1

Nanosystems

Prof. Markus Becherer

Technische Universität München

School of Computation, Information and Technology

Professorship of
Chip-based Magnetic Sensor Technology

Discover the world of

Nanoelectronics

Today’s lecture addresses further film fabrication methods.

 2

L01 Kick-off + ERD roadmap

Nanofabrication Nanometrology/Patterning Applications Magnetic devices

L02 Conversion + PVD
L03 CVD + Others
L04 Nanometrology I - Electrical
L05 Nanometrology II - Optical and SPM
L06 Nanometrology III - Beams of Electrons/Ions
L07 Patterning - Lithography
L08 Metallic Nanowires I
L09 Metallic Nanowires II
L10 Quantum LEDs with Si NP
o CVD
→ Quite some theory L11 Introduction (nano)magnetism
L12 Magnetization dynamics
→ Epitaxy and ALD
L13 Spin-wave devices
o Others
L14 Nanomagnetic Logic
→ Cleaning procedures L15 Nanomagnetic Devices 2.0
→ Spin coating
→ Spray coating
2

Last time, we discussed conversion (e.g., oxidation of silicon and doping of silicon by im-
plantation of ions) and physical vapor deposition ( ‘boiling’ of solids in a vacuum chamber
and, hence, evaporation and the sputtering of material in a low-pressure plasma). Today,
we move on to chemical vapor deposition (CVD) and some other methods that have been
heavily used at the Chair of Nanoelectronics (TUM), where I did part of my PostDoc
career.
 3

Chemical vapor deposition (CVD)

3
 4

Thin film generation ….

„Conversion“
„Umwandlung“ Deposition
Abscheiden

PVD
PVD CVD
CVD
Others
Physical vapor deposition
physical vapor deposition Chemical vapor deposition
chemical vapor deposition

Thermische
Thermal oxidation Electrochemical
Oxidation Evaporation
Aufdampfen CVD
CVD
Elektrochemisch

LPCVD
LPCVD
Spin-on
Spinn-on

Doping
Dotierung of Semicond.
Sputtering
Sputtern

- -Implantation
Implantation Printing
Epitaxy
Epitaxy
- -Diffusion
Diffusion

Spray-Coating
ALD: Atomic Layer Deposition
… the classical approach 4

Having a closer look, CVD is one sub-branch of deposition. A film is deposited on

a substrate without (or very, very little) change of the underlying film or substrate.
There are many flavors of CVD methods depending on the reactor geometry, pressure
conditions, plasma enhancement, chemicals used, etc. ...
 5

CVD-Oxides --- low temperature processes

C
Low Temperature Oxidation (LTO) SiH 4 + 2O2 ⎯400
⎯⎯ → SiO2 + 2 H 2O Silane-Pyrolyses

APCVD: Advantage: high growth rate R < 100nm/min, + low temperature

Disadvantage: poor edge coverage - porous, low electrical load capacity
LPCVD: Disadvantage: small growth rate R < 10nm /min, - generates Si-particles in the gas stream
Advantage: better edge coverage

C
Plasma Enhanced Oxidation (PEO) SiH 4 + 4 N 2O ⎯350
⎯⎯ → SiO2 + 2 H 2O + 4 N 2
TEOS or SiH4 + Plasma low temperature, medium oxide quality, non-conformal, particles

Ozone Enhanced Oxidation (OEO) Si(C2 H5O)4 + 8O3 ⎯400

⎯⎯ C
→ SiO2 + 8CO2 + 10H 2O
TEOS or SiH4 (+BPSG) + O3
low temperature, medium oxide quality, non-conformal,

A typical application of CVD: deposit insulating SiO2 fabricated at a lower temperature

than what we have seen in the last lecture by thermal oxidation of Si. There, the typical
process temperature was 1000 ◦ C. Here, the chemical reaction can be ‘tuned’ to 400 ◦ C
and below by chemical enhancements. There are some examples here, and there are many
more. One of the reaction products in all examples is SiO2 , but the quality, conformity,
growth rate, etc., varies.
 6

Chemical Vapor Deposition (CVD) is a process in which a solid layer is formed from gas molecules
(precursors) via chemical reactions (in the gas and/or on the substrate surface).
By appropriate selection of precursors and process parameters different materials can be deposited. The crystal structure
can be monocrystalline, polycrystalline or amorphous.
Gas reactions

Gas inlet Gas outlet

Convectional main gas flow

Concentration cGas
Flow velocity vGas

Thickness  of the boundary

Diffusion of
layer,
the decrease of the reactants to
reactants concentration csurface,
crystalline substrate decrease of the flow velocity to
vsurface = 0 (the atoms adhere)
Resistive heater
→ Radiation
Adhesion (deposition), Germ formation Desorption of volatile
surface diffusion, reaction products
surface reaction 6

CVD: a solid layer is formed from gas molecules (precursors) via chemical reactions (it
the gas and/or on the substrate surface). Now, let’s identify the components in the given
reaction chamber. A glass tube or vacuum chamber is the ‘controlled’ environment for
the CVD process. The precursor gas, often combined with inert purge gases, is led into
the chamber, generating the main gas flow. Perpendicular to it, there is diffusion taking
place (along the concentration gradient), and the precursor gas maintains the deposition
process at the surface. Adhesion of molecules, germ formation, and desorption of volatile
reaction products is taking place. An external resistance heater controls process temper-
ature by radiant heat. Please note: at the sample surface, the gas flow velocity equals
zero. Chemicals only diffuse to the surface; there is no lateral flow except for surface
diffusion due to elevated temperatures.
 7

CVD Chemical Vapor Deposition --- it’s complex

External parameters : - Gas pressure (particle concentration)
- Gas flow (flow velocity)
- Heating (temperature)
- Reactor geometry

- In gas (convection flow along flow lines)

Mass transport - In the gas (diffusion against the concentration gradients)
- At the surface (diffusion)

Chemical Reactions: - In gas (formation of new molecules, law of mass action)

- On the surface (adsorption, binding, nucleation)

Heat transfer: - Radiation, convection, heat conduction

All processes are generally temperature-dependent, geometry-dependent and pressure-dependent.

7

It’s a discouraging slide. How do you control so many parameters at the same time? By
splitting the task into smaller problems, the overall goal can be achieved. Nevertheless,
it is one of the main challenges to keep track of the most important parameters in a
CVD processing chamber. Often, one dedicated tool is used for a single process only. As
soon as you start to run different processes, cleaning cycles are needed. You can imagine
having one machine for each task is rather costly, but actually, it is often the standard
in industry.
 8

The balance between diffusion and reaction determines the step coverage
T
3
2  Q 
D k r  exp− 
p  k BT 
diffusion >> reaction diffusion << reaction
-> low pressure (LPCVD) -> high pressure (APCVD)
-> low temperature -> high temperature or
reaction term diffusion term low temperature + plasma activation
limits the growth limits the growth
 Q 
3
2
T
Reaction limited k r  exp−  D Diffusion limited
 k BT  p

gas atoms everywhere, reaction waiting for delivered atoms,

reactions deposits equal on every surface favourable and non-favourable positions

Fully conformal Shadowing effects

e.g. Atmospheric Pressur CVD

e.g. Low Pressure CVD
Plasma Enhanced CVD 8

There are two regimes in CVD to be distinguished. Diffusion and reaction limited growth.
Do you remember the oxidation conditions in thermal oxidation? The same happens
here. Imagine that diffusion is very strong (e.g., in a furnace with very low pressure)
— left branch. This is achieved by low pressure p (and pressure is easily decreased by
orders of magnitude!). At the same time, the temperature is rather low. The chemical
reaction will be slow because the minimum activation energy for the chemical reaction
Q is in the numerator, and temperature T is in the denominator. −→ reaction rate is
the limit. The opposite behavior is shown in the path on the right: diffusion is hindered
by atmospheric pressure or plasma conditions. Temperature is sufficient to immediately
trigger a chemical reaction. −→ diffusion (lack of reactants) is the limit. One important
consequence is the level of conformity of the deposited layers.
 9

CVD mass transport

Theoretical deposition rates from the reaction:

, p =1mbar
SiH 2 ⎯T⎯ ⎯⎯→ Si + H 2
Activation energy depo of monocrystalline Si : Q ~ 1.6 eV

3
2 Diffusion-limited deposition is
T 1
vDiff = D   very geometry-dependent (by )
 p 

k r = k0  exp − Q  Reaction-limited deposition is

 kT  very temperature-dependent

after Hansch, W.
Lecture@TUM 2008 9

The discussed growth conditions are not just a theoretical gimmick. It can be directly
visualized in a clever plot, comparing CVDs for SiO2 . Knowing that the proportionality
for a temperature-activated chemical reaction is an exponential term with − Q T as argu-
ment, why not plot the growth rate on a logarithmic scale vs the inverse temperature?
As soon as we see a straight line, it means that we have the expected behavior, and we
can even fit the activation energy of the chemical reaction −→ blue-dashed-line. For high
temperatures, the diffusion (= mass transfer) limits the process as we have seen before.
Hence, we see the saturation behavior −→ red-dashed-line.
 10

after Hansch, W.
Lecture@TUM 2008
CVD Conformity
In nanofabrication, trenches and step-like structures often have to be uniformly / conformably filled or covered.

Example: To create the inner capacitor plate of an (older) DRAM

technology, the trench must be filled with highly doped poly-Si.
For this, a conformal deposition is necessary:
(conformal = same layer thickness everywhere)

Principle layer geometries for conformal deposition

Step 1 Step 2 Step 3

1 µm If the same layer thickness is deposited everywhere, a vertical trench

grows until completely closed (except for a seam in the middle).
10

Here, a very old cross-sectional image of a DRAM is depicted. Deep trenches have to be
filled with highly doped poly-Si. In principle, it means the seamless filling of high-aspect
structures. It’s impossible with PVD, tricky but doable with CVD. Let’s qualitatively
learn the basics of conformity.
 11

after Hansch, W.
Lecture@TUM 2008
CVD conformity: typical velocities and mean free path
Kinetic gas theory
T K 
vconvection ~ 1 m/sec v=
8 kT
 v = 146  m / sec
 m M mol 

The (non-directional) thermal velocity (diffusion) is much higher than the

directional gas velocity (convection).

Vtherm ~ 500 m/sec

Typical arrangement of wafers in a wafer boat:

  p  10 −4 m  mbar 
10 −4 m  mbar 
(1000mbar = APCVD)  (1000 mbar )   10 −7 m
1000 mbar
(1mbar = LPCVD)  (1mbar )  10 −4 m  0.1mm
wafer Even at low pressures (LPCVD), the free path length of the gas molecules between
two collisions is significantly smaller than the geometric distances in the reactor
Distance: ~6 mm (length, wafer distance) -> always isotropic, undirected particle motion

11

We have to distinguish between convection and diffusion in the gas flow. (1)The main
gas flow is normally driven by convection with a typical velocity vconvection = 1 m s−1 .
There is an overlying isotropic thermal movement with v̄ = 500 m s−1 . 500x larger than
convection. Calculating the mean free path α is mainly a function of pressure and can be
estimated by the equation in (2). Note that in CVD reactors, wafers are often organized
in a boat, they are placed vertically with a small distance of a couple of millimeters
to allow batch processing −→ movement of the gases between the wafers is (a) mainly
my diffusion, not by convection and (b) particles scatter a lot and hence, isotropic gas
molecule movement is dominant.
 12

after Hansch, W.
Lecture@TUM 2008
CVD conformity: angular distribution of incoming flux
Since even at low-pressure CVD (1mbar) the reactor geometries are larger than the mean free path lengths, the directional
distribution from the gas space to the substrate surface is isotropic, due to the scattering of the gas molecules.

Substrate 2
If one sits on a place on the substrate and observes the number
of the incoming gas molecules (the flow F) for different angles of
Scattering events of gas molecules incidence, one can determine the angular distribution by a
cosn(Θ) function:

Distribution of the arrival angle

F ~ cos n ()
many particles at steep angles on the surface

Incoming flux F
few particles at
flat angles
Substrate 1

n = 1: Isotropic distribution, from all directions, the same amount of

anisotropy is present
n > 1: Increasingly directed (anisotropic), more and more arriving vertically -90° 0° 90°

Θ 12

We see an isotropic gas movement between the wafers because of scattering and the
small mean-free path of 0.1 mm @ 1 mbar. But simultaneously, we deal with a laterally
extended wafer surface (substrate). However, below a thin sheet on the length scale of
the mean free path on the wafer surface, one can observe a distribution with flat angles
having less flux than steep ones. With Θ being the angle of incidence measured towards
the wafer normal, we see a F ∝ cosn Θ distribution, with F being the flux of incoming
molecules and n a measure of anisotropy. In a CVD process, we see n = 1, in contrast
to, e.g., PVD evaporation, where n > 1 due to a single point source of flux.
 13

after Hansch, W.
Lecture@TUM 2008
CVD conformity: Viewing angle + surface diffusion
small high percentage
percentage from many
Viewing angle directions
from few
directions
top
The viewing angle determines the
proportion of the incoming flux directly
available for built-in.
bottom As a rule, upper areas are hit much more
frequently than lower areas.
Surface Diffusion

Corners and edges are places with a high

surface energy. Arriving molecules are
preferentially installed here to planarize the
surface and thus reduce the surface
energy.

1 + 2 leads to a preferential attachment (built-in) at the upper openings of trenches. 13

Often, we are not dealing with flat substrates but patterned surfaces with trenches or
steps. In (1), a deep trench is considered, where due to simple geometry, the incoming
flux at the bottom is reduced, whereas close to the surface, higher flux is expected,
explained by the spatial angle. This we can call ‘viewing angle’ . We have to consider
surface diffusion (2) as well. Corners and edges are preferably covered, as surface energy
shall be minimized. This is more effectively accomplished at edges and corners.
 14

after Hansch, W.
Lecture@TUM 2008
CVD Conformity: Sticking Coefficient
Sticking coefficient Adsorption Desorption
Deposition
Diffusion

The physical processes of adsorption, Deposition Diffusion

surface diffusion, desorption, and Adsorption
reaction (built-in) occur on the surface.
Substrate Desorption

The sticking coefficient Sc

describes the proportion of the
deposited molecules (atoms) to the
arriving molecules (atoms).

Deposition rate
Sc = -------------------------
Incoming rate

The desorption retakes place under a cosn(Θ)-distribution. High sticking coefficient Low sticking coefficient
There is no "memory" of the angle of incidence. Sc = 1 Sc < 1 14

As a last consideration for CVD conformity, let’s look at the sticking coefficient (3).
The process of adsorption and surface diffusion is followed by either desorption (not
successful built-in) or deposition (successful built-in). This can be expressed by relating
the deposition rate to the incoming rate (measuring those rates is a different story, but
discussing it is easy). Hence, a sticking coefficient of Sc = 1 says that every incoming
molecule is successfully built in; a lower value says that it needs more attempts. It is
superfluous to mention that a conformal coverage is achieved with Sc << 1. A summary
of that is on the next slide.
 15

@ Sc = 1 n = 1, isotrop n=3 n = 10

Variation of the incoming distributions cosn ()

Source: Simulator SPEEDIE

Sc = 0.1
@n=1
Variation and Sc = 0.2
of
Sc Variation of the edge steepness
The more directed the incidence (n > 1), the worse the edge coverage. Even with n = 1, the
coverage could be better. (compare with evaporation and sputtering).
Sc = 0.01 Low sticking coefficients significantly improve conformity.
Flat edge slopes significantly improve conformity. 15

A summary of simulation results performed with SPEEDIE, a technology simulation tool

(of older days). Blue arrow, variation of ‘anisotropy’ in the process with n = 1; n =
3; andn = 10 @ fixed Sc , −→ no sidewall coverage of the trench for a highly anisotropic
process (e.g. large mean free path and single point source as in evaporation). Red arrow:
Variation of the Sticking Coefficient Sc @ fixed anisotropy exponent n −→ low sticking
coefficient leads to almost perfect conformal deposition. Last, setting the slope of the
sidewall of a trench to smaller values than perpendicular is a compromise such that at
much higher sticking coefficients (duration of deposition is a big issue in the industry), the
successful filling of the trench can be achieved. Simulation of parameters like anisotropy,
sticking coefficient, and slope angle of the trench can help develop a fast and robust
process for CVD or PVD processes.
 16

Epitaxy:

atomically
precise
CVD
for
crystalline
growth

http://www.povray.org/preview/irtc-cd3/stills/19980831/view/epitaxy.htm

16
 17

after Hansch, W.
Growing crystals on crystals by epitaxy Lecture@TUM 2008

Epitaxy means "the order“; crystal growth means further growth of the crystal lattice directed at the crystal
nucleus (or substrate). If the same material continues to grow, one speaks of homoepitaxy, if two different
materials grow on each other, one speaks of heteroepitaxy. Heteroepitaxy requires that the crystal lattices
have similar lattice constants (mismatch < 1%). The lattice mismatch leads to high mechanical strains, which
are relieved by the formation of defects.
Material B Compression of B
thick
thin layer
layer

Material A (strained layer → pseudomorph) (relaxed layers, → relocations

a b
Critical thickness: hc  L Dislocation distance L=
2 b−a a,b: lattice constant
17
The formulas above are a rough estimate only! It can be very different depending on the materials.

If you aim to grow crystalline films, you need a suitable substrate. One crucial pre-
condition: Material B should be very similar to material A. You can imagine that for
Epitaxy high-quality substrates, very clean substrates are needed. Therefore, Epitaxy is
a costly process, even though the schematic given here looks simple. Take two similar
materials and material B will ideally grow on A. As soon as we have a mismatch in lattice
constant, compressive or tensile mechanical strain will lead to permanent stress. If the
stress becomes too high, relocations and relaxed layers are formed. To calculate a critical
thickness, a simple formula is given, but this is only a rough estimate. The experiment
will show if your material system of choice works. Examples? SiGe on Si works for low
Ge content.
 18

after Hansch, W.
Lecture@TUM 2008
Crystal Relaxations
Dislocations

Epitaxial overgrowth of a platinum single crystal with

strontium titanium oxide.
Formation of twin crystals (twins) (Schmidt et al, Univ. of
California, 2006, paper 1144) 18

With the advent of Transmission Electron Microscopy (TEM), spatial imaging of epitax-
ially grown layers became possible. Two material systems are presented here, and from
both, I neither know their technical relevance nor if the results are ‘good’ or ‘bad’ .
But at least the images are impressive, and dislocations can be visualized in real space.
However, be aware of the tiny section of the sample with a field-of-view smaller 20 nm;
it is a little section from a wafer, and there is no statistical information of defects or
‘problems’ on a wafer scale. It is a tiny little zoom-in only.
 19

19

There is another way to deposit thin films, sometimes more precise than conventional
Epitaxy: Atomic Layer Deposition (ALD). I found it for you, and already, the abstract is
challenging to read. How many technical fabrication and metrology terms of the abstract
do you know already after two lectures?
 20

Atomic Layer Deposition: ALD as special case of CVD

key lock principle

Schematic representation of ALD using

self-limiting surface chemistry and an
AB binary reaction sequence.

George S. M.,Chem. Rev. 2010, 110, 111–131 20

ALD is a subbranch of CVD, and its main characteristics are like this: it uses cycles
of different precursor chemicals (one might even call it the key-lock principle, following
biological processes having this working mechanism). A is locked to the surface, forming
one monolayer of molecules, then B matches A on the surface, and the next monolayer is
formed, and afterward, we again do A. Please do not ask me about the chemistry behind
it, and most of the turnaround for companies in this business is done by developing the
fitting precursor. They can be really, really costly indeed.
 21

21

From one manufacturer/reseller, I found the thin-film periodic table of ALD precursors.
It seems to work for Fluorides, Nitrides, Oxides, Selenides, Sulfides, and Tellurides, and
also pure elements. ALD is very powerful, as you can imagine.
 22

ALD results

SEM cross-section of 3D cavity coated with conformal Al2O3 / TiO2 (ATO)

George S. M.,Chem. Rev. 2010, 110, 111–131 nano-laminate ALD film. Courtesy of O. Poncelet, L. Francis, Univ. de
Louvain – Deposited in Fiji® system. 22

Let’s move on to the results of ALD processes. As we talked about the difficulty of
conformal deposition before, with ALD, you can fill deep trenches (left) and cavities
(right). The crossectional images of the structures are very impressive. You can imagine
that ALD is very interesting, especially for research institutions: The tools are flexible,
precise, and, compared to a big batch processing epitaxy tool or ion implanter, quite
affordable. We run one tool already in our labs, deposition rate: 1 nm h−1 of Al2 O3 , and
one of the precursors can more or less be used as rocket fuel – honestly.
 23

Others (spin-coating and spray-coating)

… if you have less money ☺

+ Cleaning of surfaces before films are deposited!

23
 24

Thin film generation ….

„Conversion“
„Umwandlung“ Deposition
Abscheiden

PVD
PVD CVD
CVD
Others
Physical vapor deposition
physical vapor deposition Chemical vapor deposition
chemical vapor deposition

Thermische
Thermal oxidation Electrochemical
Oxidation Evaporation
Aufdampfen CVD
CVD
Elektrochemisch

LPCVD
LPCVD
Spin-on
Spinn-on

Doping
Dotierung of Semicond.
Sputtering
Sputtern

- -Implantation
Implantation Printing
Epitaxy
Epitaxy
- -Diffusion
Diffusion

Spray-Coating
ALD: Atomic Layer Deposition
… the classical approach 24

The most extended branch for deposition is ‘Others’ . But I will shorten it to some
examples in our labs.
 25

after Hansch, W.

Cleaning of Si surfaces --- by wet chemistry Lecture@TUM 2008

RCA-clean HF-Dip
(Fa.RCA) or Huang A and Huang B
1 part HF + 10 parts H2O
SC 1: 1 part NH4OH + 1 part H2O2 + 5 parts H20
(dissolves lower organic compounds) When a Si substrate is immersed in a
hydrofluoric acid solution, the Si
surface is passivated for a short time
SC 2: 1 part HCl + 1 part H2O2 + 5 part H2O (~15 minutes). This prevents the
(dissolves metals and ions) formation of the natural oxide and the
adsorption of contaminations.
W. Kern and D. Puotinen, RCA Rev., 31, 187 (1970).

It is used to remove SiO2 from Si

Process time: approx. 40 min at 75°C surfaces.

High temperatures increase the etching rate

Typically, before film deposition, expensive

process (many high-purity chemicals in flow), Si
can be etched (if the temperature fluctuates).
25

First, an important aspect neglected so far. To form reliable thin films, you need clean
surfaces and control of the interface. For Si, recipes over 50 years old still apply. The
RCA-clean sequence was first applied in the Radio-Company-America — hence the name
RCA. In Si technology, also HF plays a crucial role, as SiO2 can be selectively etched.
The chemical equations look simple, but they are typically highly toxic and dangerous
— HF being pretty dangerous. The chemicals used are of the highest purity with, e.g.,
‘Selectipur’ ® being orders of magnitude cleaner than what you probably used in high
school classes.
 26

after Hansch, W.

Cleaning surfaces --- by wet chemistry Lecture@TUM 2008

Piranha-clean: also called Caro-clean)

2 parts H2SO4 (97%) + part H2O2(25%)

Oxidizes organic molecules to more water-soluble components such

as carboxylic acid COOH. Viscous products can be formed by good
rinsing or subsequent RCA-cleaning.

Process time: approx. 25 min at 90°C

High temperatures increase the etching rate

Typically used to remove stubborn resist strip after the process (e.g.
plasma etching), usually followed by SC1.

26

Another classical cleaning ‘solution’ is Caro-clean, colloquially termed Piranha-clean. It

acts like Piranhas in a bath of bloody water. We typically use it when large amounts of
polymers (after lithography) decide to reside on chromium masks or Si wafers, depicted
in the graph on the right.
 27

An extract of our lab course on fabrication of semiconductor devices:

We start with several diced pieces of SI wafers. Rinsing the wafers with plenty of DI water
removes coarse impurities and particles. The supply of kinetic energy in an ultrasonic
bath removes even stubborn particles. Generously long rinsing in DI water also helps to
remove particles. A subsequent so-called RCA1 step, a basic cleaning solution

+ removes organic substances,

+ complexes some heavy metals like Cu, Ag, Ni, Cd

– attacks Si when H2 O2 is decomposed,

– and is only suitable for thin organic layers.

This is followed by a so-called RCA2 step, which is an acidic cleaning solution of hy-
drochloric acid and hydrogen peroxide

+ heavy metals that were not completely removed in the first step are bound,

+ ionic impurities such as NNa+ and Ca2 + are neutralized to salts, but it

– may introduce heavy metals like Cu, Ag, Ni, Cd,

– attacks Si when H2 O2 is decomposed,

– is suitable ONLY for cleaning thin organic layers.

Furthermore, hydrofluoric acid (HF, attention, one of the most dangerous acids for hu-
mans!) is used to remove the storage oxide on the back of the wafer, as well as the native
oxide on the front (these oxides are not shown in the cross-sectional images), because

+ with HF can be selectively etched SiO2 ,

– terminate the Si surface with hydrogen atoms for 20-30 minutes,

i.e., it is possible to protect the Si surface to charge a process furnace or vacuum systems
without water film and attached gas molecules. At the time of termination, no disordered,
native SiO2 grows up either.
*** Translated with www.DeepL.com/Translator (free version) ***

That’s all about wet-chemical cleaning, even though one could spend hours on it.
 28

Cleaning surfaces --- plasma cleaning

Schematic representation of a microwave plasma asher and a reactive-ion etcher.

a. Microwave plasma: The process gas is excited by microwave radiation coupled into the chamber. Microwave plasma usually
provides isotropic plasma etching/activation.
b. Reactive-ion etcher (RIE): The process gas is excited by a strong alternating electromagnetic field. Ions move unidirectional
towards the negatively charged substrate, resulting in a non-isotropic etch profile.

T. Häberle, PhD thesis, 2017 27

Especially polymers and organic compounds are heavily used in patterning films. As in
cleaning procedures, such solids are often diluted only (not entirely removed); polymers
might stick at the surface and remain as ultra-thin layers at the interface. To overcome
this, plasma cleaning with oxygen content is a viable way to clean a surface from organic
compounds. The ions are highly reactive and form CO2 as volatile gas that can be
pumped out easily. The big brother of a simple microwave plasma furnace is a reactive
ion etcher with much better control of gas flow, bias voltages, and power levels. You
cannot imagine a semiconductor lab without plasma etching and reactive ion etching
tools.
 29

Spin coating of films

Schematic for the spin coating process that employs the following steps: (a) applying the
dissolved material, which is commonly done by drop coating; (b) Rotation at high speed to
spread the material over the substrate; and © drying to evaporate the residual solvent.

Often used (but not limited to) to “deposit” resists for lithography processes.

M. Bobinger, PhD thesis, 2018 28

After the excursion on cleaning, a straightforward way of thin-film formation is the so-
called spin coating process. A dissolved solid or polymer is dropped on the substrate,
which is placed on a rotating disc. Depending on the viscosity of the polymer, one can
achieve thicknesses in the lower 2-digit nm- up to 2-digit µm-range. Typical applications
are the planarization of patterns on a wafer or photo-active resists in microelectronic
lithography.
 30

Spin-coating

Source: www.MicroChemicals.com → there you get nice brochures to debug lithography :-)
29

The spin-coating process looks simple. However, it is good to read the application notes
of the polymer manufacturer. A typical thickness vs. spin-speed curve is given on the
right. The faster the spin speed, the thinner the film thickness, and we see a tfinal ∝ √1ω
dependence with the rotational speed ω and the final film thickness tfinal . Please also
note that the spinning process is divided into ramp-up, plateau, and ramp-down times,
typically to be set in the spin-coater tool.
 31

Spin-coating – typical problems

Source: www.MicroChemicals.com 30

But simple processes like spin-coating can generate quite some trouble. Here, I recom-
mend brochures distributed via the webpages of the spin-coater manufacturers or the
chemical industry as distributors of the spin-coating materials. A very helpful brochure
is that of Microchemicals, but numerous others are available. Please read it before being
disappointed in doing it.
 32

A ‘small’ overview ☺ on resists

31

To give you a feeling, how many different polymers for lithography tasks exist, and this
is the list of a relatively small distributor only. If you decided to work in nanotechnology,
sooner or later, you would likely come across those tables. It is not only the resist; it is
also the developers (after applying light patterns in lithography, you develop the resist)
and removers (after lithography, you typically sacrifice the resist patterns). Much more
information (thermal stability, etch resistance, dry-etch stability, toxicity, etc. ) can be
found in the datasheet of the polymers.
 33

Spray-coating of
resist

Source: www.MicroChemicals.com 32

Sometimes, spin-coating is not a good choice, especially regarding uneven, 3D, or non-
circular substrates. In such cases, spraying the polymer is an option, even though film
formation might not be as smooth as spin-coating. But with some effort, it can work
out, and it is the perfect connecting passage to the last topic of today: spray coating at
the Chair of Nanoelectronics.
 34

Spray coating at NANO: Evolution over the years

33

It started with a very simple airbrush applied in a fume hood. Mounting the brush to
a post gives some stability, and the results might be more scalable. Last, a commercial
spraying robot was purchased, and by programming recipes and using it in a controlled
environment, the results were significantly improved.
 35

Spray coating at NANO: home-made setup for fumehood

34

A cross-section of an advanced airbrush is given here. It would be best to have the

proper nozzle and, more critically, the suitable solvent for your sprayed material. We,
e.g., sprayed polymers, carbon nanotubes, Copper, and Silver nanowires regularly. The
non-uniform spray pattern is a significant drawback that can be overcome by moving the
nozzle or the hot plate with the sample.
 36

Inside the spray gun and patterning by stencil mask

The spray gun is set to a well-defined distance from a hotplate

where the substrate lays on a hotplate. A stencil can be used
to define a particular shape. In the chamber, the atomization is
obtained through a pressurized air flow, which impacts the
material in the atomization nozzle.
A. Falco, PhD Thesis, 2016 35

A close-up view of the spray nozzle. Quite some parameters have to be tested to get
satisfactory results. There is one subtlety worth mentioning: using spray-coating (similar
to PVD evaporation), one can apply a so-called stencil mask – a thin ‘perforated’ metal
sheet, where you can pattern your devices ‘on-the-fly’ (or if you like ‘on-the-spray’ ).
Resolution is limited by the stencil patterning process, where cutting-lasers achieve down
to 50 µm quickly. A drawback: no closed structures like toroids (circular rings) can be
fabricated in a single step.
 37

Spray coating at NANO: an automized spray robot

36

Last, the benchtop dispenser with computer control and boxed housing. I guess the idea
is clear – more reliable processing by some level of automation and parameter control.
 38

Spray coating at NANO: the spray process of Cu NWs

Photos for the spray deposition process using a handheld airbrush: (a) Hot plate at a temperature
of 70 ◦C with a glass substrate on the top side. (b) Spray deposition with the handheld airbrush. The spray-
cone with the CuNW-dispersion, as well as the nozzle-to-sample distance, are sketched. (c) CuNW-film
homogeneously spray-deposited to a glass substrate.
M. Bobinger, PhD thesis, 2018 37

This brings me back to a topic we will discuss in the application part of this lecture,
namely Ag and Cu nanowires for thin-film heaters and antennas. The most critical steps
in the fabrication of nanowire films are given here. It looks like poor man’s (and poor
woman’s) technology, but compared to running simulations on a PC, it is not for free at
all ;-). There is still some investment, and running the chemistry lab produces monthly
costs.
 39

Topography and Carbon Nanotube (CNT) distribution

38

On this last slide of film fabrication, we see the different ways to qualify nanoscale films,
i.e., a spray-coated carbon nanotube (CNT) film. An optical microscope can image
the film with and without dispersant and give a first overview of the micrometer scale
(northeast images). But no chance to look at the nanowires in detail. You don’t see them
in optical microscopes. For that, better resolution in imaging from a scanning electron
microscope is needed (south-west images). With an atomic force microscope (AMF),
we get information on the thickness and distribution of the nanowires (northeast), and
in the best case, the length distribution can be accessed. Indeed, we need a lecture on
metrology and imaging of nanostructures soon.
 40

Q 1:
Thermal evaporation and (in most cases) sputter deposition are PVD methods. Explain
the term CVD and sketch the cross-section of a typical furnace for a CVD process. Show
a chart for convection-diffusion-adsorption followed by deposition and desorption of gases
in the furnace.
Q 2:
Why is it possible to form CVD SiO2 -layers at lower temperatures than in thermal
oxidation? Do you necessarily need a Si-wafer for CVD SiO2 deposition?
Q 3:
What are reaction limited and diffusion limited CVD processes? Write down the propor-
tionality of the rate/magnitude. Hint: D ∝, kr ∝. Which parameters influence the rate
of reaction/diffusion?
Q 4:
Why is it convenient to plot the logarithm of thin-film growth rate versus 1 × 103/T(K)?

Q 5:
Discuss the occurrence of a cosn (Θ) distribution for the incoming flux in CVD processes.
Sketch viewing angle, surface distribution, and the sticking coefficient when filling a
trench in the substrate.
Q 6:
What is the meaning of relaxation when performing heteroepitaxy? What are typical
lattice mismatches acceptable for heteroepitaxy?
Q 7:
ALD is using a chemical key-lock-principle. Explain in your own words.
Q 8:
How to clean a Si substrate? Mention three typical wet-chemical procedures. Sketch a
tool for plasma cleaning of planar (microelectronic) surfaces. How can you remove the
natively grown oxide on a silicon surface?
Q 9:
Spin coating is often used for deposition of resists. Explain the basic steps. Plot the
achieved film thickness versus the rotational speed of the spin coater.
Q 10:
Sketch a typical setup for automated spray-coating of thin films.
Q 11:
Sketch a stencil mask from metal sheets able to fabricate a 50 µm long and 10 µm wide Ni
Hall-bar that can be connected with probe-card needles of 50 µm diameter. How could
one fabricate such a mask?
 41

Q 12:
Why is an optical microscope still of use when imaging nanowire films, even though the
wires are too small to be resolved?