Bachelor Project

Czech
Technical
University
in Prague

F3 Faculty of Electrical Engineering

Department of Microelectronics

Atomic Layer Deposition

Karolína Veselá

Supervisor: doc. RNDr. Jan Voves, CSc.

Field of study: Open Electronic Systems
May 2021
ii
 Acknowledgements Declaration

I would like to thank Jan Voves for the I declare that I completed the presented
opportunity of being part of the ALD re- thesis independently and that all used
search group, for his kind guidance and ex- sources are quoted in accordance with
pert management of this thesis. I am also the Methodological instructions that cover
grateful to Alexandr Pošta for his help and the ethical principles for writing an aca-
patience in showing me around the lab- demic thesis.
oratory. Furthermore, I must thank my
amazing classmates Erik Rapp and Mar-
tin Šimák for their friendship and support In Prague, 18. May 2021
throughout my studies.
Prohlašuji, že jsem předloženou práci
vypracovala samostatně, a že jsem
uvedla veškeré použité informační zdroje
v souladu s Metodickým pokynem o do-
držování etických principů při přípravě
vysokoškolských závěrečných prací.

V Praze, 18. květen 2021

iii
 Abstract Abstrakt

Development in the microelectronic indus- Rozvoj mikroeletrotechnického průmyslu

try leads to the scaling down of semicon- vede ke zmenšování rozměrů polovodi-
ductor devices; hence, a need for thin, čových součástek, a proto vyvstává po-
perfect layers of materials arose. The ptávka po materiálech z tenkých vrstev.
application of atomic layer deposition Využití metody depozice atomárních vrs-
(ALD) has sparked a good deal of in- tev přineslo vlnu zájmu díky jejím výho-
terest due to its benefits compared to dám oproti tradičním depozičním techni-
other traditional thin film deposition tech- kám. ALD je chemická metoda depozice
niques. ALD is a chemical gas phase de- z plynné fáze založená na opakovaných
position method based on sequential, self- samo-nasycovacých povrchových reakcích,
saturating surface reactions, which grad- které postupně formují jednotlivé vrstvy
ually forms monolayers of new material. nového materiálu. Tato bakalářská práce
The thesis introduces ALD with its basic představuje ALD, její základní principy
principles and applications. It focuses on a aplikace. Dále se zaměřuje na komerční
a commercial ALD system manufactured ALD stroj vyráběný firmou SENTECH
by SENTECH Instruments GmbH and Instruments GmbH a jeho mechaniku. V
its mechanics. In the experimental part, experimentální části je proveden růst ten-
the growth of Al2 O3 and SiO2 thin films kých vrstev Al2 O3 a SiO2 na křemíkovou
is performed on Si wafers and on an alu- polovodičovou desku a na hliníkovou elek-
minum electrode. The deposited samples trodu. Vytvořené vzorky jsou zkoumány
are examined by Raman spectroscopy and Ramanovou spektroskopií a AFM.
AFM.

Klíčová slova: depozice atomárních

Keywords: atomic layer deposition, vrstev, epitaxe z plynné fáze,
vapor phase epitaxy, nanotechnology, nanotechnologie, růst tenkých vrstev
thin film growth

Překlad názvu: Depozice atomárních

Supervisor: doc. RNDr. Jan Voves, vrstev
CSc.
Faculty of Electrical Engineering,
Czech Technical University in Prague,
Technická 2,
Praha 6

iv
 4.3.3 Hardware and Software
Contents Control . . . . . . . . . . . . . . . . . . . . . . 22

1 Introduction 1 4.3.4 Real Time Monitor . . . . . . . . . 23

2 Basics of Atomic Layer 5 Experimental Part 25

Deposition 3

5.1 Preparation . . . . . . . . . . . . . . . . . . 25
2.1 A Typical ALD Cycle . . . . . . . . . . 3

5.2 Trial PEALD Growth of Thin

2.2 Characteristics and Advantages of Oxide Layers . . . . . . . . . . . . . . . . . . . 27
ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

5.3 Characterization Methods . . . . . 28

2.3 Precursors and Co-reactants . . . . 7

5.3.1 Raman Spectroscopy . . . . . . . 28

2.4 Modified ALD Techniques . . . . . . 9

5.3.2 Atomic Force Microscopy . . . 29

3 ALD Applications 11

5.4 Results and Discussion . . . . . . . . 30

3.1 FinFETs and high mobility FETs 12

5.5 Capacitor Experiment . . . . . . . . . 32

3.2 Non-volatile Memory (NVM)
Devices . . . . . . . . . . . . . . . . . . . . . . . . 13
6 Conclusions 37

4 SENTECH SI PEALD LL System 15

Bibliography 39

4.1 Brief View on ALD Market . . . . 15

Project Specification 43
4.2 SENTECH’s ALD Performance . 17

4.3 SI PEALD LL Description . . . . . 18

4.3.1 Reactor Unit . . . . . . . . . . . . . . 20

4.3.2 Precursor and Gas Cabinet . . 21

v
 4.6 Direct draw (left) and bubble draw
Figures (right) precursor containers and their
installation. [13] . . . . . . . . . . . . . . . . 21
2.1 Schematics of one ALD cycle. [3] 4
4.7 Reactor chamber and plasma
source schematics. [11] . . . . . . . . . . 22
2.2 ALD coating of trenches. [2] . . . . 5

4.8 ALD system software example.

3.1 Schematic drawing of FinFET. [4] 12 [10] . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.2 Different gate design structures 4.9 PEALD with ellipsometer and the
with the use of ALD. (a) FinFET principle of ellipsometry analysis.
transistor, (b) Omega-gate structure [10] . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
wrapping around a Ge channel, (c)
Pi-gate surrounding a Si nanowire.
4.10 Real time monitoring of TiO2
[1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
deposition. [11] . . . . . . . . . . . . . . . . . 24

3.3 (a) Schematic drawing of NC

memory and (b) TEM image of 5.1 Front view of the glovebox
HfAlO/W/HfAlO structure of NCs connected to PEALD. [2] . . . . . . . . 26
embedded in HfAlO. [4] . . . . . . . . . 14

5.2 Side view of PEALD. [2] . . . . . . . 26

4.1 Properties of thermal ALD (a)
layer thickness, (b) refractive index 5.3 (a) Renishaw’s Raman
uniformity. [10] . . . . . . . . . . . . . . . . . 17 spectrometer [22], (b) NTEGRA
atomic force microscopy device. [25] 28
4.2 Properties of PEALD (a) layer
thickness, (b) refractive index 5.4 Laser beam deflection for AFM.
uniformity. [10] . . . . . . . . . . . . . . . . . 17 [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.3 Linearity of growth. [10] . . . . . . . 18 5.5 (a) original Si wafer, (b) wafer with
Al2 O3 thin film (c) wafer with SiO2
thin film. . . . . . . . . . . . . . . . . . . . . . . 30
4.4 SI PEALD LL system. [10] . . . . 19

5.6 Raman spectrum of Al2 O3 film

4.5 SI PEALD LL Reactor unit with
measured with 633 nm laser. . . . . . 31
RTM. [11] . . . . . . . . . . . . . . . . . . . . . 20

5.7 Raman spectrum of plain Si wafer

measured with 532 nm laser. . . . . . 31

vi
5.8 AFM diagram of (a) real surface
scan of Al2 O3 and (b) flattened
Tables
diagram of the surface in 3D form. 32
2.1 Overview of the materials prepared
5.9 AFM diagram of (a) real surface by ALD. [5] . . . . . . . . . . . . . . . . . . . . . 7
scan of SiO2 and (b) flattened
diagram of the surface in 3D form. 32
5.1 PEALD dosing and waiting times 27
5.10 The border between Al and Al2 O3
with tape impurities under regular 5.2 Samples’ PEALD characteristics 33
microskope: (a) for 20 nm sample (b)
for 100 nm sample. . . . . . . . . . . . . . . 33

5.11 AFM diagram of (a) the Al2 O3

layer and (b) the border between Al
and Al2 O3 with tape impurities. . . 34

5.12 Height profile of the surface of the

20 nm capacitor. . . . . . . . . . . . . . . . . 34

5.13 Height profile of the surface of the

100 nm sample with Al2 O3 coating. 34

5.14 The four samples after the second

Al layer deposition. . . . . . . . . . . . . . 35

5.15 (a) Photo of the sample with 20

nm functional capacitor. (b) The 20
nm capacitor under Raman
microscope. . . . . . . . . . . . . . . . . . . . . 35

vii
 Chapter 1

Introduction

Atomic layer deposition (ALD) is a vapor phase technique capable of deposit-

ing a variety of nano-thin film materials. At first, ALD was not expected to
play a significant role in the electronic industry, mainly due to its low growth
rate. However, since the requirements for smaller and more sophisticated
structures have not ceased, ALD-grown thin films have slowly attracted
the attention of many technological studies. Nowadays, ALD is considered
one of the most promising techniques for nanoscale device development and
fabrication.

Thin films are an unreplaceable part of any modern technology, concerning

applications like surface coatings or nanoelectronics fabrications. Such films
must be of high electronic and structure quality and usually require vapor-
phase deposition techniques. The growth of polycrystalline and amorphous
films began in the 1980s with methods as chemical vapor deposition (CVD)
and physical vapor deposition (PVD). CVD is a technique involving chemical
reactions. Two or more volatile chemicals called precursors decompose at
the material’s surface, leaving a thin film and byproducts. The material is
usually referred to as a substrate and must be heated as the reactions are
driven thermally.

These vapor phase epitaxy methods were improved over time into, e.g.,
metalorganic CVD (MOCVD) or molecular beam epitaxy (MBE). MOCVD,
however, cannot deposit layers in the units of nanometers. MBE is a reliable
thermal evaporation process where the grown film thickness can be controlled
in real-time, yet it requires an ultra high vacuum environment, making it
operationally inefficient.

1
1. Introduction .....................................
ALD was invetned by Dr. Tuomo Suntola in 1970s in Finnland. At that
time, the method was called atomic layer epitaxy and held the revolutionary
idea of depositing crystalline materials one atomic layer at a time. It was
mainly developed for the production of flat panel displays. The method
caught interest in the late 1990s, when the deposition of metal oxides was
cultivated and used in semiconductor research; the name ALD was established
as well. Over the past years, the variety of ALD materials has significantly
expanded, including also nitrides, carbides, and even organic materials.

The unique ALD properties come from the deposition process that is based
on self-limiting chemical reactions of precursors and substrate, identifying
ALD as a self-assembly technique. ALD procedures were partially developed
from the known CVD processes, but the exposure of precursors was altered,
leading to notable decreases in needed process temperature. ALD also offers
exceptionally conformal thin film production, including thickness control at
the atomic scale. Due to these benefits, ALD has become the first choice in
many applications. It is able to fulfill the ever-growing requirements of the
microelectronics industry, which other methods fail to meet.

This thesis is intended to introduce the reader to the basics of ALD and
highlight selected applications, including high-κ gate dielectric transistors.
The second part of the thesis is dedicated to an existing ALD system from
the SENTECH company recently installed at the laboratory at FEE, CTU.
Several experiments will be introduced as well in the experimental part,
confirming the correct installation of the ALD system.

2
 Chapter 2

Basics of Atomic Layer Deposition

A thin ALD film is built up in cycles. In each cycle, a submonolayer of

material is deposited onto the surface of the substrate. In one cycle, the
substrate is exposed to several precursor chemicals, each containing different
elements of the final material.

Precursor chemicals are introduced to the reactor space separately, enabling

them to react with the surface in a self-limiting way. In other words, the
precursor molecules react with the surface chemical groups as long as these
are available. This process will eventually lead to saturation, leaving the
surface with a new uniform layer. This sequential exposure to precursors
gives ALD its unique characteristic. It prevents uncontrolled and unwanted
chemical reactions and byproducts and constructs the exceptionally conformal
material layers.

2.1 A Typical ALD Cycle

A typical ALD cycle occurs in a vacuum reactor. As described in the

schematics in Figure 2.1, one cycle consists of four steps: (1) the precursor
exposure, where the precursor, a gaseous chemical with a metal compound,
is pulsed into the reactor for a designated time, allowing it to fully react
with the surface of the substrate. Next is (2) the purge period, where an
inert carrier gas (usually N2 ) flows through the reactors and removes any
unreacted precursor or reaction by-products. This is followed by (3) exposure

3
2. Basics of Atomic Layer Deposition ...........................
to a second precursor, also referred to as co-reactant (usually oxidant). At
this time, a monolayer of the desired film is formed on the surface. The ALD
cycle is completed with the second (4) purge period.

Figure 2.1: Schematics of one ALD cycle. [3]

The most widespread example of ALD is Al2 O3 film fabrication. It is made

with Trimethylamine (TMA) as the first precursor that leaves the surface
covered with AlCH3 . As co-reactant, vapor H2 O is used. The total chemical
equation of this process is

3 1
Al (CH3 )3 + H2 O → Al2 O3 + 3CH4 . (2.1)
2 2
The precursor and co-reactant chemicals must be chosen well as they have
to react with the substrate but must not react with themselves or with the
surface molecules they create (more on precursors in Chapter 2.3). The
pulses of precursors must also be long enough to allow chemical reactions to
saturate, leaving the whole surface covered. Reactions between precursor and
co-reactant molecules are prevented by purge periods that do not leave such
components on the material’s surface or in the gas phase and are also timed.

This cycle is repeated until the desired film thickness is achieved. The
typical growth per cycle (often referred to as GPC) rate is about 1 Å and
the time range of one cycle goes from one to several seconds depending on
the ALD system and process design. Note that some films can also be grown
using three or even more precursors, creating a so-called supercycle with
more than four steps leaving the product with a wider variety of films. It is
even possible to create new artificial materials with unique features this way,
making ALD a powerful technology for nanotechnology research.

In general, the process is conducted at temperatures lower than 350 ◦ C and

requires both the substrate and the reactor to be heated. The temperature
range for which the correct ALD behavior is obtained is known as the
temperature window. Temperatures outside the window lead to a poor
growth rate and can altogether change the principle of deposition. Lower
temperatures may cause condensation of precursors on the surface. They could
also decrease the reactivity of molecules with the surface preventing reaction
saturation and lowering GPC. High temperatures also bring unwelcomed

4
........................ 2.2. Characteristics and Advantages of ALD

features such as decomposition of precursors, which leads to a CVD-like

growth or desorption of the surface material and hence lower growth. In
conclusion, the temperature window is a range where the GPC should be
only slightly or not at all temperature-dependent. That is how the saturation
of reactions together with the desirable ALD deposition is ensured.

2.2 Characteristics and Advantages of ALD

ALD’s unique characteristics are based on the self-saturating surface-controlled

film growth and the sequential process. Leading metrics are an excellent
conformality of the film, followed by precise thickness control and altogether
low deposition temperature.

Let us start with conformality, which is the main factor in choosing

ALD over CVD and similar technologies. Conformality derives from the
self-limiting adsorption of precursor. It enables highly uniform coatings over
the surface of the substrate as well as equal thickness in 3D structures as
trenches or pores. There are two main requirements to reach the even film
thickness over such structures. It is a sufficient pulse time and molecule
flux of precursors allowing them to reach all substrate areas. When such
conditions are ensured, the ALD is able to cover even deep trenches evenly,
as demonstrated in Figure 2.2, where the coating of semiconductor memory
devices is shown. When working with complex nanostructures, the degree of
saturation may differ on top and at the bottom of the substrate. Molecule
flux decreases going to the bottom and may lead to incomplete saturation.
Also, purging periods must be carefully timed to prevent CVD growth deeper
in the structure.

Figure 2.2: ALD coating of trenches. [2]

5
2. Basics of Atomic Layer Deposition ...........................
The self-limiting nature of ALD and the possibility of controlling the number
of cycles enable the film’s exact thickness and composition control.
Moreover, considering this concept of layer-by-layer deposition, thickness
control is available on the atomic level. Therefore ALD comes into play
whenever an extra-thin film is needed, e.g., for capacitor insulators, gate
oxides in MOSFETs, and many more, as described in Chapter 3. The
continuous uniform growth also provides a pinhole-free coating which is
another feature that all high-quality insulators need.

However, when controlling thickness on the atomic level, the final film
measurement is usually not a direct sum of all the monolayers of the deposited
material. For most ALD cases, the GPC is a bit smaller than the growth rate
of a single monolayer. It is caused by the steric hindrance of molecules (larger
parts of molecules hinder small molecules causing slow or no chemical reaction).
It is possible to predict the growth rate based on the steric hindrance and
the number density of molecular adsorption sites. Factors such as substrate
materials and growth temperatures also affect the result. Another unpleasant
phenomenon that must be considered is the formation of islands in the initial
growth stages. This also affects the growth rate and causes non-linear growth
of the film, preventing a smooth surface. Island formation is observed in
some material systems, with metal deposition on oxide substrate being the
frequent example.

Another characteristical advantage is the generally low temperature

of the deposition process that still maintains its quality. Nevertheless, the
chemical reactions must still be thermally controlled. Therefore heating of
the substrate, the reactor, and the precursor tubes is necessary. For some
materials, ALD can produce a relatively pure film at temperatures under
100 ◦ C. The previous example of Al2 O3 made with TMA and water is a
model low-temperature process possible at 33 ◦ C. Al2 O3 deposition was even
achieved at room temperature when replacing water with ozone, O3 . [4]
The low-temperature deposition widens the substrate choices to materials
with a low melting point, such as polymers. Plastic substrates might play
a key role in the development of flexible electronic devices, e.g., flexible
displays. Metal ALD usually requires higher growth temperature, partially
because of a lower reactivity of the chemical process, and thus growth below
100 ◦ C is rarely executed. One such rare example is a deposition of Pd at
80 ◦ C using bis(hexafluoroacetylacetonato)palladium [Pd(C5 HF6 O2 )2 ] and
molecular hydrogen. [4] To lower the temperature more significantly, the
Plasma Enhanced ALD (PEALD) was developed. This method utilizes plasma
for better activation of reactants and will be more described in Chapter 2.4.

Before concluding this chapter, one more drawback of ALD must be men-
tioned. The layer-by-layer fabrication, the pulsing and purging periods, and

6
............................. 2.3. Precursors and Co-reactants

generally a long cycle time induce slow deposition rates. Most ALD rates
range between 100-300 nm/h. [1] The reactor design and structure of the
substrate play the primary role in the deposition time. Bigger volume of the
reactor chamber and larger or more complicated shape of the substrate (e.
g., trenches) lead to longer pulsing and purging periods. In order to suppress
this drawback, a new method called Spatial ALD was developed and will be
mentioned further below.

2.3 Precursors and Co-reactants

A wide selection of materials has been grown by ALD and reported in

numerous studies. [1, 2, 3, 4] New materials and ALD processes are being
actively developed as well. The research has taken a significant leap from the
very first ALD demonstration in the mid-1970s, which used elemental zinc
and sulfur to grow ZnS. In Table 2.1, an overview of all materials prepared by
ALD that were known in 2019 is described and structurally written down into
a periodic table. The most common types of precursors nowadays are oxides
followed by nitrides, sulfides, and pure elements. The final product of ALD
can be crystalline or amorphous and varies from metals and insulators to
semiconductors. Materials deposited from three or more elements also gained
interest, their properties winning over the complexity of the ALD process
scheme (also called a supercycle). Some of the well-known compounds are
SrTiO3 or Hf Six Oy . [3]

Table 2.1: Overview of the materials prepared by ALD. [5]

7
2. Basics of Atomic Layer Deposition ...........................
As mentioned before, there is a wide selection of ALD materials. However,
it does not encompass all the materials, the main limitation being the un-
availability of effective reaction pathways. In order to avoid uncontrollable
reactions and maintain the self-limiting growth in the gas-phase process and
overall stability, precursors must fulfill several requirements:

. sufficient volatility but no decomposition at the deposition temperature

. sufficient and preferably fast reactivity with the surface sites and towards
the other precursor leading to growth saturation

. neither the reactants nor the reaction products should dissolve or damage
the substrate, the reactor, or the growing film

. preferably low-toxicity and easy handling (liquids are ideal)

Finally, the material availability (or possible synthesis) and cost must be
taken into account. Choosing the best precursor is not an easy task and
usually consists of balancing more and less desired qualities together with
economic possibilities.

The first precursor to be introduced to the reactor is usually a metal

reactant. Metal precursors can be either inorganic: elemental, halides,... or
organic: alkyls, amides, beta-diketonates,... Let us look at some benefits
and drawbacks of a few groups from this extensive collection of materials.
Halides such as HfCl4 are very reactive, have high deposition rates, low steric
hindrance, and are inexpensive. However, they have reactive by-products
that may be incorporated into the film at low deposition temperatures.
Furthermore, products such as HX (X being F, Cl, Br, I) may cause corrosion
and etching of the film. Alkyls do not suffer from this problem, are still
highly reactive, and can use H2 O as a co-reactant. They are not as well
thermally stable, and their decomposition at higher temperatures is limited.
Despite that, they are beneficial for the deposition of oxide materials, the
previous example of Al2 O3 being probably the most studied ALD precursor
of all time. Another typical alkyl example is zinc and ZnO thin film. [2]
Similar properties also have cyclopentadienyls. Even though they are available
for only a small number of materials, they can be used for the production
of integrated circuits, e.g., the production of HfO2 by using (C5 H5 )2 HfCl2 .
In general, for IC applications like HfO2 or ZrO2 , metal precursors from
halides and nitrates are more conventional. [4] The last group of precursors
in this shortlist are β-diketonates. They are known for their volatility and
good thermal stability and can be utilized for a wide range of elements.
Unfortunately, their reactivity and growth rate are not very high, and they
demand more reactive co-reactants as O3 .

8
.............................. 2.4. Modified ALD Techniques

The choice of the second precursor/co-reactant is also important and

mirrors the features of the precursors above. Co-reactants are usually small
molecules with good volatility and reactivity at a given temperature, adding
a second component to the final film. The broadest range of co-reactants is
related to metal oxides. The most commonly used ones are H2 O, O3 , O2 ,
N2 O4 , H2 O2 , and O− from a plasma source. Out of this group, O3 and O−
are the most reactive, enabling lower deposition temperatures. However, they
are more prone to oxidize the substrate surface and creating an unwanted
material layer. H2 O is more gentle towards the surface and is, despite its
lower reactivity, most frequently applied. The deposition of pure metals is
made possible mainly by H2 , though at a higher temperature. Zinc vapor,
O2 , and more generally reducing agents were also successfully applied. [2, 1]
This list of co-reactants shall be concluded with nitrides deposition. For a
clean surface reaction, the deposition requires a nitride to be both the first
precursor and co-reactant. In this way, NH3 compounds must be used twice
in the process to grow e. g. TiN or Ta3 N5 thin film. [2]

2.4 Modified ALD Techniques

To fix some of the ALD shortcomings, various techniques were added to the

basic ALD process. One of these modifications mentioned earlier in the text
is the Plasma-enhanced ALD (PEALD). This technique creates highly
reactive co-reactants from a plasma source, the leading example being O2
with H2 and NH3 in tow. The main advantage of PEALD is the capability
to reduce the temperature range of a general ALD process and thus widely
broaden the selection of precursors. However, the coating quality of vertical
edges is lowered, and sometimes only one side of the substrate can be coated
via PEALD.

Plasma-enhanced ALD can be sorted under a general modification called

Energy-enhanced ALD, which also includes ozone-based ALD. [6] The
main goal of this method is to obtain more reactivity by using species with
short lifetimes. Such precursors must be produced in situ by applying a
certain form of energy (electrical discharge, thermal cracking). Thanks to the
short lifetime of reactants, long purge periods are not necessary. However,
the precursors must be placed closer to the reactor chamber.

In Chapter 2.2, the disadvantage of slow deposition rates was mentioned,

especially for the coating of complex 3D structures. To improve this drawback,
Spatial ALD was introduced. Spatial ALD does not separate the steps of
an ALD cycle in the time domain but instead in space. A spatially resolved

9
2. Basics of Atomic Layer Deposition ...........................
head is placed into the reactor and creates several reaction zones. Exposure
to precursors can occur in different areas that are separated by purging gasses.
Such deposition can happen when the head moves around the substrate, or
the substrate moves past stationary precursor nozzles. Overall such spatial
techniques increase the deposition rate to 3600 nm/h.

Another interesting modification is Area-selective ALD which, as the

name suggests, is a deposition of a film at specific locations of the substrate.
This selective process needs to begin with either deactivation of the unwanted
area or activation of the area where the film will be deposited. Photolithog-
raphy or etching are used for this and are followed up by the precursor
adsorption on the chosen growth area. It is, however, quite challenging to
obtain high selectivity due to imperfect growth initiation with minor defects
and impurities. [7]

10
 Chapter 3

ALD Applications

The main applications of ALD lie in the microelectronics industry. From

semiconductors and memory devices to solar cells and energy storage, every-
thing takes advantage of the conformal, pinhole-free, thickness-controlling
deposition. The very first breakthrough in ALD manufacturing dates back
to the mid-1980s. It was not surprisingly in Finland, where the production
of TFEL displays begun and continued for over 20 years. [2] The research
on ALD utilization in semiconductors development started in the 1990s as a
potential new method that could possibly guarantee Moore’s law conservation
in the future. Especially when the industry started focusing on the use of
high-κ dielectrics as a transistor gate oxide, the remarkable traits of ALD
increased the technology demand. In 2007, Intel introduced ALD into their
mass production in order to reduce transistor gate oxide thickness. [1]

Before some of the more specific microelectronics applications will be

introduced, it must be mentioned that the ALD also reaches beyond electronic
devices. The silver jewelry industry uses ALD coatings as a way of preventing
the tarnishing of silver products. The coatings also have anti-corrosion and
even decorative purposes on collector coins and in watchmaking. An ALD
bioactive layer found a role in medical implants like dental joints or surgical
fixators. [15]

11
3. ALD Applications ...................................
3.1 FinFETs and high mobility FETs

The reduction of the gate oxide thickness in transistor development, although

desirable for down-scaling of the devices, brought new performance limitations.
Hence a search for more innovative alternatives to the traditional transistor
concept began. The planar structure was changed in order to increase the
transistor channel area covered by the gate oxide. Instead of a planar form,
the semiconductor channel partially protrudes from the bulk crystal, forming
a fin that is rising over the rest of the surface, as shown in Figure 3.1.
This irregular structure naturally needed to be covered with a gate oxide of
excellent conformity and without pinholes. This task appeared to be made
for ALD and definitely owes its wide production to the existence of such a
technique.

Figure 3.1: Schematic drawing of FinFET. [4]

Many more new generations of FinFETs are being researched. The main
idea is to spread the gate oxide to encompass even more of the semiconductor
surface. One variant is modeling the gate oxide into an omega shape, under-
cutting the gate a bit on the fourths side. Another version is called a Pi-gate,
which covers the fin even below the surface.

Figure 3.2: Different gate design structures with the use of ALD. (a) FinFET
transistor, (b) Omega-gate structure wrapping around a Ge channel, (c) Pi-gate
surrounding a Si nanowire. [1]

12
......................... 3.2. Non-volatile Memory (NVM) Devices

Another approach to balance the down-scaling limitations with the perfor-

mance efficiency is to use high mobility channel materials instead of silicon.
The increase in the electron and hole mobility leads to an increase in the
devices’ speed and power reduction; the saturation current is also improved.
A good example of such material is III-V compound semiconductors with
GaAs taking the lead with excellent electron mobility. On the other side,
Ge is a material with outstanding hole mobility. These materials offer ex-
citing possibilities, however, a high-κ dielectric is required to adapt to these
materials, and that is the challenging part of the process.

Successful deposition of Al2 O3 and HfO2 for Ge was reported. [1, 4] A

critical part of the coating is a good passivation layer on top of the Ge prior
to the ALD oxide growth. The interface is rid of the native surface oxides,
and the electrical characteristics of the gate oxide are visibly improved. The
fabrication of GaAs MOSFET also proved to be difficult, mainly due to the
poor electrical properties of the GaAs-oxide transition. It was a TMA and
water deposition of Al2 O3 as a gate oxide that brought good quality results
at last. Another ALD material, HfO2 from HfCl4 and water, was successfully
applied as a gate oxide to the GaAs device. [8] The study of ALD high-κ
dielectrics continues for a GaAs channel, still electrical properties comparable
to Si MOSFET are yet to be determined.

3.2 Non-volatile Memory (NVM) Devices

One of the more traditional ALD applications after gate oxides is in the
field of dynamic random access memory (DRAM) devices. [3] The main
development trends changed over time from down-scaling to low operating
voltage, high speed, and non-volatility. The DRAM might be replaced by
new memory devices, including nanocrystal (NC) memory, resistive switching
memory (ReRAm), and phase-change memory (PRAM).

The NC memory is the most promising candidate out of the three mentioned
variants. The nanocrystal memory is more scalable and operates under lower
voltage than conventional memory devices. For a correct implementation,
high interface quality and good tunneling oxide are expected. Again, high-κ
materials are preferred over SiO2 , leading to improving charge loss properties
due to a smaller electron barrier height. ALD can be used for a fine and
reliable deposition of the tunneling oxide.

13
3. ALD Applications ...................................

Figure 3.3: (a) Schematic drawing of NC memory and (b) TEM image of
HfAlO/W/HfAlO structure of NCs embedded in HfAlO. [4]

One of the first ALD depositions used HfALO grown by HfCl4 and H2 O
as a tunneling oxide for the NC memory. [4] More variants were explored,
generally leading to good performance with a data retention window of 0.7 V
for 10 years. NC with a large memory window over 3 V were obtained for
ALD deposition of ZrO2 as well. These ALD grown oxides were compared
with the SiO2 , showing higher charge tunneling probabilities and thus higher
stored charge density. [9]

The future of ALD shows promise for applications in emerging electronics

where the utilization of ALD is swiftly expanding. It plays a key role in the fab-
rication of lower dimensionality semiconductors with unique shapes like tubes
or wires and structures like graphene or carbon nanotubes. The ALD oxide
coating does not significantly disrupt even sophisticated materials properties.
The precise thickness control on nano-scale and the excellent uniformity,
together with other characteristics, secure ALD technology as a critical tool
for nanotechnology and microelectronics research and development.

14
 Chapter 4

SENTECH SI PEALD LL System

SENTECH Instruments GmbH is a German private company located in Berlin

that engages in business fields like Thin Film Metrology, Plasma Process Tech-
nology, and Atomic Layer Deposition. [10] The company manufactures a wide
range of respective systems. Leading the low temperature deposition branch
is Inductively coupled plasma enhanced chemical deposition (ICPECVD) and
Plasma-enhanced atomic layer deposition (PEALD). The ALD machine that
will be presented in this chapter and later used in the experimental part is
SENTECHS’s SI ALD LL.

4.1 Brief View on ALD Market

The ALD systems are developed and manufactured by companies worldwide

with perhaps a higher density in the USA. Two ALD systems commonly
mentioned in ALD experimental articles are the PICOSUN ALD system
and Veeco (Cambridge Nanotech) Fiji series. PICOSUN is a company from
Finnland, the origin land of ALD, with a solid and historical position on the
market providing ALD systems for both research and mass production. The
American Fiji ALD is mostly research-oriented, similarly to Sentech, which is
designed for research and small-scale production.

All three companies offer a plasma-enhanced variant of their ALD systems.

The temperature range of the substrate processing is also relatively similar.
SENTECH offers a 50 - 400 ◦ C range with an optional variant up to 600 ◦ C.

15
4. SENTECH SI PEALD LL System ............................
[12] PICOSUN can perform between 50 - 500 ◦ C (650 ◦ C optional). [17] And
Fiji (with no lower boundary presented) is able to go from standard 500 ◦ C up
to 800 ◦ C if desired. [18] Regarding the number of precursors, the differences
are slightly bigger. Picosun has 6 separate precursor inlets enabling up to 12
different precursors sources. Fiji offers 4 - 6 separate inlets while SENTECH,
though also capable of having 6 different precursors, presents only 4 inlets
(but has up to 7 gas lines). These numbers are not as much evidence of the
machine quality but more of a way to more widely cover the market’s demand.
All of the systems are able to work with liquid, solid, and gas precursors as
well as with ozone.

A more vital factor is, however, deposition uniformity. The wide variety of
ALD processes was in this thesis already stressed enough that it must be clear
to the reader now that uniformity of deposition alters for different materials
and dosing times. Yet, it seems that a standard metric is used, and it is the 1σ
uniformity of Al2 O3 deposition. PICOSUN presses down thermal deposition
non-uniformities to only 0.13 %, [15] SENTECH closely follows with 0.15
% thermal and 0.58 % plasma deposition. [10] Fiji performs both thermal
and plasma deposition with a defect of 1.5 %. [18] SENTECH’s performance
will be more thoroughly described in the next section. Let it be noted for
now that it can clearly keep up with the top-of-the-market ALD systems.
Before a price-wise point of view, the comparison will be finished with a slight
mention of additional options that companies offer with their systems. Both
SENTECH and Fiji present themselves as suitable candidates for research
applications. It is not surprising then that both of them offer a variety of
deposition observation and measuring techniques, namely Ellipsometry, Mass
Spectrometer, or Quartz Crystal Microbalance, which can usually confirm the
quality of the sample in-situ. PICOSUN (also containing in-situ analytics)
focuses more on the up-to-date coating of complex 3D structures [16] and
offers a diffusion enhancer, improving the deposition of deep trenches or
porous structures.

For a price comparison, public data from several Czech universities were
used. PICOSUN ALD system is steadily more expensive than others; the
basic version costs around 12 million CZK. The Veeco Fiji system ranges
between 6 (basic) - 11 million (advanced) CZK, only slightly higher or equal to
SENTECH. The prices elevate with additional options like plasma equipment,
heaters, in-situ analytics, or glovebox. In the end, SENTECH was chosen by
the CTU’s Faculty of Electrical Engineering due to a good experience between
the German company and another company overseeing the construction of
the new laboratory, where the ALD was supposed to operate. The total sum
was around 10.5 million CZK, including all components of the ALD system.

16
............................ 4.2. SENTECH’s ALD Performance

4.2 SENTECH’s ALD Performance

Before a more thorough description of the machine, its hardware and soft-
ware, system performance will be shown in several figures. It will again be
demonstrated on Al2 O3 films deposited on 200 mm wafer from TMA and
H2 O by thermal ALD and TMA and O2 plasma by PEALD. Beginning with
thermal deposition, Figure 4.1 (a) shows a graph of layer thickness quality.
The grown layer was 49.9 nm thick and reported only 1.2 % error. Figure 4.1
(b) examines the film’s uniformity using refractive index statistics and leads to
the previously mentioned 0.15 % non-uniformity. The substrate temperature
was 200 ◦ C, and the growth rate was 0.8 Å/cycle with a duration of 2 s/cycle.

Figure 4.1: Properties of thermal ALD (a) layer thickness, (b) refractive index
uniformity. [10]

Plasma-enhanced growth is captured in Figure 4.2. Graph (a) on the left

reports again layer thickness quality. The grown layer was 26.8 nm thick
and reported a slightly higher error of 1.6 %. Figure 4.2 (b) shows that
non-uniformity has also increased in this case to 0.58 %. The substrate
temperature was 200 ◦ C, and the growth rate was 1.1 Å/cycle with a duration
of 10.5 s/cycle. Although PEALD might seem weaker now due to this lower
performance, its possibilities for broader film and precursor selection still
prevails.

Figure 4.2: Properties of PEALD (a) layer thickness, (b) refractive index
uniformity. [10]

The film thickness of both thermal and plasma-enhanced processes shows a

linear dependence on the number of cycles, as seen in Figure 4.3. It can also

17
4. SENTECH SI PEALD LL System ............................
be observed that PEALD indeed has a higher growth rate over thermal ALD.
Overall, SENTECH presents its ALD as a flexible system build for a wide
range of processes with very good uniformity and conformality that can be
further optimized by in-situ analysis.

Figure 4.3: Linearity of growth. [10]

4.3 SI PEALD LL Description

As mentioned before, SENTECH’s ALD is designed for research, use in

universities, and small-scale production. Hence it can provide a wide range
of deposition modes using flexible system architecture. The deposition of
oxides, nitrides, metals, and other materials is possible by thermal and plasma-
enhanced processes. The plasma source is included in the upgraded PEALD
model, and the system can be additionally expanded with in-situ monitoring
called Real Time Monitor, with additional precursor lines and other options.
The system can be separated into the following modules:

. Reactor unit containing reactor, load lock, and electronics

. Gas and precursor cabinet and supply system
. Computer with SENTECH software
. Real Time Monitor
. Pumping system
. Mains connection box

18
.............................. 4.3. SI PEALD LL Description

Figure 4.4: SI PEALD LL system. [10]

In the standard deposition process, all chambers are under vacuum. This
vacuum is maintained utilizing foreline and turbo molecular load lock pumps.
The function of the pumping system is vital in the handling of the wafer,
which is conducted between pumped chambres, and for removal of possible
dangerous gasses. For better integration of the sample into the ALD loading
chamber, an external glovebox can be used. The wafer handling needs
to be done manually before using the automatic load lock. The glovebox
workstation filled with nitrogen ensures that this process is done in an inert
atmosphere and later protects the newly growth film.

The mains connection box contains a power supply unit that distributes
electrical power within the ALD system. The front panel of this box also
includes the main switch button, other buttons and light indicators, and
coded sockets for components, e. g. pumps. The rest of the modules will
be introduced in separate sections as they deserve more space for a proper
description.

19
4. SENTECH SI PEALD LL System ............................
4.3.1 Reactor Unit

The heart of the reactor unit is an inner cylindrical process chamber. This
chamber is made from a monolithic, seamless ingot, ensuring low leakage.
The chamber is built out of aluminum AlMgSi and has additional flanges:
upper flange for plasma source, side flange for load lock, optional flanges for
in-situ monitoring, and bottom flange for the vacuum system. A substrate
electrode is located in this cell. The stainless steel electrode is equipped with
an integrated heater and thermocouple sensor. The main goal of the electrode
is the heating the substrate material (also called a wafer). This wafer can
have a diameter of up to 200 mm and is carried on a substrate plate with 220
mm diameter. The electrode can heat the material from anywhere between
50 to 400 ◦ C (500 ◦ C optional). Reactor wall temperature can be set up to
150 ◦ C. However, with the additional isolation for plasma source and with
Real Time Monitor, the maximum temperature drops to 100 ◦ C.

Single wafers are loaded into the reactor via a vacuum load lock. The load
lock is made from the same material as the reaction chamber, has a transparent
lid and a pick-and-place mechanism which enables clean and careful handling
of the substrate. It is connected to the reactor by a rectangular gate valve
32 x 222 mm. The automatic (un)loading is done by a pneumatic transfer
mechanism, and part of the process is also an evacuation of containing gas
and a purge with nitrogen. The load lock works under base pressure lower
than 10−1 mbar and has a maximum leakage rate 10−3 mbar·l/s.

Figure 4.5: SI PEALD LL Reactor unit with RTM. [11]

All the vacuum needed for the correct system process is supplied by the
vacuum system. It is equipped with a dry roots pump resistant against

20
.............................. 4.3. SI PEALD LL Description

corrosive gases. It is also water-cooled, nitrogen purged, carries several

sensors and control features, and overall provides reliable pressure operation.
The safety of the operator is moreover ensured by the programmable features
of the reactor unit. In both the load lock and the reactor, purging cycles
are set via system software and, besides the secured safety, contributes to
chamber cleanliness.

4.3.2 Precursor and Gas Cabinet

The precursor and gas cabinet contains different precursor pots, precursor
lines with several cut-off valves in their way, and mass flow controllers (MFCs).
MFCs control all gas lines, purge/carrier gases as well as a plasma source,
and provide highly constant flow rates. Moreover, they are equipped with
particle filters. Up to three MFCs can be provided for precursor delivery.
The control values of the MFCs are set accordingly to the ALD process and
can be altered from the user interface.

Figure 4.6: Direct draw (left) and bubble draw (right) precursor containers and
their installation. [13]

There are two different methods for the delivery of precursors with different
properties into the reactor. A direct draw is used for chemicals with high
vapor pressure; for low vapor pressure chemicals, a bubbling draw is utilized.
The precursors are supplied in cylinders or bubblers fitted with one or two
manual valves, respectively. These containers’ sizes can vary from 50 to 600
ml, with the standard pot being a 300 ml stainless steel bubbler from STREM
Chemicals. Pots are attached to the precursor line and can be optionally
equipped with heating devices. The standard offer includes 4 precursor lines
with separate inputs into the reactor installed in the precursor magazine.
Two more precursors can be added however they are added into a series

21
4. SENTECH SI PEALD LL System ............................
with a shared line into the reactor. The lines are made from electro-polished
stainless steel with a 6 mm diameter. All lines are separately heated (up to
200 ◦ C) and connected to the lid of the reaction chamber.

The plasma source, or more accurately capacitive coupled plasma source

(CCP), is attached to the upper flange of the reactor. This remote source is
driven by a 13.56 MHz generator and uses an external box for the 50 Ohm
output impedance of the generator to match the plasma load. The power
supply equals 300 W, and the pressure range of the cell is 0.07 - 1 mbar.
The CCP source verges on a gas line (with MFC); another 4 non-corrosive
gas lines can be added if desired. The gas line leads to the reactor chamber,
where the high flux of reactive gas species is provided at the substrate surface.

Figure 4.7: Reactor chamber and plasma source schematics. [11]

4.3.3 Hardware and Software Control

The system is controlled by hardware and software architecture. A reliable

remote field controller (RFC) secures basic safety interlocks and enables real-
time control of all components. This control is done via the serial field bus; bus
nodes are located in the control rack and the reactor unit. The architecture
enables quick manipulation with components and error diagnostics while
being noise-immune, fast, and with a stable, experienced protocol. Additional
Windows PC is supplied and used for process control, better visualization,
and advanced data logging. The PC communicates with the RFC by Ethernet
and is also equipped for LAN communication with the clients.

22
.............................. 4.3. SI PEALD LL Description

The SENTECH software offers a user-friendly interface that enables com-

fortable process development by writing recipes and the following monitoring
in an ALD process. Here, RFC provides additional safety control, independent
of the software for jeopardizing error commands or input data. The deposi-
tion process can be done manually or automatically via a recipe. The recipe
mode is produced from a predefined sequence of steps that are performed
automatically. Moreover, operations like opening the precursor lines can only
be done automatically, as the movement must be swift. The program further
allows the operator to control other command executions like opening the
dry pump’s valves or setting reactor and precursors’ temperature. Graphical
data logging is displayed together with the current state of the system. The
software employs a device called a sequencer for the organization of ALD
cycles. The sequencer is able to call up hardware components like valves and
MFCs within one cycle. Accordingly to chosen time intervals, it controls flow
rates of chemicals with a minimal time step of 10 ms and repeats the cycles
as defined in the recipe.

Figure 4.8: ALD system software example. [10]

4.3.4 Real Time Monitor

Real Time Monitor (RTM) can monitor each step within a single ALD
cycle and measure the thickness changes between deposited layers, all in-situ
with real-time values. The analysis can be done with different methods:
mass spectroscopy, quartz crystal microbalance. However, the most desirable
methods used are optical methods like ellipsometry as they do not influence the
ALD process, are surface sensitive, and fast. The principle of the ellipsometer
is captured in Figure 4.9. and its real implementation in Figure 4.5.

The key factor in ellipsometry is a measurement change of polarization

upon reflection of the incident radiation after it interacted with the material.

23
4. SENTECH SI PEALD LL System ............................

Figure 4.9: PEALD with ellipsometer and the principle of ellipsometry analysis.
[10]

The polarization change is quantified by the amplitude ratio Ψ and the phase
difference ∆. [14] The reflected signal depends on the material properties,
like refractive index, and their measurement can be transferred to thickness
(see Figure 4.9). Furthermore, the data are able to determine precursor
pulse, co-reactant pulse, and purge times in between. The ALD growth and
saturation can be consequently observed in every cycle.

Figure 4.10: Real time monitoring of TiO2 deposition. [11]

Thanks to the RTM dynamic substrate monitoring, a deeper understand-

ing of surface reaction can be gained as well as better optimization of the
ALD process by saving purging and dosing time and precursor consumption.
RTM has its own software that communicates with the ALD software and
contributes to the confirmation of a correct ALD regime.

24
 Chapter 5

Experimental Part

The experimental part of this thesis is dedicated to the SENTECH SI PEALD

LL system that was recently bought by the Faculty of Electrical Engineering,
CTU. The machine is stationed in a new laboratory operated by the Depart-
ment of Microelectronics. In this chapter, the very first steps of the ALD
machine are captured. First, there will be a brief description of the ordered
SENTECH ALD system, followed by its first trial deposition of Al2 O3 and
SiO2 thin films on Si wafer. The samples will be examined by AFM and
Raman spectroscopy. These methods will be described here as well, with a
short theoretical resume in the beginning. Finally, an experiment involving a
capacitor with a dielectric layer of PEALD grown Al2 O3 will be introduced.

The ALD system purchased by CTU is plasma-enhanced with RTM cor-

responding to the description in Chapter 4. Additionally, a glovebox from
MBRAUN was installed together with a CLEANSORB scrubber (CS Clean
Solution) for the adsorption of waste gases. There is a photo of the system
installed in the laboratory in Figures 5.1 and 5.2. Figure 5.2 shows a side
view of PEALD with open precursors and MFCs cabinets.

5.1 Preparation

The initialization of SI PEALD LL is fairly straightforward. The machine

and RTM must be switched on manually, but after that, they are accessible
via the supplied software. After ensuring that all pumps and the scrubber are

25
5. Experimental Part...................................

Figure 5.1: Front view of the glovebox connected to PEALD. [2]

Figure 5.2: Side view of PEALD. [2]

in running mode, the operator can open valves via the software, connecting
them with the system. Manual calibration of the ellipsometer is also needed
and is done with support from the RTM software.

With the ALD system running, several checks need to be done before the
first deposition and after a certain amount of time as part of the maintenance
cycle of the machine. One of the important tests is the water test for nitrogen
(N2 ). Nitrogen is the main carrier gas of the system, enabling precursors
dosing and purging. The amount of water in the gas needs to be checked, and
the correct flow rate of the gas ensured. When a new precursor is installed,
a protocol must be run to equilibrate pressure in the reactor chamber, the
precursor line, and the container. This is usually done by repeated opening

26
....................... 5.2. Trial PEALD Growth of Thin Oxide Layers

of respective valves until there are no pressure peaks observed. Before the
deposition, it is also desirable to activate the plasma source and adequately
heat the system. Specifically, to heat the reactor chamber, the precursor
containers, and the dosing and purging lines. After inserting the substrate
from the load lock to the reactor chamber, enough time must be given for
the substrate electrode to heat the sample, as this is usually the highest
temperature component in the system.

5.2 Trial PEALD Growth of Thin Oxide Layers

Having the PEALD system running and prepared as written above, the
deposition process could begin. A silicon wafer (with diameter of 100 ± 0.5
mm) was chosen as a substrate and placed into the reactor via the glove box
and load lock. For the Al2 O3 thin film, TMA was used as a precursor; for the
SiO2 , the precursor was a chemical called SAM24, bis(dimetylamino)silane.
Both processes used O2 plasma with 200 W power as a co-reactant and
nitrogen with a flow rate of 40 SCCM as a purge gas. The substrate was
in both cases heated to 200 ◦ C, TMA precursor was utilized via a direct
draw, whereas SAM24 needed a bubble draw, with bubbler at 60 ◦ C. The
PEALD cycle was designed by SENTECH, which supplied the appropriate
recipe. The recipe was loaded into the software platform, and the RTM was
synchronized, monitoring the growing film thickness. Table 5.1 shows the
time of each step of the PEALD cycle defined by the recipes.

Al2 O3 SiO2
Precursor 0.06 s 0.2 s
Purge 2s 4s
Plasma 3s 1s
Purge 5s 5s
Table 5.1: PEALD dosing and waiting times

The Al2 O3 film was grown into a 70 nm thick layer, the SiO2 film into 50
nm. The deposition went through successfully with no in-between system
errors. The samples were then evacuated from the reactor, and after cooling
down, subjected to the Raman spectroscopy and AFM.

27
5. Experimental Part...................................
5.3 Characterization Methods

5.3.1 Raman Spectroscopy

Raman Spectroscopy is a molecular spectroscopy technique that uses the

interaction of light with matter to gather chemical and structural information.
Observed light scattering is used for the measurement of molecular vibration
of the surface and the interior of the sample. Collected data can then be
formed into molecular (Raman) ’fingerprint’, and used to identify a substance.

The key factor of this spectroscopy method is Raman scattering. When

light interacts with molecules in a material, most photons are scattered at
the same energy as the incident photons (described as Rayleigh scattering).
But in a rare event, Raman scattering occurs when a small number of these
photons (1 in a 10 million [20]) will scatter at a different frequency. The
difference between the energy of the incident photon and the scattered photon
is called the Raman shift. With these data, a Raman spectrum is rendered.
The shape of an observed Raman spectrum is then compared to spectrums of
known materials, and the measurement result is concluded.

Raman spectrometer used in the experiment is the inVia Qontor from

Renishaw (Figure 5.3). The system offers three lasers with wavelengths of 532
nm, 633 nm, 830 nm; and has additional options like conducting an averaged
spectrum from a large number of measurements.

Figure 5.3: (a) Renishaw’s Raman spectrometer [22], (b) NTEGRA atomic force
microscopy device. [25]

28
.............................. 5.3. Characterization Methods

5.3.2 Atomic Force Microscopy

Atomic force microscopy is a technique that enables imaging of almost any

type of surface profile (semi/conductors, polymers, ceramics, etc.). It consists
of a sharp tip with a diameter from 10 to 20 nm attached to a cantilever,
both usually made from Si or Si3 N4 . [23] The tip is scanned over the surface,
interacting with it and using a feedback loop to adjust parameters for better
imaging and tip protection. The tip-sample reactions are mapped with a laser
beam deflection system (Figure 5.3), and with this, many force interactions
can be measured: wan der Waalse, electrical, magnetical, etc.

Figure 5.4: Laser beam deflection for AFM. [2]

The AFM has two basic operation modes. In the contact mode, the tip is in
continuous contact with the surface. This mode is used for a lesser amount of
applications than other modes. The tapping mode and the semi-contact mode
both offer a different approach: The cantilever carrying the tip is vibrated
above the sample surface in such a way that the tip comes into contact with
the surface intermittently. By moving the tip across the sample, the atom
at the apex of the tip reacts with individual atoms on the surface and forms
chemical bonds with them. These interactions slightly alter the vibration
frequency of the tip, which is detected via the laser beam.

Hence, the AFM can provide a 3D surface profile with almost zero sample
preparation requirements and under ambient atmosphere. However, the
scanning of the surface is sensitive to higher surrounding vibrations, and the
microscope must be placed on sufficient vibration isolation (e.g., marble base).
The AFM device used in the experiment is NTEGRA by NT-MDT Spectrum
Instruments (Figure 5.4).

29
5. Experimental Part...................................
5.4 Results and Discussion

The human eye can not distinguish the 70 nm and 50 nm layers of Al2 O3
and SiO2 , neither was a change of the surface structure observed. However,
what could have been seen after the deposition process was a slight change of
the wafers’ colors. The original silicon wafer was silver, Al2 O3 colored the
substrate with dark blue, and the SiO2 substrate had a brownish reflection.

Figure 5.5: (a) original Si wafer, (b) wafer with Al2 O3 thin film (c) wafer with
SiO2 thin film.

Both samples were examined by Raman spectroscopy. Raman spectrums

were measured with all three lasers with almost identical results. For pic-
ture representation, a laser that measured the ’nicest’ spectrum (smoothest,
slightest noise) was usually chosen. However, since the thin layer of oxide
is practically negligible in the mass of silicon, the Raman spectroscopy did
not recognize the deposited layers. Some scientific articles admit that, e.g.,
the Al2 O3 layer does not exhibit Raman signal [28] and use this method only
to compare the intensity of the spectrum for films with different thicknesses.
But according to [27], slight peaks at 428 cm−1 , 613 cm−1 , and 665 cm−1 can
be assigned to the oxide material. Those peaks could have been only very
slightly seen after the data were rid of the main peak. The original Raman
spectrum of Al2 O3 is in Figure 5.6 and is also almost identical to the SiO2
spectrum. For comparison, the Raman spectrum of plain Si wafer is also
shown (Figure 5.7); consequently, the overall difference in amplitude can be
observed. In conclusion, Raman spectroscopy verified that the samples were
unperturbed, and no damage was done to them.

The analysis by atomic force microscopy demonstrated the quality of the

grown thin films better. By imaging the sample in semi-contact mode, a
measurement of the surface was taken and graphically illustrated. Figure
5.8 is dedicated to the Al2 O3 layer. The diagram on the left represents
the original measurement in planar form. A gradual but smooth change in

30
............................... 5.4. Results and Discussion

Figure 5.6: Raman spectrum of Al2 O3 film measured with 633 nm laser.

Figure 5.7: Raman spectrum of plain Si wafer measured with 532 nm laser.

height can be observed, with a maximum difference of about 100 nm. This
continuous disproportion can only mean that the sample was tilted during
the measurement. In Figure 5.8 (b) there are the same data as in figure (a),
but flattened by the AFM software and hence devoided of the undesirable
tilting. They are for better visualization modeled in 3D graphics. A fairly
smooth surface is displayed with only a 2 nm difference at most.

The AFM measurement of SiO2 was evaluated analogically to Al2 O3 . Figure

5.9 (a) shows again a continuous change in height with a maximum variation
of 50 nm. The flattened 3D graphics reports a surface inequality that is not
higher than 0.8 nm. Figure 5.9 (b) also shows that surface is corrugated in a

31
5. Experimental Part...................................

Figure 5.8: AFM diagram of (a) real surface scan of Al2 O3 and (b) flattened
diagram of the surface in 3D form.

regular pattern. These waves are not considered an error of the deposition
as they are very small and hence negligible. They were most likely caused
by the feedback loop of the AFM system that could have induced frequency
on the sample in coincidence with the scanning frequency. Altogether both
measured thin films exhibited very good uniform coating, SiO2 film slightly
better than Al2 O3 .

Figure 5.9: AFM diagram of (a) real surface scan of SiO2 and (b) flattened
diagram of the surface in 3D form.

5.5 Capacitor Experiment

As an additional demonstration of the PEALD growth, an attempt to create a

parallel-plate capacitor was formed, specifically four capacitors with different
thicknesses of the dielectricum. The main purpose of this experiment was
to deposit a PEALD layer without SENTECH’S supervision correctly; the

32
................................ 5.5. Capacitor Experiment

precision of the creation of capacitor electrodes was henced less significant in

this experiment. PEALD was used to deposit a dielectric layer of Al2 O3 onto
an aluminum electrode. The final product was again subjected to Raman
spectroscopy and AFM, and the capacitance of the complete capacitor was
measured on LCR Bridge HM8118 from the HAMEG company.

The process began with four pieces (rectangles of approx. 3.5 cm × 2.5 cm
surface) of glass being wholly coated with aluminum Al by a thermal evapora-
tion system. The system used for the evaporation was Q150T Turbomolecular
pumped coater by Quorum. Then one half of each piece was covered with
thermal-resistant tape. Together, these samples were put into the PEALD
reactor and were taken out one by one after four different deposition. A
thin film of Al2 O3 was deposited onto the Al electrodes in the same way
as in Chapter 5.2 via the same recipe. Table 5.2 shows how many cycles
were run for each sample and a computed final thickness of the layers. After

Number of cycles Thickness

50 5.5 nm
182 20 nm
455 50 nm
909 100 nm
Table 5.2: Samples’ PEALD characteristics

the PEALD process, the samples were let to cool down, and the tape was
removed. The tape left a trace at the edge of the newly coated layer that can
be seen in Figure 5.10.

Figure 5.10: The border between Al and Al2 O3 with tape impurities under
regular microskope: (a) for 20 nm sample (b) for 100 nm sample.

All these samples were examined by AFM, the 100 nm layer piece was cap-
tured on Figure 5.11 (a), on the Al2 O3 coated part. The majority of the
surface reports about 100 nm in height differences with singular peaks with up
to 300 nm difference. The second measurement was taken on the borderline
of Al and Al2 O3 with the hope of seeing the thickness increase on the oxide
part. Indeed, in the middle of Figure 5.11 (b), where the border is clearest, a

33
5. Experimental Part...................................
change of color and thus height can be observed.

Figure 5.11: AFM diagram of (a) the Al2 O3 layer and (b) the border between
Al and Al2 O3 with tape impurities.

With the same AFM software, a height profile was created for the 20 nm and
the 100 nm sample. The 20 nm profile shows a slight elevation of about 10-20
nm; however, the 100 nm profile shows only 30 nm elevation. Moreover, when
conducting this measurement, it was found out that the elevated part of the
sample switched to the other side when the imaging direction was changed.
Therefore, the measurement is most likely invalid, for the tape remnants
highly deflected the tip scanning process.

Figure 5.12: Height profile of the surface of the 20 nm capacitor.

Figure 5.13: Height profile of the surface of the 100 nm sample with Al2 O3
coating.

The next step in the experiment was a deposition of the upper electrode. The
four samples were covered with tape on all sides, so only small squares of the
oxide layer were left visible. The samples were again put into the thermal
evaporation system, and according to the system software, a 58 nm thick

34
................................ 5.5. Capacitor Experiment

layer of Al was deposited onto them. The tape was removed from the samples,
leaving them in a final form, as seen in Figure 5.14.

Figure 5.14: The four samples after the second Al layer deposition.

All samples underwent a capacitance measurement. Only one sample

with capacitance emerged out of this measurement; the rest came out as
conductors. The functional capacitor was the 20 nm layer sample, and the
capacitance was 4 pF. This number sadly does not match the capacitor
parameters. Considering the standard capacitance formula
S
C=ε , (5.1)
d
the permittivity and the capacitance are both determined and known, leaving
the S and d parameters for discussion. If the proper implementation of
PEALD leading to 20 nm thickness is considered, the area S equals 7·10−10 m,
which is much smaller than the real area of the upper electrode, approximately
24 mm2 . A proposed conclusion is that the second aluminum layer was not
uniformly applied, leaving grains of Al on the oxide surface. According to
the previous calculation, such grain would be 30 µm wide in diameter, which
could fit into the physical reality.

Figure 5.15: (a) Photo of the sample with 20 nm functional capacitor. (b) The
20 nm capacitor under Raman microscope.

35
36
 Chapter 6

Conclusions

Electronic devices are becoming smaller and increasingly structured with

complex 3D shapes, needing a controllable deposition of conformal thin films.
ALD can meet these demands with sequential self-limiting reactions while
being one of the most effective methods on the market. ALD processes have
been exploited for a wide variety of materials, and yet new materials and
technologies like spatial ALD, allowing faster depositions speeds, are being
actively developed. The demand for precise deposition techniques has never
been higher, identifying ALD as a crucial developing instrument for emerging
nanofabrications.

In this thesis, the basics principles of ALD were described, and many
advantages of ALD were highlighted. An overview of a broad range of ALD
applications was given, focusing on the transistor and non-volatile memory
devices. A commercial ALD system from SENTECH was introduced in
more detail and operated in the experimental part, where a successful ALD
deposition of thin films was conducted. Grown Al2 O3 and SiO2 films were
examined, demonstrating the ALD characteristical features. From another
partially successful experiment, one functional, though underperforming,
parallel-plate capacitor emerged.

While proposing an exciting future of ALD worldwide, the ALD system at

the Faculty of Electrical Engineering laboratory has its own bright future as
well. The new laboratory also includes a plasma-etching system, presenting
a powerful duo with ALD in creating original-shaped structures with broad
coating possibilities. In cooperations with research teams in the Department
of Microelectronics and other Czech academic institutions, the growth of

37
6. Conclusions .....................................
thin high-κ dielectrics films, protecting passivation coating of electronic parts,
sensory layers based on ZnO, TiO2 , and many more applications will be
investigated and developed.

38
 Bibliography

[1] RICHARD W. JOHNSON, ADAM HULTQVIST, STACEY F. BENT, A

brief review of atomic layer deposition: from fundamentals to applications.
Materials Today, vol. 17, Issue 5, 2014, pp. 236–246, ISSN 1369-7021.

[2] ARTO PAKKALA, MATTI PUTKONEN, 27 - Atomic Layer De-

position,. William Andrew Publishing, 2010, pp. 364-391, ISBN
9780815520313.

[3] H.C.M. KNOOPS, S.E. POTTS, A.A. BOL, W.M.M. KESSELS, Chapter
8 - Atomic Layer Deposition, Handbook of Crystal Growth (Second
Edition). North-Holland, 2015, pp. 1101-1134, ISBN 9780444633040.

[4] KIM HYUNG JUN, LEE HAN-BO-RAM, W.-J- MAENG, Applications

of atomic layer deposition to nanofabrication and emerging nanodevices.
Thin Solid Films, vol. 517, Issue 8, 2009, pp. 2563–2580, ISSN 0040-6090.

[5] ERWIN KESSELS, Overview of all materials prepared by

atomic layer deposition (ALD). [Online, accessed from
https://www.atomiclimits.com/wp-content/uploads/2019/02/
Periodic-table-ALD-Feb2019.pdf February, 2019].

[6] S.E. POTTS, W.M.M. KESSELS, Energy-enhanced atomic layer deposi-

tion for more process and precursor versatility, Coordination Chemistry
Reviews, Volume 257, Issues 23–24, 2013, pp. 3254-3270, ISSN 0010-8545.

[7] ADRIE MACKUS, Approaches, challenges and opportunities for area-

selective ALD, Handouts, Eindhoven University of Technology, Tutorial
ALD 2017.

39
6. Conclusions .....................................
[8] M. M. FRANK, G. D. WILK et al., HfO2 and Al2 O3 gate dielectrics
on GaAs grown by atomic layer deposition, Appl. Phys. Lett. 86, 2015,
152904

[9] P. PUNCHAIPETCH, Y. URAOKA et al., Enhancing memory efficiency

of Si nanocrystal floating gate memories with high-κ gate oxides, Appl.
Phys. Lett. 89, 2006, 093502

[10] BERND GRUSKA, Low temperature deposition of thin passivation layers

by plasma ALD, Handouts to CTU, SENTECH Instruments GmbH,
Germany.

[11] B. GRUSKA, H. Gargouri, M. ArensReal Time True Surface Monitoring

for ALD Processes, Handouts from Workshop on ALD, 7. 10. 2014,
SENTECH Instruments GmbH, Germany.

[12] SENTECH Instruments GmbH SI ALD LL Atomic Layer Deposition

with Load lock, Product description, Handouts to CTU, SENTECH
Instruments GmbH, Germany.

[13] SENTECH Instruments GmbHAtomic Layer Deposition System SI ALD

(LL), Operation Manual, Version 1.4, May 2015, SENTECH Instruments
GmbH, Germany.

[14] SENTECH Instruments GmbHALD Real Time Monitor, User Manual,

October 2015, SENTECH Instruments GmbH, Germany.

[15] PICOSUNT M Industrial ALD Product Line, Brochure, 2016,

PICOSUNT M , Finnland.

[16] PICOSUNT M , Picosun’s ALD technology enables record capacitance

density (...) [Online, accessed from https://passive-components.eu/
picosuns-ald-technology-enables-3d-silicon-integrated-micr_
ocapacitors-with-unprecedented-performance/ 20. 4. 2020].

[17] PICOSUNT M , PICOSUN R-200 Advanced [Online, accessed from https:

//www.picosun.com/product/r-200-advanced/ 7. 5. 2021].

[18] Veeco, Fiji - Plasma Enhanced ALD for R&B [On-

line, accessed from https://www.veeco.com/products/
fiji-plasma-enhanced-ald-for-rd/ 7. 5. 2021].

[19] SENTECH Instruments GmbH, Atomic Layer Deposition Systems

[Online, accessed from https://www.sentech.com/en/ALD-Systems_
_2492/ 7. 5. 2021].

[20] MITTLER TOLEDO, Raman Spectroscopy. [Online, accessed

from https://www.mt.com/au/en/home/applications/L1_AutoChem_
Applications/Raman-Spectroscopy.html#overviewaf 8. 5. 2021].

40
..................................... 6. Conclusions

[21] EDINBURGH INSTRUMENTS, What is Raman Spec-

troscopy?. [Online, accessed from https://www.edinst.com/blog/
what-is-raman-spectroscopy/ 8. 5. 2021].

[22] RENISHAW, inVIa Qontor confocal Raman microscope. [On-

line, accessed from https://www.renishaw.com/media/thumbnails/
320wide/45f0e05befac4cedae5ab894c1b55078.jpg 8. 5. 2021].

[23] NT-MDT Spectrum Instruments, NTEGRA. [Online, accessed from

https://www.ntmdt-si.com/data/media/images/products/ntegra/
prima/integra_new.jpg 8. 5. 2021].

[24] nanoScience Instruments, Atomic Force Microscopy. [Online,

accessed from https://www.nanoscience.com/techniques/
atomic-force-microscopy/ 8. 5. 2021].

[25] nanoScience Instruments, Atomic Force Microscopy. [Online, accessed

from https://www.nanoscience.com/wp-content/uploads/2018/04/
Laser-beam-deflection-for-Atomic-Force-Microscopes.png 8. 5.
2021].

[26] SUPRAKAS SINHA RAY, 4 - Techniques for characterizing the structure

and properties of polymer nanocomposites. Series in Composites Science
and Engineering, Environmentally Friendly Polymer Nanocomposites,
Woodhead Publishing, 2013, pp. 74-88, ISBN 9780857097774.

[27] YUNPING LAN, YONGGANG ZOU, XIAOHUI MA, LI XU, LINLIN

SHI AND JIABIN ZHANG, Fabrication of amorphous Al2 O3 optical
film with various refractive index and low surface roughness. Materials
Research express, vol. 7, number 8, 2020, pp. 086405,

[28] C. WIRTZ, T. HALLAM, C. P. CULLEN, N. C. BERNER, M. O’BRIEN,

M. MARCIA, A. HIRSH and G. S. DUESBERG Atomic layer deposition
on 2D transition metal chalcogenides: layer dependent reactivity and
seeding with organic ad-layers. Chemical Communications, Issue 92 2015,
pp. 16553 —16556.

41
42
 ZADÁNÍ BAKALÁŘSKÉ PRÁCE

I. OSOBNÍ A STUDIJNÍ ÚDAJE

Příjmení: Veselá Jméno: Karolína Osobní číslo: 483890

Fakulta/ústav: Fakulta elektrotechnická
Zadávající katedra/ústav: Katedra radioelektroniky
Studijní program: Otevřené elektronické systémy

II. ÚDAJE K BAKALÁŘSKÉ PRÁCI

Název bakalářské práce:
Depozice atomárních vrstev

Název bakalářské práce anglicky:

Atomic Layer Deposition (ALD)

Pokyny pro vypracování:

1. Seznamte se s obecnými principy růstu vrstev pomocí ALD a jejich konkrétními aplikacemi v elektronice.
2. Seznamte se s parametry a obsluhou konkrétního zařízení od firmy Sentech.
3. Proveďte růst testovací tenké oxidové vrstvy a charakterizujte její povrchové vlastnosti pomocí optického mikroskopu
a mikroskopu AFM.

Seznam doporučené literatury:

[1] R. W. Johnson, A. Hultqvist, S. F. Bent: A brief review of atomic layer deposition: from fundamentals to applications,
Materials Today, sv. 17, č. 5, str. 236-246, 2014
[2] H. Kim, H. Lee, W.-J. Maeng: Applications of atomic layer deposition to nanofabrication and emerging nanodevices,
Thin Solid Films, sv. 517, str. 2563–2580, 2009
[3] H. Li et al.: Enhanced electrical properties of dual-layer channel ZnO thin film transistors prepared by atomic layer
deposition, sv. 439, str. 632-637, 2018

Jméno a pracoviště vedoucí(ho) bakalářské práce:

doc. RNDr. Jan Voves, CSc., katedra mikroelektroniky FEL

Jméno a pracoviště druhé(ho) vedoucí(ho) nebo konzultanta(ky) bakalářské práce:

Datum zadání bakalářské práce: 22.01.2021 Termín odevzdání bakalářské práce: _____________

Platnost zadání bakalářské práce: 30.09.2022

___________________________ ___________________________ ___________________________

doc. RNDr. Jan Voves, CSc. doc. Ing. Josef Dobeš, CSc. prof. Mgr. Petr Páta, Ph.D.
podpis vedoucí(ho) práce podpis vedoucí(ho) ústavu/katedry podpis děkana(ky)

III. PŘEVZETÍ ZADÁNÍ

Studentka bere na vědomí, že je povinna vypracovat bakalářskou práci samostatně, bez cizí pomoci, s výjimkou poskytnutých konzultací.
Seznam použité literatury, jiných pramenů a jmen konzultantů je třeba uvést v bakalářské práci.

.
Datum převzetí zadání Podpis studentky

CVUT-CZ-ZBP-2015.1 © ČVUT v Praze, Design: ČVUT v Praze, VIC