4. Nanometer-thin ﬁlms - Acclab h55.it.helsinki.fiknordlun/nanotiede/nanosc4nc.pdf · 4.1. Growth - There is a wide variety of methods to manufacture very thin ﬁlms. - Most are

4. Nanometer-thin films

[Poole-Owens 6, 7, Wolf ch 7 etc.]

- Thin films have been manufactured for a long time, and some of the methods are very traditional.

Even ordinary painting can be considered a variety of thin film deposition...

- When thin film thicknesses enter the nanometer range, they often do get different or even new

kind of functionality. Then it is justified to describe them nanoscience.

- Also many of the thin film manufacturing methods inherently work on the nanometer scale, making

their manufacturing and analysis a variety of nanoscience.

- Conventional Si-based integrated electronics is now in the nanometer scale. The basic solutions

are so far largely based on downscaling of existing solutions, making it doubtful whether it should

be called nanoscience.

- On the other hand methods originating from Si IC technology are widely used in nanoscience studies

(e.g. to make the interconnects to nanotubes), so at least basic knowledge of IC manufacturing

Introduction to Nanoscience, 2005 JJ J � I II × 1

techniques is almost necessary for a nanoscientist. But on this course we do not have time to go

into that.

- But many aspects of nanometer thin film use are inherently nanoscale effects.

- Hence on this course we give a very brief overview of a few of the most important thin film

manufacturing methods, then present examples of their properties.


4.1. Growth

- There is a wide variety of methods to manufacture very thin films.

- Most are growth methods, i.e. one grows a new layer of one material on top of another.

- The growth methods of nanoscience interest share the common characteristic that the growth

occurs by deposition of single or a few atoms at a time.

- Growth atom layer by atom layer

- Allows for arbitrarily thin films to be manufactured, down to less than a monolayer of atoms, by

controlling the amount of material deposited and/or the deposition time.

- This has also the advantage that multilayers can be manufactured by changing the deposited

material every now and then.

- Downside: it is slow...

- In the following we briefly describe a few of these methods to give you a flavour of how they work.


- Share the characteristics that they are usually referred to with 3-letter acronyms

4.1.1. PVD, Physical vapour deposition

- In PVD methods a thin film is grown by depositing single atoms at a time directly on a surface.

The deposition occurs in a vacuum, typically in the range 10−3 to 10−6 mbar.

- The individual atoms can be made in several ways:

- evaporation: source of coating material is heated close to melting point so that it starts emitting

atoms

- sputtering: source of coating materials is bombarded with atoms so that it emits energetic atoms,

which are then directed to the coated sample. Typically a magnetron is used since from this kind

of sputter source lots of sputtered atoms can be obtained.

The setup principle is easy [Schneider et al, J. Phys. D: Appl. Phys. 33 (2000) R173]. The parts between the atom

source on top and sample at bottom is a rf coil which ionizes the sputtered atoms, allowing for

increasing the plasma density and thus the deposition efficiency.


and here is a picture of a sputtering PVD setup [http://sb2.epfl.ch/instituts/akarimi/pvd.html]: it need not be

too big.


- Pulsed laser deposition (PLD): focused lasers beams with very short (ns or even ps) time pulses

are used to decompose the material.

- Ion implantation: a low-energy ion accelerator is used to shoot energetic atoms at a

sample. Has the advantage that the atoms may go below the surface, which sometimes

allows for making different bonding structure than surface deposition would.


- E.g. for amorphous C deposition an ion energy of about 50 eV has been shown to

make optimally high sp3 bonding fractions, giving the film diamond-like properties.

- Plasma immersion ion implantation (PIII): the sample is immersed in a plasma with a voltage

gradient close to the sample surface. This results in an ion implantation process. Advantage is that

also very irregular-shaped objects can be coated everywhere

Draw picture

4.1.2. CVD, Chemical vapour deposition

- There are numerous varieties of CVD methods (for a list of some 10 different see e.g.

http://www.azom.com/details.asp?ArticleID=1552).

- In CVD a precursor gas of the deposited material is delivered into the reaction chamber. They

pass over the sample to be coated, which is at a high temperature (typically 800 - 1000 ◦ C). When

the molecules come into contact with the surface, they decompose and some part of them stick to

the surface.

Here is a nice sequence of images illustrating this, from [http://cape.uwaterloo.ca/che100projects/vapourdep/]



- Examples of precursor gases: TiCl4, SiH4, Ni(CO)4, metal organic compounds etc. etc.

- The surface temperature is critical to what the end result is.

- CVD often produces hardness and good wear and corrosion resistance.

- Downside is that the high temperature is necessary, limiting use to materials which can tolerate it.

Also the reaction byproduct (what remains from the precursor) may be toxic or flammable.

4.1.3. MBE, Molecular beam epitaxy

[http://www.knovel.com/knovel2/Toc.jsp?BookID=140, http://www.ece.utexas.edu/projects/ece/mrc/groups/mbe.html]

- MBE a ’molecular beam’ of a material impinges on a substrate.

sometimes unkindly referred to as ”Mega Buck Evaporator” or ”Mostly Broken Equipment”

- It is especially important for semiconductor deposition; it was originally developed for GaAs, and

still is the standard way to deposit compound semiconductor thin films. But it can also be used for

e.g. High-Tc superconductors and metals.

- The source material (e.g. Ga, Al and As) are evaporated from individual ovens (one for each


material) into a ultra-high vacuum (UHV) and directed onto a surface. The surface is at a high

temperature, e.g. about 600 ◦ C for GaAs.

- The atoms come one by one or in small molecules which break up at the surface to form the

structure of interest.

Schematic of an MBE source arrangement, with multiple sources for different materials:

[http://www.latrobe.edu.au/ee/mbe/aboutmbe.php]:

- MBE allows for growth of very high quality epitaxial layers, and

thanks to the UHV very pure ones


- Also dopants may be deliberately added

- Downside is that the method is slow and expensive, and requires

suitable surface chemistry to work at all. It is also very expensive.

- But because of its high quality, extremely important method for

surface science.

Fig shows GaAs growth with As:Ga ration 10 from [Mur04]

4.1.4. ALD, Atomic layer deposition

[http://www.helsinki.fi/~eorkm_ww/text/thinfilms/thinfilms.html,http://www.icknowledge.com/misc_technology/Atomic%20Layer%20Deposition%20Briefing.pdf

- In ALD atomic layers are deposited one at a time using subsequent gas pulses. The idea is to

select gas precursors which act in such a way that the process stops by itself after one monolayer

has been deposited, and nothing more will happen before a next gas pulse is fed in. Thus the

process is self-limiting.

- This way in principle one can obtain perfect layer-by-layer growth.

- Here is an example of one cycle for ZrO2:


1-2.◦ ZrCl4 gas is introduced into the chamber. It sticks to the surface, but has the property

that it can not stick on itself. Thus the process stops when one full layer has been

grown.

2.b.◦ (not shown): ZrCl4 gas is removed from the chamber with an inert gas.

3-4◦ H2O gas is introduced into the chamber. The O in it reacts with the Zr, forming ZrO2

while the remainder (H and Cl) vaporize into the gas. Additional oxygen can not stick

onto the top oxygen (remember O is a dimer gas and does not form a covalent solid at

normal pressure). Thus the end result is a perfect monolayer of ZrO2.


4.b.◦ (not shown): the remaining gases (H, Cl and extra H2O) are pumped away.

5.◦ Return to step 1, or quit.

- ALD is usually carried out in the 200 ◦ C – 400 ◦ C temperature range.

Too high T => chemical bonding not sustained or density of reactive sites is reduced. Too low T => chemisorption and film

forming reaction rates decrease, making for slower growth

- Although ALD thus obviously requires the existence of suitable precursors, a wide variety of them

has been found enabling growth of a multitude of materials, such as Ta, Ni, SiO2, Al2O3, WN,

TiSiN, TiN, and HfO2

- The method is particularly interesting for the semiconductor industry to form very thin

layers on Si chips, e.g.

- Ta as a diffusion barrier between Cu and Si

- HfO2 (ot other high-k materials) to replace the SiO2 as the gate oxide in transistors

k is here the dielectric constant, giving D = kε0E for an insulator


- Originally the method was often called atomic layer epitaxy, ALE. Although by the simple

description of the method given above it should indeed produce perfectly epitaxial films, in practice

for most materials defects are eventually introduced and the layers are no longer epitaxial (note that

the sample temperature is lower than for MBE, increasing the possibility to have stable defects).

Because of this the term atomic layer deposition is nowadays preferred.

- Incidentally, the method was originally developed in Finland in 1974 by Dr. Tuomo Suntola and

co-workers, then working at Instrumentarium Datex [T. Suntola, ”Methods for producing compound thin films”,

US patent 4058430], after Russian scientists had in 1965 published the TiCl4 + H2O on SiO2 reaction

[Shevyakov, A.M.; Kuznetsova, G.N.; Aleskovskii, V.B., Khim. Vysokotemp. Mater., Tr. Vses. Soveshch., 2nd (1965), Meeting Date 1965,

149.]

[http://sc.el.utwente.nl/tdr/Projects/HighKOxide/history.html

http://www.sci.fi/ suntola/, Now has his own theory of relativity and cosmology....

- Quick google gallup of the relatice importance of the methods:

- ”Atomic layer deposition” gives 20800 hits


- ”Molecular beam epitaxy” 266000 hits

- ”Chemical vapor deposition” 322000 hits

- ”Physical vapor deposition” 51300 hits.


4.2. Growth modes

[http://www.courses.vcu.edu/PHYS661/pdf/04Growth041.ppt]

- In most cases when thin film growth is desired, one wants smooth layer-by-layer growth

- Except for ALD, this is not trivial to achieve.

- Since the atoms land on random sites, just placing atoms on top of each other at random sites

would lead to a rough surface.

Draw picture

- However, in most cases the growth temperature is high enough that atoms are mobile on surfaces

- There are many different possible surface diffusion mechanisms:


- After arrival a single atom on the surface (adatom ) will migrate on the surface with some

activation energy Ed.

- If it finds another adatom, it may nucleate an island

- If a mobile adatom migrates to an existing island, it is likely to stick there.

- An adatom on top of an existing island can migrate down from it. Note, however, that there may

be a barrier making it difficult to come down from the island. This is called an Ehrlich-Schwoebelbarrier


- (Islands themselves may be mobile, but this can usually be neglected)

Draw schematic Epot diagram for such

- In some cases it is actually possible that it is energetically favourable to migrate up along a barrier.

- Depending on the migration rates, different growth modes can result:


- If the migration rate is high compared to the rate at which new atoms come in, and there is no

(or only a small) Ehrlich-Schwoebel barrier, it is possible to achieve layer-by-layer growth.

- Figure of Au growth on NaCl at constant temperature:


- If the rate of incoming atoms is high compared to the migration rate, and/or there is a high

Ehrlich-Schwoebel barrier, island growth will occur.

- The intermediate case of Stranski-Krastanov growth is possible in a heteroepitaxial system: the

diffusion energetics may then be different on the first monolayer compared to the additonal ones.

- The Volmer-Weber or Stranski-Krastanov growth modes may, and often actually are, utilized

intentionally to grow quantum dots, i.e. nanosize atom mounds on surfaces

- For thin film growth the goal, however, always is layer-by-layer growth. Sometimes this is difficult

to achieve.

- A possible solution is to use a low-energy ion beam during a conventionaly deposition processes.

The idea is that the energy given to the lattice by the ion beam breaks up islands, thus creating

more nucleation sites and facilitating layer-by-layer growth. This method is called ion beam assisted

deposition (IBAD).


4.3. Mechanical properties

[Poole-Owens 6.1.4]

- One of the main application areas of thin films is in the field of hardness.

- Hard coatings are utilized in a wide variety of areas, ranging from antscratch coatings in eyeglasses

to machine part coatings with the aim to improve on their wear resistance.

- Many of the hard coatings involve multilayers of thin films, i.e. one kind of material on top of the

other. Usually they come as multiple alternating sequences of two materials (of same or different

thickness).

- The simplistic explanation of why they may have good properties: as explained above, very hard

materials tend to be brittle, while ductile materials tend to be softer. In principle by joining materials

of the two kinds one could get a combination of ductility and hardness.

- A more scientific explanation is that when a multilayer is formed of two materials with a different

lattice constants, the interface between them acts as a barrier to dislocation motion. This pins


down dislocations, and improves on the hardness of the material as dislocations can not move and

multiply.

- Example: measurement of the hardness of multilayered TiN/NbN thin films:

- Hardness of thin films is usually measured with so called nanoindenters, which are a nanoscale

version of conventional indenters. These simply press some hard material against the measured


material, and measure how deep the indenter penetrates in a material for a given load. The tip is

of some very hard material, typically diamond

- In nanoindenter the tip is of nanoscale size, and the loads applied also much smaller. The

nanoindenter is interesting from a scientific point of view because it measures hardness of a very

small area of the material, small enough that it is not likely to contain dislocations or grain

boundaries. Thus the nanoindentation results may be interpretable from a purely atomistic model

of hardness.


4.4. Electronic properties

Thin films can have a multitude of interesting electronic properties, too many to go through them

all. Here we mention only two of the most important ones.

4.4.1. Insulators in semiconductors

[http://www.itrs.net/Common/2004Update/2004_06_FEP.pdf

The basic MOSFET transistor which forms the core operating block of all modern processors is of

the following type:


- The so called gate oxide is conventionally made of silicon dioxide, and manufactured simply by

allowing the Si to oxidize.

(This is actually the basic reason why Si is used instead of Ge: Ge does not form a native oxide

- But when the chip sizes have been continuously decreasing, so has the oxide thickness. Now

(2005) we are at oxide thicknesses of almost exactly 1 nm which means the oxide is only a few

atom layers thick! [http://www.itrs.net/Common/2004Update/2004_06_FEP.pdf p. 27

- The leakage current over the oxide is increasing and will become

intolerable in a couple of years


- Needs to be replaced by some so called high-k material which

means some very good insulator.

- E.g. Intel has announced they can replace the SiO2 with a

high-k material in 2007 (they don’t disclose what it is except to

say ”most of the high-K gate dielectrics investigated are Hf-based

and Zr-based”. ) [http://www.physorg.com/news80.html]

- The high-k material will be deposited with some of the methods

described above, and most likely this will be ALD which has been

shown to be able to make good HfO2 and ZrO2 films.

4.4.2. Quantum wells

- Another major application area of thin films is manufacturing of so called quantum wells

- We will later during this course discuss the electronic functionality of QW structures

- Here only atomic structure is discussed

- The quantum well structures are multilayers of compound semiconductors, which are chosen such

that they have different band gaps and/or lattice constants.


- The electrical functionality arises from this difference.

- A typical example of a QW heterostucture is alternating thin layers of AlGaAs and GaAs. These

are used as the basic operational layer of semiconductor diodes and lasers.

- Layer thicknesses typically of the order of 100 nm, and there may be tens of them.

- This is a schematic of a typical GaAs semiconductor laser structure (with only a few layers)

[http://britneyspears.ac/physics/fplasers/fplasers.htm]

Also uses as light detectors


- Even more layers are rtoutinely used in so called vertical cavity surface emitting lasers (VCSEL’s).

In these one of the key components is a mirror called a distributed Bragg reflector (DBR), which is

formed by multiple alternating semiconductor layers.

[http://britneyspears.ac/physics/vcsels/vcsels.htm]


4.5. Magnetic properties

[Poole-Owens 7.6, http://www.stoner.leeds.ac.uk/research/gmr.htm]

A major interest into thin films rises from the development of the read heads in hard disks.

- A hard disk stores information as magnetized bits in the disk itself.

- To read it, one thus needs something which convertes a magnetic field into a measurable electric

signal

- Initially this was done using traditional magnetic coils

- One alternative possible way of doing that is to use magnetoresistance (MR), the effect where a

magnetic field changes the resistance of a material: R = R(B)

- Magnetoresistance does exist in conventional materials, but is usually a very small effect, much

less than 1 %.

- This changed dramatically in 1988, when so called Giant magnetoresistance or GMR materials

were discovered. In these the MR is as large as a few %, whence the word giant was taken into use


- Since then people found materials with MR of tens of %:s, and later with 100’s of %’s, and started

to run out of superlatives: after Giant MR the terms Huge MR and Colossal MR were taken into

use... Nowadays the terms GMR and CMR persist.

- With GMR and CMR a much better sensitivity can be achieved than in traditional

coil-based read heads, and they have been in industrial use since around 1996

[http://www.research.ibm.com/research/demos/gmr/index.html]

- The GMR effect exists in alternating layers of a ferromagnetic and nonferromagnetic material.

- The effect was first discovered in Fe and Cr, but since then numerous other alternatives have been

found. E.g. combinations of Co and Cu display much higher effects.

- Also the geometry can be varied between multilayer and embedded nanoparticle systems:


(a) the basic simple case

(b) Randomly oriented Co nanoparticles embedded in Cu

(c) Co nanoparticles in an Ag matrix, sandwiched between NiFe magnetic layers with alternating

magnetization


- All of these cases produce CMR

- GMR or CMR is related to how electrons of a different spin state interact with magnetic layers.

- It is known that conduction electrons with a certain spin state scatter more easily from a magnetic

layer with magnetization oriented against their own spin state.

- Hence a simplistic view of GMR is as follows:


- On the left hand side electrons with spin up can pass both layers, since their spin is the same as

that of both layers.

- On the right side neither kind of electrons can pass easily, making for a different resistance!

“To get a little more technical, the band structure in a ferromagnet is exchange split, so that the density of states is not the same

for spin up and down electrons at the Fermi level. Fermi’s golden rule states that scattering rates are proportional to the density of

states at the state being scattered into (in this case the Fermi level), so the scattering rates are different for electrons of different

spin. These ideas were used as early as 1936 by Sir Neville Mott to explain the sudden decrease in resistivity of ferromagnetic

metals as they are cooled through the Curie point.”

- The relation to an outside magnetic field and the multilayers can in a highly simplified form be

understood as follows: an external magnetic field M will reverse easily the magnetization of the

non-ferromagnetic layer, while the ferromagnetic layer keeps it magnetization longer. Hence the

external field can make half of the multilayer system change spin states, i.e. switch between the left

and right part of the figure above. Thus the resistance change comes about!

- There is an inherent reason to why the layers involved need to be in the nanometer regime: the


mean free paths of electrons are of the order of a 100 nm, so if the layers are very thick the electrons

scatter and change direction several times within one layer, and the effect no longer works.

The above is my own idea, is it really true?. Yes, see J. Phys. Condens. Matter 11 (1999) 5717

- The MR is also dependent on the layer size in much more complicated ways related to at what

rate the magnetic fields actually change directions, in ways we will not discuss on this course.

- The effect depends clearly on layer thickness:


( A nice set of animations on GMR read heads from IBM: http://www.research.ibm.com/research/demos/gmr/index.html)


Documents

4. Nanometer-thin ﬁlms - Acclab h55.it.helsinki.fiknordlun/nanotiede/nanosc4nc.pdf · 4.1. Growth - There is a wide variety of methods to manufacture very thin ﬁlms. - Most are