Tuning defect density in single

layer graphene by ion

implantation technique

Lee Yan Long

A0111488E

National University of Singapore

Supervisor: Dr. Pattabiraman Santhana Raman

Co-Supervisor: A/Prof Jeroen Anton van Kan

A thesis submitted in partial fulfilment for the degree of

Bachelor of Science (Honours) in Physics

2017 April

Abstract

Current technologies in water desalination always face the issue of be-

ing hard to implement in less affluent countries due to the high cost

involving membrane replacement and power requirements for com-

mercial use. Single layer graphene membrane has been highlighted to

be a suitable substitute [1] for the current polymer membrane that is

lower in cost yet higher efficiency.

In this report, we attempt to adjust the defect density on single layer

graphene supported on silicon oxide by ion implantation in order to

investigate the defect density that the layer can support. We plan

to optimise different parameters such as energy and dosage of ions to

maximise the defect density. Such result can then be applied onto free

standing graphene on gold grids to help save cost. We then analyse

and characterise the sample after ion beam irradiation using Raman

spectroscopy.

They were later separated into their respective carbon amorphisation

regime and this is used to determine unwanted results (i.e. highly

amorphous carbon). The carbon regime is also used to calculate their

defect density to determine the optimised dosage and energy. Within

experimental limits, we observe that maximal defect density is about

10keV while remaining highly sp2 hybridised and that the upper limit

for the dosage is identified to be about 1016 ions/cm2. The trend of

the defect density was explained using stopping power in the linear

cascade and were checked using simulation from Kalypso. Overall

results align quite closely with theory but remain limited due to small

energy range used in this work.

Acknowledgements

Firstly, I would like to express my heartfelt thanks to my supervisors,

for the opportunity to work with one of the most interesting mate-

rial, graphene. It was mainly from Professor Jeroen and Dr Raman

oversight of the project that made it all work out in the end.

I would also like to express my greatest gratitude to Dr. Tanmoy Basu

who was with me throughout my entire lab work, clarifying my doubts

and being the role model researcher. The laboratory technicians Raj

and Mdm Ng Soo Ngo were also crucial in this project by assisting me

in clearing up the issues I have caused when using their equipment.

Of course, a call-out to all the other members of the CIBA lab; Huei

Min, Sarfraz and those who were not named for all their kind support

during this project. Thank you so much.

This project has also been made memorable with other Final Year

Project students that had to struggle through their own projects.

Andrew Seah, Zhang Danwei and Stella Ng whose presence had made

this journey more enjoyable than it would have been. I wish them all

the best for their projects.

Most importantly, I would like to thank my family members who were

with me through thick and thin throughout my entire student life.

Contents

List of Figures vi

List of Tables vii

1 Introduction 1

1.1 Motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.1 Advantages of Single-Layer Graphene (SLG) . . . . . . . . 2

1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Fabrication methods 5

2.1 Methods of Fabrication . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2 Ion Beam Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3 Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3 Related Theories 11

3.1 Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . 11

3.1.1 Peak features . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.2 Type of defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.2.1 Defect density calculation . . . . . . . . . . . . . . . . . . 18

4 Results & Analysis 20

4.1 Simulation method . . . . . . . . . . . . . . . . . . . . . . . . . . 20

4.2 Experimental result . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4.3.1 Ordering in graphene . . . . . . . . . . . . . . . . . . . . . 25

iv

CONTENTS

4.3.2 Defect density . . . . . . . . . . . . . . . . . . . . . . . . . 28

References 32

A List of Spectra peaks 35

A.1 Data table for all peaks of collected spectra . . . . . . . . . . . . 35

v

List of Figures

1.1 Schematic diagram of a conventional desalination membrane . . . 1

1.2 Diagram of water flow . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Side-by-side comparison of Multilayer graphene and SLG . . . . . 3

1.4 Flowchart of the membrane control . . . . . . . . . . . . . . . . . 4

1.5 Image of expected final product . . . . . . . . . . . . . . . . . . . 4

2.1 Plasma irradiation of the graphene membrane . . . . . . . . . . . 5

2.2 ISTB in Centre of Ion Beam Application (CIBA) . . . . . . . . . 6

2.3 ISTB schematics (CIBA) . . . . . . . . . . . . . . . . . . . . . . . 7

2.4 ISTB interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.5 Film carrier method . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.6 TEM graphene holder . . . . . . . . . . . . . . . . . . . . . . . . 10

3.1 3 different type of scattering of photons in a material . . . . . . . 12

3.2 Classical illustration of Selection rules . . . . . . . . . . . . . . . 14

3.3 Band diagram of the transition for Raman peaks . . . . . . . . . . 15

3.4 Raman Spectrum of Graphene. . . . . . . . . . . . . . . . . . . . 16

3.5 Graphene defects . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4.1 Kalypso simulation visualizer . . . . . . . . . . . . . . . . . . . . 20

4.2 Logarithm graph of the number of atoms to the atomic displace-

ment from origin in the z-direction . . . . . . . . . . . . . . . . . 21

4.3 Sputtering yield result . . . . . . . . . . . . . . . . . . . . . . . . 22

4.4 Typical stopping power during ion implantation . . . . . . . . . . 23

4.5 Stages or regime of graphite . . . . . . . . . . . . . . . . . . . . . 26

4.6 Images from AFM . . . . . . . . . . . . . . . . . . . . . . . . . . 28

vi

4.7 Defect density vs Energy . . . . . . . . . . . . . . . . . . . . . . . 29

4.8 Defect density vs Dosage . . . . . . . . . . . . . . . . . . . . . . . 29

List of Tables

4.1 Chart of D peak position, I(D)/I(G) ratio and 2D peak FWHM . 27

A.1 Table of intensity peaks for all the spectra . . . . . . . . . . . . . 35

A.2 Table of peak shifts for all the spectra . . . . . . . . . . . . . . . 36

vii

viii

1

Introduction

Most desalination filters are based on the idea of reverse osmosis based on polymer

films. It is made by having a large polyester base as a support and a polysulfone

layer as a substrate with a thin layer of polyamide layer that serves as the actual

filter.

Figure 1.1: Schematic diagram of a conventional desalination membrane

- The layer of polyamide grown through interfacial polymerisation (IP) [2]. Image

adapted from Sydney Water Corporation.

This membrane is then rolled into a tube where differential pressure is applied

from one end of the tube to the other. As the water passes through the membrane,

desalination occurs as salt were trapped and unable to permeate into the inner

columns. Water has to pass through several columns of such permeable membrane

into the inner column where the water is considered desalinated.

However, it has been discussed that a graphene based material will be better

suited for such an application instead.

1

1. INTRODUCTION

Figure 1.2: Diagram of water flow - The amount of portable water obtained

is shown in the stream of water at the tube centre. Salt is rejected away from the

core and remains stuck in the outer column. Image adapted from Sydney Water

Corporation.

1.1 Motivations

In a paper published in 2013, a Massachusetts Institute of Technology (MIT)

research team has completed their Density Functional Theory (DFT) calculation

and proved that it was possible for a graphene membrane to sustain average pores

of 7A ideal for desalination purpose [1].

1.1.1 Advantages of Single-Layer Graphene (SLG)

It is expected that the graphene membrane would improve conventional desalina-

tion processes since reverse osmosis requires high pressure which is not required

by the graphene filter [3]. Also, conventional methods requires high maintenance

since the membrane has to be changed regularly in every few months or even

weeks if desalination occurs in brackish water [2] due to the absorbance of salt

particle on the membrane.

There are also many other possible material such as a multilayer stacked

graphene oxide membrane which could also form nanochannels for filtration [4].

While such layers are easy to manufacture, this filtration technique is less

direct since the channels for water flow is from between the gap of each layer

that can potentially be equally hard to control. Furthermore, direct pores on

graphene is a much more efficient method since the layer is much thinner, leading

to a higher diffusion of water molecule.

2

1.2 Objectives

Figure 1.3: Side-by-side comparison of Multilayer graphene and SLG -

(a) shows the multilayer graphene while (b) shows SLG. The water molecules are

depicted by the red and white spheres were allowed to pass the membrane while

the rest of the larger ions such as Na+ (purple sphere) and Cl− (green sphere) are

impermeable. Image adapted from [4].

1.2 Objectives

As mentioned in the abstract, the key objective of this report is to mainly obtain

a maximised defect density of Single-Layer Graphene (SLG) on silicon substrate

and extend the result onto free standing graphene. The rationale is to minimize

cost since operating directing on the TEM grids for the free standing graphene can

be expensive and requires a high level of technical knowledge. After optimizing

the defect density, we will then proceed onto the TEM grid to evaluate the size

of the pores we create and fine-tune the pore through the use of other gas source.

While it is possible to use the alternative Ar+2 ions, the wien filter which is simply

a velocity selectoer did not have sufficient E/B field strength to filter through the

larger projectile.

Figure 1.4 shows a flowchart of the project.

Upon collecting the base data which optimised the defect density, we would

be characterising the type of defects under Transmission Electron Microscope

(TEM) that is observed from the ion beam irradiation and investigate the average

pore sizes across the graphene. Do note that Transmission Electron Microscope

is best performed under low voltages of 60kV to avoid reconstruction of the

3

1. INTRODUCTION

Figure 1.4: Flowchart of the membrane control - This is the rough sequence

of the project. First, we start trying to obtain a baseline comparison for the defect

density formed on the supported graphene before moving to further stages.

graphene surface as will be mentioned later in section 3.2. However, this will not

be explored in this phase of the project due to time constraint.

The final step before passing over the project to the engineers to improve

and implement the membrane is to achieve the ideal pore size and test the ion

transport across the membrane.

Figure 1.5: Image of expected final product - The graphene layer is grafted

onto a polyester tube. Pressure will be exerted from the top and water will flow

into the tube of diameter 2R and filtered through pores of diameter 2a where the

ideal size of 2a is about 7A. Choice of polyester will be based on compatible surface

free energy with graphene. Image adapted from [1].

4

2

Fabrication methods

2.1 Methods of Fabrication

There were many potential methods meant for formation of such defects. Oxygen

plasma and beam irradiation are highlighted to be best methods.

For plasma, the key idea is to form a plasma above the sample and use the

oxygen to oxidise the carbon and liberate it from the sample.

Figure 2.1: Plasma irradiation of the graphene membrane - The above

schematics shows how the experiment was carried out. The blue material illustrates

SiN supported on Si. A circular hole of 2micron diameter is drilled using Focused

Ion Beam drilling before the graphene layer is grown on top of it. Image adapted

from [5].

While some samples were found to be water permeable with sufficiently ideal

pore sizes, the pores were found to be distributed fairly randomly and large

5

2. FABRICATION METHODS

spread of pore sizes was observed [5]. Furthermore, success rates was also low

as water permeability was only observed in 20% amongst 140 samples [5]. Due

to the irregularities in pore formation through such methods, there is a need of

exploring more concrete methods of pore formation.

Electron beam irradiation have also been tested in some research paper but

the observed pore size was observed to be too small (∼3A) [6]. It was believed

that the low mass of the electron was the cause.

Thus, for a controlled means of pore formation, the most direct method would

be ion sputtering. We would be using the Ion Source Test Bench (ISTB) which

operates in the high vacuum (∼10−6 mbar). This ion beam machine allows an

indirect control over the dosage via monitoring the current and has multiple

adjustable beam size set by the aperture within. The source of the beam is

generated through Radio Frequency (RF) voltages which requires lower pressure

to sustain a plasma.

Figure 2.2: ISTB in Centre of Ion Beam Application (CIBA) - Equipment

and setup was designed and assembled by CIBA.

6

2.2 Ion Beam Sputtering

2.2 Ion Beam Sputtering

To ensure proper alignment of the beam spot, we calibrate the beam spot for each

energy and aperture position on a sliver plate before performing the sputtering

on our sample. The energies of the Ar+ ions are filter by a wien filter (velocity

selector). Only ions of selected charge and velocity, controlled by the potential,

are allowed to pass into the aperture undeflected.

The beam then passes through 2 aperture that can be manipulated to align

and focus the beam. A copper foil is attached to the last aperture and blocks the

stray beams of argon from reaching the sample.

Figure 2.3: ISTB schematics (CIBA) - The beam focus is monitored by the

probe and extraction potential. The ion is then accelerated through the potential

column until the desired energy. The first aperture aligns the beam while the wien

filter deflect ions of different energy. The second aperture helps control the beam

size. Red line indicates the direction of beam travel.

We can also monitor the dosage via the use of the current. Dosage here refers

to total dosage per unit area (cm2) henceforth. Once the Ar+ ions collide into

7

2. FABRICATION METHODS

the graphene sample, it will transfer this charge onto it. Since the sample is

relatively conductive as the only non-conductive part of the sample is the thin

layer of silicon oxide layer, current will be able to flow from the charge integrator

to neutralize the charged sample. This current is monitored and calculated to

serve as our indirect measurement of the dosage. Sometimes, secondary electrons

could also be emitted from the sample during sputtering and thus a voltage of

∼+90V is applied on the sample to suppress the emission.

Figure 2.4: ISTB interface - This is the interface used to control all the pa-

rameters of the beam. The interface was designed by CIBA members. The photo

shows the parameters used for optimising the beam (maximising the current) for

15keV.

Energies of 5keV, 10keV, 15keV Ar+ ion beam are selected to obtain a rough

trend of the defect density. The choice of the energies is arbitrary and based on

the energy range of the ISTB. We expected that at high energies, we could get

higher defect density with the same dosage. The beam are all aligned to sputter

at the centre of the graphene where defects were observed to be minimal during

the transfer of graphene onto the substrate [7].

8

2.3 Sample preparation

2.3 Sample preparation

Although graphene samples are grown in the C2D centre, it has to be discussed

here since the graphene transfer method could potentially induce defect before

irradiation.

The graphene sample is grown via chemical vapor deposition (CVD) on typ-

ically copper substrate. This is chosen for its low solubility with carbon that

prevents diffusion of the carbon atom into the substrate. The substrate is pre-

treated with argon and hydrogen gas and later annealed. This ensures common

orientation that maximises homogeneous and defect-free growth. The thickness of

the substrate was also noted to be 25micron where layer-by-layer growth occurs.

In CVD growth of graphene, we used methane gas as precursor since it can

decompose at the lower temperatures (∼ 1200K) even without catalysis. The

product of hydrogen gas after adsorption will be carried out by the gas flow. The

growth occurs for 30minutes at about 1000K (copper serves as the catalysis as

well).

Figure 2.5: Film carrier method - Crystal size was found to have millimeter

single crystal domain in this method. However, quality is still inferior to the scotch-

tape method. Image adapted from [8].

Here, we use the carrier film method to transfer the film over from onto

silicon. We spin-coat poly(methyl methacrylate) (PMMA) onto the sample and

9

2. FABRICATION METHODS

later dissolve them into solution of iron chloride (FeCl3) to remove the copper

substrate. The graphene layer held by the PMMA is then further treated by

acetone to remove any copper residue and soft-baked onto the silicon substrate.

We then remove the PMMA with toluene. This method is the easiest by far

but the crucial downside to this method is that the PMMA residue cannot be

entirely removed, thus induces a lot of contamination. Nevertheless, we attempt

to conduct our spectroscopy in regions of low contamination as seen under optical

microscopy (see Figure 4.6). In addition, we do note that the Full Width Half

Maximum (FWHM) for 2D is quite close to theoretical value from the AFM,

indicating little defects during transport.

For the latter portion of the experiment, we had to analyse the size of the

pore through TEM and thus, the free-standing sample was prepared and placed

on gold grids instead for characterization.

Figure 2.6: TEM graphene holder - (a) shows the sample used in this work.

The faint contrast shows the graphene layer on the blue silicon substrate. The size

of the graphene is about 5mm by 5mm. Choice of substrate (Si with Thin layer

of SiO2) is primarily to enhance Interference Raman Scattering effect. (b) shows

the sample for TEM characterisation for future work. The TEM grid that holds

the graphene sample is at row 10. The holder is sealed and vacuumed to reduce

contamination. The grid sizes are 1micron placed 2micron apart while the entire

grid has a diameter of 3mm and is 250micron thick. Gold is chosen due to its inert

properties yet high conduction to suppress noise signal.

10

3

Related Theories

The primary aim of the project involves identifying the defect density from the

sample. Thus the most appropriate method would be Raman spectroscopy. This

method is non-destructive and would not affect the sample too much especially

at low power of 3mW for short durations (scan time: ∼300seconds) [9]. 514nm

wavelength was used instead of the standard 532nm for better resolution of the

peak [10].

In graphene, there are several features in the spectrum that could be observed.

To understand these spectra, we would need to first understand the mechanics of

Raman spectroscopy.

3.1 Raman Spectroscopy

Raman spectroscopy operates on the key principles of excitation of the molecules.

As the photon from the laser strikes the sample, it can scatter inelastically with

the various modes of the crystal array, resulting in higher or lower wavelength

based on the transition. For instance, when it interacts with the ground state, it

could excite it to a virtual state before de-exciting to a vibrational or rotational

state. This is known as the Stokes scatter and is the most commonly observed

transition since most of the atoms are in the ground state during characteriza-

tion. Fundamentally, only vibrational and rotational states that causes the dipole

moment to change will be pronounced in the spectrum. Thus, not all possible

11

3. RELATED THEORIES

states can be seen through this spectroscopy. Raman sensitive states are deter-

mined by Group Theory. (E2g and A1g seen in literature refers to the group of

the vibrational states.)

Figure 3.1: 3 different type of scattering of photons in a material -

Rayleigh scattering (in green) is the elastic scattering of the photon whereas Stokes

and Anti-Stokes scattering are inelastic. Only virtual states that correspond to

rotational and vibrational mode will be observed in the Raman spectrum. Image

adapted from wikipedia.

As such, the shift in the wavelength of the outgoing laser will correspond to

the energy between the energy level transition. However, the mechanism can be

much more complex. Ideally, in a perfect crystal, these transitions have to obey

selection rules based on the conservation of the energy and crystal momentum

with the crystal phonons of the optical branch shown as such:

k1 = k2 + q (3.1)

k1 = k2 + q1 + q2 (3.2)

Equation 3.1 and 3.2 show the selection rule for one phonon and two phonon

processes respectively where q refers to the phonon contribution. Note that there

are more than one two-phonon processes than can be allowed. For the one phonon

12

3.1 Raman Spectroscopy

process, the recombination must occur k1=k2 for the process to be Raman sensi-

tive, hence q is necessarily zero. However in the presence of a defect state, this is

no longer obeyed (q 6= 0).

3.1.1 Peak features

In short, the Raman spectra give the various energies of the vibrational and

rotational mode within the sample. Thus, if defects are introduced, these modes

will be disturbed and we can observe different modes.

When a laser acts on the sample, the photon from the laser excites and create

and electron-hole pair that emits phonons before recombining into a photon of

lower energy. This recombination must also occur at the initial position [11].

One of the common peak is known as the G peak. It reflects the stretching

mode of the C-C atom in the graphene and is activated by electrons near the

zone centre. For a perfect crystal, since there are no dispersion by the phonons

(q=0) at the zone centre due to the conservation of crystal momentum, this peak

is always fixed at 1580cm−1, independent of excitation wavelength. The width

and peak position of this peak is sensitive to the strain and deformation of the

zone centre and can be observed via the peak shift in G band [12].

The next two frequently observed peaks are known as D and 2D peaks. This

two peaks arise due to the breathing mode of the carbon rings. In the breathing

mode, carbon atoms in the hexagonal ring expands outwards and then inwards

periodically [13].

Graphene has a dirac cone structure when following the phonon dispersion

along K point to the K′ point. Likewise, the photon can produce an electron-hole

pair. If this electron emit a phonon along K to K′ point, another phonon will

be emitted to back-scatter that electron from K′ to K. Suppose the graphene is

defect-free, the total conservation of crystal momentum will have ensured that

the net momentum of the phonon must be zero (q=0) this is only allowed if the

phonon are of opposite wavevector. The shift in the wavenumber would then be

based on the initial amount of energy the electron has which is proportional to

the wavelength of the laser. This shift is about (∼2700cm−1) when using 514nm

green laser.

13

3. RELATED THEORIES

Figure 3.2: Classical illustration of Selection rules - The green line shows

the phonon scattered while the solid black arrow shows the electron path. The

lightning and explosion symbol represents electron-hole production and recombi-

nation respectively. All of these are two phonon processes (since we have 2 green

lines for each image). (i) demonstrated a process that happens when q1 + q2 6= 0

in a perfect crystal (the sum of the 2 phonon wavevector do not cancel). While

momentum is conserved, the electron hole pair cannot recombine to form the pho-

ton to emit for a wavelength shift to be detected (i.e. incoming laser 514nm, no

outgoing beam [recall: Raman shift needs a change in wavelength so you need both

an incoming and outgoing photons to detect the change]). (ii, iii and iv) demon-

strates the allowed transition. For (iv), the black spot illustrates a defect that can

scatter the electron or hole resulting in an allowed recomination even though net

momentum is not zero. Image adapted from [12].

14

3.1 Raman Spectroscopy

If we now consider a defect, the crystal structure is destroyed and crystal

momentum need not be conserved. This meant that the electron can scatter

a phonon that can be back scattered by a defect state (or vice versa). Since

the phonon momentum is not longer strictly observed as in Equation 3.1 and

3.2, more transitions are allowed [14]. These defect states manifest itself as the

D peak and with increasing defects, it further relaxes momentum conservation,

leading to a wider D peak.

However, note that not all the defect type will be reflected in the intensity

or width of the D peak but also the presence of another peak known as D’ peak

(∼1370cm−1) or additional shifts in the G peak (if it causes strain).

Figure 3.3: Band diagram of the transition for Raman peaks - Green and

red line illustrates absorption and emission of photons respectively while the solid

line illustrate the scattering by the electron. The electron transition must also

obey conservation of crystal momentum as seen from the G and 2D band where

the total phonon momentum is zero. While the energy transition can vary, only

those transition that correspond to an actual vibrational state will resonate. Thus,

only at some energy value, we see a Raman peak. However, this conservation is

not observed in the D band due to the defect state. Image adapted from Institute

of Optics, University of Rochester.

Core details are as follow: 2D peak will always be observed in both highly

symmetrical graphene or low defect graphene while the D peak will only be present

in defect graphene.

15

3. RELATED THEORIES

Figure 3.4: Raman Spectrum of Graphene. - The key features reflecting

the mechanism mentioned is shown here. D’ peak is the intravalley case of D peak

where the transition is only about K or K’ instead of K to K’ or vice versa. D’ peak

is also sensitive to several type defect that may not be obtained from the intensity

of the D peak such as the zigzag edge. Image adapted from William Marsh Rice

University.

3.2 Type of defects

Since the order of the pore size we are looking for is in the order of nm, the

defects are mostly localised thus in the this section, the primary focus would be

on point defects rather than 1D defects.

Graphene can easily reconstruct into non-hexangonal rings even when there

are no point defects. The most commonly known would be the Stone-Wales defect

where the rotation of one the C-C bond would cause 2 pentagons and 2 heptagons.

This defect is generally negligible at room temperature but it can be very stable

after ion beam once formed under ion beam irradiation [15]. Nevertheless, since

there are no dangling bonds or vacancies formed from such defect, it is not the

defect we are primarily concerned with since the end goal involves pore formation.

Specific to our case, for singly charged ions, most of the defects are expected

to be single or double vacancies [16]. For single vacancies, one atom is removed

from the lattice and the graphene is reconstructed to form a 5 membered ring

and a 9 membered ring. This is known as the V1(5-9) defect.

Even though the formation energy is high (∼7.5eV), the migration energy is

very low (∼1.5eV) and thus will be an important factor in the redistribution in

16

3.2 Type of defects

the pores formed. When 2 of such defect combines through migration, it can

lead to a double vacancy site where the graphene will reconstruct again to form 2

pentagons and 1 octagon [V2(5-8-5 defect)]. This defect has a formation energy

of 4eV per atom which meant this double vacancy is more favored than the single

vacancy. By rotating the bonds in the octagon, it can be transformed to the

V2(555-777) defect of 3 pentagons and 3 heptagon.

Figure 3.5: Graphene defects - Structures of reconstructed graphene (a)

V2(5-8-5) (b) V2(555-777) (c) V2(5555-6-7777) This defect is formed via ro-

tating a bond from a heptagon in V2(555-777) (d) V1(5-9). Image Adapted from

[17].

When many atoms are removed at once, such as during sputtering, the re-

moval of surface will also led to pores of dangling unsaturated bonds. However,

investigation of the type of defects would not be possible without the use of TEM

and hence this portion is mainly a build-up for the future progression of the

project.

Nevertheless, such information also meant that at visible light range (∼ 2eV)

and at low power (∼ 2mW), defects are only able to migrate at best. This has

been supported by other papers [15][17].

17

3. RELATED THEORIES

3.2.1 Defect density calculation

Since each of the modes can be reflected on the spectrum due to resonance from

the highly symmetrical band structure, the general consensus have been to use

those peak intensities to evaluate the corresponding defect density.

As the sample becomes increasingly amorphous, a D peak appears with in-

creasing I(D)/I(G) ratio where I(X) refer to the intensity of the X peak. D′ peak

slowly appears as the peaks starts to broaden due to less stringent constraint on

the selection rule due to increasing defects [18]. At low defect regime, we can

interpret I(D) as the average defect probed by the laser over the area of the laser

while I(G) is proportional to the area itself.

I(D)

I(G)∝ (Ll)

2/(Ld)2

(Ll)2=C(λ)

Ld

2

(3.3)

Ld refers to the average length between defects and Ll refers to the length

covered by the laser while C(λ) is some constant related to the excitation energy.

This equation is known as the Tuinstra Koenig relation [19].

At higher disorder to the state of amorphous carbon, I(D) will start to drop

as the amount of sp2 hexagonal rings began to be distorted while I(G) stays the

same since the relative motion of the sp2 atoms remains mostly unchanged [18].

I(D)

I(G)= C ′(λ) ∗ L2

d (3.4)

The values of the constant has been computationally determined [20] and

fitted to many different paper [21][12][1][19]. We did not managed to verify such

proportionality ratio due to the lack of excitation energies we could use since

we have only characterised our sample under 514nm. After substituting the

excitation wavelength into (3.3) and (3.4), it reduces the former equation into

this:

For low defect regime (Ld > 10nm)

L2d(nm

2) = 126× [I(D)

I(G)]−1 (3.5)

n2D(cm−2) = 2.5 ∗ 1011 × [

I(D)

I(G)] (3.6)

18

3.2 Type of defects

For high defect regime (2.5nm < Ld < 10nm)

L2d(nm

2) = 183× I(D)

I(G)(3.7)

n2D(cm−2) = 1.7 ∗ 1013 × [

I(D)

I(G)]−1 (3.8)

nD refer to the defect density which is related to the length between defect

(Ld). While the actual equation is much harder to derive, this simpler model

fitting should do well to fit the regime. However, even the high defect regime will

no hold (Ld ∼ 2nm) [14].

19

4

Results & Analysis

4.1 Simulation method

In order to attain some expectation of defects formed expected from the various

parameter, we used Kalypso [22] to simulate the sputtering on the sample based

on the energies and species of the ion. Kalypso is a molecular dynamics (MD)

simulation that operates on the leapfrog algorithm for non-equilibrium conditions

such as sputtering and collision calculation in the classical framework.

Figure 4.1: Kalypso simulation visualizer - The diagram shows the time

evolution in one of the runs at 15keV. The timestamp demonstrates the sputtering

from t=0.65fs onwards. The sample was viewed from the sides for the top two

image while the last image was from the top view.

20

4.1 Simulation method

Kalypso is used to calculate the sputtering yield we expected to obtain for the

different energies that we are sampling. Using the coulomb potential obtained

from Li [23], we explore the sputtering yield of a singly charge argon atom on

a free standing graphene. One key assumption in doing so is that the Ar+ ions

interact independently between one another. Yet it saves computational time.

The target is a graphene monolayer of 1×105 atoms at 300K. 1000 trials were

recorded.

All the potential used are empirical potential (i.e. potential based on ex-

perimental fitting). The repulsive potential between the C-Ar, C-C interactions

follows the Ziegler-Biersack-Littmark (ZBL) while that of the attractive potential

follows the Brenner potential [23].

When using the program we also varied the timestep based on each parameter

used in order to allow the layer to achieve local energy minimization. Total time

for each simulation was also obtained by based on the timestep. The sputtering

yield was obtained by analyzing the final position of all the atom cumulatively and

the average value of the 1000 trials for each energy assigned is used to represent

the yield.

Figure 4.2: Logarithm graph of the number of atoms to the atomic

displacement from origin in the z-direction - This is the final record position

of all atoms across all run for 15keV. As seen, most of the atoms remains within

4.5A which is used as an indicator. Any atoms beyond 4.5A will be deemed to

have been sputtered and liberated. There is an increasing trend in number of atom

between 1.5A and 4.5A due to the dislocation perpendicular to the surface caused

by the target.

21

4. RESULTS & ANALYSIS

From this simulation, we can observe how the yield varies at different energy

per atom basis.

Figure 4.3: Sputtering yield result - Plot of the number of sputtered atoms

per graphene atom at the various energies.

In Figure 4.2, we obtain a trend of the atomic position of all the atoms after

each run. However, the atoms would be displaced from the z-axis and only at a

sufficient far distance from the plane would the atom be considered removed or

fully dislodged. The final yield over all the energies are then collated and plotted

in Figure 4.3.

At low energies, nuclear stopping is much more dominant and that causes the

most amount of damage in graphene. However, that starts to drop as electronic

stopping starts to dominate. This is embedded within the ZBL potential [24].

For our simulation, this peaks at 5keV where the damage is maximised. While

stopping power of Ar+ in such setup were not found in known studies, Stopping

Ranges of Ion in Matter (SRIM) program [10] demonstrates that the stopping

power at lower energies for Ar+ projectile on amorphous C is mostly nuclear in

nature (Se/Sn ∼0.2).

22

4.2 Experimental result

Figure 4.4: Typical stopping power during ion implantation -

This figure demonstrates how the stopping power for a heavy element be-

haves empirically. Since the speed of our ions are slow, nuclear stop-

ping is more dominant then electronic stopping. Image adapted from:

http://www.jhaj.net/jasjeet/tcad/Learn3/l3d.htm.

4.2 Experimental result

23

4. RESULTS & ANALYSIS

24

4.3 Analysis

The above spectra are results that were obtained with cooperation from In-

stitute of Material Research and Engineering (IMRE) and sorted based on the

different energy of the ion.

They show the general trend of increasing disorder in the graphene across

the different energies. It can be seen across the peaks that the disorder start to

increase at increased dosages, the D peak appears and D′ peak appears later as all

the peaks broadens. The peaks broaden due to the increased in allowed modes of

phonon as previously discussed. This continues as the G and D’ peak converges

since they became so wide that join as a combined peak at about 1600cm−1 [12].

4.3 Analysis

4.3.1 Ordering in graphene

In carbon related materials, A.C. Ferrari and J. Robertson has classified the

stages of the material into 3 stages as shown in Figure 4.5.

We also need to check that the sample remains largely monolayer even after

sputtering for application purposes. As mentioned, the 2D peak Full Width Half

Maximum (FWHM) can be used to monitor the thickness of the carbon layer

since increasing the thickness allows more modes, resulting in a wider, shorter

peak of higher frequency. Furthermore, width of any peaks are also defect induced

so it can signal defect states as well. Since most of the peaks for 2D are single

peaks, using the FWHM would be sufficient as an indication of layer number.

Alternative method would be to use the actual intensity of G peak [21]. However,

the 2D peak is much more useful for this case since could also be used to deduce

the level of hybridisation since the peak arises mainly due to the breathing mode

(indicating hexagonal-arranged atoms).

Literature record [11] reveals monolayers to have FWHM of 30cm −1 while

double layers to be about 50cm−1. Despite having a larger FWHM as compared

to literature value, Atomic Force Microscopy (AFM) analysis has shown that it

is largely monolayer even at 38cm−1. The step edge is found to be about 0.84nm

[26] which is about the step edge of a graphene monolayer. Figure 4.6 shows

supporting images.

25

4. RESULTS & ANALYSIS

Figure 4.5: Stages or regime of graphite - The notation used: NC, a and

ta refers to nanocrystalline, amorphous and tetrahedral amorphous. This classi-

fication has not only been used in graphite but also in graphene [14]. Stage 1 is

also considered the low defect regime while Stage 2 is the high defect regime for

Equation 3.6 and 3.8 respectively. Image adapted from Ferrari and Robertson.

26

4.3 Analysis

Dose Energy G peak position I(D)/I(G) 2D FWHM

ions/cm2 (keV) (cm−1) (cm−1)

0 0 1586.8 0 38.1

1E13 5 1596.2 0.07 39.6

1E14 5 1580.6 0.40 47.7

1E15 5 1589.9 0.68 not observed

1E13 10 1593.1 0.19 36.2

1E14 10 1586.8 1.17 43.4

1E15 10 1589.9 2.86 60.7

1E16 10 1577.5 0.95 983.5

1E13 15 1577.5 0.12 not observed

1E14 15 1586.1 0.51 36.6

5E14 15 1583.7 0.85 35.3

1E15 15 1596.2 0.72 39.9

Table 4.1: Chart of D peak position, I(D)/I(G) ratio and 2D peak

FWHM) - The values are obtained from curve fitting using the Origin program.

At 10keV and 1014 dose, there is an anomaly where I(D)/I(G) exceed 2. This

anomaly has also been observed in other works [25]. The large FWHM at 10keV

of 1016 dosage indicates many spots having 2 or more layers graphene are stacked

from sputtering. To fully investigate it, we would need to deconvolve the peak.

27

4. RESULTS & ANALYSIS

Figure 4.6: Images from AFM - (a) shows an image from the optical micro-

scope where the blue contrast on the left is the graphene layer and silicon oxide on

the right. Note that there are few contaminant (black specks). This is the graphene

sample before ion beam irradiation. (b) is the phase image for the material. The

contrast shows the differing material. (c) shows the height profile of the silicon

dioxide and graphene layer. These are averaged and used to calculate layer depth.

The weak colour contrast shows that the height difference is minimal, indicating

the thin graphene layer does not have much height difference.

Almost all the above result lies within stage 1 and stage 2 amorphisation

which meant very low sp3 hybridisation involved. Only at 1016 dosage and 10keV

demonstrates the likelihood of that sample to be at stage 3 amorphisation based

on the exceptionally broad FWHM and along with combined result from the G

peak position and I(G)/I(D). Thus it can be mostly concluded that any dosage

on the order of 1016 ions/cm2 should not be applied on the sample since it would

be mostly sp3 hybridised.

4.3.2 Defect density

Using equation 3.6 and equation 3.8, the defect density are calculated and plotted

on a graph.

Dosage has been demonstrated to have increased the amount of defect density

until 1016 dosage until dosage above has resulted in unwanted amorphisation as

shown in Table 4.1. The peak in energy was expected to be about 5keV but

actual data reveals a higher energy peak of the defect density. This in fact was

also observed in other experiments [27].

28

4.3 Analysis

Figure 4.7: Defect density vs Energy - Data were collected from different

spectrum and collated into the graph. The peak defect density for the lower energies

seem to be maximal at ∼10keV

Figure 4.8: Defect density vs Dosage - The general trend shows the increasing

dosage increases the defect density

29

4. RESULTS & ANALYSIS

Some papers [28] has advised that this deviation could be due to the effects of

a substrate as presence of the substrate could backscatter the projectile, further

allowing interaction or damage with the graphene sample. At higher energy,

the substrate atoms can even be liberated and damage the graphene further.

However, displacement energy for graphene carbon about is quite high (23.6eV)

[29], making this seems less unlikely. Further testing is required see if we can

reproduce this deviation.

From Figure 4.8, we noted that there is a saturation in the dosage where

defect density at lower energies of 5keV and 10keV indicates signs of saturation.

Nevertheless, more data should be obtained before anything conclusive can be

mentioned.

We also observed an outlier in the 1015 dose of 5keV. This is mainly attributed

to the different stage of graphene amorphisation the sample is in. Calculation for

this point performed was at the higher regime as compared to the rest which were

at the lower regime. This meant that the transition will result in very different

defect profile especially at higher energies. However, this do not actually show

that higher energy will increase the density of defects since we are comparing dif-

ferent regimes. We were also reminded that higher energies of the same dosage do

not necessarily conform to a higher carbon amorphisation regime as experiments

in the MeV range has demonstrated stage 1 amorphisation [28]. Nevertheless,

the energy range between the 15keV to the 1MeV range should be investigated

as defect density in stage 2 ought to yield higher defect density.

Within the lower defect regime and stage 1 amorphisation, ∼ 10keV would be

the optimised energy but for stage 2 amorphisation, this data alone is insufficient

to tell.

Overall, one deduction based on our observation is that the overall dose should

not exceed 1016 or stage 3 amorphisation would occur. Within the experimental

limits, 10keV is the optimised beam energy for stage 1 amorphisation but more

experiments is needed to confirm this. Higher energies also have to be investigated

to further maximise defect density at possibly higher (stage 2) defect regime.

30

4.4 Further works/improvement

Currently, the pore sizes are being determined through TEM samples using sim-

ilar parameter at 15keV ion beam since the focus for the beam was better com-

pared to 10keV.

Similar to the suggestion of Prof. Jeroen Anton van Kan, alternative use of

cluster instead of singly charged ions could also be used to stimulate larger pore

size since the cluster would be more likely to create larger pores although it will

be much harder to control and manipulate due to the lower stability of a cluster.

One future work that has to be done is to test for the ion transport across

the graphene pores. This can be done by transferring the graphene layer onto

a membrane support and monitor the potential of the membrane to test for ion

transportation and selectivity of the membrane towards NaCl.

5

Conclusion

In this work, I have managed to learn how to operate a sputtering machine and

became aware of how a general ion beam sputtering machine work even though

some of these learning experiences are derived from damaging the equipment and

samples. By manipulating the beam properties, I was able to induce different

densities of defects on supported graphene and characterize the various sample

using techiques such as Raman spectroscopy and AFM. The defect density was

then calculated and found to be optimized at around 10keV. To further verify

the result, simulations on Kalypso was done and result is found be to be close to

experiments.

31

References

[1] David Cohen-Tanugi and Jeffrey C. Grossman. Nanoporous graphene as a

reverse osmosis membrane: Recent insights from theory and simulation. De-

salination, 366:59–70, 2015. ii, 2, 4, 18

[2] Lawrence K Wang, Jiaping Paul Chen, Yung-Tse Hung, and Nazih K Shammas.

Membrane and desalination technologies. Springer, 2010. 1, 2

[3] Arash Aghigh, Vahid Alizadeh, Hin Yong Wong, Md Shabiul Islam, Nowshad

Amin, and Mukter Zaman. Recent advances in utilization of graphene for

filtration and desalination of water: A review. Desalination, 365:389–397, 2015. 2

[4] Yi You, Veena Sahajwalla, Masamichi Yoshimura, and Rakesh K Joshi.

Graphene and graphene oxide for desalination. Nanoscale, 8(1):117–119, 2016.

2, 3

[5] Sumedh P. Surwade, Sergei N. Smirnov, Ivan V. Vlassiouk, Raymond R. Uno-

cic, Gabriel M. Veith, Sheng Dai, and Shannon M. Mahurin. Water desalina-

tion using nanoporous single-layer graphene. Nature nanotechnology, 10(5):459–64,

2015. 5, 6

[6] Michael D Fischbein and Marija Drndic. Electron beam nanosculpting of

suspended graphene sheets. Applied physics letters, 93(11):113107, 2008. 6

[7] Irina Veniaminovna Antonova. Chemical vapor deposition growth of graphene

on copper substrates: current trends. Physics-Uspekhi, 56(10):1013, 2013. 8

[8] Xinran Wang and Yi Shi. Fabrication Techniques of Graphene Nanostructures.

2014. 9

[9] Huanping Yang, Hailong Hu, Yingying Wang, and Ting Yu. Rapid and non-

destructive identification of graphene oxide thickness using white light contrast

spectroscopy. Carbon, 52:528–534, 2013. 11

[10] David Tuschel. Selecting an Excitation Wavelength for Raman Spectroscopy.

2016. 11, 22

32

REFERENCES

[11] AC Ferrari, JC Meyer, V Scardaci, C Casiraghi, Michele Lazzeri, Francesco

Mauri, S Piscanec, Da Jiang, KS Novoselov, S Roth, et al. The Raman

fingerprint of graphene. arXiv preprint cond-mat/0606284, 2006. 13, 25

[12] A C Ferrari and D M Basko. Raman spectroscopy as a versatile tool for

studying the properties of graphene. Nat Nanotechnol, 8(4):235–246, 2013. 13, 14,

18, 25

[13] J. Maultzsch, S. Reich, and C. Thomsen. Double-resonant Raman scattering in

graphite: Interference effects, selection rules, and phonon dispersion. Physical

Review B - Condensed Matter and Materials Physics, 70(15):1–9, 2004. 13

[14] Andrea C. Ferrari. Raman spectroscopy of graphene and graphite: Disorder,

electron-phonon coupling, doping and nonadiabatic effects. Solid State Commu-

nications, 143(1-2):47–57, 2007. 15, 19, 26

[15] Lili Liu, Miaoqing Qing, Yibo Wang, and Shimou Chen. Defects in graphene:

generation, healing, and their effects on the properties of graphene: a review.

Journal of Materials Science & Technology, 31(6):599–606, 2015. 16, 17

[16] K B Dybyspayeva, A Zhuldassov, A Ainabayev, A F Vyatkin, K Alekseev, and

Z Insepov. Multiscale simulation of ion beam impacts on a graphene surface.

Journal of Physics: Conference Series, 751(October):012029, 2016. 16

[17] Florian Banhart, Jani Kotakoski, and Arkady V Krasheninnikov. Structural

defects in graphene. ACS nano, 5(1):26–41, 2010. 17

[18] Axel Eckmann, Alexandre Felten, Artem Mishchenko, Liam Britnell, Ralph

Krupke, Kostya S Novoselov, and Cinzia Casiraghi. Probing the nature of

defects in graphene by Raman spectroscopy. Nano Letters, 12(8):3925–3930, 2012.

18

[19] Isaac Childres, Luis A Jauregui, Wonjun Park, Helin Cao, and Yong P Chen.

Raman spectroscopy of graphene and related materials. New developments in

photon and materials research, pages 1–20, 2013. 18

[20] Ryan Beams, Luiz Gustavo Cancado, and Lukas Novotny. Raman characteri-

zation of defects and dopants in graphene. Journal of Physics: Condensed Matter,

27(8):083002, 2015. 18

[21] L Gustavo Cancado, A Jorio, EH Martins Ferreira, F Stavale, CA Achete,

RB Capaz, MVO Moutinho, A Lombardo, TS Kulmala, and AC Ferrari. Quan-

tifying defects in graphene via Raman spectroscopy at different excitation en-

ergies. Nano letters, 11(8):3190–3196, 2011. 18, 25

33

REFERENCES

[22] MA Karolewski. Kalypso: a software package for molecular dynamics simu-

lation of atomic collisions at surfaces. Nuclear Instruments and Methods in Physics

Research Section B: Beam Interactions with Materials and Atoms, 230(1):402–405, 2005.

20

[23] Weisen Li, Li Liang, Shijun Zhao, Shuo Zhang, and Jianming Xue. Fabrication

of nanopores in a graphene sheet with heavy ions: A molecular dynamics study.

Journal of Applied Physics, 114(23):234304, 2013. 21

[24] F Gruner and F Bell. First-principles-simulation of both charge state and

stopping power of swift heavy ions in solids. Nuclear Instruments and Methods in

Physics Research Section B: Beam Interactions with Materials and Atoms, 245(1):15–18,

2006. 22

[25] MS Dresselhaus, A Jorio, AG Souza Filho, and R Saito. Defect characteriza-

tion in graphene and carbon nanotubes using Raman spectroscopy. Philosophical

Transactions of the Royal Society of London A: Mathematical, Physical and Engineering

Sciences, 368(1932):5355–5377, 2010. 27

[26] P Nemes-Incze, Z Osvath, K Kamaras, and LP Biro. Anomalies in thickness

measurements of graphene and few layer graphite crystals by tapping mode

atomic force microscopy. Carbon, 46(11):1435–1442, 2008. 25

[27] EH Ahlgren, SK Hamalainen, O Lehtinen, Peter Liljeroth, and J Kotakoski.

Structural manipulation of the graphene/metal interface with Ar+ irradiation.

Physical Review B, 88(15):155419, 2013. 28

[28] S Mathew, TK Chan, D Zhan, K Gopinadhan, A Roy Barman, MBH Breese,

S Dhar, ZX Shen, T Venkatesan, and John TL Thong. Mega-electron-volt pro-

ton irradiation on supported and suspended graphene: A Raman spectroscopic

layer dependent study. Journal of Applied Physics, 110(8):084309, 2011. 30

[29] AJ McKenna, T Trevethan, CD Latham, PJ Young, and MI Heggie. Threshold

displacement energy and damage function in graphite from molecular dynam-

ics. Carbon, 99:71–78, 2016. 30

34

Appendix A

List of Spectra peaks

A.1 Data table for all peaks of collected spectra

Table A.1: Table of intensity peaks for all the spectra

35

A. LIST OF SPECTRA PEAKS

Table A.2: Table of peak shifts for all the spectra

36