CHAPTER 1
Overview of Chemical Vapour Deposition
ANTHONY C. JONESa
AND MICHAEL L. HITCHMANb
a
Department of Chemistry, University of Liverpool, Crown Street, Liverpool,
L69 7ZD, UK; b
Thin Film Innovations Ltd., Block 7, Kelvin Campus, West of Scotland
Science Park, Glasgow, G20 0TH, UK
1.1 Basic Deﬁnitions
In the broadest sense chemical vapour deposition (CVD) involves the formation of a thin solid ﬁlm
on a substrate material by a chemical reaction of vapour-phase precursors. It can thus be distinguished
from physical vapour deposition (PVD) processes, such as evaporation and reactive
sputtering, which involve the adsorption of atomic or molecular species on the substrate. The
chemical reactions of precursor species occur both in the gas phase and on the substrate. Reactions
can be promoted or initiated by heat (thermal CVD), higher frequency radiation such as UV
(photo-assisted CVD) or a plasma (plasma-enhanced CVD). There is a sometimes bewildering
array of acronyms covered by the overall cachet of CVD and the interested reader is referred to
several reviews.1–4
Some of the more commonly used acronyms are deﬁned below.
Metal-organic chemical vapour deposition (MOCVD) is a speciﬁc type of CVD that utilizes
metal-organic precursors. In the strictest sense a metal-organic (or organometallic) compound
contains a direct metal–carbon bond (s or p) (e.g. metal alkyls, metal carbonyls). However, the
deﬁnition of MOCVD has broadened to include precursors containing metal–oxygen bonds (e.g.
metal-alkoxides, metal-b-diketonates) or metal–nitrogen bonds (e.g. metal alkylamides), and even
metal hydrides (e.g. trimethylamine alane).
Metal-organic vapour phase epitaxy (MOVPE) or organometallic vapour phase epitaxy
(OMVPE) is an MOCVD process that produces single crystal (i.e. epitaxial) ﬁlms on single crystal
substrates from metal-organic precursors. In MOCVD and MOVPE gas-phase reactions can
sometimes play a signiﬁcant role in the deposition chemistry.
Plasma-assisted or plasma-enhanced CVD (PECVD) is a technique in which electrical energy
rather than thermal energy is used to initiate homogeneous reactions for the production of chemically
active ions and radicals that can participate in heterogeneous reactions, which, in turn, lead to
layer formation on the substrate. A major advantage of PECVD over thermal CVD processes is that
1
Chemical Vapour Deposition: Precursors, Processes and Applications
Edited by Anthony C. Jones and Michael L. Hitchman
r Royal Society of Chemistry 2009
Published by the Royal Society of Chemistry, www.rsc.org
deposition can occur at very low temperatures, even close to ambient, which allows temperaturesensitive
substrates to be used.
Atomic layer deposition (ALD), sometimes called atomic layer epitaxy (ALE), alternativelypulsed
CVD, or atomic layer chemical vapour deposition (ALCVD), is a modiﬁcation of the CVD
process in which gaseous precursors are introduced sequentially to the substrate surface and the
reactor is purged with an inert gas, or evacuated, between the precursor pulses. The chemical
reactions leading to ﬁlm deposition in ALD occur exclusively on the substrate at temperatures
below the thermal decomposition temperature of the metal-containing precursor and gas-phase
reactions are unimportant.
Chemical beam epitaxy (CBE) is high vacuum CVD technique that uses volatile metal-organic
precursors and gaseous co-precursors. The closely related technique of metal-organic molecular
beam epitaxy (MOMBE) uses volatile metal-organic precursors and co-precursor vapour derived
from the solid element. In CBE and MOMBE the chemical reactions occur only on the substrate,
leading to single crystal ﬁlms and so gas-phase reactions play no signiﬁcant role in ﬁlm growth.
Section 1.3 gives a more detailed description of these processes.
1.2 Historical Perspective
In common with many technologies, developments in CVD have largely arisen out of the
requirements of society. These developments have been most rapid when other thin ﬁlm deposition
technologies have proved problematic or inadequate, for instance in the production of multiple thin
ﬁlms, as in modern semiconductor devices, or when the coating of large surface areas is required, as
in large-scale functional coatings on glass. Several excellent reviews describe the historical development
of CVD processes,2,5,6
and the published literature from the earliest days to the mid-1960s
is covered by a comprehensive review by Powell et al.7
Therefore, this section gives only a brief
description, highlighting some key advances.
Probably the earliest patent describing a CVD process was taken out by a certain John Howarth,
for the production of ‘‘carbon black’’ for use as a pigment. Unfortunately, due to rather lax health
and safety standards, the process only succeeded in burning down the wooden plant.8
The early
electric lamp industry provided another early impetus for CVD, and a patent issued in 1880 to
Sawyer and Mann describes a process for the improvement of carbon ﬁbre ﬁlaments.9
However,
these proved too fragile and later patents describe CVD processes for the deposition of various
metals to produce more robust lamp ﬁlaments.10,11
One of the earliest examples of the CVD of metals is the deposition of tungsten, reported as
early as 1855. Wo¨ hler used WCl6 with hydrogen carrier gas to deposit tungsten metal.12
Later
in the century (1890), the famous Mond Process was developed. This describes the deposition
of pure nickel from nickel tetracarbonyl, Ni(CO)4,13,14
and was used for the reﬁnement of
nickel ore.15
The ﬁrst reports of the deposition of silicon by CVD by the hydrogen reduction of SiCl4 appear
as early as 190916
and 1927,17
and the widespread use of thin silicon ﬁlms in the electronics industry
is anticipated by the CVD of Si-based photo cells18
and rectiﬁers19
just after World War II.
During the late 1950s, triisobutylaluminium, [But
3Al] began to be used extensively to catalyze the
polymerization of oleﬁns by the Ziegler–Natta process. At around the same time, it was found that
the pyrolysis of [But
3Al] gave high purity Al metal (499 at.%). This led to its use in the early 1980s
as a CVD precursor to Al metal for very large scale integration (VLSI) applications.20,21
In patent
literature of the late 1960s, aluminium trihydride (AlH3, alane) was found to be useful for plating
Al ﬁlms from the vapour phase and by electroless deposition,22–24
which led to the much later use of
alane adducts such as [AlH3(NMe3)] as CVD precursors for high purity Al thin ﬁlms.25
The reader
is referred to Chapter 7 (Section 7.3) for recent developments in Al CVD.
2 Chapter 1
Another important development in the history of CVD was the introduction of ‘‘on-line’’ CVD
architectural coatings by Pilkington (now NSG Group). These coatings are deposited on a very large
scale by atmospheric pressure CVD on a ﬂoat glass production line.26
By applying the coating
directly to the ﬂoat glass manufacturing line, economies of scale and production are achievable that
are not possible with ‘‘oﬀ-line’’ deposition processes such as sputtering. Perhaps the most notable of
these is ﬂuorine-doped tin oxide, [SnO2:F] developed by Pilkington in the mid-1980s (‘‘Pilkington KGlass’’).
This is a low thermal-emissivity (low–E) coating on windows, which prevents heat loss from
the home and is essential to modern ecological energy saving eﬀorts (Chapter 10, Section 10.1.1). It
can be deposited using precursors such as [Me4Sn], [SnCl4] with halo-ﬂuorocarbons or HF (Chapter
10, Section 10.2.2). A much more recent commercial product of Pilkington is ‘‘self-cleaning’’ glass.
This has been coated on-line with a thin transparent ﬁlm of TiO2, and this chemically breaks down
dirt by photocatalysis in sunlight (Chapter 10, Section 10.6).
Despite the various developments in CVD described above, the major impetus to the technology
has undoubtedly been provided by the rapid development of the microelectronics industry since the
mid-1970s. This has led to a requirement for very thin high purity ﬁlms with precise control of
uniformity, composition and doping.
Thin epitaxial ﬁlms of n- or p-doped Si are the basic requirement for all Si integrated circuit
technology. One of the earliest reports of silicon epitaxy was the closed tube transport of SiI4
produced by heating solid Si in the presence of iodine.27
Epitaxial Si ﬁlms were later produced in the
1970s on a large commercial scale by the pyrolysis of monosilane (SiH4) in H2.28
Interest in the use of metal-organic compounds for CVD applications began in the early 1960s.
The ﬁrst reported preparation of a III-V material from a Group III metal-organic and a Group V
hydride was by Didchenko et al. in 1960, who prepared InP in a closed tube by the thermal
decomposition at 275–300 1C of a mixture of [Me3In] and liquid [PH3].29
Next, in 1962, Harrison and
Tomkins produced InSb in a closed tube by heating a mixture of [Me3In] and [SbH3] at 160 1C, and
they also produced GaAs by heating a mixture of [Me3Ga] and [AsH3] at 200 1C.30
In 1961 and 1965
patent applications by the Monsanto Co. claimed methods of depositing III-V compounds ‘‘suitable
for use in semiconductor devices’’.31,32
The processes involved the pyrolysis of volatile Group III and
Group V compounds in an open tube system on a cubic crystal substrate to produce epitaxial ﬁlms.
However, the Monsanto applications were of a rather general nature, listing a large range of
volatile Group III compounds, and the few speciﬁc process examples given mainly involved Group
III trihalides. In 1968, Manasevit and co-workers at the Rockwell Corporation gave the ﬁrst clear
description of the use of metal-organic compounds for the chemical vapour deposition of III-V
materials. The ﬁrst publication describes the deposition of GaAs by pyrolysis of a gas phase mixture
of [Et3Ga] and [AsH3] in an open tube system using H2 as the carrier gas.33
Manasevit named the
technique metal-organic chemical vapour deposition (MOCVD) and a patent was later ﬁled for the
MOCVD of a range of III-V materials and wide band-gap compound semiconductors.34
The emphasis in Manasevit’s early work was on growth of non-epitaxial ﬁlms on insulating
substrates such as sapphire and spinel. However, in 1969 the growth of epitaxial GaAs on a GaAs
substrate by metal-organic vapour phase epitaxy (MOVPE) was demonstrated.35
Subsequently, a
wide range of III-V compounds were deposited by MOCVD (or MOVPE), including AlGaAs,36
InP,
InAs, InGaAs,InAsP,37,38
GaN and AlN,39
although semiconductor device quality III-V materials
still had not been produced. This was due largely to low purity precursors (often obtained from
commercial suppliers of metal-organics for catalysis applications) and non-optimized MOCVD
reactors and processes. In 1975, however, high-purity device quality GaAs ﬁlms were grown40
that
had a low residual carrier concentration of n ¼ 7 Â 1013
cmÀ3
and high electron mobility (m77K ¼
120 000 cm2
VÀ1
sÀ1
) (Section 1.7.2.4). Conventional techniques for the deposition of III-V materials
such as liquid phase epitaxy (a combined melt of the components) proved incapable of producing the
very thin multilayer structures required for eﬃcient III-V devices and so MOVPE technology
developed with ever increasing pace, and state-of-the-art GaAs photocathodes and ﬁeld eﬀect
3Overview of Chemical Vapour Deposition
transistors (FETs) were soon produced,41
as well as complex multilayer structures such as AlGaAs/
GaAs/AlGaAs double heterojunction lasers.42
A particularly signiﬁcant advance in III-V technology
was the discovery of how to p-dope GaN-based semiconductors grown by MOVPE43
(Chapter 6,
Section 6.6.4.1) as this spawned the growth of a large industry in full-colour high-brightness light
emitting diodes for energy eﬃcient displays (Chapter 13, Section 13.7).
The reader is referred to Chapter 6 for a detailed description of the MOCVD of III-V compound
semiconductor materials, including (Section 6.3.1) further details on the historical development of
the technology. In many ways this is now a mature technology, more the province of salesmen and
chemical buyers than development scientists, and commercial developments in this technology are
described in Chapter 13.
The discovery of high-Tc superconducting oxides in the mid-1980s led to intense eﬀorts to prepare
these materials as thin ﬁlms. This led to the development, beginning in the late 1980s,44
of
MOCVD techniques for the deposition of oxides such as YBa2Cu3O7Àd, and other superconducting
oxides.45
The diﬃculty in transporting low volatility metal precursors was largely responsible for
the introduction of liquid injection MOCVD (Section 1.5 Figure 1.15). There has also been a great
deal of eﬀort devoted to the MOCVD of ferroelectric oxides such as Pb(Zr,Ti)O3 and SrBi2Ta2O9,
and early reports date back to the early 1990s.46
More recent advances in the MOCVD of a range of
ferroelectric oxide materials are given elsewhere47
and in Chapter 8 (Section 8.4).
The rapid recent advances in Si integrated circuit technology have largely been achieved by
aggressive shrinking or ‘‘scaling’’ of devices such as metal oxide semiconductor ﬁeld eﬀect transistors
(MOSFETS) and dynamic random access memories (DRAMs) (Chapter 8, Section 8.3, and Chapter 9,
Section 9.2.2 and Figure 9.2). This has led to a requirement for new high-permittivity (or high-k) oxide
insulating materials to replace the conventional SiO2 insulating or capacitor layers. PVD techniques
can not give the desired deposition control of the very thin ﬁlms required, or the necessary step coverage
in high aspect-ratio device structures such as trench- and stack-structured DRAMs. Therefore,
over the last few years there has been an intense eﬀort to develop CVD processes for the deposition of
high-permittivity metal oxides, such as Al2O3, ZrO2, HfO2, Zr- and Hf-silicate and the lanthanide
oxides, and many CVD developments in this area are also detailed in Chapter 8 (Section 8.3).
Shrinking device dimensions also make it necessary to modify existing multilevel metallization
technologies. This has led to recent eﬀorts to deposit Al and Cu by MOCVD (Chapter 7, Sections
7.3 and 7.4), as well as stimulating research into the MOCVD of diﬀusion barriers such as TiN and
TaN (Chapter 9, Sections 9.3.1 and 9.3.2).
Atomic layer deposition (ALD) was ﬁrst introduced by T. Suntola and co-workers in the early
1970s,48,49
and was initially used for the manufacture of luminescent and dielectric ﬁlms required in
electroluminescent displays (Chapter 4, Section 4.5.1).50,51
More recently, ALD has been used to
deposit the very thin conformal oxide ﬁlms required as gate insulators in CMOS technology and in
DRAM capacitor layers; see Chapters 4 (Section 4.5.3) and 8 (Section 8.3).
It is impossible to do justice here to the huge volume of research and development carried out on
CVD over the past 100 years or so, but hopefully this brief survey gives a ﬂavour of the great
advances made.
1.3 Chemical Vapour Deposition Processes
1.3.1 Conventional CVD Processes
CVD processes are extremely complex and involve a series of gas-phase and surface reactions. They
are often summarized, though, by overall reaction schemes, as illustrated in Scheme 1.1.
An overall reaction scheme tells us little about the physicochemical processes and the gas-phase
and surface reactions involved. A more informative illustration of a CVD process is illustrated by
4 Chapter 1
the simple schematic52
for an MOCVD reaction carried out at moderate pressures (e.g. 10–760
Torr) shown in Figure 1.1. A signiﬁcant feature of the process is the presence of a hot layer of gas
immediately above the substrate, termed the ‘‘boundary layer’’, and at these pressures gas-phase
pyrolysis reactions occurring in the layer play a signiﬁcant role in the MOCVD deposition process.
A more detailed picture of the basic physicochemical steps in an overall CVD reaction is illustrated
in Figure 1.2, which indicates several key steps4
:
1. Evaporation and transport of reagents (i.e. precursors) in the bulk gas ﬂow region into the
reactor;
2. Gas phase reactions of precursors in the reaction zone to produce reactive intermediates and
gaseous by-products;
3. Mass transport of reactants to the substrate surface;
4. Adsorption of the reactants on the substrate surface;
5. Surface diﬀusion to growth sites, nucleation and surface chemical reactions leading to ﬁlm
formation;
6. Desorption and mass transport of remaining fragments of the decomposition away from the
reaction zone.
In traditional thermal CVD, the ﬁlm growth rate is determined by several parameters, the primary
ones being the temperature of the substrate, the operating pressure of the reactor and the composition
and chemistry of the gas-phase. The dependence of ﬁlm growth rate on substrate temperature is
typiﬁed by the growth of GaAs by MOCVD using [Me3Ga] and [AsH3]53,54
(Figure 1.3).
[Me3Ga](g) + [AsH3](g)
500 700 C− °
→ GaAs(s) + 3CH4(g)
[SiH4](g)
500 800 C− °
→ Si(s) + 2H2(g)↑
[SiH4](g) + O2(g)
350 475 C− °
→ SiO2(s) + 2H2(g)↑
2[Ta(OEt)5](v) + 5H2O(v)
350 450 C− °
→ Ta2O5(s) + 10EtOH(v)
[NbCl5](v) + 5/2H2(g) + ½xN2(g)
960 C°
→ NbNx(s) + 5HCl(v)
↑
↑
↑
Scheme 1.1 Overall reaction schemes for a variety of CVD processes.
Partly pyrolysed precursor
molecules in the gas phase
Stagnant boundary layer
Surface reactions
Substrate
Figure 1.1 Simple schematic representation of the MOCVD process. (After ref. 52, Copyright John Wiley &
Sons Limited, 1992. Reproduced with permission.)
5Overview of Chemical Vapour Deposition
Figure 1.2 Precursor transport and reaction processes in CVD. (After ref. 4, p. 31, Copyright Wiley-VCH
Verlag GmbH & Co. KGaA, 1997. Reproduced with permission.)
Figure 1.3 Plot of the normalized MOCVD growth rate of GaAs as a function of growth temperature.
(After ref. 53, Copyright John Wiley & Sons Limited, 1987. Reproduced with permission.)
6 Chapter 1
In this plot, of log of growth rate vs. reciprocal thermodynamic temperature, three regions are
apparent. At lower growth temperatures the growth rate is controlled by the kinetics of chemical
reactions occurring either in the gas-phase or on the substrate surface. This region is generally
termed the region of kinetic growth control and the ﬁlm growth rate increases exponentially with
substrate temperature according to the Arrhenius equation:
Growth rate / expðEA=RTÞ ð1:1Þ
where EA is the apparent activation energy, R is the gas constant and T is the temperature. As the
ﬁlm growth rate is controlled by chemical kinetics, uniform ﬁlm thickness can be achieved by
minimizing temperature variations over the substrate surface, and this is the region utilized in hotwall
batch reactors used for the commercial production of Si epitaxial wafers by low pressure CVD.
As the temperature increases, the growth rate becomes nearly independent of temperature and is
controlled by the mass transport of reagents through the boundary layer to the growth surface, and
this is termed the region of mass transport or diﬀusion-controlled growth.
At even higher temperatures, the growth rate tends to decrease, due to an increased rate of
desorption of ﬁlm precursors or matrix elements from the growth surface and/or depletion
of reagents on the reactor walls due to parasitic gas-phase side reactions. Gas-phase reactions
become increasingly important with increasing temperature and higher partial pressures of the
reactants.
Notably, Figure 1.3 is rather misleading in that it suggests there are sharp changeovers between
the various regions. This is because of the nature of the plot where the log of the growth rate is used.
If the growth rate itself were plotted then a much more gradual transition from one rate controlling
step to the next would be seen.2
An Arrhenius type plot as in Figure 1.3 has to be used with caution
to obtain an activation energy for a kinetic process since there will be a contribution from mass
transport. The slope of the ‘‘kinetic region’’ will not give a true value of the activation energy, but a
lower value. Also, the ordinate contains the growth rate that is precursor concentration dependent
and this may vary with temperature so, again, a true value of the activation energy will not be given
from the slope of the ‘‘kinetic region’’.
A crucial factor that determines the relative importance of each regime is the pressure of the
CVD reactor. From atmospheric pressure (760 Torr) to intermediate pressures (e.g. 10 Torr)
gas phase reactions are important and, in addition, a signiﬁcant boundary layer is present. Kinetics
and mass transport can both play a signiﬁcant role in deposition. As the pressure falls gas
phase reactions tend to become less important, and particularly at pressures below 1 Torr layer
growth is often controlled by surface reactions. At very low pressures (e.g.o10À4
Torr) mass
transport is completely absent and layer growth is primarily controlled by the gas and substrate
temperature and by desorption of precursor fragments and matrix elements from the growth
surface.
A much more detailed discussion of the modelling of CVD processes is given in Chapter 3, and in
Chapter 6 there is a detailed description of the chemistry (Section 6.3.2), thermodynamics, kinetics
and hydrodynamics (Section 6.3.3) involved in III-V MOCVD technology.
1.3.2 Variants of CVD
As the brief historical overview of CVD given in Section 1.2 shows, CVD has gone through wideranging
developments over the years. Since most CVD reactions are thermodynamically endothermic
and they also have a kinetic energy of activation then energy has to be supplied to the
reacting system. Traditionally, CVD processes have been initiated by a thermal energy input, which
7Overview of Chemical Vapour Deposition
can be inputted by several methods, e.g.:
 direct resistance heating of the substrate or substrate holder;
 rf induction of the substrate holder or susceptor;
 thermal radiation heating;
 photoradiation heating.
The use of thermal CVD can, however, be disadvantageous. For example, heat input can result in
damage to temperature-sensitive substrates and so alternative forms of energy input have been
developed which allow deposition at lower temperatures.
One way of reducing growth temperatures is to use plasma-assisted or plasma-enhanced CVD
(PECVD).55
With this technique deposition can occur at very low temperatures, even close to
ambient, since electrical energy rather than thermal energy is used to initiate homogeneous reactions
for the production of chemically active ions and radicals that can participate in heterogeneous
reactions, which, in turn, lead to layer formation on the substrate. In non-thermal plasmas,
which are typically generated by electrical discharges in the gas phase, the electron temperature is
much higher than the gas temperature and inelastic collisions of the electrons with precursor
molecules form the chemically active species. In addition, surfaces in the plasma can be bombarded
with ions, electrons and photons, leading to changes in surface chemistry. Although PECVD
usually allows lower temperature deposition than thermal CVD, the plasma bombardment of a
surface often causes some substrate heating. PECVD processes have been widely used for the
deposition of a large range of materials with standard and novel properties; both inorganic and
organic materials, as well as polymers, have been prepared by the technique. There are complications,
though, with PECVD, including plasma damage of the substrate and the growing ﬁlm,
and a strong process dependency on several parameters such as rf power and frequency, gas
pressure, reagent ﬂow rate, reactor geometry, etc. Chapter 12 reviews PECVD technology and its
applications.
Another method of inputting energy to a CVD process is to use high energy photons. The process
of photo-assisted CVD56
involves interaction of light radiation with precursor molecules either in
the gas phase or on the growth surface. Precursor molecules must absorb energy, and since traditionally
simple inorganic precursors have been employed this necessitates the use of UV radiation.
If more complex molecules are used as precursors, then photosensitizing agents may need to
be added. The use of organometallic precursors (with p- and s-bonded moieties) opens up the
possibilities for a wider range of wavelengths, but this can lead to an increased potential for carbon
incorporation. Photo-assisted CVD has similar potential advantages to those of PECVD; namely,
low temperature deposition, modiﬁcations of properties of grown layers, e.g. dopant incorporation,
and independent control of substrate temperature and dissociation of precursor. In addition,
though, with masking or laser activation it is possible to achieve selected area growth. Chapter 11
considers photo-assisted CVD processes in detail.
A rather diﬀerent variation of CVD is atomic layer deposition (ALD) and the specialist version
atomic layer epitaxy (ALE).57
In this modiﬁcation of CVD, gaseous precursors are introduced
alternately to the reaction chamber, where they reach a saturated adsorption level on the substrate
surface. Introduction of the precursors is separated other by an inert gas purge, which removes any
excess precursor molecules and volatile by-products from the reaction chamber, thus preventing
unwanted gas phase reactions. In marked contrast to traditional thermal CVD, which involves
pyrolysis of precursor molecules, ALD proceeds through surface exchange reactions, such as
hydrolysis, between chemisorbed metal-containing precursor fragments and adsorbed nucleophilic
reactant molecules. A typical growth cycle in the ALD of a lanthanide oxide from a precursor
[LnL3] (e.g. L ¼ Cp, OR) occurs in the sequence H2O pulse/[LnL3] pulse/N2 purge/H2O pulse/N2
purge (Figure 1.4).
8 Chapter 1
In the ﬁrst step a pulse of H2O gives a reactive [OH]-terminated surface. A pulse of the LnL3
precursor then leads to a chemisorbed [(L)2Ln–O–] or [(L)Ln–O2–] surface species, and the liberated
LH species are removed by a N2 purge. The surface is then eﬀectively terminated with unreactive
L groups and growth self-limits. The next H2O pulse removes the remaining L groups and
regenerates a reactive [OH]-terminated surface.
Under optimum conditions, ﬁlm growth proceeds through self-limiting surface reactions of a
saturated adsorbent in which one ALD cycle produces one monolayer of material. However, due
to steric hindrance or lack of reactive surface sites, the growth rate per cycle is often considerably
less than one monolayer. This is illustrated in Figure 1.4 where, for steric reasons, the large
LnL3 molecules are unable to react with all the surface-bound OH groups. Nevertheless, the growth
rate per cycle is constant, so the thickness of the thin ﬁlm can be controlled simply and accurately
by varying the number of deposition cycles. Because ALD reactions occur exclusively on
the substrate surface, the process can give superior step-coverage (or improved conformality)
to traditional CVD, and so ALD has become the technique of choice for ﬁlm deposition on
very high-aspect ratio substrates. In ALD it is important that surface reactions predominate and
that thermal decomposition of the precursor is minimized or avoided altogether, otherwise selflimiting
growth will break down. ALD processes are, therefore, generally carried out at substrate
temperatures in the region 200–350 1C, which is below the thermal decomposition temperature
of most precursors. Chapter 4 (Section 4.2.1) gives a much more detailed discussion about
ALD processes.
Chemical beam epitaxy (CBE) is a rather specialized CVD technique. This is a high vacuum
process that uses volatile metal-organic precursors (e.g. a Group III metal alkyl) and gaseous coprecursors
(e.g. AsH3 or PH3). The closely related technique of metal-organic molecular beam
epitaxy (MOMBE) uses volatile metal-organic precursors and a co-precursor vapour derived from
the solid element (e.g. As, P).58,59
In CBE and MOMBE the chemical reactions occur only on the
substrate surface and so gas-phase reactions play no signiﬁcant role in the growth process. The use
of CBE has perhaps declined somewhat in recent years, but its aims are to combine the advantages
Figure 1.4 Schematic of an ALD growth cycle for deposition of a lanthanide oxide thin ﬁlm from a
lanthanide precursor [LnL3] (e.g. L ¼ Cp, OR) and H2O.
9Overview of Chemical Vapour Deposition
of metal-organic vapour phase epitaxy (MOVPE) with those of molecular beam epitaxy (MBE).
Since CBE is an ultrahigh vacuum (UHV) technique a potential advantage over conventional CVD
is the ability to use vacuum in situ diagnostic techniques (e.g. RHEED, AES, MBMS) that provide
real time analytical information on the growth process.
The growth kinetics of the CBE growth process are shown schematically in Figure 1.5, where it is
compared to MOVPE and MBE processes.60
Since CBE is an UHV technique, precursor desorption
from the surface limits CBE growth processes to less than about 700 1C. It is therefore
necessary to pre-pyrolyse thermally stable precursors. For example, in GaAs growth, thermally
stable [AsH3], has to be pre-decomposed to form the active surface species, As2. A consequence of
this is that there is a lack of active [AsHx] species, which remove carbon-containing fragments in
MOCVD,61
and this leads to problems of carbon contamination in CBE.
There are still many other variants of CVD, but whatever the variant it is very apparent that, for
a process to occur, starting materials, or precursors, are required, as is some form of reactor. The
next two sections give a brief tour of these topics.
Figure 1.5 Schematic representation of the growth kinetics involved in MOVPE, CBE and MBE (molecular
beam epitaxy, a PVD technique). (After ref. 60, Copyright John Wiley & Sons Limited, 1988.
Reproduced with permission.)
10 Chapter 1
1.4 CVD Precursors
1.4.1 Precursor Requirements
Whatever form a CVD process takes, the same precursor requirements generally apply. The characteristics
of an ‘‘ideal’’ CVD precursor can be summarised as follows:
 Adequate volatility to achieve acceptable growth rates at moderate evaporation temperatures.
 Stability so that decomposition does not occur during evaporation.
 A suﬃciently large temperature ‘‘window’’ between evaporation and decomposition for ﬁlm
deposition.
 High chemical purity.
 Clean decomposition without the incorporation of residual impurities.
 Good compatibility with co-precursors during the growth of complex materials.
 Long shelf-life with indeﬁnite stability under ambient conditions, i.e. unaﬀected by air or
moisture.
 Readily manufactured in high yield at low cost.
 Non-hazardous or with a low hazard risk.
Although these features are common for most CVD precursors, sometimes the precise precursor
requirements can depend on the speciﬁc nature of the CVD process. For instance, in a traditional
CVD processes it is a strong advantage if the precursor is relatively air-stable and not susceptible to
reaction with water. This is a major disadvantage, though, in ALD, as it is essential that the surface
exchange reaction with [OH] is facile, otherwise very low growth rates would result.
Another contrast between the requirements of a precursor for CVD in general and ALD is the
size of the ligands attached to the central atom. In MOCVD, it is often advantageous to use very
bulky ligands in the precursor, as this can often increase its vapour pressure, and also render it less
air/moisture sensitive. In ALD, however, the presence of bulky or sterically demanding ligands
around a precursor metal centre can lead to impractically low growth rates.
Also, a very high thermal stability (e.g. as in metal halides) can be a severe disadvantage
in MOCVD, especially in microelectronics applications where low deposition temperatures
(o500 1C) are often required. In contrast, for ALD high thermal stability can be an advantage, as
long as the precursor is reactive to H2O or surface [OH] species, as this will lead to exchange
reactions predominating over thermal decomposition reactions, as required for self-limiting
growth. For examples of the very diﬀerent requirements for MOCVD and ALD precursors
see ref. 62.
The high vacuum environment of CBE leads to the added precursor requirement that it
should ideally decompose cleanly by a unimolecular process, without incorporating excessive
amounts of carbon in the ﬁlm. This tends to lead to the use of Group II and Group III metal
alkyls with alkyl ligands containing an easily-eliminated b-hydride atom (e.g. ethyl, iso-propyl,
etc.).
Throughout this volume the characteristics of precursors will be referred to in a wide range of
contexts, and most of the requirements outlined above will become apparent. Here, though, some
particular aspects of precursor chemistry mentioned above are considered in more detail.
1.4.2 Precursor Volatility
The main requirement of any precursor is that it has an adequate volatility. With precursors
that occur naturally in a gaseous state (e.g. silane, diborane, ammonia, etc.) this is not a
problem. However, for precursors that are liquids or solids, volatility often has to be enhanced by a
reduction of the intermolecular forces that lead to dimer, oligomer or polymer formation. Good
11Overview of Chemical Vapour Deposition
examples of how this is achieved are given in Chapter 8 (Section 8.2) and Chapter 5 (Section 5.6.1)
for metal oxide precursor chemistry where neutral complexes [MLn] are designed to have monoanionic
ligands that completely ﬁll the coordination sphere of the metal.
The introduction of new chemicals into production processes can be an expensive, time consuming
exercise. Therefore, a key parameter to ensure optimum process design is reliable precursor
volatility data. A good idea of the temperature required to evaporate the precursor can be obtained
by thermogravimetric analysis (TGA), in which the weight loss of a precursor is measured as a
function of increasing temperature. Since many precursors are air sensitive they must be transferred
to the TGA apparatus in an inert atmosphere, and hence the apparatus is generally enclosed in a
nitrogen box. An example of the usefulness of TGA is illustrated by Figure 1.6, which shows the
TGA data for three diﬀerent precursors used in ALD, [Hf(NEtMe)4], [(MeCp)2HfMe2] and
[(MeCp)2Hf(OMe)Me] (Chapter 8, Section 8.3.1.2).
The TGA data show that the [Hf(NEtMe)4] complex evaporates in the temperature range B100–
225 1C, leaving a small amount of residue (B2.5%), possibly due to trace air ingress during sample
handling. The organometallic complexes [(MeCp)2Hf(OMe)Me] and [(MeCp)2HfMe2] are clearly
less volatile, evaporating in the ranges B150–300 1C and 150–320 1C, respectively. TGA data also
provide information on precursor stability, with the presence of large amounts of residue indicating
decomposition during evaporation.
However, although TGA is a useful indicator of precursor volatility and stability, it is no substitute
for accurate vapour pressure determination. It is not a trivial excersise, though, to obtain
accurate vapour pressure data, as shown by the large variation in literature values for even wellknown
precursors.63–65
During vapour pressure measurements it is essential that the precursor is
fully vapourized and that condensation of the precursor within the measurement equipment
upstream of the pressure measurement device is avoided; this can be diﬃcult, especially for lowvolatility
compounds. It is important too that the precursor is free from gaseous contaminants,
deriving from precursor decomposition or from the inert gas blanket in the precursor storage
container. It is also essential to calibrate the data for the precursor against a standard with a known
accurately determined vapour pressure (e.g. naphthalene66
). Figure 1.7 shows a schematic diagram
MeCp2HfMe2 (a)
MeCp2Hf(OMe)Me (b)
Hf (NEtMe)4 (c)
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600
Temperature (°C)
(a)
(b)
(c)
Weight(%)
Figure 1.6 Thermogravimetric analytical (TGA) data for [Hf(NEtMe)4], [(MeCp)2HfMe2] and [(MeCp)2
Hf(OMe)Me].
12 Chapter 1
and photograph of a vapour pressure measurement system used by a commercial precursor
manufacturer.66
The instrument is constructed entirely of stainless steel with high integrity joints
throughout. The key sections of the instrument are located inside an oven to prevent precursor
condensation and vapour pressure is measured using a Baratront pressure gauge attached to a
vacuum manifold.
The vapour pressure of the compound is measured at various temperatures (see Figure 1.8 for
data for [Ta(NMe2)5]) and from a plot of log10P against 1/T (in K) vapour pressures over a wide
range of temperatures can be determined. The vapour pressure of [Ta(NMe2)5] is given by Equation
Figure 1.7 Apparatus used by a commercial manufacturer to accurately measure the vapour pressure of
metalloorganic MOCVD and ALD precursors. (Reprinted from ref. 66, with permission from
Elsevier.)
13Overview of Chemical Vapour Deposition
(1.2),66
which allows the calculation of vapour pressure at any speciﬁed temperature:
Log10P ¼ À4124:9=T þ 11:265 ð1:2Þ
Vapour pressure data for several metal oxide precursors are given in Chapter 8 (Table 8.5), while
data for a selection of III-V semiconductor precursors are shown in Chapter 6 (Table 6.4).
1.4.3 Precursor Thermal Stability
As has been indicated earlier, thermal stability is also a crucial factor in determining the suitability
of a precursor for CVD applications, and it is now routine for a commercial precursor manufacturer
to carry out extensive thermal stability tests on their products. First, the precursor should
be suﬃciently stable for long-term storage at room temperature, and then it should not undergo
any appreciable decomposition at the evaporation temperatures necessary to achieve adequate gasphase
transport. For safety reasons it is also essential to ensure that rapid thermal decomposition of
a precursor does not occur by misuse or by accident, particularly when the precursor is manufactured
and used in large multi-kilogram or ton quantities. In the case of some precursors, such as
alane adducts (e.g. [AlH3(NMe)3]) – see Chapter 7 (Section 7.3.1.3) – poor storage practices can
lead to decomposition, generating large volumes of gaseous products and pressures of hundreds of
atmospheres in a sealed container.
Diﬀerential scanning calorimetry (DSC) is a common method of determining the thermal stability
of precursors. DSC measures the diﬀerence in heat ﬂux between a sample and a reference
material as a function of temperature. Both the sample and reference material are subjected to a
controlled temperature programme, and endothermal and exothermal changes in enthalpy can be
observed. The heat changes reﬂect physical and/or chemical processes, such as phase transitions or
chemical recations. Figure 1.9 shows a typical DSC curve for a CVD precursor, illustrating four
common features commonly observed in DSC experiments.67
It can be seen that DSC provides information about the temperatures at which melting and
decomposition begin (Tonset,) and when the thermal event ends (Tend). Tmax and Tmin are the
temperatures at the maximum of the exotherm and the minimum of the endotherm, respectively.
The area under the exothermic and endothermic peaks are related to the total heat of reaction.
These DSC parameters depend strongly on the rate of sample heating, with sample size being less
critical, so that a low rate of heating (B2 1C minÀ1
) is generally used. Although it is diﬃcult to
1
10
100
25 30 35 40 45 50 55 60
Temperature (C)
VapourPressure(mTorr)
Figure 1.8 Vapour pressure data for [Ta(NMe2)5]. (Reprinted from ref. 66 with permission from Elsevier.)
14 Chapter 1
derive absolute thermochemical values from DSC data, general trends in the thermal stability of
CVD precursors can be established.67
1.4.4 Precursor Purity and Precursor Analysis
In any chemical process the purity of the reactants is of paramount importance. For thin ﬁlm
deposition by CVD this is especially true. The inﬂuence of precursor purity on the properties of
the deposited ﬁlm can be illustrated by considering the doping of III-V compound semiconductors
(e.g. GaAs, InP), for which much work has been done on correlating impurity levels in the precursor
with those in MOCVD-grown ﬁlms, as described in an early review.68
The presence of volatile
trace metal impurities in the Group III metal-organic precursor (e.g. [Me3Ga], [Me3In]) leads to the
incorporation of shallow ionized impurities in the semiconductor layer. The energies of these
impurities place them in the band gap of the III-V layer, and when introduced intentionally they
lead to various sophisticated III-V devices, such as GaAs/AlGaAs/GaAs double heterojunction
lasers and p/n doped GaN-based LEDs.
However, with unintentional doping the impurities can seriously degrade the properties of a
III-V semiconductor by lowering the electron mobility through scattering,69
and decreasing the
photoluminescence by providing non-radiative recombination pathways.70
GaAs and InP
have intrinsic free charge carriers of o1013
cmÀ3
, and it has been found that as little as 1 ppm of
an unintentional dopant impurity in the layer corresponds to an impurity concentration of
2.5 Â 1016
cmÀ3
, which is unacceptable for most device applications.
Unintentional dopants are generally introduced during precursor synthesis, as is the case of
volatile metal-organic impurities in [R3Ga] or [R3In] precursors.71
For instance, [Me3Ga] can be
synthesized from the reaction between a Grignard reagent, such as [MeMgI], and [GaCl3] (Chapter
6, Section 6.5.1). Magnesium is thus a common impurity in the [Me3Ga], along with Si and Zn,
Figure 1.9 Typical DSC trace, illustrating four diﬀerent types of transition: (I) a baseline shift caused by a
glass transition; (II) an endothermal eﬀect caused by melting; (III) an endothermal eﬀect caused by
dissociation of the compound; (IV) an exothermal eﬀect caused by decomposition of the compound.
(After ref. 67, Copyright John Wiley & Sons Limited, 1997. Reproduced with permission.)
15Overview of Chemical Vapour Deposition
which originate in the Mg and/or the GaCl3. Table 1.1 shows some common dopants in III-V
materials.
As shown in Table 1.1, these impurities generally act substitutionally, occupying a lattice site
normally belonging to a Group III or V element, which has the net eﬀect of adding an electron
(n-type charge carrier) or hole (p-type charge carrier). The concentration of these impurities in the
III-V ﬁlm is often a strong function of growth temperature, for instance the level of the n-type
dopants Si and Sn tends to increase with increasing growth temperature, whilst the level of others,
such a the p-type dopants Zn, Mg and Cd, tends to decrease.
To minimize unintentional doping, highly sophisticated analytical techniques have been developed
for the determination of trace metal impurities in metal-organic precursors, which are often
highly reactive or even pyrophoric (e.g. [Me3Ga], [Me3In], and [Me3Al]). The techniques include
ﬂameless atomic absorption spectrophotometry,72
direct injection inductively coupled plasma
optical emission spectroscopy (ICPOES) and ICP-mass spectrometry (ICPMS).73
These techniques
can detect impurities at the ppb level (see examples for [Me3Al], Chapter 6, Tables 6.5 and 6.6).
Table 1.2 shows a correlation between impurities in the metal-organic precursor and the electrical
properties of a III-V layer.68
As well as trace metal impurities aﬀecting layer characteristics, carbon and oxygen contaminants
also play a crucial role in determining layer properties. Carbon impurities are inherent to
MOCVD and CBE growth processes, originating from the thermal decomposition of the organic
groups in the precursor, and the level of carbon incorporation is strongly dependent on the
molecular structure of the precursor. Oxygen may be introduced at various stages of an MOCVD
process, through, for example, volatile oxygen-containing impurities in the precursor
(e.g. [Me2AlOMe] in [Me3Al]), or from leaks in the source container or reactor inlet lines. The presence
of oxygen is a particular problem in optical devices containing Al in the active layer, as this
can seriously degrade their optical eﬃciency. Organic impurities are, perhaps, more diﬃcult to
detect and quantify than metallic impurities in precursors. Mass spectrometry is useful for
identifying hydrocarbon impurities,68
whilst Fourier- transform (FT) NMR has been used to
detect oxygen-containing organic impurities at the low ppm level (Chapter 6, Section 6.5.2 and
Table 6.7). Finally, hydrogen, present in the metal-organic precursor and in the MOCVD process
gases, can also inﬂuence semiconductor layer properties, by passivating intentional dopants such as
Mg in GaN.
Table 1.1 Common dopants in III-V semiconductors.
Impurity
elements Group Site occupied Electrical behaviour
Typical dopant
precursors
Be, Mg II A III A (Al, Ga, In) p-type [Et2Be], [Cp2Mg],
[(MeCp)2Mg]Shallow ionized
acceptor
Zn, Cd II B III A (Al, Ga, In) p-type [Me2Zn], [Et2Zn]
[Me2Cd]Shallow ionized
acceptor
C IV A V A (As, P) p-type [CCl4], [CBr4]
(CH3
d
)radicals from
[Me3M] precursor
Shallow ionized
acceptor
Si, Sn, Ge IV A III A (Al, Ga, In) n-type [SiR4], [SiH4], [R4Sn],
[R4Ge]Shallow ionized
donor
S, Se, Te VI A V A (As, P) n-type [H2S], [H2Se], [R2S],
[R2Se], [R2Te]Shallow ionized
donor
16 Chapter 1
1.4.5 Precursor Puriﬁcation Techniques
The critical eﬀect of precursor purity on layer properties, particularly in III-V MOCVD technology,
has led to an intense eﬀort to purify precursors to levels of up to 99.9999% purity (on a metal
basis); the improvement of precursor purity, especially in the 1980s, played a critical role in the
development of the compound semiconductor industry. Classical puriﬁcation techniques74
include
sublimation, recrystallization, fractional distillation (or rectiﬁcation),75
preparative chromato-
graphy76
and, for III-V precursors, distillation of the metal-organic compound from a reactive melt
such as gallium77
or sodium/potassium alloy.78
The removal of trace metal and organic impurities
from an organometallic compound can often be diﬃcult using techniques such as distillation,
especially when the contaminant has a similar boiling point to the organometallic compound (e.g.
[R4Si], [R2Zn] etc.). This led to the development some years ago of ‘‘adduct-puriﬁcation’’
techniques,
which involve the formation of an involatile (Lewis acid–Lewis base) adduct as an intermediate
in the puriﬁcation process.79,80
An example of this is the puriﬁcation of trimethylindium by
the formation of the [(Me3In)2 Á DIPHOS] adduct (DIPHOS ¼ 1,2-bis(diphenylphosphino)ethane)
(Figure 1.10).81
Volatile impurities such as diethyl ether solvent and trace metal contaminants such
as [R4Si] and [R4Sn] which do not form adducts with DIPHOS can easily be removed under
vacuum. Pure [Me3In] is then obtained by thermal dissociation of the relatively weak [Me3In]—
[DIPHOS] bond and isolation by vacuum distillation80
(Chapter 6, Section 6.5.1 and Figure 6.10).
Table 1.2 Eﬀect of precursor purity on the electrical properties of a GaAs layer.68
Group III
precursor
ICP-ES analysis
a
(ppm)
III-V material
[growth temp.
(1C)]
Electron carrier
concentration b
[Z77K (cmÀ3
)]
Electron mobilityb
[m77K (cm2
VÀ1
sÀ1
)]
[Me3Ga] Si 1.2, Zn 2.0 GaAs (650) 1.1 Â 1015
42 600
Batch 1
[Me3Ga] Sio0.03 GaAs (650) 1.4 Â 1014
137 000
Batch 2 Zno0.2
a
No other impurities detected.
b
Low Z values and high m values indicate that the layer is of high purity with few extrinsic impurities.
In
P
Figure 1.10 Crystal structure of the [(Me3In)2 Á DIPHOS] adduct used to purify [Me3In]. (Courtesy of the
Cambridge Structural Database, see ‘‘The Cambridge Structural Database; a quarter of a million
crystal structures and rising’’, F.H. Allen, Acta Crystallogr., Sect. B: Struct. Sci., 2002, 58, 380.)
17Overview of Chemical Vapour Deposition
The adduct puriﬁcation process was also applied very successfully to [Me3Ga] using adducts with
high boiling ethers.71,79
1.5 CVD Reactors
A very large range of reactors and several diﬀerent precursor delivery systems have been used in the
various areas of CVD technology. These are dealt with in detail in Chapter 2, but it may be useful
to begin with a brief and general overview of CVD reactors. Irrespective of the variations in CVD
processes, all thermal CVD reactors have common features:
 Precursor sources.
 Gas handling system to control input of precursor gases or vapours to the reaction zone.
 A reaction zone, usually within an enclosed cell, with a holder that can accommodate the
substrate, and which is heated by a surrounding oven or furnace, or by external radiofrequency
or infrared radiation.
 An exhaust system, which may include a vacuum pump for low pressure operation, to remove
waste products and a waste treatment facility with any necessary waste monitoring devices.
A schematic of a typical MOCVD reactor used to deposit a III-V material from a Group III trialkyl
compound and a Group V hydride is shown in Figure 1.11 (taken from Chapter 6, Figure 6.8).
The volatile precursors are contained in separate containers, commonly called ‘‘bubblers’’,
usually made of stainless steel (although quartz or Pyrex glass can also used). The bubbler is
equipped with inlet and outlet bellows-sealed or diaphragm-sealed valves and a dip-tube attached
to the inlet valve (Figures 1.12 and Ch. 6, Figure 6.9). Precursor transport is controlled by passing
an inert carrier gas (e.g. H2, Ar, N2) through the precursor in the bubbler. A mass ﬂow controller
upstream of the bubbler can accurately control the rate of precursor transport. Ideally, the carrier
gas should become fully saturated with precursor vapour, but this is not absolutely essential as long
as the pick-up by the carrier gas is reproducible. To facilitate this, the bubblers are held in temperature-controlled
baths. For the most careful deposition control, reactors are generally operated
in ‘‘vent-run’’ mode, in which a vent line allows the ﬂow of reactants to be stabilized before entering
the reactor. When the reactant is required the manifold very rapidly switches the reactant ﬂow into
the reactor, and this can allow the growth of complex multilayer materials.
The ﬂow into a reactor can be in a vertical or horizontal manner and schematic illustrations of
some common reactor types are shown in Figure 1.13.53
As mentioned earlier, various modes of
heating can be employed.
Precursors used for some growth processes, such as III-V and II-VI semiconductors generally have
high vapour pressures (41 Torr), at moderate source temperatures, typically o501C.82,83
However,
for other processes, such as metal oxide MOCVD, many precursors have only very low vapour
pressures of oo 1 Torr close to room temperature, so that high evaporation temperatures are needed
for eﬃcient deposition. Several strategies have been employed to deal with this situation. The
chemistry-based approach, mentioned earlier, is to design precursor molecules with reduced intermolecular
forces to overcome oligomerization. An equipment-based approach is to have not just the
precursor container heated but also the pipe-work to the reactor heated to a temperature high enough
to prevent precursor condensation; this would typically be 20–30 1C above the bubbler temperature.
Pipe-work can be heated by using co-axial tubing with a heat transfer reagent (e.g. oil) passing
through the annular region (Figure 1.14),84
or simply by wrapping heating tapes around the pipes.
Both approaches have been successfully used for the growth of several complex ferroelectric
oxides such as PbTiO3, Pb(Zr,Ti)O3 and high-Tc superconducting oxides.84,85
A problem remains,
though, in that some metal oxide precursors, such as metal alkoxides or b-diketonates, have
18 Chapter 1
Gas cabinet
exhaust
H2
hydride
precursordopant
precursor
Load
Lock
Reactor
particle trap
MOCVD reactor system
load
lock
reactor
chamber
exhaust
process
pump
PC
MFC
MFC MFC
gate valveGroup III
precursor
MO precursors
PC
pump
MFC
Pd cell purifier
N2
injection manifold
to chamber (run)
dopant
precursor
vent
effluent
system
Figure 1.11 Schematic of a typical MOCVD reactor used for the deposition of III-V semiconductor ﬁlms.
Taken from Chapter 6 (Figure 6.8).
Figure 1.12 A commercial ‘‘bubbler’’ used to evaporate metal-organic compounds for MOCVD. (Courtesy
of SAFC Hitech.)
19Overview of Chemical Vapour Deposition
insuﬃcient thermal stability to withstand heating for long periods. They can slowly decompose in
the bubbler or in the inlet pipe-work, leading to poor oxide layer uniformity and reactor blockages,
or undergo annealing and crystallization, so that carrier gas pick-up rates change with time. These
problems associated with low volatility precursors can be largely overcome by the use of liquid
injection MOCVD, in which the precursor is dissolved in an inert solvent, usually an ether (e.g.
tetrahydrofuran) or a hydrocarbon (e.g. heptane or nonane).47
The precursor solution is kept at
room temperature and when required it is delivered at a precisely controlled rate and quantity into
a heated evaporator and transported by a carrier gas into the reactor zone. For multi-component
layers, separate precursor solutions can be used for growth, or several precursors can be dissolved
Figure 1.13 Some common reactor cell geometries used in MOCVD: (a) rotating disc, gas inlet at top,
(b) horizontal reactor, (c) ‘‘T’’ reactor with rotating susceptor, (d ) reactor with inverted wafer
mounting, (e) horizontal reactor with substrate normal to gas ﬂow, (f ) multiple barrel reactor,
( g) barrel-type reactor with separate inlets, (h) inverted stagnation point ﬂow reactor,
(i ) chimney reactor, ( j ) large-scale barrel reactor. (After ref. 53, Copyright John Wiley & Sons
Limited, 1987. Reproduced with permission.)
20 Chapter 1
in an appropriate molar ratio in a single solution to form a ‘‘precursor cocktail’’. Figure 1.15 shows
a schematic of a liquid injection MOCVD system.47
A potential disadvantage of liquid injection MOCVD is that the precursor solution is evaporated
from a single heated evaporator, unlike in conventional MOCVD where each bubbler is separately
heated. In liquid injection MOCVD it is therefore often necessary to select, or design, co-precursors
that evaporate at similar temperatures.
In ALD a similar precursor delivery system to conventional CVD is generally used, but valves are
added to shut oﬀ the precursor pulse or oxygen reactant pulse, and an inert gas purge line is added
to remove materials desorbed during each ALD cycle. A full explanation of an inert gas valving
system is given in Chapter 4 (Section 4.4).
With CBE, although ﬁlm growth occurs in a UHV environment, the condensed phase metalorganic
Group III precursors and gaseous Group V precursor are again held in containers external
to the system. In this case, though, reagent ﬂow into the growth chamber is in the molecular regime,
allowing retention of the use of shutters, and in situ masks. An advantage of CBE is the use of in situ
analytical equipment such as RHEED, mentioned above, and also of residual gas analysis for the
detection of impurities. A typical CBE reactor is shown in Figure 1.16.58,59
Chapter 2 gives a highly detailed description of the large range of CVD reactors, injection systems
and ancillary equipment, together with a discussion of gas dynamics, delivery systems and
reactor cells.
Figure 1.14 Schematic of a conventional MOCVD reactor with a heated delivery line used to deposit ferroelectric
oxides [e.g. Pb(Zr,Ti)O3] using low vapour pressure precursors. (After ref. 84.
Reproduced by permission of the MRS Bulletin.)
Figure 1.15 Schematic diagram of a simple liquid injection MOCVD reactor. (After ref. 47, reproduced by
permission of The Royal Society of Chemistry.)
21Overview of Chemical Vapour Deposition
1.6 Materials Deposited by CVD and Applications
A very extensive range of materials has been deposited using conventional CVD and its variants.
These materials ﬁnd applications in the following main areas of technology:
 microelectronics
 optoelectronics
 protective and decorative coatings
 optical coatings
Table 1.3 give examples of the classes of CVD-grown ﬁlms, classiﬁed according to their chemical
nature, together with applications and typical precursors used. The table is by no means exhaustive,
but gives a good idea of the wide variety of precursors used in the various application areas of
CVD. Although some precursors are more suited to a particular CVD technology (e.g. ALD rather
than MOCVD), many of these precursors are utilized in all areas of the technology.
Detailed discussions of CVD grown materials are given in Chapters 4 and 5–12.
1.7 Materials Properties
1.7.1 Layer Morphology
Layer morphology is important in determining the physical characteristics and usefulness of a ﬁlm.
It is determined by surface diﬀusion and nucleation processes during deposition, which in turn are
Figure 1.16 Schematic diagram of a typical CBE reactor system. (After ref. 60, Copyright John Wiley &
Sons Limited, 1988. Reproduced with permission.)
22 Chapter 1
Table1.3SelectionofﬁlmsgrownbyCVD,theirapplicationsandtypicalprecursorsused.
GeneralclassofmaterialSpeciﬁcmaterialApplicationsTypicalCVDprecursors
DielectricoxidesTiO2,ZrO2,ZrSixOy,HfO2,
HfSixOy
High-kgatedielectriclayersin
CMOStechnology
Metal-b-diketonates(e.g.[M(b-dik)4]
(M¼Zr,Hf),[Ln(OR)3],[Ln(b-dik)3](Ln
¼lanthanide)+O2/O3
Ln2O3,LnSixOy,LnAlO3Dielectriccapacitorlayersin
DRAMs
Metalalkoxides,(e.g.[M(OR)4](M¼Si,Ti,
Zr,Hf),[M(OR)5](M¼Ta,Nb)+O2/O3
SiO2,Ta2O5,Nb2O5Dielectriccapacitorlayersin
DRAMs
OpticalcoatingsMetalalkylamides[M(NR2)4](M¼Ti,Zr,
Hf)+O2/O3
Organometallics(e.g.[Cp2MMe2],
[(RCp)2MMe2](M¼Zr,Hf)+O2/O3
[LaCp3],[Ln(RCp)3]+O2/O3
FerroelectricoxidesSrTiO3,(Ba,Sr)TiO3,Pb(Zr,Ti)O3,
SrBi2(TaxNb1Àx)2O9,Bi4Ti3O12,
Pb(Sc,Ta)O3,Pb(Mg,Nb)O3
DRAMs,NVFERAM,computer
memories,infrareddetectors,
microelectromechanicaldevices,
transducers,ceramiccapacitors,
Metal-b-diketonates([Ba(thd)2],[(Sr(thd)2],
[Ti(OPri
)2(thd)2],[Bi(thd)3]
[Pb(thd)2],[Sc(thd)3],[Mg(thd)2]+O2
Metalalkoxides([Bi(OR)3],[Ta(OR)5],
[Nb(OR)5])+O2
Ferrites(Ni,Zn)Fe2O4,(Mn,Zn)Fe2O4Recordingmedia,highfrequency
readheads
Metal-b-diketonates([Ni(thd)2],[Zn(thd)2],
[Mn(thd)2],[Fe(thd)3])+O2
SuperconductingoxidesYBa2Cu2O7Àx,Bi-Sr-Ca-Cu-OJosephsonjunctions,bolometers,
SQUIDS
Metal-b-diketonates([Y(thd)3],[Ba(thd)2],
[Sr(thd)2],[Cu(thd)2],[Ca(thd)2])+O2
Conductingoxides(La,Sr)CoO3,(La,Mn)O3FerroelectriccapacitorelectrodesMetal-b-diketonates+O2
RuO2,SrRuO3[Ru3(CO)12],[Ru(b-dik)3],[Ru(RCp)2]
(R¼H,Et)
Low-emissivityand
conductingoxides
F-dopedSnO2Architecturalcoatingsonﬂatglass
(solarcontrol,anti-reﬂective),dis-
playcoatingsonglass
[SnCl4]/[NH4F]/O2
Sn-dopedIn2O3[InCl3]/[SnCl4]/O2,[In(thd)3]/[Bu2Sn
(acetate)]/O2,[In(acac)3]/[Sn(acac)2]/O2
Electrochromicandphoto-
chromicoxides
WO3Architecturalcoatingsonﬂatglass
(displays)
[WCl6]/O2,[W(CO)6]/O2
MoO3[Mo(CO)6]/O2
(Continued)
23Overview of Chemical Vapour Deposition
Table1.3(continued).
GeneralclassofmaterialSpeciﬁcmaterialApplicationsTypicalCVDprecursors
ThermochromicoxidesVO2Architecturalcoatingsonﬂatglass[VCl4]/O2,[VOCl3]/O2
Self-cleaningcoatingsTiO2Architecturalcoatingsonﬂatglass[TiCl4]/[Ti(OPri
)4]/O2
GarnetsY3Fe5O12Microwaveelements,magneto-optic
recording
Metal-b-diketonates+oxygen
ElementalsemiconductorsSiMicroelectronicdevices[SiH4],[Si2H6],[SiHxCl4Àx](x¼0–3)
Doped-Si[SiH4]/[PH3]/[AsH3]/[B2H6]
Ge[GeH4]
Diamondanddiamond-likecarbonSemiconductordevices[CnH2n12]
III-VCompound
Semiconductors
GaAsSolarcells,LEDsGp.IIItrialkyls,[R3M](M¼Ga,In,Al;R
¼Me,Et)
+
GroupVtrihydrides([AsH3],[PH3],[NH3]or
organometallics([But
AsH2],[But
PH2],
[R3Sb])
GaAs/AlGaAsHeterostructurelasers,solarcells,
HEMTS,HBTs,FETs
GaPRedLED,photocathodes
GaAsP,InGaPRedLEDs
AlGaInPYellow/greenLED
GaNBlueLED
InGaNGreenLED
InGaP/AlGaInPRedlaserpointers
InPGunndiodes,radardevices
InP/InGaAsDetectorsinopticalﬁbretechnology
InGaAsPEmittersinopticalﬁbretechnology
GaSb/AlGaSbThermalimagingdevices,environ-
mentalsensors
II-VIcompound
semiconductors
ZnSBluephosphors,TFELsGp.IIdialkyls,[R2M](M¼Zn,Cd;R¼Me)
ZnSe,ZnSSe,ZnMgSSeBlue/greenLEDsandlasers+
ZnCdS,CdS/CdTeSolarcellsGpVIdihydrides,[H2Se],[H2S]or
CdTe,CdHgTeInfrareddetector,thermalimaging
systems
GpVIdialkyls,[Et2Se],[Et2S],[Et2Te],
[Pri
2Te]
MetalsAlInterconnectsinmicroelectronic
devices,metallizedpolymers,gas
[Bui
3Al],[AlH3(NR3)]
24 Chapter 1
diﬀusionbarriers,opticalcoatings,
adhesionaid
WMetallizationintransistors,inter-
connectsinICs,wearanderosion
protection
[WF6],[WCl6],/H2;[W(CO)6];W-alkyls
CuInterconnectsinICsCu(II)b-diketonates/H2;[(hfac)Cu(I)L](L
¼[PMe3],[1,5-COD],alkyne,VTMS)/H2
Au,AgMetallizationinULSIMe2Au(b-diketonate)];[Ag(b-diketonate)]
Pt,Pd,NiMetalcontactsinmicroelectronic
devices,catalysts,protectiveand
decorativecoatings
[Pt(acac)2],[PtMe3];[(1,5-COD)PtMe2];
[Pd(allyl)2];[CpPd(allyl)];[Ni(CO)4];
[NiCp2];[Ni(MeCp)2]
TiAdhesionlayers,metalfoils,corro-
sionresistantcoatings
[TiI4],[TiBr4]/H2
CrCorrosionprotection[CrF2],[CrCl2],[Cr(CO)6],[CrPh2]/H2
MoICcontactandgatemetallization,
wearresistantcoatings,infrared
reﬂector,lasermirrorcoating
[MoF5],[MoCl5],[Mo(CO)6]
RuICcontactandgatemetallization,
diﬀusionbarriers
[Ru3(CO)12],[RuCp2],[Ru(acac)2]/H2
TaCapacitors,resistors,corrosion
resistantcoatings
[TaCl5],[TaF5],[Ta(CO)5]/H2
MetalnitridesAlNSurfaceacousticwavedevices,
packagingmaterialinmicroelec-
tronicdevices
[AlBr3]/[NH3],[AlMe3/NH3];[Al(NMe2)3]/
[NH3];[AlMe3(NH3)]adduct
Si3N4Chemicalpassivation,encapsula-
tionofsiliconbipolarandMOS
devices
[SiH4]/[NH3];[Si2Cl6]/[NH3];
[Si(NMe2)4ÀnHn]
TiNWear-resistantandfrictionreducing
coatings,transparentopticalﬁlms,
energyeﬃcientwindows,dec-
orativecoatings,lowresistant
contacts,diﬀusionbarriers,and
gateelectrodesinmicroelectronic
devices
[TiCl4]/[NH3],[TiI4]/[NH3];[Ti(NR2)4]/
[NH3]
ZrN,HfNHardcoatingsformachinetools,
diﬀusionbarriersandgateelec-
trodesinmicroelectronics
[M(NR2)4]/[NH3];[M(NEt2)4]/[Me2NNH2];
[M(NR2)4/N2](plasma-assisted)
TaN,NbNDiﬀusionbarriersforCuinICs;
gateelectrodes
[TaCl5]/[NH3],[TaBr5]/[NH3];[Ta(NMe2)5]/
[NH3],[Ta(NEt2)5]/[NH3];[Ta(NEt2)3(N-
But
)]/[NH3];[Ta(NEt2)3(NCMe2Et)]/[NH3]
(Continued)
25Overview of Chemical Vapour Deposition
Table1.3(continued).
GeneralclassofmaterialSpeciﬁcmaterialApplicationsTypicalCVDprecursors
[NbCl5]/[NH3];[NbCl5]/[Me2NNH2];
[Nb(NR2)5]/[NH3],[Nb(NR2)4]/[NH3]
WNDiﬀusionbarriers,gateelectrodesin
microelectronicdevices
[WCl6]/[NH3],[WF6]/[NH3];[W(CO)6]/
[NH3];[W(NBut
)2(NR2)2]/[NH3];
[(RN)WCl4(NCR)]/[NH3];[W2(NMe2)6]/
[NH3]
MoN[MoF6]/[NH3],[Mo(CO)6]/[NH3];
[Mo(NMe2)4]/[NH3];
[Mo(NBut
)2(NMe2)2]/[NH3]
MetalcarbidesTiCHardcoatingforcutting,milling,
formingandstampingtools
[TiCl4]/[CH4];[Cp2TiCl2]/[H2]
ZrCCoatingfornuclearfuelparticles[ZrCl4],[ZrBr4]/[CH4]or[C3H11]
HfCOxidation-resistantcoatingfor
composites,coatingfor
superalloys
[HfCl4]/[CH4]
Cr7C3,Cr3C2Corrosion-andwear-resistant
coatings
[CrCl3]/n-butane;[Cr(C6H5-i-Pr)2];[CrBut
2]
WC,W2C,W3CToolcoatings,catalysts[WF6]/cyclopropane;[WCl6]/[CH4]
V-carbideWear-andcorrosion-resistant
coatings
[VCl4]/[CH4];[VCp2]
Ta,Nb-carbideProtectivecoatings[TaCl5]/[MeCl]/[H2];[NbCl5]/[CCl4]/[H2]
SiCBlueLEDs,heatsinks,protective
coatings
[MeSiCl3]/[H2]
GeCOpticalcoatings[CnH2n12]/[GeH4]
Abbreviations:b-dik¼b-diketonate,acac¼acetylacetonate,thd=tetramethylheptanedionate,hfac¼hexaﬂuoroacetylacetonate(Ch.8,Table8.2),Cp¼C5H5,RCp¼RC5H4,
1,5-COD¼1,5cyclooctadiene.
26 Chapter 1
signiﬁcantly inﬂuenced by CVD process parameters such as growth temperature, partial pressures
of gaseous species and the total pressure of the system. A detailed description of how such parameters
aﬀect the microstructure of CVD-grown layers is not yet available, but a few useful reviews
describe the current state of understanding.86–89
Three main types of morphology are brieﬂy discussed
below. Various analytical techniques are used to determine layer morphology. Scanning
electron microscopy (SEM) and atomic force microscopy (AFM) give information about crystallinity
to the sub-micrometer level, whilst crystalline orientation can be determined by X-ray diffraction
(XRD), which also allows the observations of the transition when amorphous layers, with
no features in XRD, crystallize to give clearly deﬁned X-ray diﬀraction peaks.
1.7.1.1 Epitaxial Layers
Epitaxial layers are single crystal ﬁlms grown by lattice matching the crystalline spacings of the ﬁlm
and underlying substrate. To achieve this it is important that the substrate is free from defects and
surface contamination. Also for epitaxy, low growth rates are required so that surface diﬀusion is
fast relative to the arrival of new growth initiating species on the surface. This allows adsorbed
species to diﬀuse to step growth sites and form a layer that replicates the underlying substrate.
Growth at high temperatures (typically 4700 1C) is generally required to promote epitaxy since this
increases the desorption of impurities as well as the surface mobility of adsorbed precursor species.
Growth on a substrate of the same material is called homoepitaxy and growth on a diﬀerent
material, but with very similar crystal structure, is known as heteroepitaxy. A good review on the
epitaxial growth of silicon by CVD has been published,28
and the importance of epitaxial growth of
compound semiconductors is illustrated in Chapter 6 of this book.
1.7.1.2 Amorphous Layers
The formation of amorphous ﬁlms is promoted by very high growth rates and low substrate
temperatures, where the arrival of ﬁlm precursors is much more rapid than the diﬀusion of surface
species. However, care has to be taken not to leave ﬁlms grown in the amorphous state in a hot
reactor for any length of time since this can lead to annealing and crystallization.
1.7.1.3 Polycrystalline Layers
Polycrystalline layers are often deposited at intermediate temperatures and growth rates between
those used for growth of single crystal and amorphous ﬁlms. Polycrystalline growth is facilitated by
polycrystalline surfaces where nucleation occurs at many diﬀerent surface sites, leading to growth
of islands that coalesce to form a polycrystalline layer. The control of the size and nature of the
crystallites is important in determining the properties of CVD ﬁlms.
In Figure 1.17, the SEM data for an amorphous data ﬁlm of Hf-silicate are compared to the SEM
data for a polycrystalline ﬁlm of HfO2. The columnar microstructure of the oxide ﬁlm is in marked
contrast to the featureless structure of the Hf-silicate ﬁlm. The transition from an amorphous to a
polycrystalline phase is illustrated by XRD data for a HfO2 ﬁlm deposited at various substrate
temperatures (Figure 1.18), from which it can be seen that the crystallinity of the material is a
strong function of growth temperature, with the polycrystalline monoclinic phase forming at 450 1C
and above.
1.7.2 Layer Properties
The aim of depositing a ﬁlm by CVD is to obtain a functional material with speciﬁc mechanical,
electrical, magnetic, optical or chemical properties, or a combination of several of these properties.
27Overview of Chemical Vapour Deposition
(a)
(b)
Figure 1.17 SEM data for (a) an amorphous HF-silicate ﬁlm and (b) a polycrystalline HfO2 ﬁlm.
10 20 30 40 50 60
2θ angle (degrees)
Intensity(Arb.unit)
*
550°C
500°C
450°C
400°C
(11-1)
(002)
(200)
(012) (221)
Figure 1.18 Variation of crystallinity of HfO2 grown by MOCVD with growth temperature.
28 Chapter 1
Layer properties can be strongly inﬂuenced by several ﬁlm characteristics common to all
CVD-grown layers.
1.7.2.1 Layer Thickness and Density
The thickness of layers deposited by CVD can vary over a wide range, from single atomic layers
(e.g. in ALD), through a few nanometres in, for example, multi-quantum well structures, to
hundreds of nanometres for optical and semiconductor applications, and even to greater than
100 mm for some applications such as wear resistant coatings. Layer thickness is dependent on both
the CVD technique employed and the range of deposition parameters.
Thick ﬁlms result from high growth rates often associated with high temperatures and pressures
of many thermal CVD and PECVD processes, whilst thin ﬁlms are obtained from the low growth
rates and nano-scale control that are a feature of ALD and CBE techniques. Layer thickness is
generally measured after deposition by optical techniques (e.g. interferometry and ellipsometry),
scattering techniques (e.g. Rutherford scattering and b back scattering), electromechanical proﬁlometry
or microscopy (using an optical microscope or scanning electron microscope after layer
sectioning). Simple weighing of the substrate using an accurate microbalance before and after ﬁlm
deposition can also be used to measure layer thickness, provided the density of the layer is known.
The measure of weight variation as a function of time using a quartz crystal oscillator, or by
deposition on a substrate suspended from the arm of a microbalance, allows the in situ monitoring
of layer growth.
Film density is a measure of the quality and potential functionality of the ﬁlm, with low density
ﬁlms indicative of high porosity and the incorporation of impurities in the crystal lattice, possibly
due to incomplete decomposition of the precursor, as often observed in ﬁlms grown at low temperature
(e.g. by ALD). Film density generally increases as the growth temperature increases, or by
high temperature annealing of the ﬁlm, and the refractive index of the ﬁlm can give a measure of
ﬁlm density, with dense ﬁlms having high refractive indices.
1.7.2.2 Layer Adhesion
For any practical use, it is essential that a CVD ﬁlm adheres well to the substrate. However, the
factors that inﬂuence ﬁlm adhesion are not totally understood, although it is generally accepted
that to promote good ﬁlm adhesion it is important that the substrate surface is thoroughly cleaned
before ﬁlm growth. The reason for this is that impurities on the surface can inhibit deposition of a
coating and can also signiﬁcantly aﬀect the mode and amount of nucleation of depositing material.
The measurement of adhesion is still very much based on empiricism, as it is diﬃcult to quantitatively
model the forces governing adhesion. Consequently, for example, qualitative assessment of
how well a ﬁlm is adhering to a substrate is made by the rather unsophisticated technique of the
‘‘scratch test’’90
and the even less sophisticated ‘‘Scotch tape test.’’91
The real test, though, of any
ﬁlm adhesion comes when the layer is put to a practical application. A good review of the qualities
necessary in eﬀective protective coatings is available.92
1.7.2.3 Layer Composition and Purity
Layer composition and purity are vital factors in determining the functionality of a CVD ﬁlm.
Layer composition can be measured by several techniques such as Auger electron spectroscopy
(AES), X-ray photoelectron spectroscopy (XPS), Rutherford back scattering (RBS), time-of-ﬂight
elastic recoil spectrometry (TOF-ERDA), energy or wavelength dispersive X-ray analysis (EDX or
WDX) and secondary ion or secondary neutral mass spectrometry (SIMS or SNMS). The ﬁrst ﬁve
29Overview of Chemical Vapour Deposition
of these techniques can determine bulk stoichiometry of a layer and impurities, such as carbon and
nitrogen, down to levels of about 1 at.%. SIMS is a much more sensitive technique, capable of
detecting impurities in the ﬁlm down to the ppm, or even ppb, levels, although quantitative analysis
needs careful calibration samples to be made. SNMS, on the other hand, can measure from the high
atomic % down to ppm much more readily.
The reason that it is important to be able to determine layer composition at very low levels is that
even very small traces of an impurity species can dramatically aﬀect layer characteristics. This is true
for layers that are grown for optical, mechanical, chemical and magnetic applications, but it is
particularly so for materials that are deposited with speciﬁc electrical properties in mind. This is
illustrated in the next section, which considers the electrical characteristics of compound semiconductor
layers.
A rather specialized technique used to investigate the thickness, composition and crystallinity of
thin ﬁlms is medium energy ion scattering (MEIS). This involves bombarding the ﬁlm with energetic
He1
ions (B200 keV), and then analysing the energy of the ions scattered from the ﬁlm. Owing
to the inelastic scattering that the He1
ions undergo as a function of depth below the sample
surface, the energy distribution gives an eﬀective depth proﬁle of the target atoms. MEIS also
provides valuable information on the extent of interaction between the ﬁlm and the substrate,
particularly valuable in assessing the stability of dielectric oxide ﬁlms on Si for CMOS and DRAM
applications, with ﬁlm crystallization often leading to an increased ﬁlm–substrate interaction.
The method used for determining the properties and functionality of a thin ﬁlm clearly depends
on the speciﬁc application. However, due to the increasing importance in microelectronics of thin
ﬁlms grown by CVD, some methods used for the characterization of two types of ﬁlm from this
area of technology are discussed below.
1.7.2.4 Electrical Characterization of Compound Semiconductor Films
The principal electrical parameters used to characterize III–V or II–VI compound semiconductor
materials are the carrier concentration, n (i.e. the number of charge carriers per cm3
) and the carrier
mobility, which is a measure of the ease with which the carriers can move under the inﬂuence of an
electrical ﬁeld. The mobility of the charge carriers in a semiconductor is a more complicated
function of temperature than the conductivity of a metal because, besides the temperature
dependent scattering processes found in metals the actual number of carriers (n) and their energy
distribution are temperature dependent.
The carrier concentration and carrier mobility in a semiconductor can be determined from the
conductivity of the ﬁlm, which is derived from resistivity measurements on the layer using the fourpoint
probe method93
and Hall measurements. The measured value of n represents the net value of
residual charge carriers and is thus a balance between the total concentration of ionized donor (ND)
and acceptor impurities (NA). Once the carrier mobility has been determined, it is possible to
determine the concentrations of ionized donor and acceptor impurities and the ratio NA/ND (the
compensation ratio) is often used as an indication of layer purity. For n-type material the residual
electron concentration is often expressed as n, although the term (ND – NA) is more accurate, and
residual hole density can be expressed as p or (NA – ND).
High purity compound semiconductor material will have a low residual carrier concentration
and a correspondingly high carrier mobility. For instance a good quality GaAs layer might have a
residual carrier concentration (n) of 1 Â 1014
cmÀ3
and a mobility (m) at 77 K of 130 000 cm2
VÀ1
sÀ1
,
with the material having a relatively low level of donor impurities (ND ¼ 1.4 Â 1014
cmÀ3
) and an
even lower level of acceptor impurities (NA ¼ 0.4 Â 1014
cmÀ3
). The low compensation ratio (NA/
ND) of 0.28 shows that the ﬁlm is high purity. By contrast, a low purity sample of GaAs may still
have a low residual carrier concentration (n) of, for example, 1 Â 1014
cmÀ3
, but a low electron
30 Chapter 1
mobility (e.g. 26 500 cm2
VÀ1
sÀ1
) at 77 K reﬂects the low purity of the sample. This is due to the
presence of high levels of donor and acceptor impurities (ND ¼ 2 Â 1015
cmÀ3
; NA ¼ 1.9 Â 1015
cmÀ3
), which lead to the very low charge mobility because of impurity scattering. In this case, the
compensation ratio of 0.96 is high.
The dramatic eﬀect of impurities on layer properties can be seen by relating carrier concentrations
to the level of impurity incorporated in the layer. For example, a GaAs layer with a carrier
mobility of 130 000 cm2
VÀ1
sÀ1
and a carrier concentration of 1 Â 1014
cmÀ3
has one carrier atom
per 2.167 Â 108
gallium atoms, or 4 Â 10À7
mol.% of carriers.
1.7.2.5 Electrical Characterization of Dielectric Thin Films
A detailed review covering the physics of metal-oxide semiconductors has been published.94
However, in view of the great importance of dielectric ﬁlms in microelectronics applications,95
[see,
for example, Chapters 4 (Section 4.5.3) and 8 (Section 8.3)] it may be useful to brieﬂy describe how
the ﬁlms are frequently characterized.
For dielectric thin ﬁlms, key parameters are the permittivity or dielectric constant (k) and leakage
current. For insulating gate oxides (e.g. SiO2, HfO2) deposited on a Si substrate, high frequency
capacitance–voltage (C-V) data on a metal/oxide/semiconductor capacitor (MOSC) can be used to
determine ﬁlm quality.96
Figure 1.19 shows the high-frequency C-V data for a [Al/HfO2/SiO2/p-Si]
MOSC structure, before and after annealing. The HfO2 dielectric layer was deposited by ALD at
360 1C using alternating pulses of [Hf(mmp)4] (mmp ¼ OCMe2CH2OMe) (Chapter 8, Section
8.3.1.2) and water.97
The C-V curves show features characteristic of a well-insulating dielectric material with a high
capacitance accumulation region at negative bias voltages, a depletion region of decreasing capacitance,
and a low capacitance inversion region at positive biases. The capacitance of the MOSC
structure is a series combination of metal oxide capacitance and the Si semiconductor depletion
layer capacitance. At negative bias voltages, the majority charge carriers (holes) in the p-type
Si are attracted towards the gate oxide. The holes accumulate at the HfO2/Si interface and
1.0
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
SAl = 2.04 x 10-7 m2
dHfO2
= 73 nm
f = 500 kHz
Al/HfO2/SiO2/p-Si(100)
as-deposited
at 360°C
annealed
+ 4′ at 850°C in N2
30′ at 450°C in N2/H2
Capacitance,nF
Bias voltage, V
-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5
Figure 1.19 High-frequency (500 kHz) C-V data for a [Al/HfO2/SiO2/p-Si] MOS structure fabricated by
ALD using [Hf(mmp)4] and H2O. (Reprinted with permission from ref. 97. Copyright 2003,
American Chemical Society.)
31Overview of Chemical Vapour Deposition
there is no depletion region. As a result, the measured capacitance is close to the HfO2 ﬁlm
capacitance and the eﬀective dielectric constant, or permittivity, (k) of the HfO2 ﬁlm is given by
Equation (1.3):
k ¼ Coxd=0A ð1:3Þ
where Cox ¼ capacitance value in accumulation region, d ¼ oxide layer thickness, A ¼ area of gate
contact (in this case 1.96 Â 10À7
m2
) and e0 ¼ permittivity of free space (8.854 Â 10À12
F mÀ1
).
As the gate voltage becomes more positive, the positive gate electrostatically repels the
holes away from the HfO2/Si interface. A depletion zone penetrates more deeply into the Si
semiconductor and the depletion capacitance becomes smaller, reducing the overall capacitance
of the MOSC stack. The depletion layer broadens, until there is an accumulation of minority
charge carriers (electrons) at the Si/insulating oxide interface. The accumulated minority carrier
layer is called the inversion layer, because the original polarity of the surface has now become
inverted. In equilibrium conditions, the minority carriers in the inversion layer inhibit further
penetration of the electric ﬁeld in the semiconductor, leading to the minimum capacitance values in
the C-V curve.
The amount of hysteresis in the C-V curve during a backwards and forwards voltage sweep
across the capacitor provides information about the amount of mobile ionic charge (e.g. H1
ions)
and/or slow states in the dielectric oxide ﬁlm. The ﬂatband voltage shift (DVFB) is the diﬀerence in
bias voltage measured at the ﬂat band capacitance in the depletion region in the C-V curve from the
ﬂatband voltage calculated for the ideal case (B–0.9 to À1.0 V), and is a measure of interface
quality, indicating the amount and nature of ﬁxed charge at the dielectric/Si interface. The C-V
data for the [Al/HfO2/SiO2/p-Si] structure shown in Figure 1.19 clearly show that the quality of the
HfO2 ﬁlm improves after annealing, with a marked decrease in hysteresis and a reduced ﬂatband
voltage shift, indicating a reduction in mobile ions and interface ﬁxed charge.
The C-V data for an n-type MOSC such as [Al/HfO2/SiO2/n-Si], in which the majority charge
carriers in the Si are electrons, are essentially a mirror image (in the y-axis) of the p-type MOSC
curve shape, with the accumulation region occurring at positive polarities and the inversion region
at negative polarities.
When comparing diﬀerent dielectric materials, and especially high-k dielectric oxide ﬁlms used in
gate dielectric and capacitor applications (Chapter 8, Section 8.3), the ﬁlms are often characterized
by quoting the ‘‘equivalent oxide thickness’’ (EOT), which is the theoretical thickness of SiO2 that
would provide the same capacitance density as the alternative high-k dielectric (ignoring the eﬀect
of leakage current and reliability).
Current–voltage (I-V) data obtained on MOSC structures are also frequently used to assess the
quality of an insulating dielectric ﬁlm. I-V data give a measure of the leakage of current through the
ﬁlm at a particular applied voltage and they are critically dependent on ﬁlm thickness, polarity of
applied voltage, gate electrode material and the type and nature of defects in the ﬁlm, with thin
ﬁlms generally giving higher leakage currents than thick ﬁlms of the same material. As an
approximate guide, leakage current densities for a thin ﬁlm (e.g. B5 nm) of a high quality dielectric
oxide are generally below B5 Â 10À5
A cmÀ2
. For instance, Figure 1.20 shows a plot of leakage
current density (J) against electric ﬁeld (Eox) for a series of HfO2 ﬁlms grown by liquid injection
MOCVD and ALD using [(MeCp)2HfMe(OR)] (OR ¼ OPri
, mmp).98
The data show that all ﬁlms
showed good electrical integrity (o2 Â 10À5
A cmÀ2
) at 1 MV cmÀ1
, although it must be noted that
the ﬁlms grown by ALD are much thinner (1.7 and 9.6 nm) than those grown by MOCVD (76 and
102 nm), indicating that they are better insulators. At higher electric ﬁelds (B2.5 MV cmÀ1
) the
leakage current density of three of the ﬁlms rose considerably to B1 Â 10À4
to 5 Â 10À3
A cmÀ2
,
indicating some breakdown in the integrity of these HfO2 ﬁlms.
32 Chapter 1
1.8 Postscript
Although it is impossible in a chapter of this length to give a comprehensive overview, it is hoped
that this brief description of some of the main elements of CVD gives a ﬂavour of the complex
interweaving threads associated with the technology. For more detailed descriptions of the many
various aspects of CVD processes read on!
References
1. G. B. Stringfellow, Chapter 1 in Organometallic Vapor Phase Epitaxy, Academic Press,
New York, 1989.
2. M. L. Hitchman and K. F. Jensen, in Chemical Vapor Deposition, eds. M. L. Hitchman and
K. F. Jensen, Academic Press, New York, 1989, Chapter 1.
3. T. T. Kodas and M. J. Hampden-Smith, Chapter 1 in The Chemistry of Metal CVD, VCH,
Weinheim, 1994.
4. A. C. Jones and P. O’Brien, in CVD of Compound Semiconductors, VCH, Weinheim, 1997, and
references therein.
5. R. L. Moon and Y.-M. Houng, in Chemical Vapor Deposition, Eds. M.L. Hitchman and K.F.
Jensen, Academic Press, New York, 1989, Chapter 6.
6. E. M. Sherwood and J. M. Blocher, J. Metals, 1965, 17, 594.
7. C. F. Powell, J. H. Oxley and J. M. Blocher, Vapour Deposition, The Electrochemical Society,
Pennington NJ, 1966.
8. J. F. Gallie, Petroleum Reﬁner, 1944, 23, 176.
9. W. E. Sawyer and A. Mann, US Patent 229,335 (1880).
10. J. W. Aylsworth, US Patent, 553,296 (1896).
11. A. de Lodyguine, US Patent 575,002 (1897).
Eox (MVcm-1)
2.5
J(Acm-2)
1e-10
1e-9
1e-8
1e-7
1e-6
1e-5
1e-4
1e-3
1e-2
1e-1
S832
S847
S834
S864
0.0 0.5 1.0 1.5 2.0
Figure 1.20 Leakage current density (J) versus electric ﬁeld (Eox) for HfO2 ﬁlms deposited by MOCVD
(K,J) (ﬁlm thicknesses 76 and 102 nm) and ALD (., n) (ﬁlm thicknesses 1.7 and 9.6 nm).
(After ref. 98, Copyright Wiley-VCH Verlag GmbH & Co. KGaA, 2007. Reproduced with
permission.)
33Overview of Chemical Vapour Deposition
12. F. Wo¨ hler and L. Uslar, Lieb. Ann., 1855, 94, 255.
13. L. Mond, C. Langer and F. Quincke, J. Chem. Soc., 1890, 57, 749.
14. L. Mond, U.S. Patent, 455,230, 1891.
15. R. Mond, Chem. Ind., 1930, 49T, 371.
16. J. N. Pring and W. Fielding, J. Chem. Soc., 1909, 95, 1497.
17. R. Hoelbling, Z. Angew.Chem., 1927, 40, 655.
18. G. K. Teal, J. R. Fisher and A. W. Treptow, J. Appl. Phys., 1946, 17, 879.
19. K. H. Storks and G. K. Teal, US Patent 2,441,603 (1948).
20. M. J. Cook, R. A. Heinecke, R. C. Stern and J. W. Maes, Solid State Technol., 1982, 25(12), 62.
21. M. L. Green, R. A. Levy, R. G. Nuzzo and E. Coleman, Thin Solid Films, 1984, 114, 367.
22. T. P. Whaley and V. Norman, US Patent 3,209,326 (1965).
23. D. R. Carley and J. H. Dunn, US Patent 3,375,129 (1968).
24. D. L. Schmidt and R. Hellmann, US patent 3,462,288 (1969).
25. W. L. Gladfelter, D. C. Boyd and K. F. Jensen, Chem. Mater., 1989, 1, 339.
26. Pilkington Brothers Ltd., Proceedings of The Royal Society of London Series A-Mathematical
And Physical Sciences, 1969, 314, 1.
27. R. Glang and B. W. Kippenham, IBM J. Res. Devel., 1960, 299.
28. B. S. Meyerson, in Chemical Vapor Deposition, eds. M. L. Hitchman and K. F. Jensen,
Academic Press, New York, 1989, Chapter 5.
29. R. Didchenko, J. D. Alix and R. H. Toeniskoettler, J. Inorg. Chem., 1960, 4, 35.
30. B. Harrison and E. H. Tomkins, Inorg. Chem., 1962, 1, 951.
31. Monsanto Co. US Patent Applications 113,108 and 129,919 (1961).
32. Monsanto Co. UK Patent 1,011,979 (1965).
33. H. M. Manasevit, Appl. Phys. Lett., 1968, 12, 156.
34. H. M. Manasevit, U.S. Patent, 4,368,098C, 1983.
35. P. Rhai Choudry, J. Electrochem. Soc., 1969, 116, 1745.
36. H. M. Manasevit, J. Electrochem. Soc., 1971, 118, 647.
37. H. M. Manasevit and W. I. Simpson, J. Electrochem. Soc., 1971, 118, C291.
38. H. M. Manasevit and W. I. Simpson, J. Electrochem. Soc., 1973, 120, 135.
39. H. M. Manasevit and W. I. Simpson, J. Electrochem. Soc., 1971, 118, 1864.
40. Y. Seki, K. Tanno, K. Iida and E. Ichiki, J. Electrochem. Soc., 1975, 122, 1108.
41. S. J. Bass, J. Crystal Growth, 1975, 31, 172.
42. R. D. Dupuis and P. D. Dapkus, Appl. Phys. Lett., 1977, 31, 466.
43. S. Nakamura, T. Mukai and M. Senoh, Appl. Phys. Lett., 1994, 64, 1687.
44. A. D. Berry, D. K. Gaskill, R. T. Holm, E. J. Cukauskas, R. Kaplan and R. L. Henry, Appl.
Phys. Lett., 1988, 52, 1743.
45. High Temperature Superconductors: Fundamental Properties and Novel Materials Processing,
eds. D. Christen, J. Narayan and L. Scneemeyer, Material Research Society Symposium
Proceeding, 1990, Vol. 169, and references therein.
46. G. J. M. Dormans, P. J. van Veldhoven and M. de Keijser, J Crystal Growth, 1992, 123, 537.
47. A. C. Jones, J. Mater. Chem., 2002, 12, 2576.
48. T. Suntola and J. Antson, US Patent 4,058,430 (1977).
49. T. Suntola, Mater. Sci. Rep., 1989, 4, 261.
50. T. Suntola, J. Antson, A. Pakkala and S. Lindfors, SID 80 Digest, 1980, 11, 108.
51. M. Leskela¨ and L. Niinisto¨ , Electroluminescence, Proc. 6th Int. Workshop on Electroluminescence,
Cino Puntos Press, El-Paso, 1992, pp. 249–246.
52. P. O’Brien, in Inorganic Materials, Chapter 9, eds. D.W. Bruce and D. O’Hare, Wiley, New York,
1992, p. 500.
53. M. R. Leys, Chemtronics, 1987, 2, 155.
54. R. H. Reep and S. K. Ghandi, J. Crystal Growth, 1983, 61, 499.
34 Chapter 1
55. D. W. Hess and D. B. Graves, in Chemical Vapor Deposition, Eds. M. L. Hitchman and K. F.
Jensen, Academic Press, New York, 1989, Chapter 7, and refs. therein.
56. V. R. McCrary and V. M. Donnelly, in Chemical Vapor Deposition, eds. M. L. Hitchman and
K. F. Jensen, Academic Press, New York, 1989, Chapter 8, and refs. therein.
57. M. Leskela¨ and M. Ritala, in Handbook of Thin Film Materials, ed. H.S. Nalwa, Academic
Press, 2002, Vol. 1, p. 103.
58. W. T. Tsang, J. Crystal Growth, 1990, 105, 1.
59. W. T. Tsang, J. Electron. Mater., 1986, 15, 235.
60. G. J. Davies and D. A. Andrews, Chemtronics, 1988, 3, 3.
61. T. F. Keuch and E. Veuhoﬀ, J. Crystal Growth, 1984, 68, 148.
62. A. C. Jones, H. C. Aspinall, P. R. Chalker, R. J. Potter, T. D. Manning, Y. F. Loo, R. O’Kane,
J. M. Gaskell and L. M. Smith, Chem. Vap. Deposition, 2006, 12, 83.
63. M. Fulem K. Ru˚ zicˇ a, E. Hulicius, T. Sˇ imecˇ ek, K. Melichar, J. Pangra´ c, S. A. Rushworth and
L. M. Smith, J. Crystal Growth, 2003, 248, 99.
64. C. L. Griﬃths, A. Staﬀord, S. J. C. Irvine, N. Maung, A. C. Jones, L. M. Smith and S. A.
Rushworth, Appl. Phys. Lett., 1996, 68, 194.
65. O. Kayser, Chemtronics, 1988, 3, 90.
66. S. A. Rushworth, L. M. Smith, A. J. Kingsley, R. Odedra, R. Nickson and P. Hughes,
Microelectronics and Reliability, 2005, 45, 1000.
67. S. Duﬀy, P. F. Nolan, S. A. Rushworth, A. B. Leese and A. C. Jones, Adv. Mater. Opt. and
Electron., 1997, 7, 233.
68. A. C. Jones, Chemtronics, 1989, 4, 15.
69. S. Fujita, Y. Uemoto, S. Araki, M. Imaizumi, Y. Takeda and A. Sarasaki, Jpn. J. Appl. Phys.,
1988, 27, 1151.
70. R. A. Smith, in Semiconductors (Cambridge University Press, Cambridge, 1978) Chapter 3.
71. A. C. Jones, G. Wales, P. J. Wright and P. E. Oliver, Chemtronics, 1987, 2, 83.
72. T. Nakanisi, T. Udagawa, A. Tanaka and K. Kamei, J. Crystal Growth, 1981, 55, 255.
73. A. C. Jones and P. O’Brien, Chapter 2 in CVD of Compound Semiconductors, VCH, Weinheim,
1997.
74. A. A. Efremov, V. A. Federov and E. E. Grinberg, translated from Vysokchiste Veschestra,
No. 3, May-June 1998, pp 5-43, Plenum Publishing Corp. 1989.
75. E. E. Grinberg and A. A. Efremov, Electron. Tekh. Ser. Mat., 1975, 7, 62.
76. V. G. Makarenko, V. A. Federov and A. A. Efremov, Electron. Tekh. Ser. Mat., 1974, 11, 71.
77. A. A. Efremov, E. E. Grinberg, Yu M. Fetisov, USSR Pat. 546,617, Byull. Izobrett., 1977, 6.
78. H. Kodakura, K. Sawara and T. Yako, U. K. Pat. Application. GB 2,183,651(A), (1987).
79. A. C. Jones, D. J. Cole-Hamilton, A. K. Holliday and M. Munir Ahmad, J. Chem. Soc., Dalton
Trans., 1983, 1047.
80. A. H. Moore, M. D. Scott, J. I. Davies, D. C. Bradley, M. M. Faktor and H. Chudzynska,
J. Crystal Growth, 1986, 77, 13.
81. D. C. Bradley, H. Chudzynska, M. M. Faktor, D. M. Frigo, M. B. Hursthouse, B. Hussein and
L. M. Smith, Polyhedron, 1988, 7, 1289.
82. A. C. Jones, Chem. Soc. Rev., 1997, 26, 101.
83. A. C. Jones, J Crystal Growth, 1993, 129, 728.
84. M. de Keijser and G. J. M. Dormans, Mater. Res. Soc. Bulletin, 1996, 21, 37.
85. I. M. Watson, Chem. Vap. Deposition, 1997, 3, 9.
86. J. Bloem, J. Crystal Growth, 1980, 50, 581.
87. G. H. Gilmer and J. Q. Broughton, Ann. Rev. Mater. Sci., 1986, 16, 487.
88. J. A. Venables, G. D. T. Spiller and M. Hanbrucka, Rep. Progr. Phys., 1984, 47, 399.
89. A. Madhukar and S. A. Chaisas, in CRC Crit. Rev. Soild State Mater. Sci., 1988, 14, 1.
90. H. E. Hintermann, Ann. C I R P, 1982, 31, 405.
35Overview of Chemical Vapour Deposition
91. S. Wolf and R. N. Tauber, in Silicon Processing for the VLSI era (Lattice Press, Sunnat Beach,
CA, 1986).
92. G. Wahl, in Chemical Vapor Deposition, eds. M. L. Hitchman and K. F. Jensen, Academic
Press, New York, 1989, Chapter 10, and refs. therein.
93. L. J. van der Pauw, Phys. Tech. Rev., 1958, 20, 220.
94. E. R. Nicollian and J. Brews, in MOS Physics and Technology (Wiley, 1982).
95. G. D. Wilk, R. M. Wallace and J. M. Anthony, J. Appl. Phys., 2001, 89, 5243.
96. S. Taylor, P. A. Williams, J. L. Roberts, A. C. Jones and P. R. Chalker, Electron. Lett., 2002,
38, 1285.
97. K. Kukli, M. Ritala, M. Leskela¨ , T. Sajavaara, J. Keinonen, A. C. Jones and J. L. Roberts,
Chem. Mater., 2003, 15, 1722.
98. R. O’Kane, J. Gaskell, A. C. Jones, P. R. Chalker, K. Black, M. Werner, P. Taechakumput,
S. Taylor, P. N. Heys and R. Odedra, Chem. Vap. Deposition, 2007, 13, 609.
36 Chapter 1
http://www.springer.com/978-0-85404-465-8