Single-layer Molybdenum disulfide photodetectors

Single-layer Molybdenum disulfide photodetectors MSc Thesis Candidate: Oriol López Sánchez Supervisor: Prof. Andras Kis February,2012 1 Abstract Two-dimensional (2D) materials are very attractive candidates for use in next-generation nanoelectronic devices. Compared to one-dimensional materials, with 2D materials is relatively easy to fabricate complex structures. 2D materials, such as molybdenum disulfide (MoS2), have attracted increasing attention for their electronic and optoelectronic particular properties and size. MoS2 is a transition-metal dichalcogenide, it exhibits unique physical, optical and electrical properties. For this thesis single layers of MoS2 (thickness, 6.5 Å) were deposited on degenerately doped silicon substrates acting as a back gate with 270-nm-thick SiO2. Two gold electrodes as drain and source, lasers of 645 nm and 561 nm and a white light are used to investigate their photo-electric properties. Different surface treatments and the device fabrication process flow have been studied to see how to improve the electrical and optical properties of our single-layer MoS2 devices. In this project, it has been observed that the device fabrication improves their electrical properties. Surface treatment for low hysteresis, for high field-effect mobility and for a good relation hysteresis/field-effect mobility have been found. For the first time, a persistent photoconductivity (PPC) has been observed in MoS2 and a surface treatment for a fast PPC decay has been found. A fast photocurrent with higher photoresponsivity than others found before and much higher than graphene-based devices is determined by the source-drain or gate voltage. Single-layer MoS2 semiconducting materials could be a good candidate for multi-functional optoelectronic device applications in future. 2 Acknowledgments I would like to thank Professor Kis Andras for giving me the oportunity to work in his laboratory, and for helping me in my research as well as in my graduate school and fellowship application processes. I aslo wish to thank Dominik Lembke, the graduate student whose project I was a member of, for making my research experience so enjoyable and instructive at the same time. Thanks also to Professor Nuria Ferrer Anglada from my home university (UPC) for her advice and support and to Professor Aleksandra Radenovic and Serena Brando of Laboratory of Nanoscale Biology (LBEN) group in EPFL for allowing me to use their optical setup and for all their help. Finally, thanks to all the other members of Andras Kis group for being so friendly and helpful. I greatly enjoyed my time as a member of the group. 3 Contents Chapter 1 Introduction…………………………………………………………………………8 1.1- Experiment Overview…….……………………………………………………..9 1.2- Thesis Outline………………………………………………………………….. 9 Chapter 2 Instrumentation………………………………………………………………..…..10 2.1- Optical microscope – Olympus-……………………………………………....…10 2.2- Manual Prober Station ………………….…………………………………..……10 2.3- Optical setup ....…………………………………………………………….....…11 2.4- Labview program …..………………………………………...……………....…..12 Chapter 3 Basics of 2D-Materials and Photo-electric effect Theory..…………….......14 Chapter 4 Device Fabrication……………………………………………………………….. 17 4.1- Basics of MoS2 Exfoliation………………………………………….……..……...19 4.2- Substrate Preparation……………………………………………….….…………19 4.3- MoS2 Deposition…………………………………………………………..……….22 4.4- Design.……………………………………………………………………………...23 4.5- Lithography and Development. …………………………………..….…………..24 4.6- MoS2 Annealing ………………………………………………….………………..25 4.7- Coverslips Preparation……………………………………………….…………...27 4.8- Final device. …………………………………………………………..…………...29 Chapter 5 Results and Discussion………………………………………………………….30 5.1 Basic Characterization of device Quality…..………………………………….....30 5.2 Electrical results ………………………….......…………………………………...31 Threshold voltage………………………………………………………………33 Hysteresis………………………………………………………………….……35 Field-effect mobility………………………………………………………….…36 5.3 Photo response……………………………………………………………………..38 Persistent photoconductivity…………………………………………………..43 Laser spot position sweep……………………………………………………..48 Photocurrent……………………………………………………………….….…49 Chapter 7 Conclusion……………………………………………………………………….…54 4 List of figures Figure 1: Approximation of how future chips could look like [1] Figure 2: Single layer MoS2 photodetector. Figure 3: Olympus optical microscope. Figure 4: a) Probe Station. b) Agilent E5270B parameter analyser Figure 5: Image of the optical setup. a) CMOS camera. b) Sample and objective. c) 561nm laser. d) 645nm laser. e) Tunable laser (770-785nm). f) Controllable stage. Figure 6: Front panel of labview program created to control the stage for the experiment. a) Camera. b) Stage. c) Saving image. d) Scan. e) Measure current. f) Stop program. g) Cursors manual or automatic. h) Cursor positions. Figure 7: a) Three-dimensional representation of the structure of MoS2 [2]. b) AFM imaging of monolayer MoS2 [2]. Figure 8: a) Graphene lack of a band gap [3] b) Single-layer MoS2 is a direct gap semiconductor where the lowest energy interband transition occurs at the K point of the Brillouin zone. [4] Figure 9: a) MoS2 source. b) Si/SiO2 substrate with exfoliated MoS2. Figure 10: Single-layer MoS2 a) KOH treatment. b) Piranha treatment. c) HF treatment Figure 11: Plasma cleaner Figure 12: Furnace for 1000ºC treatment with a quartz suspector. Figure 13: a) Chip Maker after Plasma cleaning. b) Chip marker after 1000ºC treatment Figure 14: Contrast versus thickness measured for MoS2, WSe2 and NbSe2 flakes deposited on 270 nm SiO2 and containing different numbers of layers identified using Atomic Force Microscopy (AFM). For all three materials, the contrast increases with increasing layer number, indicating that optical imaging can be used to distinguish flakes with differing numbers of layers. [5] Figure 15: a) Front picture of final device connected holder b) Top picture of final device with the coverslip connected. 5 Figure 16: a) CAD design of a chip. b) Source drain contacts designed. c) Source drain contacts designed Improving efficiency of devices per chip. Figure 17: Vistec EBPG5000, electron beam lithography system [6] Figure 18: LAB600H Evaporator a) front b) back [6] Figure 19: a) After e-beam resist coating b) Developed with MIBK c) Evaporated d) After liftoff Figure 20: Furnace for annealing. a) Pump. b) Metal suspectors. c) Gas supply. Figure 21: a) Clean glass cover slip 24x60 mm. b) Cover slip exposed and developed with the positive AZ1512 photoresist. c) Final Coverslip. Figure 22: Final device connected on the Probe Station Figure 23: Mean number of usable flakes/chip for the different treatment combinations. Figure 24: Electrical characterization of MoS2 monolayer. Structure of a monolayer MoS2 photodetector together with electrical connections used to electrically characterize the device. [2] Figure 25: Substrate treated with HF and 1000°C treatment. Vg vs Is sweeps for the different steps to elaborate the final device Figure 26: Mean and standard deviation of the threshold voltage for the different combination of treatments. Figure 27: Threshold voltage mean of devices after annealing for the different treatments. Figure 28: a) Mean and standard deviation of the hysteresis for the different combination of treatments. Figure 29: Average hysteresis of devices after annealing for the different treatments. Figure 30: a) Mean and standard deviation of the field-effect mobility [𝑐𝑚2𝑉−1s−1] for the diferent treatments. Figure 31: Field-effect mobility mean of devices after annealing for the different treatments. 6 Figure 32: Photo-electrical characterization of MoS2 monolayer. Structureof monolayer MoS2 photodetector together with electrical connections and light source used to characterize the photoresponse of the device. [2] Figure 33: Optical setup build by Serena Brando and Aleksandra Radenovic of LBEN group in EPFL. Figure 34: A first look with a camera of the contacts without the filter. Images of the laser reflection. So we can’t really determinate where exactly the device is. Figure 35: With the filter and now using the CMOS camera much more sensitive with an exposure time of 0.4 seconds (the plasmonic resonance has a very low intensity). a) The device. b) The spot of the laser. Figure 36: a) Intensity X-profile of the red laser (645 nm) beam at the sample. b) Intensity Top-profile of the red laser (645 nm) beam at the sample. Figure 37: a) Absorbance and b) photoconductivity as a function of photon energy on MoS2 single layer and bi-layer [7]. Figure 38: Faraday cage on the optical setup. Figure 39: On/Off red laser (645nm) response for a constant Vg for differents Vs. Figure 40: ln [ ln(I(0))- ln(I(t)) ] versus ln(t) of the current response decay of the same device for the red and green laser. Surface treated with piranha and 1000ºC treatment. Constant Vg=0 and Vs=1. Figure 41: Time decay for devices of different treatments for Vg= 0, Vs= 1 V and Vg= 20 V, Vs= 1 V. Figure 42: Results obtained for different surface treatments. Colored are the best treatment for each electric or optic characteristic analyzed. In dark blue is the best treatment for good electrical properties in dark green the best treatment for a faster decay of the PPC. Figure 43: a) Vs sweeps with the b) spot of the laser in different positions on the device Figure 44: FFT of the electrical response of the device excited by a 645 nm laser beam modulated at frequency of 629 Hz. Figure 45: Photocurrent vs chopper frequecy for two diferent devices. 7 Figure 46: Photo current for different Vg of two diferent devices. Figure 47: a) The Fermi level of MoS2 is set at 4.7eV and the Fermi level of Au is 5.1eV. In the images we can see how they change for different Vg [8] Figure 48: Photo current for different Vs of two diferent devices. 8 Chapter 1 Introduction Since the invention of internet the new communication revolution has increased the load of the networks exponentially. Nowadays we need optical networks to afford the transmission of these high amounts of data. Optical fibers used before for long distances are getting closer to the user. Even before this, an older revolution started with the first transistor following Moore’s law is providing faster and more sophisticated computers and phones. Every year processors contain more and smaller transistors with sizes getting closer to the atomic scale. One day in close future each chip will process huge amounts of information and they will need light to be also the way to transfer information between the chips (fig.1). Figure 1: Approximation of how future chips could look like [1] Photodetectors are an important part of these predicted future chips [1]. If every year the components are getting smaller, two-dimensional (2D) materials compatible with CMOS fabrication which offer huge amount of applications could be the next step. Furthermore, they could be a door for flexible electronics. Graphene which has yielded in 2010 a Nobel Prize in physics to Andre Geim and Koustantin Novoselev [9] it is a good candidate for 2D electronics because of its unique electrical properties. Using graphene as a photodetector is however not easy because of the lack of a band gap. There are other 2D materials interesting for this purpose such as molybdenum disulfide (MoS2) with its single layer presents a direct band gap1 of 1.8 eV. For this reason, this is the material we used in this thesis. 1 Band gap: The energy difference between the top of the valence band and the bottom of the conduction band in insulators and semiconductors. 15 Single-layer MoS2 could be interesting as a semiconducting analogue of graphene, which does not have a bandgap in its pristine form (fig.8a). Bandgaps up to 400 meV have been introduced by quantum mechanical confinement in patterned [16] or exfoliated [17] graphene nanoribbons, but always at the price of significant mobility reduction (200 𝑐𝑚2𝑉−1s−1 for a 150 meV bandgap) [18], loss of coherence [19] or increased off-state currents due to edge roughness [20]. This is comparable to the mobility of 250 𝑐𝑚2𝑉−1s−1 found in 2 nm thin strained silicon films [21]. MoS2 monolayer has similar mobility but a higher bandgap than graphene nanoribbons [17], and a smaller thickness than the thinnest silicon films fabricated to date [21]. Bulk MoS2 has an indirect bandgap of 1.2 eV [22] and has also attracted interest as photovoltaic and photocatalytic material [23]. Whereas the band gap increases with decreasing thickness below 100 nm due to quantum confinement [24]. Single-layer MoS2 is a direct gap semiconductor with a bandgap of 1.8 eV (fig.8b) [7]. Other features that could make MoS2 interesting for nanoelectronic applications include the absence of dangling bonds5 and thermal stability up to 1,100 ºC. Figure 8: a) Graphene lack of a band gap [3]. b) Single-layer MoS2 is a direct gap semiconductor where the lowest energy interband transition occurs at the K point of the Brillouin zone. [4] The band gap is important to explain the Albert Einstein's mathematical description of how the photoelectric effect was caused by absorption of photons (eq.1). Was in one of his 1905 papers "On a Heuristic Viewpoint Concerning the Production and Transformation of Light" where his simple explanation in terms of absorption of discrete quanta of light explained the features of the phenomenon and the characteristic frequency. Einstein's explanation of the photoelectric effect was rewarded by the Nobel Prize in Physics in 1921. It says: 5 Dangling bonds: In chemistry, is an unsatisfied valence on an immobilised atom. a) b) 16 𝐸𝑝ℎ𝑜𝑡𝑜𝑛=ℎ𝑐 𝜆>𝐸𝑏𝑎𝑛𝑑𝑔𝑎𝑝 (1) Appling this theoretical explanation of photoelectric effect to single-layer MoS2, the incident photon energy must be greater than the energy gap (Eg) around 1.83 eV in single-layer MoS2 [4]. 𝐸𝑝ℎ𝑜𝑡𝑜𝑛=ℎ𝑐 𝜆> 1.83𝑒𝑉=> 𝜆𝑝ℎ𝑜𝑡𝑜𝑛<676𝑛𝑚 (2) Only those incident photons with wavelength < 676 nm (eq.2) can excite electrons from the valence band (VB) to the conduction band (CB) in the single-layer MoS2, generating the photocurrent when the drain voltage is applied. So far, electrical characterizations of single-layer MoS2 have shown n-type conductivity. Such n-type doping might come from the impurities, such as halogen (Cl or Br) atoms, which could replace S atoms in the natural MoS2 crystals or exist as the interstitial atoms in the interlayer gap of MoS2. Doping with rhenium could also lead to n-type doping. This increases the total electron concentration of the host MoS2 system resulting in an n-type doping for MoS2 [8]. The device resistance can increase during storage at ambient conditions. This could be attributed to absorption of oxygen and/or water from the environment [2]. 17 Chapter 4 Device fabrication The scotch tape-based micromechanical exfoliation [14] allows to transfer MoS2 monolayers from commercially available crystals of molybdenite (SPI Supplies Brand Moly Disulfide) to cleaned degenerately doped silicon substrates covered with 270-nm-thick SiO2 (Fig.9b) [2]. An optical microscope (Olympus) is used to locate the target MoS2 sheets. Afterwards, electron-beam lithography and metal deposition by evaporation are used to obtain devices with gold source drain contacts. After lift-off with acetone, the fabricated phototransistors are annealed for 2 h in a vacuum tube furnace with 100sccm Ar flow at 200 ºC (similar recipe than the used for transistor MoS2 in [2]). Step Process description Cross-section after process 01 Clean Si/SiO2 Substrate: Acetone, IPA, water 02 Surface treatment 1: KOH, PI or HF 03 Surface treatment 2: Plasma treatment or 1000°C treatment 04 Exfoliate MoS 2 : Scotch tapebased method 18 Finally, devices are transferred on a transparent coverslip also with gold contacts and the electrical properties of the phototransistor are tested at room temperature in air. 05 Ebeam Resist: 100nm MMA EL6, 100nm PMMA A2 06 Ebeam: Vistec EBPG5000 07 Develop: 3minutes 1MIBK:3IPA 08 Evaporate: 90nm Au 09 Lift-off: Acetone, IPA. 10 Annealing: 200ºC, Ar atmosphere 19 4.1 Basics of MoS2 exfoliation Exfoliated MoS2 is made by placing bulk samples of high-quality MoS2 into contact with a substrate (fig.9b). The adhesive force between the MoS2 and substrate may in some cases be great enough to peel layers of MoS2 off the bulk sample (fig.9a). Because of the low probability of only a single layer of MoS2 being peeled off in this process, the traditional method for exfoliation is to use a tape covered with MoS2 to contact the substrate, so a reasonable number of single layer MoS2 flakes can be obtained. Figure 9: a) MoS2 source. b) Si/SiO2 substrate with exfoliated MoS2. 4.2 Substrate preparation The first step of the fabrication process is the preparation of substrates to transfer MoS2. To exfoliate MoS2 on top of SiO2 chips they have to be clean. Samples are first sonicated for 10 min with acetone to remove all organic residue, then isopropyl alcohol (IPA) to remove the acetone and finally de-ionized water (DI-water) to remove the IPA. Samples are then dried with a nitrogen gun. Afterwards, there are different treatments to the substrate to improve the I-V characteristics of the final devices [25]. KOH treatment The first treatment tried after cleaning the chips is KOH treatment. The substrates are put in 30% vol. KOH for 30 min. This etches several nm away from the top of the SiO2 layer covering the samples. Afterwards, they are rinsed with DI-water and dried with a nitrogen gun. a) b) 20 Piranha treatment The next treatment used is Piranha (PI) treatment. The substrates are immersed in 60ml of sulfuric acid with 20ml of peroxide for 45 min. Afterwards they are rinsed with DI-water and dried with the nitrogen gun. HF treatment Finally the most aggressive treatment used is HF. The chips are left inside 2ml of 50% vol. HF and 70ml of DI water for just 30 seconds. Afterwards they are rinsed with DI-water and dried with the nitrogen gun. With this 30 sec 50 nm of the initial substrate have been removed. This explains the difference in the optical image contrast with the other two treatments (brighter images, fig.10c) Figure 10: Single-layer MoS2 with a) KOH treatment. b) Piranha treatment. c) HF treatment After these cleaning treatments there are other treatments that improve even more the electrical properties of the final devices. Furthermore, the probability of finding more singlelayer flakes after the exfoliation can increase with these treatments. Plasma6 treatment Some substrates were placed inside the plasma cleaner (fig.11) and exposed to oxygen plasma at 270 RF power and 30 sccm atmosphere flow for 20 minutes. 6 Plasma: A partially ionized gas generated by introducing gas into a vacuum chamber and exposing it to an electromagnetic field. 10µm 10µm a) b) c) 10µm 21 An important characteristic of low-pressure plasma is its extraordinary penetration capability. Narrow spaces where liquids cannot penetrate present no problem for gas plasma permitting parts with complex shapes and even micrometric cracks to be treated. Plasma cleaning removes organic contamination from surfaces primarily via chemical reactions with physical ablation being secondary. In the case of oxygen plasma, excited oxygen species decompose low molecular weight organic molecules by breaking the C-H and C-C bonds to form water vapor and CO2 which are evacuated from the chamber via the vacuum pump. Continual elimination of the gas eliminates cross contamination. Figure 11: Plasma cleaner 1000 °C treatment In this case, the chips are exposed to 1000ºC temperature for 5 min in ambient atmosphere inside a furnace (fig.12). The furnace includes a processing chamber configured to store a substrate; a susceptor located in the processing chamber to load the substrate and an auxiliary heater to heat the substrate. There are different susceptors for the different temperatures. The quartz susceptor is used for this 1000ºC treatment (fig.12). 22 Figure 12: Furnace for 1000ºC treatment with a quartz suspector. The differences of these last two treatments starts in their optical images (fig.13). With the 1000ºC treatment we observe how all the markers present a green color (fig.13). Figure 13: a) Chip Maker after Plasma cleaning. b) Chip marker after 1000ºC treatment 4.3 MoS2 deposition After the SiO2 substrate is prepared, thin layers of MoS2 are taken from a bulk (fig.9a) with a tape. Then the tape with MoS2 is pressed on the chip and once is taken off different layers of MoS2 are left on the device. Exfoliated MoS2 produce high quality single layer flakes in terms of electrical properties, but it is difficult to produce them in large quantities, and the placement of the MoS2 on the substrate is uncontrollable by the person doing the exfoliation. After the exfoliation singlelayer MoS2 are localized with the optical microscope Olympus (presented in instrumentation chapter fig.3). a) b) 23 The reason of the oxide thickness used is because of previously work done by the Laboratory of Nanoscale Electronics and Structures (LANES) at EPFL. They found this thickness optimal for optical detection of single-layer MoS2. They had established the correlation between contrast and thickness of various transition-metal dichalcogenides (TMDC) (fig.14) [5]. Figure 14: Contrast versus thickness measured for MoS2, WSe2 and NbSe2 flakes deposited on 270 nm SiO2 and containing different numbers of layers identified using Atomic Force Microscopy7 (AFM). For all three materials, the contrast increases with increasing layer number, indicating that optical imaging can be used to distinguish flakes with differing numbers of layers. [5] 4.4 Design Once the best flakes are localized they are selected in terms of dimensions, how clean and easy to contact they are. The contacts are designed with the software Design CAD express 18. The devices will have to be protected from the water due to the water-immersion objective (see instrumentation chapter) by a thin transparent glass cover slip to let pass the light (fig.15b). To be able to measure the electrical properties of the final photo excited device glass coverslips with gold contacts are connected to the contacts of the device (fig.15). 7 AFM: Consists of a cantilever with a sharp tip (probe) at its end that is used to scan the specimen surface. When the tip is brought into proximity of a sample surface, forces between the tip and the sample lead to a deflection of the cantilever according to Hooke's law. Typically, the deflection is measured using a laser spot reflected from the top surface of the cantilever into an array of photodiodes. 24 Figure 15: a) Front picture of final device connected holder b) Top picture of final device with the coverslip connected. The previous images show how the source and drain contacts on the chip need to go till the end of the chip to be connected with the coverslip contacts. Hence, not just the device but also the coverslip with contacts need to be built to connect the device. Furthermore, the efficiency in terms of number of devices per chip can be improved by building more than one device contacts in one flake if it is big enough (fig.16c). Figure 16: a) CAD design of a chip. b) Source drain contacts designed. c) Source drain contacts designed Improving efficiency of devices per chip. 4.5 Lithography, development Once the devices are designed they are ready -after spin coating with the ebeamresist- to be exposed with Vistec EBPG5000 electron beam lithography system (fig.17). 10µm 10µm a) b) b) c) a) 31 Figure 23: Mean number of usable flakes/chip for different treatment combinations. 5.2- Electrical results Electrical characterization of the device is performed at room temperature using the prober station mentioned previously in the instrumentation (fig.4a, b). The MoS2 flake which is 6.5-Å- thick is first characterized by applying a drain–source bias Vds to the gold contacts (drain and source) and a back gate voltage Vbg to the silicon substrate (fig.24). Figure 24: Electrical characterization of MoS2 monolayer. Structure of a monolayer MoS2 photodetector together with electrical connections used to electrically characterize the device. [2] This structure has been previously studied by LANES group, as they said the source current versus source bias characteristics is linear in the +50 mV range of voltages, indicating that 0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 KOH/plasma treatment KOH/1000°C treatment Piranha/plasma treatment Piranha/1000°C treatment HF/plasma treatment HF/1000°C treatment Mean usable flakes/chip KOH Piranha HF 32 the gold contacts are ohmic, excluding the possibility that the field-effect behavior is dominated by Schottky barriers at the contacts [2]. The device was characterized after different parts of the process. This way if there is a step that degrades the device and find a way to improve it are: - Before annealing, once the gold contacts are already connected to the flake (chapter 4.5). - After annealing, the device should present a better quality because the water is evaporated and the contacts resistance is lower [2] (chapter 4.6), - Final device, once the chip is connected with the coverslip (chapter 4.7, 4.8) so it is ready for the optical setup. The sweeps of one device with the substrate treated with HF treatment and 1000°C treatment are plotted in the same graph (fig.25). The low-field-effect mobility increases in the process of making this final device. Furthermore, the hysteresis seems to be reduced but it looks like the threshold voltage it has been moved far away from zero. Figure 25: Substrate treated with HF and 1000°C treatment. Vg vs Is sweeps for the different steps of the process flow. The previous image is just representative of one device. More devices need to be compared. For this purpose the mean and standard deviation of the threshold voltage, hysteresis and Vds= 10 mV V g (V) Is (nA) Vg(V) I s (nA) V ds = 10 mV 50 40 30 20 10 33 low-field-effect mobility is measured for the different combination of treatments in the different parts of the process. 94 devices were analyzed before annealing, after annealing and the final device. Not all of them worked on these three steps. Some devices broke in the process (most of them in the last step) and some did not present any response before annealing and they did after. At the end, 31 devices with different surface treatments and analyzed on the different steps in the process were used for the characterization of how the process flow affects the devices in terms of the electrical properties. To compare the different surface treatments, 82 devices analyzed after annealing are used. Threshold voltage Threshold voltage is defined as the gate voltage above which the transistor is turned on. It should be located in the range of the voltages present in the electronic circuit [26]. The ideal value for digital electronics would be zero. In figure 26 the average of the threshold voltage seems to be usually shifted to lower voltages as proceed with the fabrication of the device. This is good if it is getting closer to zero but not when the threshold voltage becomes negative. Considering just the mean, our final devices present a better threshold voltage in general than at the beginning although looking at the standard deviation (stdev) the variability of the threshold voltage normally increases after annealing (fig.26). Maybe, a more important factor than the threshold voltage is the hysteresis of this threshold voltage and knowing how it can change in the same device. Related to comparing the different treatments (fig.27), analyzing the devices after annealing for the different treatments, cannot be concluded there is a best for a desired threshold voltage. The treatments with good threshold voltage have a high deviation and the ones with a short deviation have an unfavorable threshold voltage. 34 Figure 26: Mean and standard deviation of the threshold voltage for the different combination of treatments. Figure 27: Threshold voltage mean of devices after annealing for the different treatments. -50 -40 -30 -20 -10 0 10 20 30 40 Pl tr 1000 MEAN PT MEAN 1000 KOH Piranha HF Plasma Plasma Plasma 1000°C 1000°C 1000°C Mean Plasm. treatment after annealing Mean 1000°C after annealing treatment Threshold voltage (V) Plasma treatment after annealing 1000°C treatment after annealing 8 Samples 5 Samples 5 Samples 6 Samples 2 Samples 5 Samples ° ° ° 35 From these measurements we can conclude that there are other, more important factors that determine the threshold voltage than the surface treatment. Hysteresis In order to measure the threshold voltage it is important to know if it is always the same or it for different sweeps or it is changing from onw sweep to the other. This hysteresis is not just important for the threshold voltage: it is important to know the error made characterizing the device. Figure 28 shows an improvement of the devices hysteresis thanks of the process flow. After annealing the hysteresis is a bit smaller in general than before the annealing but the deviation of the hysteresis is higher. The important difference is after the devices are wired with the coverslip, the hysteresis is reduced and the deviation is lower, presumably because the coverslip could prevent the absortion of humidity. In terms of hysteresis and related to the device fabrication procedure we can affirm that the the hysteresis is reduced in our final devices. Figure 28: a) Mean and standard deviation of the hysteresis for the different combination of treatments. Related to comparing the different treatments, analyzing the hysteresis of devices after annealing (fig.29), the different treatments can be ordered from the best to worst in terms of hysteresis: PI/1000ºC, KOH/1000ºC, HF/1000ºC, KOH/plasma, HF/plasma, PI/plasma. 8 Samples 5 Samples 5 Samples 6 Samples 2 Samples 5 Samples ° ° ° 36 Figure 29: Average hysteresis of devices after annealing and for different treatments. It’s interesting how 1000ºC treatment reduced the hysteresis when compared to the plasma treatment. Field-effect mobility The other important flake quality factor is the charge mobility. The low-fileld field-effect mobility (eq.3) can be extracted with the values mentioned previously (eq.4). The field-effect mobility of the devices is lower than ∼200 cm2V−1s−1 obtained from the topgate FET with high-κ gate dielectric of HfO2. The reason could be that the trap/impurity states that exist at the SiO2 surface in the bottom gate FETs, and the scattering from these charged impurities degrades the device mobility [2, 27, 28]. In figure 30a, b shows how the process flow improves the mobility of the devices. After annealing the average mobility is already much higher for all the different treatments than before the annealing and even higher for the final devices. 0.25 0.50 1.00 2.00 4.00 8.00 16.00 32.00 64.00 Pl tr 1000 MEAN PT MEAN 1000 KOH Piranha HF Plasma Plasma Plasma 1000°C 1000°C 1000°C Mean Plasm. treatment after annealing Mean 1000°C after annealing treatment Hysteresys (V) Plasma treatment after annealing 1000°C treatment after annealing 37 The deviation also increases in each step of the fabrication and is highest for the final device. This deviation is around much higher values than the original devices which makes them better than those at the beginning of the process in terms of electrical properties. Figure 30: a) Mean and standard deviation of the field-effect mobility [𝒄𝒎𝟐𝑽−𝟏𝐬−𝟏] for the different treatments. Related to the different treatments fig.31 shows the different values of the mobility obtained after annealing. The relation between the standard deviation and the mean can order the treatments from highest to lowest field-effect mobility as follows: KOH/plasma, KOH/ 1000ºC, PI/plasma, PI/1000ºC, HF/plasma, HF/1000ºC. The data shows that in terms of mobility the plasma treatment is better than the 1000ºC and KOH is better than PI which in turn is better than HF. In conclusion, the final devices glued on the cover slip are better in terms of electrical properties. From the surface treatment point of view, it depends what are we looking for: mobility, hysteresis or threshold voltage. For the purposes of this project it is more interesting the optimal treatment for photodetectors the next chapter 5.3 answers the question. 8 Samples 5 Samples 5 Samples 6 Samples 2 Samples 5 Samples ° ° ° 38 Figure 31: Field-effect mobility mean of devices after annealing for the different treatments. 5.3 Photo response Fig.32 shows how the phototransistor has a similar configuration of the field-effect-transistor (FET). Light sources used to generate a photo response have different spectrum and power. Chapter 3, with the eq.1 and eq.2, shows the wavelength of the laser has to be lower than 676 nm (energy of the photons higher than 1.83eV –fig. 37b-). This is the reason for using the 645 nm and 561 nm lasers. The third light source was a broad white light of Olympus KL 1500 LCD (see annex) of 0.073 W/cm2 . 0.0078125 0.015625 0.03125 0.0625 0.125 0.25 0.5 1 2 4 8 16 32 64 Pl tr 1000 MEAN PT MEAN 1000 KOH Piranha HF Plasma Plasma Plasma 1000°C 1000°C 1000°C Mean Plasm. treatment after annealing Mean 1000°C after annealing treatment Mean Mobility [I.S] Plasma treatment after annealing 1000°C treatment after annealing 39 Figure 32: Photo-electrical characterization of MoS2 monolayer. Structure of a monolayer MoS2 photodetector together with electrical connections and light source used to characterize the photoresponse of the device. [2] In the optical setup fig.33 the laser light interacts with the sample after going thru a lens, high pass filters (HPFs - transmit wavelengths higher than a given value and reflect the rest-), optical fibers (OF), dicroic filters (reflect the unwanted portion of the light and transmit the remainder) and the objective. Figure 33: Optical setup build by Serena Brando and Aleksandra Radenovic of LBEN group in EPFL. Before the camera, there is a filter to remove the reflection of the red laser. In this way the plasmonic resonance of the gold contacts (fig.35) and the markers we can be seen when they are illuminated with the red laser (645 nm). The plasmonic resonance of the gold (fig.35) 40 tells where exactly the gold contacts are. With the reflection of the laser (fig.34) the device cannot be localized. Figure 34: A first look with a camera of the contacts without the filter. Images of the laser reflection. Figure 35: With the filter and using the CMOS camera much more sensitive with an exposure time of 0.4 seconds (the plasmonic resonance has a very low intensity). a) The device. b) The spot of the laser. After all lenses, filters and the objective, the light from different lasers that we used, reaches the sample with a power much lower than at the laser source. In order to determine the spot size of the lasers and the light power on sample, the Thorlabs Beam Profiler BP104 with the Beam Analyzing Software (fig.36a,b) is used. a) b) 47 Figure 41: Time decay for devices of different treatments for Vg=0,Vs=1V and Vg=20V,Vs=1V I think that, the most aggressive treatment (HF) results in the smallest impurity concentration which would produce faster decays while the less aggressive treatment (KOH) results in more impurities the slowest decays. The deviation is also decrease, as seen in fig.41 it is even smaller for higher Vg. The PPC effect in single layer MoS2 is been observed, which can be well described by the stretched-exponential function. Furthermore, HF surface treatment has faster responses than with PI and with PI faster than with KOH. The following figure 42 summarize all the results obtained for different surface treatments. The best in terms of good relation of mean and standard deviation is KOH/1000°C treatment for good electrical properties and HF for a fast decay of the PPC. 32 64 128 256 512 1024 0 1 2 3 4 5 6 7 KOH PI HF KOH MEAN PI MEAN HF MEAN Vg=0V, Vs=1V Vg=20V, Vs=1V τ (sec) 48 KOH Piranha HF O2 plasma (13 samples) 1000°C (12 samples) O2 plasma (24 samples) 1000°C (15 samples) O2 plasma (8 samples) 1000°C (10 samples) Vt (V) 0.30 Min= -42.9 Max= 27.45 Stdev= 19.30 10.88 Min= -10.91 Max= 20.7 Stdev= 11.23 2.51 Min= -29.25 Max= 18.5 Stdev= 14.57 17.75 Min= -22.55 Max= 25.5 Stdev= 11.58 14.58 Min= 7.665 Max= 21.8 Stdev= 4.13 21.33 Min= 16.05 Max= 24.55 Stdev= 2.33 Hysteresis (V) 8.39 Min= 2.8 Max= 25.3 Stdev= 7.38 7.15 Min= 3.7 Max= 15.94 Stdev= 3.17 13.12 Min= 0.6 Max= 40.96 Stdev= 10.70 3.63 Min= 0.9 Max= 17.7 Stdev= 4.27 11.30 Min= 1.3 Max= 30.74 Stdev= 8.99 7.01 Min= 1.5 Max= 13.2 Stdev= 4.26 Mobility (cm2V-1s-1) 6.10 Min= 0.02 Max= 13.2 Stdev= 4.33 8.73 Min= 0.73 Max= 21.7 Stdev= 7.42 14.27 Min= 0.43 Max= 33.5 Stdev= 16.30 4.58 Min= 0.02 Max= 50.4 Stdev= 13.34 4.42 Min= 0.06 Max= 10.9 Stdev= 3.86 0.83 Min= 0.01 Max= 2.67 Stdev= 0.88 Decay time (sec) (6 samples) 322.84 Min= 90.84 Max= 556.9 Stdev= 278.95 (5 samples) 164.96 Min= 57.17 Max= 449.31 Stdev= 151.58 (4 samples) 121.66 Min= 56.52 Max= 437 Stdev= 137.92 Figure 42: Results obtained for different surface treatments. Colored are the best treatment for each electric or optic characteristic analyzed. In dark blue is the best treatment for good electrical properties in dark green the best treatment for a faster decay of the PPC. Laser spot position sweep The controllable stage and the camera are used to sweep the laser spot along the device to measure if there is a difference in photo response in relation to the position of the laser on the device. Vs sweeps for a constant Vg are made in different positions of the laser spot with 30 min difference for each sweep to make sure the persistent photocurrent is gone. After 10 sweeps with steps of 1µm position a change of the photo response is seen with a maximum when the laser spot is on the contacts and a higher response on the flake than next to the device (fig.43). This difference could may be higher if the spot size of the laser is decreased closer to the theoretical value. Changing position of a 1mm laser spot the intensity of the light changes in each position but the laser spot is illuminating the device in all steps. 49 Figure 43: a) Vs sweeps with the b) spot of the laser in different positions on the device. Photocurrent The PPC is different than the actual photocurrent of the devices. The on/off photocurrent observed in reference [8] is really fast (50 ms) and has an amplitude of 1.2 nA for Vs= 1 V Vg= 50 V which is much lower than the values observed with the PPC, 2 orders of magnitude higher. However, is still lower than that from graphene -tens of picosecondsas the carrier transport in graphene is ballistic and very fast [41]). This photocurrent is behind this PPC but the values are much lower and cannot be seen. To characterize the photo switching characteristic of single-layer MoS2, the laser light was modulated with a chopper at room temperature which is common practice when studying materials for photodetectors [42]. The Fast Fourier Transform (FFT) of the response to the modulated light (fig.44), shows a component at the frequency of the chopper. This peak is constant in amplitude which means the intensity of this photo response is not changing in time and that the response is fast because the light is modulated at relatively high frequencies. From this, it seems that we can get rid of the PPC by rapidly switching the light source on and off. b) a) 2 3 1 1 2 3 V g = 0 V Vs= 1 V I s (nA) Vs (V) 50 Figure 44: FFT of the electrical response of the device excited by a 645 nm laser beam modulated at the frequency of 629 Hz. When this photocurrent is extracted for different modulation frequencies, different Vg and different Vs there is not a frequence dependence (fig.45) the difference is lower than the error. There is however an important depencdence on Vg or Vs (fig.46,48) Figure 45: Photocurrent vs chopper frequecy for two diferent devices. V g = 0 V Vs= 1 V dB V rms 2 V g = 0 V Vs= 1 V I ph (pA) 51 The photocurrent generation efficiency can be further enhanced by increasing the gate voltage (fig.46). Figure 46: Photo current for different Vg of two diferent devices. The gate voltage-dependent photoresponsivity in single-layer MoS2 is attributed to the n-type doping of MoS2. If the applied gate voltage is positive, The Fermi level approaches the Conduction Band (CB) of MoS2 (fig.47), forming a smaller barrier between the CB of MoS2 and the Fermi level of the Au electrode as compared with the state under zero gate voltage. This leads to the photo-generated charges which drift efficiently to the external circuit to produce a high photocurrent.[8] Figure 47: a) The Fermi level of MoS2 is set at 4.7 eV and the Fermi level of Au is 5.1 eV. In the images we can see how they change for different Vg [8]. b) Vg = 0 V. c) Vg < 0 V. d) Vg > 0 V. The drain-voltage-dependent photocurrent generation (fig.48) indicates that some photo generated charge carriers cannot be converted to the photocurrent when the applied drain a) b) c) d) V s = 1 V Freq= 600 Hz I ph (pA) Vg (V) 52 voltage is low. This is reasonable, since a larger drain voltage can better drive photo generated charges to electrode, or suppress photo generated charges from the recombination. Figure 48: Photo current for different Vs of two diferent devices. Photoresponsivity, is a critical parameter to evaluate the performance of these devices. Photoresponsivity = Iph Plight (19) Iph is the generated photocurrent and Plight is the total incident optical power on the MoS2. For a Vg=50V, Vs=1 with the red laser as the light source (0.013W/cm2) in this device of 3µ𝑚2 we can calculate a photoresponsivity of 30.8 mA/W Plight = (0.015 ± 0.00068 )𝑊 𝑐𝑚2 𝑥 3µ𝑚2= (0.45 ± 0.020)10−9𝑊 (20) Photoresponsivity = 1.2𝑛𝐴 0.45·10−9𝑊=267 𝑚𝐴 𝑊 (21) |𝛿Photoresponsivity|= Photoresponsivity ��𝛿Plight Plight �2+ �𝛿Iph Iph �2= 10 𝑚𝐴 𝑊 (22) Photoresponsivity = (267 ±10) 𝑚𝐴 𝑊 (23) V g = 10 V Freq= 600 Hz I ph (pA) V s (V) 53 This value is much higher than 7.5 mA/W the one obtained for single-layer MoS2 [8]. It is low compared to the reported phototransistors based on ZnO nanowires (1.29·104 A/W) or vertical Si nanowirearrays (∼ 105 A/W) [43, 44]. It is higher than the one reported in the 2D graphene based devices, that show a photoresponsivity of∼1 mA/W at a gate voltage of 60 V [41]. 54 Chapter 6 Conclusion For the first time, it has been observed persistent photoconductivity (PPC) on mechanically exfoliated single-layer MoS2. The PPC decay of single-layer MoS2 is explained with the stretched-exponential and Large Lattice Relaxation (LLR) model. Furthermore, aggressive surface treatment such HF improve the decay of PPC. The electrical properties of single-layer MoS2 have been studied in relation to the surface treatments and the device fabrication process flow. The device fabrication improves the device electrical properties. The best surface treatment for good hysteresis is PI/1000ºC treatment and the best for high field-effect mobility is KOH/plasma. The best treatment used in terms of reaching a good relation hysteresis/field-effect mobility is KOH/1000ºC treatment. The photocurrent generation is extracted with a pulse modulated light. The photocurrent increases with source-drain and gate voltage. A photoresponsivity of (267 ± 10) mA/W is obtained which is higher than others previously observed values for single-layer MoS2 (7.5 mA/W) [8] and graphene devices (1 mA/W). Next steps will be to make similar measurements done with the white light with a laser light for different surface treatments. The photoresponse should be measured on more devices to confirm if the behavior observed is persistent. Finally, would be interesting to find the way to obtain devices without PPC. This project shows how optoelectronic devices based on single-layer MoS2 semiconductors can be made and why they could be used in future for many applications. 55 Chapter 7 Apendix Fabrication protocol • Chip characteristics : o Substrate of silicon with 270 nm of SiO2. • Cleaning procedure: 1- Sonicate in acetone (10 min) 2- Sonicate in IPA (10 min) 3- Sonicate in DI-water (10 min) 4- Dry with nitrogen gun • Surface treatment 1: o KOH 1- 30 min in 30% vol of KOH 2- Rinse DI-water 3- Dry with nitrogen gun o PI 1- 45 min in 60ml sulfuric acid with 20 ml peroxide 2- Rinse DI-water 3- Dry with nitrogen gun o HF 1- 30 sec 2 ml 50% vol of HF with 70 ml of DI-water. 2- Rinse DI-water 3- Dry with nitrogen gun • Surface treatment 2: o Plasma treatment 1- 20 min at 270 RF power and 30 sccm atmosphere flow in the plasma cleaner. o 1000°C treatment 1- 5 min at 1000ºC in ambient atmosphere in a furnace. • Exfoliate MoS2: 1- Scotch tape-based method • Ebeam process 1- Spin 100nm of MMA EL6 film (4000 rpm, time= 60 sec) 2- Bake 5 min 180ºC 3- Spin 100nm of PMMA A2 film ( 1500 rpm, time= 60 sec) 4- Write contacts with Vistec EBPG5000. 56 5- Develop for 3 minutes in MIBK (25%) with IPA (75%) 6- Rinse with IPA 7- Dry with nitrogen gun • Evaporate: 1- 90nm of Au with LAB600H. 2- 8 hours in acetone for the lift-off. 3- Rinse with IPA 4- Dry with nitrogen gun • Annealing: 1- 2h at 200ºC in Ar atmosphere inside the furnace • Coverslip procedure: o Cleaning procedure : 1- Rinse with acetone. 2- Rinse with IPA 3- Rinse with DI-water 4- Dry with nitrogen gun o Photolitography procedure: 1- Spin coat 1 µm of photoresist AZ1512 2- Expose photoresist with SUSS MJB4 Mask Aligner 3- Develop 1 min with CD-64 4- Rinse DI-water 5- Dry with nitrogen gun o Evaporation and lift-off: 1- Evaporate 10nm of chromium and 90 nm of gold. 2- Lift-off with 10 min in acetone sonication. 3- Rinse with IPA 4- Rinse with water 5- Dry with nitrogen gun. Labview block diagram The labview program is divided in 6 different blocks: Saving, Stage, Camera, Cursors, Scan, Labview electrical setup. 63 18. Jiao, L., et al., Narrow graphene nanoribbons from carbon nanotubes. Nature, 2009. 458(7240): p. 877-880. 19. Sols, F., F. Guinea, and A.H.C. Neto, Coulomb Blockade in Graphene Nanoribbons. Physical Review Letters, 2007. 99(16): p. 166803. 20. Yoon, Y. and J. Guo, Effect of edge roughness in graphene nanoribbon transistors. Applied Physics Letters, 2007. 91(7): p. 073103. 21. Aberg, I., et al. Electron and hole mobility enhancements in sub-10 nm-thick strained silicon directly on insulator fabricated by a bond and etch-back technique. in VLSI Technology, 2004. Digest of Technical Papers. 2004 Symposium on. 2004. 22. Roxlo, C.B., et al., Bulk and surface optical absorption in molybdenum disulfide. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films, 1987. 5(4): p. 555-557. 23. Kubiak, C.P., L.F. Schneemeyer, and M.S. Wrighton, Visible light driven generation of chlorine and bromine. Photooxidation of chloride and bromide in aqueous solution at illuminated n-type semiconducting molybdenum diselenide and molybdenum disulfide electrodes. Journal of the American Chemical Society, 1980. 102(22): p. 6898-6900. 24. Neville, R.A. and B.L. Evans, The Band Edge Excitons in 2H-MoS2. physica status solidi (b), 1976. 73(2): p. 597-606. 25. Nagashio, K., et al., Electrical transport properties of graphene on SiO2 with specific surface structures. Journal of Applied Physics, 2011. 110(2): p. 024513. 26. Guo, B., et al., Doping effect on shift of threshold voltage of graphene-based fieldeffect transistors. Electronics Letters, 2011. 47(11): p. 663-664. 27. Chen, J.H., et al., Charged-impurity scattering in graphene. Nat Phys, 2008. 4(5): p. 377-381. 28. Ghatak, S., A.N. Pal, and A. Ghosh, Nature of Electronic States in Atomically Thin MoS2 Field-Effect Transistors. ACS Nano, 2011. 5(10): p. 7707-7712. 29. Mooney, P.M., N.S. Caswell, and S.L. Wright, The capture barrier of the DX center in Si-doped AlxGa1-xAs. Journal of Applied Physics, 1987. 62(12): p. 4786-4797. 30. Lin, J.Y., et al., Relaxation of persistent photoconductivity in Al0.3Ga0.7As. Physical Review B, 1990. 42(9): p. 5855-5858. 31. Jiang, H.X., G. Brown, and J.Y. Lin, Persistent photoconductivity in II-VI and III-V semiconductor alloys and a novel infrared detector. Journal of Applied Physics, 1991. 69(9): p. 6701-6703. 32. Semaltianos, N.G., et al., Persistent photoconductivity and photoionization of deep electron traps in Ga-doped Cd1-.xMnxTe. Physical Review B, 1993. 47(19): p. 12540- 12549. 64 33. Lee, Y.C., et al., Persistent photoconductivity in Ho-doped InGaAsP epitaxial layers. Solid State Communications, 2002. 123(10): p. 421-425. 34. Shimakawa, K., Persistent photocurrent in amorphous chalcogenides. Physical Review B, 1986. 34(12): p. 8703-8708. 35. Mizuta, M. and K. Mori, Characterization of the DX center in the indirect AlxGa1-xAs alloy. Physical Review B, 1988. 37(2): p. 1043-1046. 36. Dmochowski, J.E., et al., Electron trapping by metastable effective-mass states of DX donors in indirect-band-gap AlxGa1-xAs:Te. Physical Review B, 1989. 40(14): p. 9671- 9682. 37. Mooney, P.M., Deep donor levels (DX centers) in III-V semiconductors. Journal of Applied Physics, 1990. 67(3): p. R1-R26. 38. Lee, Y.C., et al., Observation of persistent photoconductivity in 2H-MoSe2 layered semiconductors. Journal of Applied Physics, 2006. 99(6): p. 063706. 39. Queisser, H.J. and D.E. Theodorou, Decay kinetics of persistent photoconductivity in semiconductors. Physical Review B, 1986. 33(6): p. 4027-4033. 40. Lang, D.V. and R.A. Logan, Large-Lattice-Relaxation Model for Persistent Photoconductivity in Compound Semiconductors. Physical Review Letters, 1977. 39(10): p. 635-639. 41. Xia, F., et al., Ultrafast graphene photodetector. Nat Nano, 2009. 4(12): p. 839-843. 42. Luo, L.-b., F.-x. Liang, and J.-s. Jie, Sn-catalyzed synthesis of SnO2 nanowires and their optoelectronic characteristics. Nanotechnology, 2011. 22(48): p. 485701. 43. Weng, W.Y., et al., A ZnO-Nanowire Phototransistor Prepared on Glass Substrates. ACS Applied Materials & Interfaces, 2011. 3(2): p. 162-166. 44. Zhang, A., et al., Ultrahigh Responsivity Visible and Infrared Detection Using Silicon Nanowire Phototransistors. Nano Letters, 2010. 10(6): p. 2117-2120.