the contact with the substrate or more remote as discussed
flow-type reactor, for example, although care has to be by Profijt et al.10
taken when materials or temperatures are used that
could lead to the decomposition of ozone. In this case, Spatial ALD reactor
the decomposition on the reactor surface can lead to low Instead of separating the ALD steps of Fig. 3 in the time
ozone fluxes downstream.9 domain, the steps can also be separated in the spatial
domain as shown in Fig. 14.29 This means that precursor
In the case of plasma ALD, the co-reactant species have a and co-reactant exposures occur at different positions
short lifetime and are generated closer to the substrate. using different reaction zones separated by purging are-
For instance, in a volume just above a showerhead to as. Therefore, in order to expose the substrate to these
distribute the co-reactant species evenly before the radi- different zones, either the substrate itself or a ‘deposi-
21
tion head’ must move. In this case, the throughput is scription of the poisoning process. Since both reac-
merely limited by the technical specifications of the reaction products and precursor compete for surface
tor (e.g. substrate loading, reactor dimensions, etc.) as sites the partial pressures of each reactive species
well as the kinetics of the surface reactions. For process- will affect the resulting growth. For the TiN process
es with typical half-reaction time-scales in the order of a discussed in Sec. 5, using TiCl4 as precursor and NH3
few milliseconds, very high deposition rates can be ob- as co-reactant, poisoning by the reaction product HCl
tained while maintaining a high film quality. can occur and uniformity and conformality can be
compromised.30
 Process interaction: When using different processes
7. ALD virtues and practicalities in sequence, for instance in a supercycle, interaction
In this section, some practicalities are addressed which between these processes has to be considered. For
are important to consider when ALD is utilized. Consider- instance TMA can react with a ZnOH surface, in
ations for uniform and conformal growth are discussed which Zn is removed and released from the surface
as well as how to control the growth or how to allow as ZnMe2.31
growth at low substrate temperatures. Some effects
manifest themselves as virtues for extending the applica- Excellent conformality has been achieved for many ALD
bility of ALD. processes as can be seen, for instance, for thermal ALD of
Al2O3 in the SEM image in Fig. 16. Here the Al2O3 film has
Uniform and conformal deposition equal thickness throughout the trench structure due to
In the case of an ideal ALD process, perfect uniformity saturation of the growth. Note, however, that conformal-
and conformality over the entire surface of the sample is ity in some cases can even be achieved for unsaturated
achieved when the ALD conditions as described in Sec. 3, conditions. Fig. 17 shows the thickness profile under
are fulfilled. However, since ALD relies on surface chem- unsaturated conditions for three different deposition
istry, several chemical side-effects can be present as well: regimes.32 In the case of reaction-limited growth (Fig.
17a), which occurs for low reaction probabilities of the
 Decomposition: A very basic effect is decomposition precursor or co-reactant, an equal thickness is observed
of precursor in the gas phase or on the surface, throughout the trench (i.e., no gradient) even when the
which can lead to a CVD component. This often be- process is not saturated. To achieve conformal and satu-
comes an issue when transferring processes from rated growth, the exposure times do not have to be ex-
single-wafer tools to batch tools. During the short tended compared to the saturated conditions on planar
exposure times used in a single-wafer tool, such a surfaces. Certain LPCVD processes behave similarly and
CVD component might be negligible, while the long- can provide conformal films. For high reaction probabili-
er exposure times in the batch tool can lead, for in- ties or for very high aspect ratio structures, growth is
stance, to clear non-uniformities for the same ALD generally diffusion-limited (Fig. 17b), and a growth front
process. is observed, moving from top to bottom, when extending
 Etching: Another chemical side-effect is etching, the exposure. Exposure and purge times have to be ex-
which, for instance, can be an issue with metal hal- tended to a large extent compared to the planar case to
ide precursors. The hydrogen halide reaction prod- achieve saturation in the bottom of the trench (Fig. 17b).
ucts (e.g., HCl) can etch the material being deposit- The initial coverage is similar to PVD growth where in-
ed, the substrate material or even the reactor sur- coming particles immediately adsorb at the top of the
faces. trench. For co-reactants such as plasma radicals and
 Surface poisoning: a chemical side-effect in which ozone, the growth can be recombination-limited (Fig.
the reaction products influence the process. In this 17c), where a decreasing thickness is observed when
case the reaction products compete with the precur- moving down in the trench. In this case, conformal coat-
sor for the surface groups created during the co- ing in very high aspect ratios can become impractical due
reactant exposure. Fig. 15 shows a schematic de- to the very long exposure times
22
Fig. 15. A schematic representation of surface poisoning during ALD. In contrast to Fig. 3, the reaction products during
the first half-cycle can also react with the initial surface and therefore compete with the precursor for surface sites. This
competition can lead to non-uniform deposition even when both half-cycles have sufficient exposure and purging.
Fig. 16. Cross-sectional SEM image of an Al2O3 ALD film with a thickness of 300 nm on a Si wafer with a trench structure.
The film follows the trench structure conformally.46
necessary. Nonetheless, when a sufficient exposure is The thermal ALD process of HfO2 using H2O, as described
applied, a conformal film is obtained. in Sec. 5, is a nice example of growth delay due to a lim-
ited amount of OH groups on the initial surface.14 The
Growth control growth proceeds through the formation of islands on
The substrate material and its treatment before ALD can isolated reactive sites, which slowly coalesce during con-
lead to a growth delay or growth enhancement. Fig. 18 secutive cycles, after which normal growth is obtained.
shows schematically, the effect on the average growth
rate of the different scenarios. In the case of an ideal When the sum of the surface energy of the material to be
process there would be an immediate constant growth deposited and of the interface energy between film and
rate, while an accelerated or delayed growth will lead to substrate is considerably higher than the surface energy
respectively higher or lower thickness after a specific of the substrate, de-wetting of the film can occur, leading
number of cycles. Issues with nucleation of the material to island or nanoparticle formation (similar to other dep-
to be deposited on a surface can hinder growth of closed osition techniques). For the thermal ALD process of Pt, as
thin films, but can also be exploited to obtain controlled described in Sec. 5, nucleation on non-noble metal sur-
growth of islands or even nanoparticles on surfaces.33 faces is slow and needs high O2 exposures to grow at all.
Fig. 19 shows the thickness as a function of the number
23
Fig. 17. Illustration of the different thickness profiles resulting from unsaturated growth conditions for the three deposi-
tion regimes: (a) reaction-limited regime, in this case growth occurs simultaneously everywhere until saturation is
reached; (b) diffusion-limited regime, a downwards moving saturation front is observed for this situation; and (c) recom-
bination-limited regime, here growth occurs everywhere but with a rate that is reduced near the bottom of the trench
where the condition of saturation is fulfilled last. In (b), the small amount of growth at the bottom due to species making
no or very few collisions has been neglected. The cases shown are representative of unsaturated growth conditions to
highlight the difference between the regimes (for saturated growth conditions the profiles would be conformal in all
three cases).32
Fig. 18. Monitoring the film thickness as a function of number of cycles during an ALD process. Three kinds of nucleation
behavior of the ALD films can typically be distinguished during the first few cycles. ALD process with: accelerated growth,
immediate constant growth rate (ideal process), and (pronounced) nucleation delay.
24
Fig. 19. Thickness as a function of the number of ALD cycles for thermal ALD of Pt. The bright field TEM images in the
figure illustrate that Pt ALD nucleation evolves from island growth, via island coalescence, to film closure. A scale bar is
given for the SEM images.34
Fig. 20. Resistivity of Al-doped ZnO films prepared with dimethylaluminum isopropoxide, [Al(CH3)2(OiPr)]2 (DMAI) and
TMA as Al precursors as a function of Al fraction (AF). A lower resistivity at a low Al content can be achieved when using
the larger molecule DMAI as an alternative Al precursor for doping instead of TMA. The precursor molecule diagrams are
shown as an inset. The Al fraction (AF) is varied by changing m in the supercycles (A1B1)m(A2B2)1 (Fig. 6) with A1 the Zn-
precursor (Zn(C2H5)2) and A2 the Al-precursor. B1 and B2 are water.39
25
Fig. 21. The increase in film thickness as a function of ALD cycles for Al2O3, TiO2, and SiO2 performed at room tempera-
ture. All plasma-enhanced ALD processes and the TMA/O3 ALD process showed a linear increase in thickness with the
number of ALD cycles.41
Fig. 22. Left: Bent woven cotton fabric coated with 3 nm Al2O3 at room temperature and 750 cycles of the ABC process
for Pt at room temperature (A is Pt(CpMe)Me3; B is O2 plasma; C is H2 gas or H2 plasma). Top right: SEM image of the
coated fabric surface, where a sharp image is obtained due to the presence of the conductive Pt. Bottom right: energy-
dispersive X-ray spectroscopy (EDX) spectra prove the presence of a Pt film on the cotton fabric.42
26
of cycles for this process on Al2O3 as a starting surface.34 purity, low growth rates or long cycle times.40 Energy-
The insets in Fig. 19 show TEM images of the surface enhanced ALD (e.g., plasma ALD), can generally be used
during various stages of the growth. Initially, the surface at lower temperatures because of its higher reactivity.
is covered with Pt nanoparticle islands. As the growing Pt Using plasma ALD, several oxides have been deposited
islands coalesce, the surface becomes completely cov- down to room temperature such as Al2O3, TiO2 and SiO2
ered with Pt and the growth achieves a constant rate. as shown in Fig. 21.41 To enable growth of Pt metal at low
This illustrates that, in some cases, ALD can be used to temperatures, an additional step can be used in an ABC
deposit either nanoparticles or closed films using the ALD cycle, where C is a reduction step using H2 gas or
number of cycles for control. plasma.42 Using this ABC cycle, Pt metal can be deposited
down to room temperature as demonstrated by coating
Besides the common goal of optimizing conditions and woven cotton fabric (Fig. 22).
processes in order to reduce the dependency on the
starting surface properties, the opposite is also desired in To attain certain crystalline phases for some materials,
some cases. Here the process can be developed such that relatively high temperatures are sometimes needed. This
selective growth occurs. In this case, there is controlled is, for example, the case for rutile TiO2, which is generally
growth on certain materials, while there is no growth on only obtained using ALD at elevated substrate tempera-
other materials. Such a process can be exploited in order tures. However, the use of substrate biasing to provide
to achieve self-assembly or, for instance, a mask-less additional ion energy allows rutile TiO2 to be grown at
process. The thermal ALD process of Pt described in Sec. 200 °C without specific nucleation layers or annealing
5, has a good selectivity of growth on noble metals, com- treatments.43
pared to oxides, when using a low O2 exposure.35 This
selectivity can be utilized to grow Pt structures on noble
metal seed patterns. This process can also be employed 8. Conclusion
to make core-shell nanoparticles, with for instance a Pd In this chapter, the method of atomic layer deposition
core and a Pt shell.36 Selective growth can also be ob- (ALD) has been introduced on an elementary level while
tained by inhibiting growth on the surfaces where no an overview of materials, precursors, reactants, typical
growth is desired, as has been demonstrated using self- surface chemistries, reactor configurations and ALD prac-
assembled monolayers.37 ticalities has been given to illustrate the diversity of ALD
and how it manifests itself in applications. A good under-
Supercycles (Sec. 3), can be used to control the dopant standing of the basics of ALD is key in recognizing its
introduction in a material. Al-doped ZnO has been depos- opportunities in current and emerging applications, as
ited by alternating AB cycles for ZnO with AB cycles for well as advancing the technology even further. Although
Al2O3. By choosing the metal precursor of the Al2O3 pro- it was more than 40 years ago when the technique was
cess the lateral dopant distance can be changed while established as atomic layer epitaxy (ALE), ALD has only
the dopant distance perpendicular to the substrate is recently prevailed as pivotal and inevitable for many
controlled by the number of ZnO cycles.38,39 For example technologies. It is therefore expected that the im-
a lower resistivity at low Al content can be achieved portance of ALD as a thin film deposition technique will
when using the larger molecule dimethylaluminum iso- continue to grow rapidly, complementing deposition
propoxide, DMAI, [Al(CH3)2(OiPr)]2, as an alternative Al techniques such as PVD, CVD and other thin-film tech-
precursor for doping instead of TMA (Fig. 20). nologies. This will be aided by current trends in ALD re-
search, for example, in the field of increasing throughput
Growth at low temperature by spatial ALD approaches, extending the range of (com-
Compared to CVD, ALD can typically achieve good mate-
plex) materials and process properties by multistep and
rial properties at lower deposition temperatures (200 –
supercycle processes, and exploring new avenues in ALD
400 °C). However, at even lower temperatures (25 -
such as area-selectivity.
100°C), thermal ALD processes can have low material
27
References (20) Tiznado, H.; Zaera, F. J. Phys. Chem. B 2006, 110,

13491–8.
(21) Kim, H.; Rossnagel, S. M. J. Vac. Sci. Technol. A
(1) Miikkulainen, V.; Leskelä, M.; Ritala, M.; Vacuum, Surfaces, Film. 2002, 20, 802.
Puurunen, R. L. J. Appl. Phys. 2013, 113, 021301. (22) Aaltonen, T.; Rahtu, A.; Ritala, M.; Leskelä, M.
(2) Puurunen, R. L. J. Appl. Phys. 2005, 97, 121301. Electrochem. Solid-State Lett. 2003, 6, C130.
(3) George, S. M. Chem. Rev. 2010, 110, 111. (23) Kessels, W. M. M.; Knoops, H. C. M.; Dielissen, S.
(4) Parsons, G. N.; Elam, J. W.; George, S. M.; A. F.; Mackus, A. J. M.; van de Sanden, M. C. M.
Haukka, S.; Jeon, H.; Kessels, W. M. M.; Leskelä, Appl. Phys. Lett. 2009, 95, 013114.
M.; Poodt, P.; Ritala, M.; Rossnagel, S. M. J. Vac. (24) Mackus, A. J. M.; Leick, N.; Baker, L.; Kessels, W.
Sci. Technol. A Vacuum, Surfaces, Film. 2013, 31, M. M. Chem. Mater. 2012, 24, 1752–1761.
050818. (25) Erkens, I. J. M.; Mackus, A. J. M.; Knoops, H. C.
(5) Suntola, T. Mater. Sci. Reports 1989, 4, 261–312. M.; Smits, P.; van de Ven, T. H. M.; Roozeboom,
(6) Leskelä, M.; Ritala, M. Thin Solid Films 2002, 409, F.; Kessels, W. M. M. ECS J. Solid State Sci.
138–146. Technol. 2012, 1, P255–P262.
(7) Wright, D. A.; Williams, D. A. Acta Crystallogr. (26) Knoops, H. C. M.; Mackus, A. J. M.; Donders, M.
Sect. B Struct. Crystallogr. Cryst. Chem. 1968, 24, E.; van de Sanden, M. C. M.; Notten, P. H. L.;
1107–1114. Kessels, W. M. M. Electrochem. Solid-State Lett.
(8) Potts, S. E.; Kessels, W. M. M. Coord. Chem. Rev. 2009, 12, G34.
2013, 257, 3254–3270. (27) Ritala, M. Science (80-. ). 2000, 288, 319–321.
(9) Knoops, H. C. M.; Elam, J. W.; Libera, J. A.; (28) Pore, V.; Knapas, K.; Hatanpää, T.; Sarnet, T.;
Kessels, W. M. M. Chem. Mater. 2011, 23, 2381. Kemell, M.; Ritala, M.; Leskelä, M.; Mizohata, K.
(10) Profijt, H. B.; Potts, S. E.; Kessels, W. M. M.; Chem. Mater. 2011, 23, 247–254.
Sanden, M. C. M. Van De J. Vac. Sci. Technol. A (29) Poodt, P.; Cameron, D. C.; Dickey, E.; George, S.
2011, 29, 50801. M.; Kuznetsov, V.; Parsons, G. N.; Roozeboom, F.;
(11) Heil, S. B. S.; van Hemmen, J. L.; van de Sanden, Sundaram, G.; Vermeer, A. J. Vac. Sci. Technol. A
M. C. M.; Kessels, W. M. M. J. Appl. Phys. 2008, Vacuum, Surfaces, Film. 2012, 30, 010802.
103, 103302–. (30) Elers, K.-E. E.; Blomberg, T.; Peussa, M.;
(12) Langereis, E.; Keijmel, J.; van de Sanden, M. C. Aitchison, B.; Haukka, S.; Marcus, S. Chem. Vap.
M.; Kessels, W. M. M. Appl. Phys. Lett. 2008, 92, - Depos. 2006, 12, 13–24.
–. (31) Elam, J. W.; George, S. M. Chem. Mater. 2003, 15,
(13) Potts, S. E.; Keuning, W.; Langereis, E.; 1020–1028.
Dingemans, G.; van de Sanden, M. C. M.; Kessels, (32) Knoops, H. C. M.; Langereis, E.; van de Sanden,
W. M. M. J. Electrochem. Soc. 2010, 157, 66–74. M. C. M.; Kessels, W. M. M. J. Electrochem. Soc.
(14) Puurunen, R. L. J. Appl. Phys. 2004, 95, 4777– 2010, 157, G241.
4786. (33) Lu, J.; Elam, J. W.; Stair, P. C. Acc. Chem. Res.
(15) Aarik, J.; Aidla, A.; Kiisler, A.-A.; Uustare, T.; 2013, 46, 1806–15.
Sammelselg, V. Thin Solid Films 1999, 340, 110– (34) Mackus, A. J. M.; Verheijen, M. A.; Leick, N.; Bol,
116. A. A.; Kessels, W. M. M. Chem. Mater. 2013, 25,
(16) Delabie, A.; Swerts, J.; Elshocht, S. Van; Jung, S.- 1905–1911.
H.; Raisanen, P. I.; Givens, M. E.; Shero, E. J.; (35) Mackus, A. J. M.; Thissen, N. F. W.; Mulders, J. J.
Peeters, J.; Machkaoutsan, V.; Maes, J. W. J. L.; Trompenaars, P. H. F.; Verheijen, M. A.; Bol, A.
Electrochem. Soc. 2011, 158, D259–D263. A.; Kessels, W. M. M. J. Phys. Chem. C 2013, 117,
(17) Park, T. J.; Chung, K. J.; Kim, H.-C.; Ahn, J.; 10788–10798.
Wallace, R. M.; Kim, J. Electrochem. Solid-State (36) Weber, M. J.; Mackus, A. J. M.; Verheijen, M. A.;
Lett. 2010, 13, G65. van der Marel, C.; Kessels, W. M. M. Chem.
(18) Nyns, L. Atomic layer deposition: nucleation and Mater. 2012, 24, 2973–2977.
growth behaviour of HfO2 dielectrics on (37) Atomic Layer Deposition of Nanostructured
semiconductor surfaces, Katholieke Universiteit Materials; Pinna, N.; Knez, M., Eds.; Wiley-VCH
Leuven, 2009. Verlag GmbH & Co. KGaA: Weinheim, Germany,
(19) Kim, H. J. Vac. Sci. Technol. B 2003, 21, 2231– 2011.
2261. (38) Wu, Y.; Hermkens, P. M.; van de Loo, B. W. H.;
Knoops, H. C. M.; Potts, S. E.; Verheijen, M. A.;
28
Roozeboom, F.; Kessels, W. M. M. J. Appl. Phys.

2013, 114, 024308.
(39) Wu, Y.; Potts, S. E.; Hermkens, P. M.; Knoops, H.
C. M.; Roozeboom, F.; Kessels, W. M. M. Chem.
Mater. 2013, 25, 4619–4622.
(40) Groner, M. D.; Fabreguette, F. H.; Elam, J. W.;
George, S. M. Chem. Mater. 2004, 16, 639–645.
(41) Potts, S. E.; Profijt, H. B.; Roelofs, R.; Kessels, W.
M. M. Chem. Vap. Depos. 2013, 19, 125–133.
(42) Mackus, A. J. M.; Garcia-Alonso, D.; Knoops, H. C.
M.; Bol, A. A.; Kessels, W. M. M. Chem. Mater.
2013, 25, 1769–1774.
(43) Profijt, H. B.; van de Sanden, M. C. M.; Kessels,
W. M. M. Electrochem. Solid-State Lett. 2011, 15,
G1–G3.
(44) Kessels, W. M. M.; Knoops, H. C. M.; Weber, M.
J.; Mackus, A. J. M.; Creatore, M. Mater. Matters
2013, 8, 117–120.
(45) Heil, S. B. S.; Langereis, E.; Roozeboom, F.; van de
Sanden, M. C. M.; Kessels, W. M. M. J.
Electrochem. Soc. 2006, 153, G956–.
(46) Ritala, M.; Leskelä, M.; Dekker, J.-P.; Mutsaers,
C.; Soininen, P. J.; Skarp, J. Chem. Vap. Depos.
1999, 5, 7–9.

Knoops Handbook of Crystal Growth 2014 Pre Print

Uploaded by

Copyright:

Available Formats

Knoops Handbook of Crystal Growth 2014 Pre Print

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Knoops Handbook of Crystal Growth 2014 Pre Print

Uploaded by

Copyright:

Available Formats

What is atomic layer deposition and what are its main applications?

What is atomic layer deposition and what are its main applications?

What are the basic steps in a typical ALD process?

What are the basic steps in a typical ALD process?

Atomic Layer Deposition

Harm C.M. Knoops, Stephen E. Potts,

Eindhoven University of Technology

Photo: Area-selective atomic layer deposition research in

Please cite this chapter as:

FOR TEACHING AND SCHOLARLY USE ONLY - DO NOT DISTRIBUTE

on a workpiece surface, the substrate. This vaporized

3. Basics of ALD As mentioned before, the various steps in a typical, ideal

for flux-controlled growth in Fig. 5. The top half of Fig. 5

reactants  The precursor must be reactive towards the surface

Material Common Co-reactants

H2O is by far the most popular, and reported, oxygen

amples of this (Eq. 5.1 and 5.2): + n CH4 (g) (5.3)

TMA half-cycle (condensation, n = 1, 2): O2 plasma half-cycle:

HfCl4 half-cycle (n = 1, 2): TiCl4 half-cycle (n = 1, 2):

s-Hf(–O–)2Hf (ads) + HfCl4 (vap) → .

(e.g., 1 Torr) can be obtained to have both efficient pre-

References (20) Tiznado, H.; Zaera, F. J. Phys. Chem. B 2006, 110,

Roozeboom, F.; Kessels, W. M. M. J. Appl. Phys.

You might also like