Slim Moly S makes hydrogen

UPTEC Q18 018

Examensarbete 15 hp Juni 2018

Martin Brischetto

Slim Moly S makes hydrogen

Layer dependent electrocatalysis in

hydrogen evolution reaction with

individual MoS

nanodevices

Teknisk- naturvetenskaplig fakultet UTH-enheten

Besöksadress:

Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0

Postadress:

Box 536 751 21 Uppsala

Telefon:

018 – 471 30 03

Telefax:

018 – 471 30 00

Hemsida:

http://www.teknat.uu.se/student

Abstract

Slim Moly S makes hydrogen

Martin Brischetto

Molybdenum disulfide (MoS2) has been demonstrated to be a potential catalyst in the hydrogen evolution reaction (HER). Due to its highly active edge site, abundance, and low cost, it rivals Pt. However, the potential activity of the MoS2 basal plane has largely been ignored.

The physical characteristics of MoS2 and its corresponding band structure change significantly with decreasing thickness, especially at the monolayer limit. Thus, an investigation on the thickness dependence may provide important insights into the MoS2 basal plane activity.

In this thesis, the layer dependent electrocatalytic performance is investigated with mono-, bi- and multilayer MoS2 based individual nanodevices. Three conclusions were reached. (1) Monolayers showed exchange current densities more than one order of magnitude higher than that of the multilayers, 0.12 mA/cm2 and 8.7 mA/cm2,

respectively. Furthermore, the onset potential of the monolayer was several hundred millivolts lower than that of the multilayer, about 0.2 V vs RHE for the monolayer versus 0.5 V vs RHE for the

multilayer. The Tafel slope of 100-200 mV/dec revealed that the rate limiting step was the adsorption of hydrogen. (2) Interestingly, the bilayer sample exhibited an increase in its exchange current density from 0.3 mA/cm2 to 8 mA/cm2 when cycled extensively. This is suspected to be caused by intercalation of hydrogen between the atomic layers. (3) Additionally, the back-gate voltage is applied to tune the Fermi level of the material and the catalytic performance.

It was found that the back-gate voltage induces an irreversible change in all samples, increasing the exchange current density by an order of magnitude.

The superior basal plane performance of the monolayers to that of the multilayers reveals a new way to optimize the performance of MoS2 as a HER catalyst. In addition, the results above illuminate the yellow brick road to potential improvements in other layered materials as well.

Tryckt av: Uppsala

ISSN: 1401-5773, UPTEC Q18 018 Examinator: Åsa Kassman Rudolphi Ämnesgranskare: Staffan Jacobson Handledare: Mengyu Yan

Acknowledgements

First and foremost I would like to thank Jihui Yang for giving me the oppor- tunity to join your group and to write this thesis on the work. The experience has given me significant insights on what I want to focus the rest of my aca- demic career on. The feeling of fascination has never left my side as I wander through the scientific field with which you have presented me. Without your trust, none of this would have been possible.

I would also like to thank Mengyu, my supervisor, who taught me everything I now know about electrocatalysis. This thesis would not have been possible without your guidance and our midnight debates in front of white boards in the haunted basement hallways of the physics building. You have been an excellent supervisor.

Behind every work of any quality, there is a Palash, forcibly pulling the author from that work. The distractions you subjected me to: going to San Fransisco, the Olympic peninsula, and just the nearest patch of grass, talking about the social implications of nationality, were indispensable for the com- pletion of this work.

The meaning of sainthood is rendered insignificant unless Sara makes the list for her patience. I would like to thank you for surviving the never ending rants on semiconductor physics in front of dive bar chalk boards, nodding ap- preciatively. I also need to thank you for showing me Seattle and all of its, to the unfamiliar eye, near invisible crevices, bars, and shows.

Although the time in the group has so far been brief, I want to thank Curt, for helping and relieving me of much of the workload in the final stages in the writing of the thesis. I have really appreciated our brief philosophical digres- sions and I hope to see you again after the summer.

Last but by no means least, I want to thank Uppsala University, University of Washington, and all the staff of the administrations making the opportunity of visiting University of Washington possible.

Slanka Moly S gör väte

Lagerberoende elektrokatalys vid generering av väte med individuella MoS

nanoenheter.

Martin Brischetto

Klimatförändringarna är på frammarsch: Jordens medeltemperatur ökar och extremväder blir allt vanligare. Det här är till en betydande grad orsakat av mänsklig användning av fossila bränslen, vilket har motiverat en bred forsknings- front i jakt på nya och förnyelsebara bränslen. De mest lovande förnyelsebara energikällorna utgörs av Gärdestadstrion: sol, vind, och vatten. Solkraft och vindkraft är väldigt väderberoende. Vattenkraft är väldigt platsberoende. Det här gör dessa energikällor otillförlitliga. För att lösa det problemet krävs till- förlitlig energilagring. Här kommer vätgas in i bilden. I vätgas är det möjligt att lagra 142 MJ/kg (motsvarande for bensin är 45 MJ/kg). När gasen förbränns i en bränslecell bildas dessutom endast vatten som biprodukt.

Vätgas kan produceras från vatten och elektricitet med en process som kallas elektrokatalys. Det innebär att man placerar två elektroder i vatten och tillför en spänning mellan dem. Det här kommer producera vätgas vid den ena elektro- den och syrgas vid den andra. I dagens läge är det svårt att producera väte med den här metoden i stor skala. Platina, det optimala materialet för den här typen av process, är extremt dyrt och sällsynt. Trycket på att hitta ett nytt material är stort. Den här avhandlingen är ämnad att dra ett strå till den stacken.

Molybdendisulfid, eller MoS₂, har använts flitigt i industrin som smörj- me- del under de senaste hundra åren. Till skillnad från platina förekommer MoS₂i riklig mängd i jordskorpan och är följaktligen billig. Under det senaste decen- niet, i kölvattnet av den revolutionerande upptäckten av grafén, har molybden- disulfid uppmärksammats för sina katalytiska egenskaper. MoS2 har, precis som grafit, en kristallstruktur som består av lager och, precis som med grafit, kan dessa lager separeras från varandra. Det här ger upphov till extremt tunna material som uppvisar extraordinära egenskaper i många fall. Det är när MoS2

kommer ner till storleken av ett fåtal antal lager som det uppvisar goda kata- lytiska egenskaper, jämförbara med platinas nivå. Kruxet är att man har funnit att det endast är kanterna av dessa lager som har dessa katalytiska förmågor, själva ytplanet är i stort sett inaktivt.

Vissa elektriska fenomen hos katalysatorer i elektrolyter har en viss ut- bredning i materialet. Så vad händer när materialet är för tunt för att förse fenomenen med det utrymme de kräver? Den här frågan ligger till grund för

arbetet i den här avhandlingen. Mer specifikt avser avhandlingen att undersö- ka frågan i fallet med elektrokatalys och MoS₂, så frågan omformuleras: Hur påverkas den elektrokatalytiska prestandan hos MoS₂ när materialet är som tunnast?

För att hitta svaret på frågan har prover av MoS₂ med olika antal lager producerats. Dessa har sedan placerats i en syra. Syror innehåller vätejoner och kan därför användas för att lättare producera vätgas. Notera, även om ex- perimentet inte producerar vätgas från vatten så kan en del av resultaten från experiment med syror överföras till applikationer med vatten. Sedan läggs en spänning över gränsen mellan MoS₂ och elektrolyten (syran). Samtidigt mäts strömmen som produceras. Det enda sättet för en ström att uppstå är om ladd- ningar överförs över gränsen mellan MoS₂och elektrolyten vilket endast hän- der om en reaktion sker. Ett genomtänkt val av spänningsområde försäkrar att reaktionen är produktionen av väte, och inte någon annan.

Figur 1. Uppmätta ström-spänning-kurvor för olika prover. De röda kurvorna visar prover bestående av endast ett lager (monolager), de gröna kurvorna visar prover be- stående av flera lager (multilager). Heldragen linje innebär att endast provets ytplan var i kontakt med elektrolyten, streckad linje betyder att elektrolyten även var i kontakt med provets kanter.

Förhållandet mellan den pålagda spänningen och den uppmätta strömmen avslöjar flera saker om prestandan. I grova drag innebär ”prestanda”, i det här sammanhanget, att ju högre ström (väteproduktion) man kan få ut från en viss spänning desto högre prestanda har katalysatorn. Figuren ovan visar ström- spännings-kurvorna. Från den kan vi utläsa hur prestandan påverkades av oli- ka antal lager och huruvida ytplanet deltog i reaktionen. För det första ska det sägas att alla negativa tecken på axlarna är konvention: En högre negativ ström innebär en hög ström, en hög negativ spänning innebär en hög spän- ning. Först jämförs proven bestående av enskilda lager (monolager, röda lin- jer) med de bestående av flera lager (multilager, gröna linjer). Det är uppenbart att vid spänningen -0.1 V vs RHE har monolagren utvecklat en mycket hög- re ström. Monolagrens ström vid den här spänningen rör sig från runt -20 till -100 mA/cm² medan multilagren knappt lämnat noll. Dessutom syns det att

monolagren där endast ytplanet är i kontakt med elektrolyten (röd, heldragen linje) har högre prestanda än multilagren där även kanterna är i kontakt med elektrolyten (grön, streckad linje).

Dessa resultat visar att monolager har bättre prestanda än multilager. De vi- sar dessutom att ytplanet hos monolagret är mycket mer aktivt än kanterna hos ett multilager. Det här motsäger det man tidigare har känt till om ytans akti- vitet. Dessutom innebär resultaten att molybdendisulfid med fördel kan klyvas ner till enskilda lager för att användas som katalysator. Elektroder baserat på detta material skulle både ha en lägre materialkostnad och vara effektivare än en elektrod baserat på materialets flerlagrade form. Men framförallt pekar re- sultaten, med stadig och öppen hand, mot frågorna som bör ställas härnäst: Hur väl presterar övriga material i samma familj (de med skiktad struktur) i sina tunnaste former?

Examensarbete 30 hp på civilingenjörsprogrammet Teknisk fysik med materialvetenskap

Uppsala universitet, Juni 2018

Contents

1 Introduction . . . . 1

2 Background . . . . 3

2.1 2D materials. . . .3

2.1.1 Molybdenum Disulfide . . . . 3

2.1.2 The Band Structure. . . . 6

2.2 Electrocatalysis . . . . 10

2.2.1 Three electrode setup . . . . 12

2.2.2 Cyclic Voltammetry. . . .14

2.3 Electrolyte-Semiconductor junction . . . . 17

2.3.1 Field effect . . . . 20

3 Experimental Details . . . .21

3.1 Device Fabrication - Electron Beam Lithography. . . . 21

3.2 Measurement - Cyclic Voltammetry. . . . 23

3.2.1 Setup . . . .23

3.2.2 Measurement . . . .24

3.3 Analysis . . . . 25

3.4 Characterization . . . . 26

3.4.1 Photoluminescence . . . .26

3.4.2 Atomic Force Microscopy . . . . 27

4 Results and Discussion. . . .28

4.1 Characterization . . . . 28

4.1.1 Photoluminescence . . . .28

4.1.2 Atomic Force Microscopy . . . . 29

4.2 Back gate voltage has an irreversible effect. . . . 29

4.2.1 Droplet versus PEC . . . . 32

4.2.2 Intercalation of Hydrogen in Bilayers. . . .33

4.3 Fewer layers improves performance. . . .34

4.3.1 The basal plane is responsible for layer dependent performance . . . .38

5 Conclusion . . . . 40

Nomenclature

Abbreviations

AFM Atomic Force Microscopy

CB Conduction Band

CV Cyclic Voltammetry

CE Counter Electrode

DFT Density Functional Theory FET Field Effect Transistor HER Hydrogen Evolution Reaction

LTMD Layered Transition Metal Dichalcogenides OER Oxygen Evolution Reaction

PEC Plastic Electrolyte Container

PL Photoluminescence

RE Reference Electrode

RHE Reversible Hydrogen Electrode SCE Saturade Calomel Electrode SEM Scanning Electron Microscope

VB Valence Band

WE Working Electrode

Chemical

MoS2 Molybdenum disulfide 1T-MoS2 Trigonal phase of MoS2

2H-MoS₂ Hexagonal phase of MoS₂ 3R-MoS2 Rhombohedral phase of MoS2

H⁺ Hydrogen ion

H₂ Hydrogen molecule

PMMA polymethyl methacrylate Variables & Parameters

A Tafel Slope [V/dec]

a Surface Area [cm²]

η Overpotential [-V vs RHE]

E_f Fermi Level [eV]

I Current [mA]

J Current Density [mA/cm²]

J0 Exchange Current Density [mA/cm²]

V_bg Back Gate Voltage [V]

V_RHE/SCE Voltage between SCE and RHE [V vs SCE]

V_{W E/RHE} Voltage between RHE and WE [V vs RHE]

V_{W E/SCE} Voltage between SCE and WE [V vs SCE]

1. Introduction

Climate change is no longer at our doorstep, but stampeding through the foyer with the alarming speed of∼ 0.12^◦C per decade¹with the devastating force of heat waves, droughts, floods, cyclones and wildfires². Climate change is largely attributed to emissions of carbon dioxide from the anthropogenic use of fossil fuels [1]. A world wide effort is being made to find alternative, renew- able, and reliable energy sources, ranging from solar to wind power. However, most, if not all, of these efforts will fall short on the reliability condition due to their intrinsic intermittent nature (e.g. the sun only shines during the day, wind only blows on windy days, etc.) unless the harnessed energy can be stored re- liably and conveniently in terms of weight and/or volume, depending on the application. A promising candidate is the odorless gas right under our noses, hydrogen. It has one of the highest specific energies among the elements (142 MJ/kg, versus 45 MJ/kg for gasoline) and only produces water when the energy is extracted using a fuel cell. It is the most abundant element in the universe but most of it is bound up in substances, organic and inorganic. This means it needs to be extracted, but it also means it can be extracted anywhere on earth.

With the use of electrocatalysis, energy from renewable sources can be stored in the hydrogen bonds during their production peaks. Unlike with lithium ion batteries, the energy can be stored continuously without reaching a maximum level, as long as there is a supply of water, and it enables the use of distribution systems analogous to oil pipelines.

The potential applications stemming from a full understanding of electro- catalytic processes do not end at hydrogen production. The greenhouse gas CO₂can be transformed into the energetically dense form of alcohols, CxHyOz

through catalytic processes. Nitrogen, the most abundant gas in the atmo- sphere, can be transformed into ammonia, NH₃, extensively used to produce fertilizers. If these processes can be made efficient enough, they can be moved from energy heavy factories to the empty patch of dirt behind the shed at the local farm, running mostly on the household voltage and sunshine. This would reduce the need for transportation of these chemicals. Figure 1.1 shows a beau- tiful schematic of a sustainable energy landscape based on electrocatalysis [2].

1Increase in globally averaged surface temperature calculated from 1951 to 2012 [1]

2Extreme events caused by climate change [1]

Figure 1.1. Schematic of a sustainable energy landscape based on electrocatalysis [2].

Currently, the best catalyst for hydrogen production is platinum [2]. How- ever, its scarcity in the earths crust makes it impossible for widespread usage and very expensive for limited usage. This has motivated extensive research directed into finding alternative and cheaper materials to take its place. Re- cently, atomically thin materials have attracted much attention due to their unique optical and electronic properties.

One of them is molybdenum disulfide, MoS2. The material is abundant in the earths crust and has recently been found to possess a promising catalytic performance when designed in certain forms. These forms get fairly complex due to the limitations of the catalytic performance. Only the edges of the 2D flakes and sulfur vacancies on the surface are catalytically active. Despite the amount of research put into this material, the full extent of its properties and, consequently, its potential, is scarcely known. This thesis aims to investigate the layer dependence on the performance. If the activity of the material can be made to increase simply by changing the number of layers the complexity of the MoS2electrodes can be significantly reduced.

2. Background

2.1 2D materials

Graphite consists of atomic carbon layers bound to each other with weak Van der Waals forces. Graphene (individual layers of graphite) and oxidized graphene has been synthesized several times since as early as 1859 [3–5]. But it wasn’t until 2004, when Geim and Novoselov synthezised it, that the authors discov- ered its remarkable electronic properties, earning the nobel prize in physics in 2010 [6]. In addition, their ”scotch tape method” to produce the material, using only graphite and household scotch tape, has the simplicity fit for kindergart- ners on a lazy, post nap, Tuesday afternoon. In the original paper they found that the charge carrier mobility in graphene was up to 10 000 cm²/Vs. This number has since then been found to be in the excess of 200 000 cm²/Vs under certain conditions [7]. In silicon, which is commonly used in modern field ef- fect transistors (FET), the electron mobility is less than 1500 cm²/Vs. A FET that was built with graphene in 2010 could operate at a frequency of 100 GHz (compare with 40 GHz for the state of the art Si FET with the same thickness) [8]. The discoveries of Geim and Novoselov turned the inquisitive eyes of the scientific community to other materials with layered structures and kick-started the research on two dimensional materials and their futuristic applications [9].

2.1.1 Molybdenum Disulfide

Molybdenum disulfide (MoS₂) is part of a family of materials called layered transition metal dichalcogenides (LTMD) . These materials have a layered structure, much like graphite, and have the generalized chemical formula MX₂, with M representing a transition metal and X representing a chalcogenide. Due to its abundance, MoS₂ is one of the most studied materials in this category.

It was patented as a dry lubricant in 1929 [10], as a catalyst in 1943 [11], and as a cathode for lithium ion batteries in 1980 [12]. As the field of nanotech- nology was born in the 1980s the layered materials were naturally of interest, due to the ease of separating the layers from each other. In 1995, this technol- ogy was used to synthesize MoS₂ nanotubes and fullerenes (large spherically shaped molecules) [13]. It was not until 2005 that Novoselov et al. synthesized the first two-dimensional materials (BN, MoS₂, NbSe₂, Bi₂Sr₂CaCu₂O_x, and graphite) [14] and then measured the astonishing properties of graphene [6], bringing forth what has been called the graphene revolution. The revolution

did not really reach MoS2until after the discovery of its direct band gap in sin- gle layers (explained in section 2.1.2) in 2010 [15] and the demonstration of its high performance as a transistor in 2011 [16]. The property of having a direct band gap is a requirement for efficiently transforming light into energy. This potential was recognized by M. Fontana et al. in 2013. They showed that mul- tilayered MoS2indeed produced a photovoltaic effect at 2.5% efficiency, with the implication that a device with monolayered MoS₂ would do much better [17]. Its performance as a transistor gives it a leg up on graphene which needs complicated procedures with trade-offs on other important properties to func- tion as a transistor [18–21]. These discoveries spurred the interest in MoS₂, more than doubling the number of papers on it between the years 2013 and 2014. In figure 2.1 the history of MoS₂research since 1964 up until 2014 is shown in the form of annual number of publications.

Figure 2.1. History of molybdenum disulfide research for last 50 years (1964–2014).

The graph shows the annual number of journal publications, of which title, abstract or keywords contains word molybdenum disulphide, molybdenum disulfide, or MoS2, based on Scopus query. Note that the number in Year 2014 (1099) is extrapolated from the number of publications till 30 September, 2014 (824). The figure and the caption are reprinted from [22].

MoS2 has three phases: 3R-MoS2, 2H-MoS2, and 1T-MoS2. The 2H and 3R phases are stable and naturally occurring in the earths crust. However, 2H is far more abundant than 3R which was not even discovered until 1957 [23, 24]. The 1T phase is metastable and needs to be synthesized, which was first achieved in 1992 [25]. The number in the notation represents how many layers are required to form a unit cell and the letter describes the crystallographic structure: R- rhombohedral, H- hexagonal, and T- Trigonal. Both 2H and 3R are semiconductors, while 1T is a metal. In all three cases, each single layer (or monolayer) consists of a layer of molybdenum atoms sandwiched between two

layers of sulfur atoms¹. The difference between them is the relative position of the three atomic layers. Each of the 2H and 3R layers have a mirror plane in the Mo layer while the sulfur atoms in the 1T phase is on opposite sides of the molybdenum atom, laterally. The positions of each atomic layer can be thought of having three lateral alternatives: a, b, and c. for the 1T phase, the atomic layers are ordered as AbC (where capital letters represent sulfur). For 2H and 3R however, they are ordered as AbA, with the two sulfur layers in the same lateral position. The difference between them is how the three-atomic- layer layers are stacked. 2H has a repetitive stacking sequence of AbA, BaB,...

while the 3R phase has a stacking sequence of AbA, BcB, CaC... This means that a 2H monolayer is identical to a 3R monolayer. One layer of 1T-MoS2

has a thickness of 0.599 nm, two layers of 2H-MoS₂has a thickness of 1.230 nm, and three layers of 3R-MoS2 has a thickness of 1.838 nm [24–26]. A single layer of three phases all have a thickness of about 0.6 nm. Most of the research throughout history has been focused on the abundant 2H phase and more recently on the 1T phase. The 3R phase has been comparatively ignored due to its similarity to the 2H phase while lacking the abundance. This thesis is no exception to this trend, as the phase of focus will be the 2H-MoS2 and how it differs from the monolayered form.

Exfoliation

There are a few ways to separate the layers from one another (called exfolia- tion), all varying in scale, quality, and complexity. One such method is chem- ical exfoliation. With this method, the bulk MoS₂is exfoliated by having ions, such as Li⁺, diffuse in between the layers and separating them by repulsive forces [27]. Another, more bottom-up, approach is to deposit a layer of MoO3

on an arbitrary surface and have it react with sulfur [28] using chemical vapor deposition. These methods, however, produce small, low quality MoS2flakes in large scale. The method used for the experiments in this thesis produces fewer, high quality flakes and is by far the simplest method. This method will be described in greater detail.

In the scotch-tape method uses mechanical exfoliation one uses household scotch tape and no chemicals. In this method a small piece of MoS2is placed on a strip of scotch tape. Then the tape is folded and peeled repeatedly, so that the piece of the material is exfoliated with each repetition. After a while the tape is covered with flakes of MoS₂ of varying thickness, including monolayers.

The tape is then carefully pressed onto the surface of a silicon wafer. If the field effect of the sample is to be tested, it is convenient for the silicon wafer to have an oxidized surface on it (field effect is explained in section 2.3.1).

The silicon chip, along with the tape strip, is then heated for about a minute

1In contrast to other chalcogenides, where a material like this would be called a trilayer because of its three atomic layers, the convention for transition metal dichalcogenides is to call it a monolayer.

at 120^◦C to make the material stick to the chip before peeling off the tape². Lastly, the monolayered pieces were found using an optical microscope. In this step, the choice of the silicon oxide thickness becomes relevant. The optical identification of the MoS2 thickness is based on thin film interference where the intensity of the reflected light depend on the thicknesses of both MoS₂and the silicon oxide. The optimal contrast between numbers of MoS2 layers is achieved at a SiO₂ thicknesses of either 78 or 272 nm [29]. The procedure is depicted in figure 2.2.

Figure 2.2. A depiction of the scotch-tape exfoliation method. MoS2 is placed on a piece of scotch tape which is then folded and peeled, repeatedly. The tape strip is pressed on a silicon wafer before heating it for about a minute and peeling it off.

Finally, the monolayers are found using an optical microscope.

The scotch-tape method produces large (∼ 20 µm²), high quality flakes, but on each chip only a few will be found. This means that the method is not scalable to mass production. Its simplicity, however, makes it ideal for research and prototypes.

2.1.2 The Band Structure

Individual atoms have discrete layers of electrons called orbitals. Electrons all have an energy specific for that orbital. Each orbital has a well defined energy and each element has a unique set of orbital energies. As atoms combine to form molecules these orbitals merge and form molecular orbitals. Two merg- ing atomic orbitals will form two molecular orbitals of energies similar to the atomic orbitals. A consequence of the Pauli exclusion principle is that only a handful of orbitals can have the same energy and that each orbital can only be occupied by two electrons (provided that they have different spin). As more and more atoms combine to form solids, the energy levels of the molecular orbitals remain similar to those of the atomic orbitals but they become packed

2The practice of heating the chip is based on advice from experienced researchers.

close and closer to each other due to the uniqueness consequence of the Pauli exclusion principle. When the electrons of the combining atoms add up to numbers in the order of 10²³the energy levels are so closely packed that they are more conveniently viewed as wide continuous energy bands³. In differ- ent materials these energy bands have different sizes and energies. At some energy levels they may overlap, at others the bands may touch without over- lapping, and at others there may be a gap between the bands. The properties of the material vary greatly depending on where in these bands the electrons are located.

The Fermi level (E_f) is the energy level at which there is a 50% chance of finding an electron, statistically. At the absolute zero temperature all the elec- trons are stacked from the lowest energy level up to as far as they can reach while still maintaining minimum energy. In this case the Fermi level would be between the highest occupied and lowest unoccupied energy levels, even though technically no electron can be found there; this is because of the sta- tistical nature of the Fermi level. As the temperature increases some electrons will be thermally excited and jump to a higher energy level and be deexcited back to the lower energy level. This produces a distribution of electrons with some of them being above the Fermi level and the rest below it. The higher the temperature is, the wider the energy region where these excitation-deexcitation occur is. However, the distribution will still give a 50% chance of finding an electron at the Fermi level. Thus, the Fermi level can be seen as a indicator of how the electrons are distributed in the bands. Metals have the Fermi level in one of the energy bands, this means that an electron can be excited to above the Fermi level with an arbitrarily low energy. Semiconductors and insulators, on the other hand, has the Fermi level within a gap between energy bands.

This means that an electron needs more energy than the band gap to be ex- cited. In these materials, the band below the Fermi level is called the valence band (VB) and the one above it is called the conduction band (CB) . An energy band diagram is shown in figure 2.3 to the left, along with electron probabil- ity distributions to the right. The three different Fermi levels and probability distributions correspond to three different types of doping.

Doping is a common method of shifting the Fermi level. It means to add a small amount of a foreign element to the material. If the addition of this element give rise to energy levels (through interactions with the host element) such that electrons can easily leave them and enter the conduction band of the main material, it is called a donor (because it donates electrons). On the other hand, if the foreign element has energy levels so that the electrons can easily leave the main material and enter the energy levels of the dopant it is called an acceptor (because it accepts electrons). In both cases, the number of available electrons in the material is changed and consequently shifts the

3Roughly this number of atoms form things at the sizes of every day objects, from a grain of salt (∼ 10¹⁸atoms) to an African savanna elephant (∼ 10²⁹atoms).

Fermi level. When electrons are added to the material, the Fermi level shifts up to higher energies and is called an n-type semiconductor (E_f,nin the figure).

When electrons are removed from the material, the Fermi level shifts down to lower energies and is called a p-type semiconductor (E_f,pin the figure). When a semiconductor has no dopants, the Fermi level is roughly centered in the band gap and is called an intrinsic semiconductor (Ef,iin the figure). Some materials are naturally n-type or p-type. MoS₂, for example, is naturally an n-type semiconductor.

Figure 2.3. An image of a band diagram (left) and electron probability distributions corresponding to different types of doping (right). The probability distributions are exaggerated for clarity, in reality its much more narrow around the Fermi level (The curves shown in the figure correspond to a temperature of 7000 K).

One can also use the field effect in semiconductors. This is done by adding a voltage to a conducting material that is isolated from the semiconducting material of interest, creating a capacitor. The electric field from the external material causes the energy of the electrons at the surface to decrease which, in turn, causes the electrons to accumulate at the surface. This effect is explained in greater detail in section 2.3.1.

For an electron to move through the material and produce a current, it needs to jump to adjacent vacant energy levels. This means that electrons at energies where there are many electrons (i.e. at lower energies) can not move as freely and do not contribute to currents as much as the electrons at more vacant en- ergies. This is why metals have a higher conductivity; because the electrons can find many unoccupied energy levels at an arbitrarily low energy above it.

In contrast, electrons in semiconductors need to jump from the valence band

to the conduction band before it finds vacant energy levels to move through, this is why it is called the conduction band⁴.

As will be explained shortly, both sections of the figure are simplified. The energy levels are different along different directions in the material, and the density of available states are not included here. In general there are more available energy states closer to the conduction/valence band edges than there are deeper into the energy bands and, obviously, there are zero available states within the band gap. The probability distribution would need to be multiplied by the distribution of states to get the true distribution of electrons. The point conveyed here is simply that there is a band gap, and that electrons are dis- tributed around the Fermi level.

Here, we will briefly depart from the simplified idea of the band structure presented in figure 2.3 to explain a property that is relevant for MoS2, before returning to the simpler model. The band structure is also dependent on in what direction the electron happens to be traveling. As the solids are formed, the atoms are arranged in a periodic lattice structure. This results in that each atom is surrounded by a set of other atoms at specific positions and distances from it. The periodically repeated structure is called the unit cell. The unit cell can be transformed, mathematically, to an inverse cell called the Brillouin zone represented in an inverse space called k-space. In this mathematical space, the distances are measured in wavenumbers with the inverse of distance as the unit.

The wavenumber can roughly be described as the number of waves that fit into a unit length (and then multiplied by 2π to make it compatible with trigono- metric functions). This space is more convenient when viewing electrons as standing waves stretching throughout the material. Due to the different bond lengths and angles between the bonds, the energies of the electrons will be de- pendent on their position in the Brillouin zone (i.e. what their standing wave looks like). This is commonly depicted in an energy band diagram, shown in figure 2.4. The ordinate shows the electron energy and the abscissa shows the wavenumber. The diagram usually only shows the energy range around the Fermi level. In the figure, the band diagrams of MoS₂ with different thick- nesses are shown. The blue line shows the highest energy levels of the valence band and the red line shows the lowest energy levels of the conduction band.

Electrons are generally excited from the energy maximum of the valence band to the energy minimum of the conduction band, as indicated by the arrows in the figure.

4The vacant energy levels (called holes) can be viewed as positively charged particles that re- quire adjacent occupied energy levels to move and produce a current. The holes in the valence band contribute about as much to the total current as the electrons in the conduction band do.

Figure 2.4. Electron energy band diagram for 2H-MoS2of different thickness: Bulk (a), 4-layer (b), bilayer (c) and monolayer (d). As the thickness is reduced the indirect band gap is increased while the direct excitonic transition remains constant. At the monolayer limit the material turns from being an indirect band gap material into being a direct band gap material. Figure reprinted from [30].

When the excitation requires a shift in the wavenumber, energy needs to be transfered between the lattice and the electron. When a band gap requires an excitation of this sort, it’s called an indirect band gap. When the conduction band minimum is directly above the valence band maximum no such transfer is required. Band gaps of this sort are called direct band gaps. This difference is significant, because the odds of a photon being absorbed by an indirect band gap material is much lower than that of a direct band gap material, due to the additional lattice-electron interaction required for the former.

Bulk 2H-MoS2has an indirect band gap and is a poor light absorber. How- ever, in its monolayered form, the material has a direct band gap [30]. This means that the monolayered material could potentially be used for photocatal- ysis.

For the rest of this thesis the band structure will be simplified to the one shown in figure 2.3, with a conduction band starting at the conduction band minimum and a valence band ending at the valence band maximum, without considering the different different directions of the electrons. The only take- away from image in figure 2.4 is the qualitative property of a band gap being direct or indirect. The behavior of the band diagram under different conditions will be explained in more detail in section 2.3 after electrocatalysis has been explained in the following section.

2.2 Electrocatalysis

Water can be split into hydrogen and oxygen gas with a device called an elec- trolyzer. It consists of an anode and a cathode. At the cathode a positive voltage

is applied, taking electrons away from the water molecules, producing oxygen gas; this reaction is called the oxygen evolution reaction (OER) . Hydrogen is produced at the anode, where a negative voltage is applied; this reaction is called the hydrogen evolution reaction (HER). The chemical formulas for these reactions and their net reaction are shown below.

H2O_(l) → 1

2O2(g)+2H⁺+e⁻ (Cathode, OER)

2H⁺+e⁻ → H2(g) (Anode, HER)

H2O_(l) → 1

2O_2(g)+H_2(g) (Net Reaction)

However, the role of the anode and cathode is more than just a current source and drain. They act as catalysts, providing sites(called active sites) on which the reactants can adsorb and more easily react with one another, hence the apt name electrocatalysis. The reaction path can take one of two forms: after an hydrogen ion has adsorbed it can then react either with a dissolved hydrogen ion (Volmer-Heyrovsky) or another adsorbed hydrogen ion (Volmer-Tafel).

H⁺+e⁻→ H^∗ (Volmer)

H⁺+H^∗+e⁻→ H2 (Heyrovsky)

2H^∗ → H2 (Tafel)

In 1977 H. Tributsch and J.C Bennett found that bulk molybdenum was a poor electrocatalyst for the HER [31]. However, in 2005 this was explained by B.

Hinnemann et. al. with theoretical density functional theory (DFT) calcula- tions. It was shown that only the edge sites on 2H-MoS₂ were catalytically active, while the sites on the basal sites were not [32]. In 2007 it was con- firmed experimentally by T. F. Jaramillo using 2H-MoS₂ nanoparticles [33].

Furthermore, the 1T phase of MoS2 has been found to be superior to the 2H phase [34]. There is, however, evidence to suggest that the superiority of the 1T phase is caused by its higher conductivity rather than the intrinsic activity of the sites [35, 36].

The mechanics at the catalyst can be optimized in two ways: Increasing the number of active sites and increasing the intrinsic activity of each site. These are not mutually exclusive and can often be optimized in parallel with great success.

The former can be achieved by activating inactive sites and/or by increas- ing the surface area of the catalyst. The surface area can be increased either by simply using a larger quantity or by using a material with a large surface area per volume and mass. For example, MoS₂ is mostly active at its edge sites, so great pains have been taken to engineer materials with long MoS2-edges:

producing it on materials with large surface areas and sharp curvatures such as nanowires and porous materials, or producing many small pieces on graphene substrates, or simply stacking MoS2 on the narrow end with the edges facing

outwards [37–41]. It is also possible to activate sites on the basal plane which would otherwise be inactive. This has been done by introducing sulfide va- cancies [42]. However, increasing the surface area of a material often means that the reactants and products need to travel further to reach their active sites.

This becomes a limiting factor at a certain point.

Increasing the intrinsic activity of each site, on the other hand, does not produce an analogous limiting factor. Increasing the activity means to design the active sites in such a way that they aid the reaction more efficiently. For example, in 2016 Hong Li et al. increased the activity by one order of mag- nitude by adding strain to sulfide vacancies in the basal plane [42]. Another method is doping; in 2016 Daniel Escalera-López et al. made Ni-MoS2hybrid nanoclusters in which the Mo and S atoms at the MoS₂edge were substituted for Ni, this increased the exchange current by a factor of three [43]. Similar results were found in 2009 by J. Bonde et al. when doping MoS₂ and WS₂ with Co [44]. Furthermore, supporting a MoS2 with different materials have been found to impact the hydrogen binding energy. When the adhesion to the support increases, the hydrogen binding energy decreases [45].

In 2017 J. Wang et al. combined the catalytic properties of MoS2 with the impact of field effect. Applying an external electric field to the material in- creased its conductivity and consequently its catalytic performance to a level comparable to that of platinum [46]. Their samples were thin 25 nm MoS₂ flakes. In the present thesis, a similar experiment will be done, but using mono- layered and bilayered samples instead.

2.2.1 Three electrode setup

One way to measure the catalytic activity is to apply a voltage between the working electrode (WE) and a counter electrode (CE). The working electrode is the material we’re interested in and the counter electrode is where the opposite reaction is occurring (i.e if there is a reduction occurring at the WE, oxidation is occurring at CE, and vice versa). These two electrodes are connected to each other through a voltage source, completing the circuit. During equilibrium the net current is zero, but when a large enough voltage is applied the redox reactions start and the current flows. Since the amount of hydrogen produced per unit of time is proportional to the number transferred electrons per unit of time, it is also proportional to the current. The magnitude of interest is how high voltage is required to produce some amount of hydrogen.

The current can be measured at the circuit connecting the WE and CE. The voltage can not be measured in this way. The reason is that the driving force for the reactions at the electrodes is the voltage between the electrolyte and that electrode and, much like the reaction at the WE, the relationship between the current and the voltage is not known at the CE. This is in fact the relationship we are interested in at the WE. In circuitry terms, the two electrode-electrolyte

junctions can be viewed as two non-ohmic resistors in series. So to measure the voltage only over the WE-electrolyte junction, a third electrode is used, the reference electrode (RE). The voltage is measured between the RE and the WE and since no current is flowing through the RE the problem of the CE is averted. The three electrode setup is depicted in figure 2.5

Figure 2.5. A schematic picture of the three electrode configuration. The counter electrode(CE), reference electrode(RE), and working electrode(WE) are submerged in an electrolyte. A voltage is applied between the CE and WE. The current between the CE and WE is measured while the voltage is measured between the RE and WE.

The redox potential of a reaction is the potential that needs to be applied be- tween an electrode and an electrolyte for that reaction to occur. With different electrodes different potentials need to be applied. Fortunately, the potential for HER has been measured against several electrodes (the potential relative to HER is denoted V vs RHE, the Reversible Hydrogen Electrode ). The one used in the experiments described in this thesis is called Saturated Calomel Electrode (SCE) . RHE (the HER potential) has a potential of -0.2415 V vs SCE. Knowing the potential between the RE and HER, V_RHE/SCE , and the voltage between the RE and the WE, V_{W E/SCE}, we can calculate the potential between the WE and the redox potential, V_{W E/RHE}, through simple subtrac- tion, as shown below.

η =−VW E/RHE =−(VW E/RHE − VRHE/SCE) (Overpotential) The potential between the WE and the redox potential is called the overpoten- tial, i.e. the potential over which the redox reaction could occur. A negative voltage needs to be applied for the HER to occur, so a negative sign is added to have the overpotential be positive when the reaction is occurring and negative when it is not. The relationships between the different units are depicted more clearly in figure 2.6. The figure also shows how the voltage at the WE changes over time in a cyclical voltammetry measurement, explained in the subsequent section.

Figure 2.6. A Schematic figure showing the relationship between the different refer- ence systems. The voltage is measured in V vs SCE. It is converted to V vs RHE by subtracting -0.2415 V. Overpotential η is simply V vs RHE with the sign reversed. The line marked with WE is the path of the voltage over time during cyclic voltammetry, explained in the subsequent section.

2.2.2 Cyclic Voltammetry

Cyclic Voltammetry (CV) is a type of electrochemical measurement in which the voltage at the WE is scanned back and forth between two values. Starting at the lower voltage, increasing, as the voltage passes the potential of some redox reaction, that reaction will produce a current. When the reactant is starting to deplete the current will reach a peak and start to decrease and finally reach zero when the reactants have been fully depleted. At this point the voltage sweep is reversed and is decreased⁵. Again, as the voltage reaches a certain potential a redox reaction will start to produce a current. As the product is depleted the current will reach a peak and then decrease to zero (the process is depicted in figure 2.7).

5In practice, the sweep is reversed before the reactants are fully depleted.

Figure 2.7. A schematic image of a cyclic voltammogram. As voltage increases and exceeds some threshold value, the reaction will commence and the current increase.

When the reactants are depleting the reaction speed and the current will peak and start to decrease, approaching zero. At this point the voltage is sweeped in the opposite direction, resulting in an analogous rise and fall in current, but at the opposite electrode and reaction. The red arrow marks what is called the polarization curve, where the current increases and reactants are plentiful.

This cycle can be repeated indefinitely in a fully reversible systems with reactions like

M_(aq)ⁿ⁺ + ne⁻↔ M(s) (2.1)

where the products of one reaction are available as reactants in the opposite reaction, and vice versa. In a water splitting system, on the other hand, the reactions are not fully reversible. Since hydrogen and oxygen is produced in their gaseous form they will leave the system and will not be available for their opposite counterparts. This is not a problem since we are only interested in the properties of the hydrogen evolution reaction and not the properties of the system as a whole. The information relevant to this project is the reaction rate when the water is plentiful, long before the peak is reached. To get this information the voltage needs to be sweeped from 0 V vs RHE to∼-0.4 V vs RHE. This curve would be the part of the curve running along the arrow shown in figure 2.7 and is called the polarization curve. Figure 2.8 is a schematic of what a polarization curve might look like. As the voltage decreases, the current density increases exponentially (in the negative direction, since the electrons are running from WE to the hydrogen ions, opposite to the source).

In a perfect world the reaction would start at 0 V vs RHE since that is the point at which the reaction is thermodynamically favorable, in theory. How- ever, in reality there are many more obstacles than the theoretical thermody-

namical limit to consider, like the energy barrier at the electrolyte-semiconductor junction (see section 2.3). This is the type of information that can be extracted from the polarization curve. The onset potential is the potential at which the reaction actually starts. It is difficult to define a potential at which the reaction starts. It is common to choose a current density and say that the reaction starts at the potential which produces that current, for example at -100 mA/cm². Af- ter the reaction has started its rate can be described by the Tafel equation shown below.

η = Alog (J

J0

)

(Tafel equation)

Ais called the Tafel slope and is the increase in potential required to increase the current density by a factor of 10. J0is called the exchange current density and is what the current would be at zero overpotential, when the current from the opposite redox reaction is equal. The exchange current density is a measure of intrinsic electron transfer rate of the electrode/electrolyte junction. It reveals information about the structure and bonding between the electrolyte and the electrode. The Tafel slope reveals information about what step in the reaction (Volmer, Heyrovsky, or Tafel, explained at the start of section 2.2) is the rate limiting step. If the Tafel slope is around 120 mV/dec it means that the Volmer step is the rate limiting step (i.e. adsorption of hydrogen is the slowest step).

If the slope is 40 mV/dec it means that the Heyrovsky step is rate limiting (i.e. adsorbed hydrogen atoms reacting with dissolved hydrogen is the slowest step). The Tafel step is rate limiting if the Tafel slope is around 40 mV/dec (i.e.

adsorbed hydrogen atoms reacting with another adsorbed hydrogen atoms is the slowest step) [47].

Figure 2.8. A schematic figure of what a polarization curve may look like. The Tafel curve is shown in the inset, together with a fitted Tafel equation (solid black line). Note that the abscissa in the inset is drawn in a logarithmic scale.

Using the polarization curve, the parameters of the Tafel equation can be extracted by plotting the overpotential versus the logarithm of the current den- sity. Then the Tafel equation can be fitted as a line to the linear part of the curve. The Tafel curve and the fitted Tafel equation is shown in the inset of figure 2.8.

2.3 Electrolyte-Semiconductor junction

The conceptual junction of the fields of electrochemistry and semiconductor physics can be found at the physical junction between an electrolyte and a semiconductor. At this junction physicists and chemists meet and attempt to translate the nomenclature of the opposite side into more convenient terms.

In this thesis, the terms of the chemists will be translated to the terms of a physicist.

An electrolyte contains molecules, each with molecular orbitals with unique energies for the electrons to occupy. The vast number of molecules produce more of a continuous energy band, much like in solids. The difference is this:

While the electrons in a solid move through a lattice of bound atoms, the elec- trons in an electrolyte move through a fluid of atoms where the state of the atoms depend on where the electrons are. The H⁺/H₂electrolyte, for example, can be viewed as a collection of occupied and vacant protons. The occupied protons take the shape of H₂ while the vacant protons take the shape of H⁺. The protons at lower energies will mostly be occupied while protons at higher energies will mostly be vacant, producing some electron energy distribution, just like in a solid. This means that there is some equivalent to the Fermi level in the electrolyte, i.e. an energy at which there is a 50% chance of finding an electron. As it turns out, this energy level is the redox potential. Electrolytes would most resemble a metal, since the redox potential is not in any sort of energy gap; the electrons can be excited by an arbitrarily low energy⁶, and the conductivity is much higher than that of a semiconductor. Much like in solids, the redox potential can shifted. This is described by the Nernst equation, shown below.

E(A/A⁻) = E^◦(A/A⁻) +kT ne

ln [A]

[A⁻] (Nernst equation) E(A/A⁻) is the actual redox potential of the electrolyte, E^◦(A/A⁻) is the formal redox potential of the redox couple A/A⁻, n_e is the number of elec- trons (per molecule of redox couple) that are exchanged during the reaction, k and T are the boltzmann constant and the temperature, and [A] and [A⁻]are the respective concentrations of the acceptor and donor species. The formula

6Just like in semiconductors, there is a density of states distribution that add some complexity.

This is, however, outside the scope of this thesis and not necessary for any arguments presented.

shows that increasing the concentration of A increases the redox potential, and conversely when increasing [A⁻]. This can be viewed as analaguous to dop- ing: Decreasing the pH level increases the number of H⁺ions (by definition) and can be viewed as doping the electrolyte with acceptors (because they ac- cept electrons), thus decreasing the redox potential. Now that the electrolyte has been translated into materials terms, the electrolyte-semiconductor junc- tion can be examined.

As an n-type⁷semiconductor is submerged into an electrolyte (say, H⁺/H2), chances are, the Fermi level of the semiconductor will be higher than the redox potential of the electrolyte. This means that the electrons in the semiconductor have higher energies than those in the electrolyte. They will therefore relax into the lower energy levels of the electrolyte until the whole system has the same Fermi level (or redox potential). This will produce an electron deficit in the semiconductor and an electron excess in the electrolyte. Since the Fermi level is in a band gap in the semiconductor, adding or removing electrons will have a much larger impact on the Fermi level than it does on the redox potential, which changes negligibly in comparison. Therefore, it is a good approxima- tion to say that only the Fermi level of the semiconductor decreases, while the redox potential remains the same. As the Fermi level falls, fewer electrons will occupy the higher energy states of the conduction band and the dopant. Since the electrons surrounding the dopants will leave and to a larger degree occupy the states of the valence band and the electrolyte, the dopants will remain as positively charged ions. The positively charged ions of the electrolyte will then repel from the positively charged surface of semiconductor and the negative counter ions are attracted to the surface and adsorb to it. This process forms what is called the Helmholtz layer in the electrolyte. The electron deficiency in the semiconductor is localized at the surface of the semiconductor. This is be- cause the field from the Helmholtz layer will repel electrons from the surface.

However, as the electrons leave the surface, the charged immobile dopants will remain and shield the field, decreasing the field strength. The decrease in the field will be quadratic until it reaches zero. This is a consequence of Poisson’s equation, shown below.

d²V

dx² =−dE dx =−ρ

ε_s (Poisson’s equation)

V is the voltage, E is the electric field, ρ is the charge distribution, and ε_sis the permittivity of the material. In the case shown in the figure, the charge distri- bution would be that of the dopants, which is roughly constant. The voltage of an electron at a point is proportional to its energy. This means that the electron energy has a quadratic relationship to the distance from the surface. The charge distribution of the Helmholtz layer is fundamentally different since the shield- ing is not done by immobile charges, but by highly mobile ions. The charge

7We will only consider the case of an n-type semiconductor for brevity, but the concepts are closely related for p-type semiconductors.

distribution will be decreasing exponentially. The before and after shots of this process in terms of energy and electron concentration are depicted in what is called a band bending diagram, shown in figure 2.9.

(a) Before contact (b) After contact

Figure 2.9. The electron energy band diagram at the electrolyte-semiconductor junc- tion before and after contact. As the semiconductor is submerged and equilibrium is reached, its energy bands bend. This causes an electric field across the junction forcing electrons further into the bulk of the semiconductor.

A band bending diagram works in the same way as a band diagram (shown in figure 2.3), the only difference is that the point of a band bending diagram is to show how the bands bend at a junction. However, in these figures there are properties that are necessary to point out. The ordinate shows the elec- tron energy, linearly. But the energy is defined in reference to the energy of an electron in vacuum (E_vac), and the vacuum energy is not drawn as a line throughout the diagram (even though that the vacuum energy is in fact con- stant, 0 eV, by definition). Therefore, the constant parameter throughout the diagram is the Fermi level (after contact and at equilibrium). This means that any vertical line throughout the diagram will have a constant probability of finding an electron, in terms of the probability function shown in figure 2.3.

Since the energy bands are consequences of the atomic structure, their energies do not change during the contact. This is reflected by the constant distance be- tween the CB/VB edges and the vacuum level. The parameter that has changed is the electron concentration in the conduction band at the surface of the semi- conductor. This is reflected by the increased distance between the Fermi level and the CB edge. Similar reasoning goes for the electrolyte, where the bulk has maintained its energy. At the surface there is a lower concentration of H⁺ due to the positive charge of the surface. As shown above with the Nernst equation, this raises the redox potential. However, as the figure is drawn with the fermi levels constant, instead the vacuum level is drawn decreasing in the electrolyte close to the surface. The width of the Helmholtz layer has been

exaggerated for clarity. In reality it is much thinner than the band bending of the semiconductor.

2.3.1 Field effect

Field effect is the main actor in a field effect transistor (FET), as the name suggests. As an external electric field is applied over a conductor, the charges within the material will rearrange accordingly. Negative charges will move opposite the field while positive charges will move along the field direction.

This arrangement of the charges produces an opposite field which cancels it.

If the conductor is a metal or an electrolyte, the charge distribution will have canceled the field at the depth of a few nanometers from the surface (in an electrolyte, this is the Helmholtz layer). If the conductor is an n-type or p- type semiconductor with immobile charges, the charge distribution will look similar to that in the band bending diagram in figure 2.9, for similar reasons.

If the field is such that the electrons are repelled from the surface, the electron energy at the surface will be higher, forcing them to either leave or settle in the valence band. This leaves the immobile charged dopant solely responsible for shielding the electric field, which is described by Poisson’s equation as shown in the section above. If an electric field is applied such that the electrons are attracted to the surface it is the electrons that shield the field, rather than the dopants. In this case the charge distribution is similar to that of an electrolyte or a metal: decreasing exponentially into the material.

In the experiments presented in this thesis, the external electric field is ap- plied by applying a back gate voltage, V_bg, to the silicon wafer in which the MoS2sample is placed. The sample is separated from the silicon with an oxi- dized SiO₂layer, which is an insulator.

This control of the electron concentration is what is used in an FET. As the electron concentration increases, so does the conductivity. In an FET the current is switched on and off by increasing and decreasing conductivity of a semiconductor. This effect can also be used in electrocatalysis, where the conductivity of the material can have a great impact on the performance. As mentioned earlier, this sort of effect was shown in thin MoS2flakes by J. Wang et al. [46].

3. Experimental Details

3.1 Device Fabrication - Electron Beam Lithography

The MoS₂ samples were exfoliated and found using the scotch tape method described in section 2.1.1 in the Background. Then the electrodes were fabri- cated using electron beam lithography.

A bilayer resist was used to produce the pattern for the deposition of the electrodes. The two polymers used for the bilayer resist were two variations of polymethyl methacrylate (PMMA): 495PMMA A4 and 950PMMA A4. They were spincoated at 80 RPM for 8 seconds, at 4000 RPM for 40 seconds and then baked at 180^◦C for 5 minutes, one by one and with 495PMMA A4 at the bot- tom. The pattern was drawn using a scanning electron microscope (SEM), ex- posing both layers. Then the bilayer resist was developed using an∼ 5^◦C, 7:3 mix of isopropanol and water. 495PMMA A4, the bottom layer, is dissolved to a larger degree than 950PMMA A4, the top layer, producing a smaller opening to a larger exposed area. This makes the removal of the PMMA easier after the deposition. Now, when the pattern is made, the electrode material is deposited with electron beam evaporation in a pressure of about 5·10⁻⁶ Torr. The chip is placed in acetone for a few hours to dissolve the remaining metal covered PMMA, leaving only the metal covered pattern. The process is depicted in figure 3.1.

Since the gold electrolyte possess catalytic properties [46], its separation from the electrolyte is important. However, for the MoS₂ to function as a working electrode it needs to be exposed to the electrolyte. To achieve this, the sample is spincoated with 950PMMA A4 again, the same way as before. Then a window to the sample is exposed with the SEM and developed like before, making sure the window does not expose the electrodes. Figure 3.2 shows microscope pictures of a sample before and after the lithography process. The sample in the figure was made without any exposed edges.

Figure 3.1. The electron beam lithography process (Left to right, up to down). Two layers of electron beam resist are spincoated on the silicon chip. Then, an electron beam is used to draw a pattern. The resist exposed to the electron beam is dissolved by a developer fluid, exposing the silicon surface along the pattern. The electrodes are made by depositing gold on top of the chip. When the electron resist is removed with acetone, only the gold on the silicon remains.

(a) Before (b) After

Figure 3.2. Microscope pictures before and after the device was made. In figure (a), the two brown pieces at the center are monolayered MoS2samples. Using electron beam lithography, gold electrodes were drawn over the samples. Then, the chip was covered with PMMA to cover the gold electrodes. Lastly, windows through the PMMA were drawn over the MoS2samples using the SEM. The finished device is shown in figure (b). This device was made without exposing any edges and two electrodes were made in case one would fail.

3.2 Measurement - Cyclic Voltammetry

3.2.1 Setup

The catalytic performance was measured in 0.5 M sulfuric acid with a three electrode configuration as described in section 2.2.1. The MoS₂ sample was used as a WE. Initially, a platinum CE was used. This was later upgraded to a graphite CE. The difference between the two is solely in the risk of compro- mising the integrity of the data, as the platinum could conceivably break off from from the CE and fall down onto the WE and thus contributing to the per- formance with its superiority and result in deceivingly improved results. The results from the different CE did not show any significant differences of this sort. The saturated calomel electrode was used as a reference electrode; it has a potential of 0.2415 V vs RHE. The measurements were done with a Biologic SP-300 Potentiostat.

The volume of the electrolyte has two significant limitations. (1) The silicon needs to be electrically isolated from the MoS₂ to produce a field effect, this means that the electrolyte needs to be contained on the surface of the chip (i.

e. it can not be submerged in electrolyte). (2) The pins connecting the device to the rest of the circuitry needs to be in contact with the uncovered gold elec- trodes, but the gold has catalytic properties and thus needs to be separated from the electrolyte. This means that the electrolyte can not extend far enough along the surface to come in contact with the electrodes. Initially, the electrolyte was placed on the sample in a form of a droplet. As the results suggested that the electrolyte was depleted during the experiment this was soon upgraded to a Plastic Electrolyte Container (PEC). The PEC was a plastic construction glued to the chip that could hold larger volumes of electrolyte on top of the sample while still keeping it separated from the electrodes.

The field effect was induced by applying a voltage (called back gate voltage) to the silicon which was accessed by scratching the SiO2with a diamond scribe;

SiO₂acts as an insulating dielectric. The setup is depicted in figure 3.3.