Adapting a graphene tunnelling
junction for use in DNA
sequencing
THESIS
submitted in partial fulfillment of the requirements for the degree of
BACHELOR OF SCIENCE
in
PHYSICS ANDASTRONOMY
Author : R´emi Claessen
Student ID : 1654683
Supervisor : Prof. Dr. J.M. van Ruitenbeek Dr. F. Galli
2ndcorrector : Prof. Dr. H. Linnartz
Adapting a graphene tunnelling
junction for use in DNA
sequencing
R´emi Claessen
Huygens-Kamerlingh Onnes Laboratory, Leiden University P.O. Box 9500, 2300 RA Leiden, The Netherlands
July 20, 2018
Abstract
New methods of DNA sequencing using tunnelling currents are being investigated, in order to speed up the process of reading out longer
strands. It has been demonstrated that a tunnelling junction can be created by bringing two twisted graphene sheets together. This junction
may provide a setup for DNA sequencing using tunnelling currents. Various improvements in steps in the preparation of the graphene electrodes are investigated to optimise their quality for further use in sequencing applications. Moreover, the junction must be immersed in a
liquid for use in DNA sequencing. Several issues arise from this, including the formation of leakage currents throughout the liquid. Various methods of reducing the leakage currents by insulating the
Contents
1 Introduction 7
2 Theory 9
2.1 Graphene 9
2.2 Scanning Tunnelling Microscopy 10
2.2.1 Quantum Tunnelling 10
2.2.2 STM working principles 10
2.3 DNA Sequencing 11
2.4 Piezo Elements 12
2.5 Ionic currents in liquids 13
3 Sample Preparation 15
3.1 Method 15
3.2 Critical steps in the method 16
3.2.1 Silicon wafers 17
4 Tunnelling junction 21
4.1 STM mount 21
4.2 Approach and Tunnelling 22
5 Junction in liquids 27
5.1 Leakage currents 27
5.1.1 Characterising the current 27
5.1.2 Insulating the electrodes 28
5.1.3 Insulating the electrical contacts 31
Chapter
1
Introduction
Graphene is a truly unique material, since it consists of a single layer of atoms but is still very rigid. Since its discovery, it has been tested for nu-merous electrical applications, due to the mobility of free electrons that it has. A possible interesting application focuses on break junctions. This consists of bringing two graphene sheets in point contact and retracting them, to obtain an atomic chain of which the electrical and mechanical properties can be investigated [1, 2]. A different application is to use graphene for DNA sequencing, which is of great importance in the fields of biophysics and biological engineering [3]. This application is the focus of this investigation.
The main idea consists of placing a strand of DNA between two sheets of graphene and investigating the changes in the tunnelling current be-tween them (this is explained in more detail in section 2.3). Since graphene consists of a single-atomic layer, bringing two sheets together and apply-ing a voltage will result in a tunnellapply-ing current (more on tunnellapply-ing can be found in section 2.2.1). The problem with this is that graphene sheets are so thin, that the probablity of bringing the two sheets exactly towards each other is extremely low. Bellunato and Vrbica found a method to do this, by twisting two sheets relative to each other so that when brought together, the two sheets must intersect at a point [4]. This would then be an intersection where two single atoms are facing each other, so the tun-nelling current should be stable enough for DNA sequencing.
The next challenge is to reduce the gap between the graphene sheets to a very small distance, such that tunnelling can occur. Previous methods which created a fixed graphene-graphene junction prevent its use for
ap-plications in DNA sequencing [5]. Therefore, Bellunato and Vrbica used the controls of a scanning tunnelling microscope (STM) to tune the size of the gap. The basics of the mechanism are to use two graphene samples in-stead of a sample and an STM tip. In their paper, they demonstrated that it is possible to reach the tunnelling regime by adjusting the gap size with this method.
However, this method is not yet ready to detect single nucleotides for DNA sequencing. Kanneworff partly attributes this to the poor quality of the graphene electrodes [6]. Despite its rigidity, graphene is rather fragile due to the fact that it consists of a single atomic layer. During the prepa-ration of the samples, there are numerous possibilities for the graphene to get damaged through folding or puncturing, or the accumulation of impurities. It is therefore necessary to optimise several steps in the prepa-ration of the electrodes, to improve the quality of the junction for single nucleotide detection.
Once the quality of the electrodes is optimal, the junction must be adapted to liquid environments before it can be applied to DNA sequenc-ing. This is because DNA molecules to be sequenced will be in a solution, causing the STM to be in a humid environment. However, Kanneworff found that the piezo elements that control the STM tip are damaged by the humidity when the junction is used for such purposes. A method must be found that enables the graphene tunnelling junction to operate in liquids, so that it can ultimately be used for single nucleotide detection for DNA sequencing.
The first step towards these goals is to repeatedly carry out the fabri-cation procedure for the graphene electrodes. This will enable the iden-tification of the steps in which the quality of the electrodes is most likely degraded. Alternative methods that would reduce the likelihood of dam-age to the electrodes can then be developed and tested.
Once this has been done, the adaptation of the junction can be carried out for liquid environments. This includes addressing the problems that would arise in such a liquid junction. Apart from the previously men-tioned damage to the piezos, other problems may arise from the fact that liquid conducts electricity. By testing different aspects of the junction in water, the problems can first be characterised before a solution is designed and adapted for the ultimate aims of the junction project.
Chapter
2
Theory
2.1
Graphene
Figure 2.1: The hexagonal structure of graphene. The black circles are carbon
atoms which are covalently bonded to their neighbours (black lines). An extra electron provides a sea of free electrons throughout the surface. (Source: Rusanov [7])
Graphene is one of the allotropes of carbon and consists of a two di-mensional hexagonal lattice of carbon atoms (as illustrated in figure 2.1). Every carbon atom in the surface is bonded to 3 neighbours, which means that each atom contributes a single free electron to the lattice. At the edge, two free electrons per atom cause graphene to have a high reactivity. The free electrons in the lattice have a high mobility (2×105cm2 V−1 s−1), in a zero-gap semiconductor band structure. The resistivity is dependent on the back gate voltage applied across a graphene sample. This is because such a voltage can increase the charge carrier density in the graphene. [7, 8]
Because graphene is a nanomaterial, it is very susceptible to contami-nation. This means that the electrical properties are quickly altered if the samples are contaminated when left outside a vacuum environment. This can lower the work function for graphene in a tunnelling junction. [9, 10]
2.2
Scanning Tunnelling Microscopy
The scanning tunnelling microscope (STM) is a method of imaging a sur-face at the nano-scale, by making use of the phenomenon of quantum tun-nelling. An STM can be used to map the wavefunction of a material.
2.2.1
Quantum Tunnelling
According to quantum mechanics, particles such as electrons can be mod-elled by a wavefunction, where the squared amplitude of the wave is a measure for the probability of the existence of the particle. When a parti-cle encounters a potential barrier (such as a gap between two electrodes), there is a finite probability that the particle can be found at the other side of the barrier (shown in figure 2.2). This is known as the quantum tun-nelling effect. The tunnelled wavefunction decreases exponentially with distance. If two atoms that are not chemically bonded to each other are placed a short distance apart, tunnelling can cause the wavefunctions of the electrons from both atoms to overlap. If a voltage is applied across this gap (known as a bias voltage), a net current will flow between the atoms, consisting of electrons tunnelling across the gap. [11, 12]
2.2.2
STM working principles
An STM makes use of this tunnelling current to image nanostructures. It usually consists of a sample (to be investigated) and a tip, which is made up of a single atom. By applying a bias voltage and bringing the tip close to the sample, a tunnelling current can be measured between the sample and the tip. Due to the exponentially decaying nature of tunnelling, the two must be brought very close to each other and the current is obviously dependent on the distance between the sample and the tip. In order to image the sample, the tip scans the surface of the sample. By using a feed-back mechanism, the tunnelling current can be kept constant (at a value known as the setpoint current) by adjusting the height of the tip according to the sample. This then results in an image of the height of the sample with respect to the tip.
2.3 DNA Sequencing 11
Figure 2.2:The squared amplitude of a particle’s wavefunction shown as a
func-tion of distance (in blue). In black, a potential barrier is shown with a finite max-imum potential. The wavefunction decreases exponentially inside the barrier, so that there is a finite probability of finding the particle inside it.
For this investigation, the tip will be replaced by a second (graphene) sample. The junction will be characterised in two ways: the bias voltage is swept and the current is recorded for each voltage (current-voltage spec-troscopy); the bias voltage can also be kept constant whilst the distance between the electrodes is varied and the current is measured as a function of distance (current-distance spectroscopy).
2.3
DNA Sequencing
DNA (deoxyribonucleic acid) forms the building block of living organ-isms. A DNA molecule consists of a long string of paired smaller seg-ments, each of which is called a nucleotide. One of the components of such a nucleotide is one of four bases, which are conventionally labelled as G, C, A and T1.
The DNA of each living organism contains a unique sequence of these four bases in a specific order. The practice of measuring such a sequence is called DNA sequencing and has many potential applications in fields such
1The chemical names of the respective bases are guanine, cytosine, adenine and
as medicine or security. Several methods of sequencing have been used in the past but new methods are being investigated to speed up the process of reading out long strands. A possible approach that seems promising con-sists of placing DNA between two electrodes between which a tunnelling current flows. When a single nucleotide passes between the electrodes and blocks the gap, the conductance of the tunnelling current is altered as the current passes through the single molecule. The current through the junction is then dependent on the conductance of the nucleotide in the gap. Since the value of the conductance of each of the four bases is differ-ent2, single nucleotides passing through the junction can be distinguished by looking at the variations in the current. This can be used to sequence DNA. An illustration of the process is depicted in figure 2.3. [1, 3, 13, 14]
Figure 2.3: An illustration of DNA sequencing with a graphene tunnelling
junc-tion. The DNA strand in the middle is seen to be composed of a random sequence of the bases G, C, A and T (respectively the green, red, orange and blue seg-ments). A voltage is applied across the graphene electrodes (grey) resulting in a tunnelling current (red arrows) flowing across the gap. As the DNA strand moves through the gap, different nucleotides block the gap, causing the tunnelling cur-rent to change according to the conductance of the correponding base. (Source: Heerema [14])
2.4
Piezo Elements
The electrodes in the STM mount are moved by piezo elements. A piezo element is a crystal that can expand when a voltage is applied across it.
2The relative conductances of the four bases with respect to that of G are respectively
2.5 Ionic currents in liquids 13
The STM uses shear piezos, of which the principles are illustrated in fig-ure 2.4.
Figure 2.4:The working principles of piezo elements. (a) A regular piezo element,
where a potential difference (p.d.) is applied across the polarisation direction of the crystal (depicted by the blue arrow). The applied electric field (red arrow) is parallel to the polarisation, causing the piezo element to expand along that axis. (b) A shear piezo, as used in the STM. Here, an electric field (red arrow) is applied in a direction perpendicular to the polarisation of the crystal (blue arrow). This causes a shear motion of the piezo, as depicted. (source: Simons [15])
2.5
Ionic currents in liquids
It will be shown later (section 5.1) that a current can flow in a liquid when a potential difference (p.d.) is applied. This has various causes. Firstly, if the liquid is ionic then applying a voltage across it will cause ions to flow between the electrodes. Positive ions would then be attracted to the cathode and negative ions to the anode. Since the ions are free to move, a current is formed. As the ions move to the respective electrodes, fewer ions remain throughout the liquid. Hence, the current exponentially de-creases with time, resulting in a capacitative effect of the electrodes in the liquid.
This can still occur in distilled water since no liquid is completely unionised at finite temperatures (as described below). However, this effect is then very small. A more important source of current is the (partial) electrolysis
of water. Electrolysis of water is the process where a voltage splits water into hydrogen and oxygen in the following chemical reaction:
2H2O −→ 2H2 + O2 (2.1)
The underlying process of this reaction can be simplified by describ-ing it as the dissociation of the water molecules (H2O) into hydrogen ions (H+) and hydroxide ions (OH−). These ions respectively move towards the cathode and the anode where they respectively combine with or give up an electron3. This would result in the following half-equations (reactions at each electrode):
2e− + 2H+ −→ H2(cathode)
2OH− −→ O2 + 2H+ + 4e− (anode) (2.2) For this to occur on a large scale, an energy barrier must be overcome for the water to fully ionise. This is equal to the electrochemical potential, Vec= 1.23V. If the applied p.d. is lower than this, full electrolysis cannot oc-cur. At a finite temperature however, this reaction can still occur at a very small scale for lower voltages. This is due to the large spread of energy of water molecules at room temperature. The resulting current is therefore very small but may be significant when compared to tunnelling currents. These currents are known as Faradaic currents. [16–18]
3This is not the conventional way to describe the half-equations of the electrolysis of
water, since this would result in unequal pH values at both electrodes. However, the details of the chemistry of oxidation-reduction reactions are beyond the scope of this investigation.
Chapter
3
Sample Preparation
In order to be able to sequence DNA with the junction, the samples them-selves need to be improved. This can be done by eliminating or altering several steps in the sample preparation procedure, such that the chances of human error are reduced in the process.
3.1
Method
The samples are prepared at the Leiden Institute of Chemistry with the help of Amedeo Bellunato, according to the recipe used in Bellunato and Vrbica [4]. The graphene itself is grown on a copper substrate by chemi-cal vapour deposition (CVD). In order to use the samples in the STM, the graphene is placed onto two (undoped) silicon wafers, which have a thin silicon dioxide layer on them (Figure 3.1a). Since graphene is very thin, it is difficult to see with the naked eye and handling it is a very delicate op-eration. To facilitate this, a thin layer of polymer is used in the fabrication process of the electrodes.
Polycarbonate (PCA) is used to coat the graphene. The coating is ap-plied by dissolving PCA beads in chloroform and is then spread with a spin-coater (a rapidly spinning device which spreads the coating). Once the polymer is applied, the copper substrate is dissolved in an ammo-nium persulphate solution, leaving a sheet of graphene coated in PCA. This sheet is placed onto two silicon wafers (cut with a diamond knife as described in section 3.2.1) in a specially designed holder (which is illus-trated in figure 3.1). The breaking of the silicon wafers occurs such that, in theory, the edges are sharp and clear-cut.
Figure 3.1: The samples in the holder. The samples are held down by two metal flaps which are fastened only by a screw. There is a slit in the bottom of the holder, used for the plasma etching. (a) The samples in the holder before the plasma etching. The samples are placed on two silicon wafers (blue) and consist of a sheet of graphene (the white hexagonal structure across the wafers) with a polycarbonate (PCA) polymer coating (the white layer on top of the graphene). The silicon has a thin silicon dioxide layer on it. (b) The samples after the plasma etching and after the polymer has been removed. The graphene that was over the gap between the wafers is completely removed during the etching process, causing the graphene samples to extend only to the sharp edges of the wafers. (source: Bellunato and Vrbica [4])
There is a small gap between the wafers and the graphene-PCA sheet is placed across this gap. The graphene is then etched away by placing it in a plasma etcher where hydrogen plasma is used. This process is depicted in figure 3.2. Hydrogen ions are shot (isotropically) at the graphene that lies over the gap between the wafers, causing the graphene to be destroyed at the gap. The polymer on top of the graphene should in principle be unaf-fected, if the etching time is not too long. In theory, the graphene sheets should now have a sharp edge which stops exactly at the edge of the sili-con wafers. The remaining polymer is removed in chloroform. Insoluble polymer residues can accumulate on the wafers if the plasma etching time is too long. This can be solved by using shorter plasma etching times that are still long enough to etch away the graphene.
3.2
Critical steps in the method
As previously mentioned, there are several steps in the method described above that may degrade the quality of the samples. This will then in turn have adverse effects on the electric quality of the electrodes in the junction, making them unsuitable for applications such as DNA sequencing.
3.2 Critical steps in the method 17
Figure 3.2: The plasma etching process. The silicon wafers are represented in
dark blue, the graphene in grey and the polymer in light blue. (a) During the etching, hydrogen ions are shot at the sample from below (through the slit in the holder), so that the ions reach the graphene that is not protected by the silicon or the polymer. The etching is done isotropically, meaning that the ions are not directed at the graphene. (b) The sample after the etching. The graphene that was across the gap between the wafers is completely removed by the plasma, whereas the rest of the sample remains unaffected. The remaining polymer can then be removed by using solvents.
3.2.1
Silicon wafers
One of the first steps in improving the graphene samples is to improve the silicon/silicon dioxide wafers. This is because a sharp edge of the wafer will result in a sharper edge of the graphene after the plasma etching. The quality of the wafers is mostly influenced in the step where the wafers are cut and placed in the holder from figure 3.1. When placing the wafers in the holder, the screws are only tightened once the wafers are well aligned. This means that the two wafers risk hitting each other and getting dam-aged in the process. This can easily be solved by designing a new holder with slits in which the wafers can be placed to be stable.
However, the current method for cutting the wafers is unreliable for producing wafers of the correct dimensions. The wafers are cut by making a groove with a diamond scribe on the edge of the silicon and then apply-ing pressure to the wafer. The silicon should then break along the groove but this can vary by up to a millimeter. Moreover, the groove should en-sure that the silicon breaks along a crystal plane. Since the groove is made by hand, this is often not the case which results in a wafer with an edge that is not clear-cut.
A new method for cutting the silicon wafers is being investigated, where a wafer is placed into a tetramethylammonium hydroxide (TMAOH) so-lution. This should etch the silicon along (111) crystal planes and form a groove that is more regular than that made with a diamond scribe (see fig-ure 3.3). This groove should ensfig-ure that the wafers are cut along a sharper edge.
Figure 3.3: A side view of the TMAOH etching method for cutting the wafers.
The silicon (blue) is protected by an enriched silicon dioxide etching mask (grey). A small strip of the silicon at the bottom of the wafer is not protected by the etching mask, so that the silicon can be etched. The TMAOH solution will etch the silicon along crystal planes, causing the wafer to be ectched along a triangular prism (as illustrated). The wafer can be broken along the dashed line by applying pressure on the two halves on either side. This should in theory produce two silicon wafers with sharp edges.
In this method, only one side of the silicon needs to be etched. To protect the other side, an etching mask of silicon dioxide (SiO2) is used. Since the natural oxide layer on silicon is very thin, this oxide layer will need to be enriched. Since silicon dioxide is not a perfect etching mask, it will itself also be etched slowly by the TMAOH, at a much slower rate than the silicon.
After the etching, the remaining oxide thickness should not be too thick or too thin, since optical interference effects in silicon dioxide enable a
3.2 Critical steps in the method 19
monolayer of graphene to be detected by the naked eye [19, 20]. When making electrical contact with the graphene for the junction, it is impor-tant that the graphene is visible on the wafer. It is therefore necessary to know what the final thickness of the silicon oxide needs to be to make the graphene visible at the wavelengths of optical light. This can be computed by considering the electric field of the light through the graphene and sili-con oxide. The model used to compute this is explained in appendix A and the results are shown in figure 3.4. For the results, known values for the di-electric constants of silicon, silicon dioxide and graphene are used1[21–23]. The human eye is most sensitive at wavelengths around 550nm [24]. Optimal graphene visibility at this wavelength corresponds with a silicon dioxide thickness around 150nm or a thickness around 300nm. The latter is more practical to fabricate, which means that the final thickness of SiO2 should be between 280nm and 320nm. This range leaves a small margin for error in the production of the wafers, by remaining in a regime where the human eye is still sensitive to the light. By comparing the etching rates of SiO2 to those of silicon, the initial thickness of the etch mask can be calculated [25]. This information is passed on to the MEMs lab in Delft where the new wafers are made.
1The dielectric constants used are 3.9 for silicon dioxide, 11.68 for silicon and 4.5 + 7.5i
Figure 3.4: The visibility of graphene on a silicon dioxide and silicon wafer. The x-axis shows the wavelength of the light and the y-axis shows the thickness of the silicon dioxide layer on a silicon wafer with a thickness of 0.56µm. The coloured lines show the contrast of the graphene with respect to the silicon diox-ide background. The greater the absolute magnitude of the contrast, the better the visibility of the graphene on the background (meaning that -60 provides bet-ter graphene visibility than 0). The exact definition of contrast can be found in appendix A. Dark blue lines show a region of greatest contrast and dark red show the least contrast, as denoted in the colour key to the right of the graph.
Chapter
4
Tunnelling junction
The junction uses the mount from a scanning tunnelling microscope (STM). As described before, both the tip and sample are replaced by graphene electrodes which are in turn twisted and tilted. The electrodes each con-sist of a silicon wafer on top of which a sheet of graphene is deposited to the edge.
4.1
STM mount
The mount is shown in figure 4.1. It consists of two slots in which the wafers can be placed. One of them is connected to a slider, which itself is connected to 6 piezo elements which move it in the z-direction (towards or away from the other electrode). The other slot is connected to a holder which is connected to a piezo element which can move it in the xy-plane (perpendicular to the axis running through the two electrodes). The z-movement can be used for fine and coarse motion. For the fine z-movement, a voltage is applied across the piezo so that it can move the slider across its full range (200nm). This is used during the current-distance (Iz) spec-troscopy when the samples are in the tunnelling regime. The coarse mo-tion uses the so-called ”slip-stick” mechanism described by Simons [15]. The idea is that if a voltage is applied across a piezo very rapidly, the piezo will move very quickly and overcome the static friction coefficient of the slider, such that it does not move. If this is followed by a slow movement of the piezo, the slider will be moved by the piezo and the process is re-peated, allowing the slider to be moved by a much larger distance than the range of the piezo. This process is used to move the samples into and
Figure 4.1:The mount used for the STM. The sample holders with the wafers are shown by the blue arrows. The slider that moves in the z-direction (by means of 6 piezo elements) is shown by the red arrow. The holder that is moved by the piezos in the xy-plane is shown by the yellow arrow. (Source: Kanneworff [6])
out of the tunnelling regime. The xy-piezo only uses fine movement and can be used to scan different points along the same sample.
4.2
Approach and Tunnelling
The graphene samples are twisted (about the z-axis) so that they intersect at one point and tilted (about the x-axis) so that the silicon wafers do not come into contact. This is illustrated in figure 4.2.
The mount is connected to a controller that sends a signal to the piezo for approach, applies a bias voltage across the junction and measures the tunnelling current through it. Before the tunnelling current is measured, it passes through a current preamplifier with a gain of 108.
The STM is placed inside a sealed metal box that acts as a Faraday cage to suppress electrical noise. This box is placed on a vibration-free island to shield it from mechanical noise. By using the procedure described by Simons (using the measured tunnelling current for the feedback mecha-nism), the coarse z-motion is used to approach the samples until they are in the tunnelling regime. Once this is achieved, Iz and current-voltage (IV) spectroscopy are carried out to characterise the tunnelling junction.
4.2 Approach and Tunnelling 23
Figure 4.2: An illustration of the junction. The two silicon wafers (dark blue) are
twisted by an angle θ about the z-axis, so that the two graphene sheets (the grey hexagonal structures on top of the wafers) intersect at one point. The wafers are also tilted about an angle of 150◦ with respect to each other, so that the silicon wafers do not touch each other. (Source: Bellunato and Vrbica [4])
electrodes. The IV-spectroscopy curve in (a) is fitted with the Simmons model [12]. The fit gives a gap size of 2.37nm between the electrodes and a tunnelling barrier of 0.7eV. Previous results from Bellunato and Vr-bica found respective values of 1.3nm and 1.4eV. The values from this in-vestigation are of the same order of magnitude and the small differences can be attributed to the large variability of the electrical properties of the graphene samples. This is because the monolayer samples are susceptible to contamination (for example by dust), which can increase or reduce the work function of the electrodes. The Iz-spectroscopy curve in (b) appears to consist of two regimes. For distances between the electrodes that are larger than 0.45nm, the current decays rapidly with the distance and re-mains small for even larger distances. For smaller distances, the current appears to fluctuate about a value that lies around 5nA.
A possible interpretation could be that the regime for large distances is one where the current is a tunnelling current and follows an exponen-tial decrease as a function of distance. The regime for smaller distances would then be where the electrodes have come into contact and the cur-rent flows through individual carbon atoms. This should give a linear current-distance relation, as the electrodes move into each other [1]. The
fluctuations for the smaller distances in the approach curve (blue) are not present in the retreat curve (green). This suggests that the fluctuations are merely due to noise and that the current-distance relation is actually lin-ear. The hysteresis between the approach and retreat curves can then be explained by bond formation between the electrodes during the contact.
Apart from the high noise levels, it can also be noted that the current values in the contact regime are up to 2 orders of magnitude greater for the results from Bellunato and Vrbica (∼ µA for 0.1V bias compared to ∼nA for 0.04V bias). Both of these differences may possibly be attributed to impurities in the samples. The time between the making of the samples and obtaining the results is three weeks, during which the samples can be degraded if not stored properly. When repeating the experiment, better results can be achieved if the samples are either stored in high vacuum or made just before they are used for tunnelling.
4.2 Approach and Tunnelling 25
Figure 4.3: Spectroscopy results for a tunnelling current between two graphene
electrodes. Both use a setpoint current of 2nA. (a) IV-spectroscopy, where the tunnelling current is measured as the bias voltage is changed. The blue curve shows the measured data and the green curve shows a fit according to the Sim-mons model. The fit corresponds to a gap size of 2.37nm and a work function of 0.7eV. (b) Iz-spectroscopy for a bias voltage of 40mV, where the current through the junction is measured as the distance between the electrodes is changed. The distance is calibrated by using the fitting values from the IV-spectroscopy. For the blue line, the electrodes are moved towards each other and for the green line they are moved away from each other. For distances greater than 0.45nm (blue curve), the current appears to decrease to low values quickly with distance, which can be explained by a tunnelling current. For smaller distances, the current appears to be fluctuating around 5nA. This can be explained by the electrodes coming into contact.
Chapter
5
Junction in liquids
Ultimately, the junction will need to be in a liquid environment in order to sequence DNA. This is because the DNA molecules need to be in solution. As described before, this can result in several issues. The most important are the fact that the piezos need to be protected from humidity and the leakage currents through the liquid. Before testing the junction in solu-tions (as will be needed for DNA sequencing), various aspects are tested in water.
5.1
Leakage currents
5.1.1
Characterising the current
Since a liquid is a conducting medium, currents can flow between the junc-tion’s electrodes when these are submerged (as explained in section 2.5). To characterise such currents, a simple experiment is carried out where two gold-plated electrodes are placed in Milli-Q distilled water. The height of the electrodes in the water is varied, such that the contact area is changed. The current through the water is then measured as a constant potential difference is applied. The setup is shown in figure 5.1 and the results are shown in figure 5.2. In the results, it can be seen that the leakage currents are of the order of µA. Although this is quite small, it is significantly larger than the expected order of magnitude of the tunnelling currents, which are expected to be between 1nA and 100nA.
As expected, the results show that a smaller contact area results in a smaller leakage current. Furthermore, an exponential decay of the current
can be noted in the results, which comes from the capacitative coupling of the electrodes in liquid (see section 2.5). It can be noted that the de-cay time of the smallest electrode area is significantly shorter than that for larger surface areas. This means that the final (and smallest) value of the leakage current is reached more quickly for smaller electrodes, hence re-ducing the time during which the tunnelling current is hindered by the leakage.
Figure 5.1:The setup for the experiment to characterise the leakage currents. Two
gold plated electrodes are attached to a multimeter that measures the resistance of the water in a four point setup. The height of the electrodes above the table can be adjusted and measured with a ruler. This enables the contact area of the electrode in the water to be changed.
Both of these reasons suggest that it is very critical that the electrodes are well-insulated from the liquid to reduce interference from ionic cur-rents. In distilled water, the leakage current must be reduced by up to three orders of magnitude. In ionic solutions, this reduction must be even greater since the resistivity of such solutions is orders of magnitude smaller than for water.
5.1.2
Insulating the electrodes
To optimally reduce the leakage currents between the electrodes, it is ideal to cover as much of the graphene as possible. Since the edge of the graphene is necessary for tunnelling, this means that a method is needed to com-pletely cover the upper surface of the graphene. Current methods of STM in liquid environments cover the sides of the tip, such that only the end conducts. This is done by applying substances such as apiezon grease or organic polymers [26, 27]. This cannot be used for the purposes of this
5.1 Leakage currents 29
Figure 5.2:The current through the water as a function of time, for different
elec-trode contact areas. The different colours show the estimated surface area of the electrodes in contact with the water. Since the electrodes are not perfectly cylin-drical, the surface area is merely an estimate that is accurate enough for an order of magnitude characterisation of the leakage current. A bias voltage of 1V is used for all curves.
investigation for a variety of reasons. The first reason is that it is difficult to apply a hydrophobic coating such as apiezon, without damaging the graphene. Another reason is the fact that by covering the sides of a pyra-midal STM tip and leaving the end uncovered, the exposed conducting area is of the order of magnitude of 10µm2 (usually around 3µm×3µm). This reduces the leakage currents down to the order of pA as mentioned in Hihath. In the graphene junction, the graphene sheet is approximately 3mm wide. This means that the edge used for tunnelling is already ∼ 0.1µm2. In order to lower the leakage currents sufficiently, the entire up-per surface of the graphene must therefore ideally be covered to the edge. It is extremely difficult to apply a hydrophobic coating to such precision.
A more favourable approach would be to use a sheet of graphene that already has an insulating film covering its upper surface. Any excess of this insulating material can be removed with a higher degree of preci-sion than that to which a coating can be applied. A possibility lies in the use of hexagonal boron nitride (hBN), which is a thin-layered material with a similar hexagonal structure to graphene. If a thick enough layer of hBN is applied on the graphene, it can greatly reduce the leakage cur-rents from the surface [28] . The great advantage of hBN is that it adsorbs to graphene’s hexagonal structure [29, 30]. This means that it can be de-posited onto the graphene using a similar method that is used to deposit
the graphene onto the silicon wafers. This involves spin-coating a poly-mer onto the hBN (which is itself on a copper substrate), removing the substrate to deposit the hBN onto graphene and then repeating the pro-cess as before. This would ensure that the graphene is completely covered by an insulating layer, to the edge. Moreover, during the plasma etching the hBN can be etched away like the graphene, which would result in a sample where the hBN continues to the edge of the graphene (as illus-trated in figure 5.3).
Figure 5.3: The plasma etching process with a hexagonal boron nitride (hBN)
insulating layer. The silicon wafers are represented in dark blue, the graphene in light grey, the hBN in dark grey and the polymer in light blue. (a) During the etching, hydrogen ions are shot at the sample from below (through the slit in the holder), so that the ions reach the graphene and hBN that are not protected by the silicon or the polymer. The etching is done isotropically, meaning that the ions are not directed at the graphene. (b) The sample after the etching. The graphene and hBN that were across the gap between the wafers are completely removed by the plasma, whereas the rest of the sample remains unaffected. The remaining polymer can then be removed by using solvents. Both the graphene and the hBN insulating layer stop at the clear-cut edge of the silicon wafer.
However, several factors may still limit the use of hBN as an insulator for the junction in liquid. These are the fact that current can still tunnel
5.1 Leakage currents 31
through the hBN and the fact that the coverage of hBN (if it is grown by chemical vapour deposition) may not be completely uniform. The former can be reduced by using multilayer hBN instead of a monolayer sample, since tunnelling currents decrease exponentially as a function of distance. The overall quality of multilayer hBN as an insulator can be tested by ap-plying a layer on a graphene sample and measuring the leakage currents.
Before testing the hBN as an insulating material, a multilayered hBN film (approximately 10-12nm thick) is deposited using the procedure de-scribed above. This is done on a large silicon/silicon dioxide wafer, on which a thin graphene strip is deposited. Once the polymer is removed from the graphene, the hBN is deposited and the polymer is later removed. Some graphene is left uncovered by the hBN to leave the possibility of adding electrical contact to test the hBN insulation. Optical microscopy images of the result are displayed in figure 5.4.
In (d) it can be seen that the edge of the hBN is very straight and that the hBN follows the shape of the graphene (no visible folds or tears on the hBN that covers the graphene). However, in (b) and (c) it can be seen that the hBN (green) is damaged as many parts have been removed. This means that this sample cannot be used to test the insulating quality of the hBN. The damage is likely due to human error in the deposition, as is also often the case for the deposition of graphene on the wafers. This therefore does not suggest that the deposition method is unsuitable for the application but rather means that the process must be repeated.
5.1.3
Insulating the electrical contacts
Apart from the graphene, leakage currents can also originate from the elec-trical contacts with the electrodes. These are currently made by attach-ing copper wires (which are insulated on the outside) to the graphene by means of silver paste. In order to improve the electrodes, these contacts are planned to be replaced by aluminium contacts on silicon wafers, with wire bonded contact to the rest of the circuit (as shown in figure 5.5).
In order to insulate these against ionic currents, these can be coated with a hydrophobic substance such as apiezon vacuum grease. The ex-periment from figure 5.2 is repeated with the electrodes covered in two different insulating coatings: apiezon grease and GE varnish. For both substances, the lower part of the electrodes are covered by a thick layer of
Figure 5.4:Optical microscopy images of a silicon/silicon dioxide wafer on which a thin strip of graphene is deposited, on top of which a boron nitride (hBN) film is deposited. The visibility of the graphene on the silicon dioxide background is difficult to see on these images. A white arrow points to the boundary between the graphene and the background in each image (the graphene is a slightly lighter shade of blue than the background). (a) The wafer before the polymer coating is removed from the boron nitride. The polymer is a golden colour. (b)-(d) Different parts of the wafer after the polymer is removed. The hBN is a green colour and is clearly damaged in some places.
the coatings, such that it is of the thickness of 1-3mm. Apiezon grease is a greasy, hydrophobic substance whereas GE varnish is a viscous varnish that hardens over time as it is left in contact with air. The results of the leakage currents are displayed in figure 5.6. It can be seen immediately that the apiezon is a better insulator since it reduces the current to almost 0nA. However, it is noteworthy that the GE varnish reduces the current to the order of nA for an applied voltage of 1V for an area of the order of 0.1mm2. Since tunnelling currents for such a bias voltage are expected to be higher (order of 100nA), this level of insulation should in principle be enough for the junction. It is observed that apiezon (since it remains greasy and does not harden) is partly removed from the electrodes when placed in water. Therefore, if apiezon is used in the junction a thick coat-ing must be applied to the conductcoat-ing surfaces and it must regularly be checked. Any of the two substances can be used for insulating the
electri-5.1 Leakage currents 33
Figure 5.5:The new wafers that should be used for the junction in the future. The
dark blue represents the silicon/silicon dioxide wafer. The light grey rectangles are aluminium contacts that are coated with titanium nitride against oxidation. The two parallel, horizontal contacts should replace the silver paste as a way to make eletrical contact with the graphene. The remaining (vertical) contact can be used as a back gate to increase the charge carrier capacity in the graphene, if the wafer is made of doped silicon. The translucent, dark grey rectangle is the graphene, that should cover the aluminium contacts. To make further electrical contact with the aluminium, wire bonding can be used.
cal contacts. Apiezon is however better to suppress the leakage currents, as long as it is applied thoroughly.
Figure 5.6:The current through the water as a function of time, for different elec-trode contact areas. The elecelec-trodes are covered in apiezon grease or GE varnish to insulate them against leakage currents. The different colours show the estimated surface area of the electrodes in contact with the water and the coating used. Since the electrodes are not perfectly cylindrical, the surface area is merely an estimate that is accurate enough for an order of magnitude characterisation of the leakage current. A bias voltage of 1V is used for all curves.
Chapter
6
Conclusions and outlook
Several steps have been taken to characterise the problems that could be faced if the graphene-graphene tunnelling junction were to be used in a liquid environment.
New methods are being investigated to improve both the quality and the efficiency of the sample preparation process. Testing the samples made by the new processes should show whether these are indeed better for an application such as DNA processing. However, it has also been shown that when using older samples for scanning tunnelling spectroscopy, the re-sults contain significantly more noise and have lower current values than the results from fresh samples. The cause of the latter is likely linked to the former, demonstrating that adequate storage of graphene samples is very important for the quality of the results.
Furthermore, it was demonstrated that ionic leakage currents have a significant value when electrodes are placed in a liquid and that this can hinder the detection of tunnelling currents in the junction. This is already the case in distilled water, which means that the effect will likely be larger if the junction is used in an ionic solution. It is therefore very important that any conducting surfaces in the junction that are not used for tun-nelling are well-insulated in order to reduce the effect of leakage currents. A possible method for insulating the graphene is to deposit multilay-ered hexagonal boron nitride (hBN) onto it and etch this with hydrogen plasma such that the insulating layer ends at the edge of the sample. It was shown that hBN adsorbs to the graphene but further investigation is needed to test whether the plasma etching method works. Moreover, the
hBN must be tested to see whether its insulating effect is adequate for the tunnelling junction, since the coverage of the graphene by the hBN may not be uniform. Any other submerged conducting surfaces in the junction can be insulated by applying either apiezon grease or GE varnish, where apiezon grease has a slightly better insulating effect.
Once a successful method of insulating the electrodes against leakage currents has been developed, a new design for the STM mount must be found for use in liquids. The main aim of this would be to shield the piezos from humidity, without affecting the approach of the electrodes. Only once this has been achieved can the junction actually be used to test tunnelling currents in liquid, so that it can ultimately be used for the se-quencing of DNA.
Acknowledgements
For this project I would like to thank several people who have helped me. Firstly, I would like to thank Prof. Dr. Jan van Ruitenbeek who has both given me the opportunity to work on this project and has helped me to advance through it. I want to thank Dr. Federica Galli and MSc Sasha Vrbica who have helped me with the STM in the lab and all its associated issues.
I also want to thank MSc Amedeo Bellunato and MSc Pauline van Deursen for putting aside some of their time to help me with the sam-ple preparation.
I want to thank people from both the van Ruitenbeek group and other research groups in the lab for all the advice they have given me. Finally, I want to thank Gijs van Weelden, who also did his project on the graphene tunnelling junction and has helped me to think about and solve many problems during my own project.
Bibliography
[1] N. Agrait, A. L. Yeyati, and J. M. van Ruitenbeek, Quantum properties of atomic-sized conductors, Physics Reports 377, 81 (2003).
[2] D. Djukic, K. S. Thygesen, C. Untiedt, R. H. M. Smit, K. W. Jacobsen, and J. M. van Ruitenbeek, Stretching dependence of the vibration modes of a single-molecule Pt-H2 -Pt bridge, Physical Review B 71 (2005). [3] N. yang and X. Jiang, Nanocarbons for DNA sequencing: A review,
Car-bon 115, 294 (2017).
[4] A. Bellunato, S. D. Vrbica, C. Sabater, E. W. d. Vos, R. Fermin, K. N. Kanneworff, F. Galli, J. M. v. Ruitenbeek, and G. F. Schneider, Dynamic tunnelling junctions at the atomic intersection of two twisted graphene edges, Nano Letters 18, 2505 (2018).
[5] K. Ullmann, P. B. Coto, S. Leitherer, A. Molina-Ontoria, N. Martin, M. Thoss, and H. B. Weber, Single-molecule junctions with epitaxial graphene nanoelectrodes, Nano Letters 15, 3512 (2015).
[6] K. Kanneworff, The viability of single nucleotide detection using a graphene nanogap, Bachelor thesis, Leiden University, 2017.
[7] A. Rusanov, Thermodynamics of graphene, Surface Science Reports 69, 296 (2014).
[8] K. I. Bolotin, K. J. Sikes, Z. Jiang, M. Klima, G. Fudenberg, J. Hone, P. Kim, , and H. L. Stormer, Ultrahigh electron mobility in suspended graphene, Solid State Communications 146, 1 (2008).
[9] C. Oshima and A. Nagashima, Ultra-thin epitaxial films of graphite and hexagonal boron nitride on solid surfaces, Journal of Physics: Condensed Matter 9, 1 (1997).
[10] H. Yuan, S. Chang, I. Bargatin, N. C. Wang, D. C. Riley, H. Wang, J. W. Schwede, J. Provine, E. Pop, Z.-X. Shen, P. A. Pianetta, N. A. Melosh, and R. T. Howe, Engineering Ultra-Low Work Function of Graphene, Nano Letters 15, 6475 (2015), PMID: 26401728.
[11] C. Chen, Introduction to Scanning Tunneling Microscopy, Oxford Uni-versity Press, 2008.
[12] J. G. Simmons, Generalized Formula for the Electric Tunnel Effect between Similar Electrodes Separated by a Thin Insulating Film, Journal of Ap-plied Physics 34, 1793 (1963).
[13] T. Ohshiro, K. Matsubara, M. Tsutsui, M. Furuhashi, M. Taniguchi, and T. Kawai, Resequencing of DNA and RNA, Scientific Reports 2 (2012).
[14] S. J. Heerema and C. Dekker, Graphene nanodevices for DNA sequencing, Nature Publishing Group 11, 127 (2016).
[15] M. Simons, Development of an STM with combined displacement and scan unit, Bachelor thesis, The Hague University of Applied Sciences, 2016.
[16] L. Chen, X. Dong, F. Wang, Y. Wang, and Y. Xia, Base-acid hybrid water electrolysis, Chemistry Communications 53, 3147 (2016).
[17] Bockris, Reddy, and K. N. Amulya, Modern electrochemistry: an in-troduction to an interdisciplinary area, pages 5–8, New York: Plenum Press, 1970.
[18] P. Atkins, Physical Chemistry, pages 227–239, New York: W. H. Free-man and Company, 6th edition, 1997.
[19] S. Roddaro, P. Pingue, V. Piazza, V. Pellegrini, and F. Beltram, The Optical Visibility of Graphene: Interference Colors of Ultrathin Graphite on SiO2, Nano Lett. 7, 2707 (2007).
[20] J. Y. An and Y. H. Kahng, Optical observation of single layer graphene on silicon nitride substrate, AIP Advances 8, 015107 (2018).
[21] P. R. Gray, P. J. Hurst, S. H. Lewis, and R. G. Meyer, Analysis and Design of Analog Integrated Circuits, page 40, 5th edition, 2009.
[22] E. A. Taft and H. R. Philipp, Optical properties of graphite, Physical Review 138, A197 (1965).
BIBLIOGRAPHY 39
[23] T. G. Pedersen, Variational approach to excitons in carbon nanotubes, Phys. Rev. B 67 (2003).
[24] E. F. Schubert, Light-Emitting Diodes, page 275, Cambridge University Press, 2006.
[25] J. Thong, W. Choi, and C. Chong, TMAH etching of silicon and the in-teraction of etching parameters, Sensors and Actuators A: Physical 63, 243 (1997).
[26] J. Hihath, F. Chen, P. Zhang, and N. Tao, Thermal and electrochemical gate effects on DNA conductance, Journal of Physics: Condensed Matter 19(2007).
[27] K. Gentz and K. Wandelt, Electrochemical scanning tunneling mi-croscopy, Chimia (Aarau) 66, 44 (2012).
[28] L. Britnell, R. V. Gorbachev, R. Jalil, B. D. Belle, F. Schedin, M. I. Katsnelson, L. Eaves, S. V. Morozov, A. S. Mayorov, N. M. Peres, A. H. Castro Neto, J. Leist, A. K. Geim, L. A. Ponomarenko, and K. S. Novoselov, Electron tunneling through ultrathin boron nitride crystalline barriers, Nano Letters 12, 1707 (2012).
[29] C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe, T. Taniguchi, P. Kim, K. L. Shepard, and J. Hone, Boron nitride substrates for high-quality graphene electronics, Nature Nanotech. 5, 722 (2010).
[30] J. Xue, J. Sanchez-Yamagishi, D. Bulmash, P. Jacquod, A. Deshpande, K. Watanabe, T. Taniguchi, P. Jarillo-Herrero, and B. J. Leroy, Scanning tunnelling microscopy and spectroscopy of ultra-flat graphene on hexagonal boron nitride, Nature Materials 10, 282 (2011).
Appendix
A
Appendix - Optical visibility of
graphene
This appendix describes the theoretical background that is used in the computation of figure 3.4. This is a summary of the calculations from Rod-daro [19] and An and Kahng [20].
The visibility of a thin layer of graphene on a substrate is the result of interference effects. The electric field from infalling light is modulated by the layers of graphene and substrate through which it passes. This affects both the phase and magnitude of the infalling field.
Assume that the sample consists of a series of parallel layers of differ-ent materials (as depicted in figure A.1) and that each layer is parallel to
Figure A.1: The sample represented as a series of consecutive layers of different
materials (in this case air, graphene, silicon dioxide and silicon). Each layer j ends at position zj and has thickness ∆j. The light travels in a direction parallel to the
z-axis and is made up of a component travelling in the positive z-direction (αj)
the xy-plane. For normal incidence of polarised light, the infalling ray can be taken along the z-axis and the electric field is taken to be polarised in the x-direction. The electric field in layer j is then given by:
Ex = αjeikj(z−zj)−iωt + βje−ikj(z−zj)−iωt (A.1) Here, αj and βj respectively denote the components of the field trav-elling in the positive and negative z-directions in the layer j (as shown in figure A.1). kj is the wavevector of the electromagnetic wave in the layer and z and zj are respectively the position at which the field is calculated and the position of the interface between layers j and j+ 1. ω is the angu-lar frequency of the wave and t is the time. If layer j has dielectric constant j, the dispersion relation for the light is given by k2j/ω2= j/c2where c is the speed of light in vacuum.
To relate α and β in adjacent layers, the Fresnel equations can be used: αj+1 βj+1 = P0+ P0 − · T+ T− T− T+ · αj βj (A.2) Where P± = e±ikj+1∆j+1 denote the propagation coefficients of the field through layer j + 1 with thickness ∆j+1. T± = 12(1 ± kj/kj+1) denote the transmission and reflection coefficients at the boundary between layers j and j+ 1.
After a total of n consecutive layers, the final electric field coefficients can be related to the initial ones by repeatedly applying equation A.2 to obtain: α n βn = M11 M12 M21 M22 · α 0 β0 (A.3) Since all the incident light comes in from the left, α0 = 1 and βn = 0 which when plugged into equation A.3 gives that β0 = -M21/M22. Given that this is the electric field that is reflected on the incident side, the re-flection coefficient (rere-flection amplitude squared) is given by R = | β0|2. By defining Rg as the reflection coefficient from the graphene (on the sub-strate) and Rsas that from the substrate background, the contrast of graphene on the background can be defined by:
C = Rs − Rg Rs
43
This gives a measure of how well the graphene is visible against the substrate background. If C is negative, the graphene is lighter than the background and it is darker if C is positive. The absolute value of C is greater for greater visibility.
