Scalable Graphene-on-Organometal Halide Perovskite Heterostructure Fabricated by Dry Transfer Qin Liang,1, 2, # Bhupal Kattel,2, # Tika R. Kafle,2 Maogang Gong,2 Mohan Panth,2 Yanbing Hou,1, * Judy Wu,2 Wai-Lun Chan2, * 1.
Key Laboratory of Luminescence and Optical Information, Ministry of Education, Beijing JiaoTong University, Beijing100044, China 2. Department of Physics and Astronomy, University of Kansas, Lawrence, KS 66045, United States # These authors contribute equally to this work. *Corresponding author:
[email protected] (W. -L. C.),
[email protected] (Y. H.) Abstract: Graphene, a single layer transparent conductor, can be combined with other functional materials for building efficient optoelectronic devices. However, transferring large-area graphene onto another material often involves dipping the material into water and other solvents. This process is incompatible with water-sensitive materials such as organometal halide perovskites. Here, we used a dry method and succeeded, for the first time to our knowledge, in stacking centimeter-sized graphene directly onto methylammonium lead iodide thin films without exposing the perovskite film to any liquid. Photoemission spectroscopy and nanosecond time-resolved photoelectrical measurement show that the graphene/perovskite interface does not contain significant amount of contaminants and sustain efficient interfacial electron transfer. We further demonstrate the use of this method in fabricating graphene-onperovskite photodetectors. Besides a better photoresponsivity compared to detectors fabricated by the conventional perovskite-on-graphene structure, this dry transfer method provides a scalable pathway to incorporate graphene in multilayer devices based on water-sensitive materials. Keyword: Graphene, Organometal halide perovskite, Charge transfer, Interfaces, Photodetectors. 1
1. INTRODUCTION Organometal halide perovskites such as methylammonium lead iodide (CH3NH3PbI3) have attracted much attention recently because of their ultra-long carrier diffusion length, solution processability and tailorable band gap, which make them high performing materials for solar cell applications.[1-5] The efficiency of single-junction CH3NH3PbI3 solar cells has already surpassed 20%.[6] Tandem cells consisting of multiple junctions with different bandgaps would potentially achieve even higher efficiencies.[7,
8]
However, it remains
challenging to design and integrate multiple semiconducting and metallic layers together into an effective multi-junction structure. One of the difficulties is to design electrodes and interconnecting layers between sub-cells,[9] which need to have high electrical conductivity but minimal optical absorption loss. Ultrathin metal layers, transparent conducting oxides or conductive polymers have been used for this purpose.[7, 9, 10] A viable alternative is single-layer or few-layer graphene, which have advantages of low optical absorption, high carrier mobility and tunable Fermi level via molecular doping.[11] In fact, increasing efforts have been made to incorporate graphene and other 2D layered materials in perovskite solar cells[12-14] and optoelectronic devices.[15] Graphene transferred on top of the perovskite can also passivate the moisture-sensitive perovskite layer to improve the long-term stability of the device.[16] One of the obstacles in using graphene with materials such as water-soluble perovskites is that it is difficult to grow or transfer large area graphene directly onto the perovskite. Because graphene requires a very high fabrication temperature, typically around 1000 °C, it cannot be grown directly on solution-processed semiconductors that usually have low melting temperatures and can degrade at elevated temperatures. Graphene can be transferred from the 2
growth substrate to other material surfaces using polymer stamps such as poly(methyl methacrylate) (PMMA).[17] However, in these “wet” transfer methods, the graphene receiving substrate needs to be immersed in water, which precludes the transfer of graphene on watersensitive perovskites unless the perovskite is first covered by another polymer layer.[13, 14] As a result, a graphene-on-perovskite interface cannot be formed by the wet method. Non-aqueous solvents have been used to transfer graphene on perovskite,[16] but attaching free-standing graphene, floating on top of a liquid, to the graphene receiving substrate is technically challenging and is difficult to scale-up. Various dry transfer methods using thermal release tape[18, 19] or polydimethylsiloxane (PDMS)[20] have been developed. Nevertheless, a single layer graphene often cracks during these dry transfer processes unless high pressure or high temperature is applied to the graphene receiving substrate during the transfer. Because it is difficult to transfer graphene on top of the perovskite, perovskite-sensitized graphene hybrid devices that require a direct contact between graphene and perovksite, such as photodetectors, are usually built with perovskite coated on graphene,[21-24] but not with a graphene-onperovskite inverted structure. This limitation can impede the use of graphene in more complex multilayer perovskite devices. More recently, stamps coated with a self-release layer[25] have been used to transfer graphene. In particular, it have been shown that polyethylene terephthalate (PET) coated with silicone adhesive can be used as a stamp for transferring centimeter-sized graphene onto smooth (either flat or curved) glass or SiO2 substrates.[26,
27]
In this method, the
PET/silicone/graphene stack is stamped onto the target substrate outside of the solution. Because of the weak adhesion between graphene and silicone, the PET/silicone stamp can be 3
peeled off from the substrate easily after the transfer. Previous studies[26, 27] find that this dry method leaves less polymer residue on the surface as compared to the widely-used PMMA method. Although these works
[25-27]
showed that graphene can be transferred onto relative
smooth and inert substrates (such as glass, SiO2/Si, ultrathin polymer films and self-assembled monolayer), using the dry transfer method with practical functional materials is not a trivial task. In the case of perovskite films, not only that the surface is relatively rough, the perovskite is also sensitive to moisture. Hence, transferring large-area and continuous graphene onto the perovskite surface is challenging, let alone, the graphene-perovskite interface needs to be clean enough for sustaining effective electron transfer in perovskite-sensitized optoelectronic devices. In this work, we succeeded in using the PET/silicone stamp to transfer graphene on top of water-sensitive and rough organometal halide perovskite thin films. This success is enabled by the high vacuum cleaning steps added to the previously reported process.[26] These vacuum cleaning steps can improve the adhesion of the graphene to the perovskite surface, which is critical to the successful transfer of graphene to organometal halide perovskite thin films with rough surfaces. Moreover, these additional steps can prevent trapping of solvent and water molecules at the graphene/perovskite interface. These trapped small molecules are elusive to many characterization techniques, but are critical and detrimental to optoelectronic process through impeding the interfacial charge transfer and even introducing artifacts to the devices.[28, 29]
The absence of the trapped impurities at the graphene/perovskite interface is reflected by
the surface and interface sensitive characterization, and the device performance data. Surfacesensitive ultraviolet and X-ray photoelectron spectroscopy (UPS and XPS) were employed to demonstrate the cleanliness of the graphene surface and the graphene/perovskite interface. 4
Nanosecond time-resolved photoelectrical measurements were carried out to verify that effective charge injection can occur at the graphene/perovskite interface. Perovskite-sensitized graphene photodetectors were made by transferring single layer graphene onto a perovskite thin film. Compared to our previous work[23] on similar photodetectors that have a reversed structure fabricated by spin-coating the perovskite layer on graphene, the graphene-onperovskite photodetector has a better photoresponsivity. The inverted graphene-on-perovskite heterostructure is remarkable not only for allowing the underneath perovskite or other solution-sensitive functional materials to be optimized at the required synthesis conditions (e.g. the morphology of the perovskite film can be optimized by the choice of the substrate), but also for enabling design feasibility in building hybrid multilayer optoelectronic devices. In particular, the transfer method allows graphene to be stacked either on top or underneath a functional material. In the case of our graphene-onperovskite photodetector, the inverted heterostructure provides additional flexibility in functionalizing the bottom and the top graphene surface with two different semiconductors. Such tandem-structured GFET sensors can potentially be used for photoresponsivity improvement, broadband light detection and even multifunctional sensors. 2. RESULTS AND DISCUSSION The procedure of the dry transfer method is illustrated schematically in Figure 1(a). This procedure is based on an earlier work,[26] but a few vacuum pumping and anti-solvent rinsing steps were added to make this method applicable to perovskite thin films. First, a cmsized graphene/Cu is attached to the PET/silicone stamp. In order to make the PET/silicone stamp fully adhere to the graphene/Cu, a mild pressure of ~ 5 kPa is applied to the sample 5
sandwiched between two glass slides for 30 minutes (Step i). After removing the graphene on the backside of the Cu by Ar ion etching in a vacuum chamber, Cu-etchant solution (FeCl3) was used to remove the Cu foil (Step ii).[30] Similar to the typical PMMA transfer, the graphene sample was rinsed/washed repeatedly with hydrochloric acid, ammonium hydroxide solution and deionized (DI) water (Step iii).[31] It was then rinsed by IPA solution. After these cleaning steps, the PET/silicone/graphene stack was dried with N2 gas. The PET/silicone/graphene stack was then placed in a high vacuum chamber (10-7 Torr) overnight to remove any residue solvent on the graphene surface before the transfer (Step iv). During the transfer process, the PET/silicone/graphene stack was pressed onto a freshly-prepared perovskite film (Step v). Then, the PET/silicone/graphene/perovskite/glass assembly was pumped inside a vacuum chamber (10-7 Torr) for ~ 1 hour to remove any trapped air at the graphene/perovskite interface. Then, the PET/silicone stamp was peeled off from the graphene/perovskite (Step vi). We found that the vacuum pumping steps not only kept the graphene/perovskite interface clean, it also enabled the successful attachment of the graphene onto the perovskite. Although it was shown that silicone leaves less residue on graphene as compared to the PMMA,[26, 27] we rinsed the graphene surface with a few drops of chlorobenzene, which is an anti-solvent for the perovskite (Step vii). The anti-solvent was spun off from the graphene surface at 5000 rpm for 10 seconds. Our photoemission study shows that this step is efficient to remove silicone residue on the graphene surface. An image of a ~ 1 cm × 1 cm graphene transferred onto a 400 nm thick perovskite film on SiO2 (300 nm)/Si is shown in Figure 1(b). Figure 1(c) shows an image of two stripes of graphene (~ 0.5 cm × 1 cm each) transferred on a perovskite (400 nm)/glass substrate. A 6
microscope image near the edge of the graphene is shown in Figure 1(d). A continuous graphene layer is observed. The method can be used to transfer graphene onto other substrates as well. An image of a cm-sized graphene transferred onto a C60(10 nm)/glass substrate is shown in Figure 1(e).
Figure 1: (a) Schematic illustration of the graphene dry transfer process. (b) Image of a ~ 1 cm × 1 cm graphene transferred on perovskite (400 nm)/SiO2 (300 nm)/Si substrate. (c) Two stripes of graphene (~ 0.5 cm × 1 cm) transferred on perovskite (400 nm)/glass substrate. (d) An optical microscope image of a graphene/perovskite (400 nm)/glass sample around the edge of the graphene. (e) Graphene transferred on C60 (10 nm)/glass substrate. (f) Optical absorption spectra of the perovskite film with and without transferred graphene on top of the surface. Because perovskite films can be sensitive to the environment, absorption spectroscopy was performed on the perovskite/graphene sample to check whether the perovskite film is degraded (e.g. transforming to PbI2) during the transfer process. Figure 1(f) compares the optical absorption spectra of the bare perovskite and the graphene on perovskite. The absorbance is defined as log (Pi/Pt), where Pi and Pt are the incidence power and the transmitted power respectively. The two samples show nearly identical absorption spectrum, which indicates that the perovskite film has not been degraded. Note that the optical absorption of a 7
single layer of graphene is an order of magnitude weaker (~ 2.3 %) than that of the perovskite films and is uniform across the whole spectral range. Therefore, the absorption from graphene is not visible in the spectrum. To ensure that there is no significant damage in the graphene layer, Raman spectra were taken on different spots of the graphene/perovskite sample. A typical spectrum is included in the supporting information (Figure S2). Three characteristic peaks of graphene can be observed: the tangential G band at ~1590 cm-1 derived from the in-plane vibration of the sp2 carbon atoms, the small disorder-induced D band at ~1355 cm-1, and its second-order harmonic 2D band at ~2715 cm-1. The intensity ratio of 2D to G band (I2D/IG) of ~1.4 and the negligible D band intensity (the intensity ratio of D to G band is ~ 0.1) indicate that the graphene is a monolayer and is of high quality.
Figure 2: The UPS spectra of the graphene on perovskite sample compared with the graphene on SiO2/Si substrate and the bare perovskite film. To avoid sample charging, the bare perovskite sample was spin-coated on an ITO-coated glass substrate. (a) The region near the SECO; and (b) The region near the valence band. To characterize the electronic structure of the graphene/perovskite interface, we performed photoemission studies. Figure 2 compares the UPS spectra of graphene/perovskite (purple), graphene/SiO2/Si (red) and bare perovskite (blue) samples. The secondary electron cut-off (SECO) region of the UPS spectra is shown in the panel (a). The work function of the samples can be determined from the SECO. The graphene/perovskite sample has a work 8
function of 4.2 eV. Within the experimental error, a similar work function (4.2 eV) was obtained from the graphene/SiO2/Si sample. This value agrees reasonably well with the reported work functions of transferred graphene (~4.0 - 4.2 eV).[32-34] By contrast, the work function of the bare CH3NH3PbI3 perovskite film is smaller (4.0 eV), which is comparable to those reported previously.[35] The graphene/perovskite sample has a sharp SECO and its work function is similar to the work function of the graphene. This indicates that majority of the sample areas are covered with the graphene. Otherwise, the intensity would begin to rise at a lower energy because regions with exposed perovskite have a lower work function. Figure 2(b) shows the UPS spectra of the valence bands. The energy is referenced with respect to the Fermi level (Ef) of the specific sample. Photoelectrons were collected along the surface normal direction. Hence, the spectra represent the band structure at the Γ point. For the graphene/perovskite sample, features originated from both graphene and perovskite can be observed. This is expected because the graphene is atomically-thin (< 4 Å). Specifically, the valence band features of the perovskite (labelled as P1, P2) can be seen in the perovskite/graphene spectrum. The P1 and P2 features correspond to valence bands contributed mainly by the I-5p and Pb-6s orbits, respectively.[36] Although graphene is a conductor, its band structure has a gap at the Γ point. The σ-band edge at the Γ point is located at ~ 4 eV below the Fermi level,[37, 38] which is labelled as G1 in Figure 2b. Again, the G1 feature can be observed in the graphene/perovskite spectrum. The valence band maximum (VBM) is at ~1.5 eV below the Ef for the bare perovskite and the graphene/perovskite samples. This is consistent with the reported VBM of the perovskite.[35] Because UPS is a surface sensitive technique, the appearance of both graphene and perovskite features in the graphene/perovskite spectrum 9
indicates that the graphene surface and the graphene/perovskite interface do not contain significant amount of polymer residues and trapped solvent molecules, respectively. The presence of organic residues can also shift the work function from that of the pristine materials, which is not observed in our samples.
Figure 3: (a) The XPS spectra for the graphene on perovskite, graphene on SiO2 (300 nm)/Si and the bare perovskite samples. For the graphene on perovskite sample, spectra for samples prepared with (solid line) and without (dotted line) the final cleaning step (step vii in Figure 1a) are shown. A high-resolution view of the (b) C-1s and the (c) Si-2s peaks. XPS was performed to determine whether polymer residues are present on the sample surface. Figure 3(a) compares the XPS scans of graphene (red), bare perovskite (blue) and graphene/perovskite (purple) samples. All spectra are collected under the similar conditions. Other than a vertical offset, the raw intensity is plotted in Figure 3(a). For the graphene/perovskite sample, we include the spectra for samples with (solid line) and without (dashed line) the final anti-solvent cleaning step (Step vii in Figure 1a). In the bare perovskite spectrum (blue), peaks for the Pb, I, and C can be identified, which is consistent with the XPS spectra reported previously.[39] In the graphene on SiO2/Si spectrum (red), in addition to the C1s peak originated from the graphene, the O-1s and the Si-2s, 2p peaks originated from the 10
underneath SiO2 can be observed. As expected, the graphene/perovskite spectrum (purple) contains features originated from both graphene and perovskite. However, without the final cleaning step (Step vii in Figure 1a), the intensity of I and Pb peaks are clearly suppressed. This can be related to the presence of polymer residues on the surface. Moreover, for the uncleaned graphene/perovskite sample, O-1s and Si-2p peaks are observed. This can be attributed to the silicone adhesive in the PET/silicone stamp which contains both Si and O atoms. After the final cleaning step, the intensities of the Si and O peaks are much reduced, which indicate that the final cleaning step is effective in removing the polymer residue on the graphene surface. Figure 3(b) shows the high resolution view of the C-1s peak. The graphene spectrum is dominated by a sharp peak at ~284.8 eV, which can be assigned to the sp2 hybridized carbon atoms in graphene. The shoulder at a higher binding energy can be assigned to the sp3 hybridized carbon atoms at defect sites.[40] It should be noted that similar spectra have been reported by others.[41] The C-1s peak for the bare perovskite film consists of a pair of peaks with the main peak located at higher binding energies (~286 eV) than that of the graphene. This location is consistent with previous XPS works on perovskite.[16, 42] The higher binding energy can be attributed to the C-N bond in the CH3NH3- ion.[43] The C-1s peak for the graphene/perovskite sample (after the final rinsing step) has a larger spectral width. It can be deconvoluted into three peaks which can be assigned to the graphene and the perovskite. To gauge the effectiveness of the final cleaning step, the high resolution Si-2p peak was measured and the result is shown in Figure 3c. For the graphene/SiO2/Si sample, the Si peak is contributed primarily by the SiO2 substrate. The Si-2p peak for silicone is expected to have a lower binding energy.[43] Therefore, the Si-2p peak in the graphene/perovskite spectra, which 11
has a lower binding energy, is assigned to the silicone adhesive. The intensity originated from the silicone is much reduced after the final cleaning step. The XPS result shows that even rinsing by merely a few drops of chlorobenzene is effective to remove the silicone residue on the graphene surface. We note that even a minute amount of solvent or water molecules trapped at the graphene/perovskite interface during the transfer process would block the interfacial charge transfer (CT). In our transfer process, we have pumped the graphene surface in a high vacuum environment to remove solvent or water molecules adsorbed on graphene during the water/solvent cleaning processes. To verify the resultant graphene/perovskite interface is clean enough to sustain efficient CT that is important for the device performance, we compare the CT properties of the graphene-on-perovskite and the perovskite-on-graphene interfaces. Compared to the graphene-on-perovskite, the perovskite-on-graphene sample can be fabricated using well-established methods. The graphene surface is cleaned by multiple solvents and subsequent high temperature annealing/outgassing (~400 °C overnight inside an ultrahigh vacuum chamber) before the perovskite deposition to ensure a clean interface. Our previous time-resolved photoemission studies have shown that the high temperature outgassing, after the solvent cleaning, can produce a graphene surface that supports CT rates comparable to that of the direct-grown CVD graphene (without any graphene transfer).[44] Hence, the perovskiteon-graphene sample is used as a benchmark to gauge the CT properties of the graphene-onperovskite sample.
12
Figure 4: (a) Schematics showing the electronic processes occur in the graphene/perovskite sample after the excitation by a fs laser pulse. (b-c) The temporal change of the graphene channel resistance after the graphene/perovskite device is exposed to a single fs laser pulse at t = 0. The signal rise within ~100 ns after photoexcitation observed in (b) is originated from the net hole injection from perovskite to graphene. The signal decay on the µs-s timescales shown in (c) is originated from the electron detrapping from the perovskite layer. To probe the CT dynamics from perovskite to graphene, the samples are excited by femtosecond (fs) laser pulses and the percentage change in the graphene resistance (∆𝑅𝑅/𝑅𝑅0 ) is measured as a function of time using a fast oscilloscope. The details of this method were
reported in our previous papers.[23, 45] From previous works on perovskite-sensitized graphene field-effect transistors (GFET),[21-23, 46] it is known that after photoexcitation, electron and hole transfer from perovskite to graphene are asymmetric. A portion of the excited electrons is trapped within the perovskite layer, while holes are transferred effectively to graphene. It is proposed that the band bending at the semiconductor/graphene interface favors hole injection, but suppresses electron injection.[21,
47]
Mobile iodide vacancies[48] typically found in
CH3NH3PbI3 would also trap electrons. The net negative charges built up in the perovskite layer creates a gating effect, which dopes the graphene. Figure 4(a) shows the schematics of 13
these optoelectronic processes. As shown in the middle panel, the carrier diffusion in the perovskite and the preferential hole transfer at the perovskite/graphene interface produces hole doping in graphene. This in turn changes the resistance of the graphene channel. The temporal change in the resistance (∆𝑅𝑅/𝑅𝑅0 ) is measured by a fast oscilloscope in our setup. The eventual
detrapping of electrons from the perovskite layer results in the decay of this electrical signal (Figure 4(a), panel on the right). Figure 4(b) compares the signal rise times (after the excitation by a fs laser pulse) observed in the graphene-on-perovskite device (red, this work) and the perovskite-on-graphene counterpart (black, results from Ref. [23]). The rise time represents how fast photo-excited carriers can be transported from the perovskite to the graphene. The graphene-on-perovskite sample (red) has a comparable but slightly slower signal rise time as compared to the perovskite-on-graphene sample (black). In both samples, the majority of the signal rise occurs within ~100 ns after the photoexcitation. Because solution-processed perovskite has a relative low carrier mobility (~ 1 cm2/V s),[49] the rise time is primarily limited by the carrier diffusion across the perovskite film. The slightly slower signal rise time observed in the graphene-onperovskite sample may indicate the existence of residual air gaps at the graphene/perovskite interfaces. This is because the perovskite surface is relatively rough (see the SEM images in the supporting information) and the graphene may not have a conformal attachment on the perovskite surface. Moreover, although the thickness of both perovskite films is similar (~ 400 nm), the films unavoidably have different morphologies and grain sizes because they are coated on different substrates (see the SEM images in the supporting information). For example, the perovskite film deposited on graphene has a larger grain size, which can result in better carrier 14
mobility and faster signal rise time. Nevertheless, the signal amplitude, hence the amount of net transferred holes, is similar for both samples. This indicates that the interfacial CT that directly impacts the photoresponse of the optoelectronic devices is comparable in the grapheneon-perovskite interface as compared to the perovskite-on-graphene interface. Figure 4(c) shows the decay of the electrical signal on the µs – s timescale. The x-axis is shown in a log scale because the signal decay spans multiple-timescales. As mentioned earlier, the signal decay is caused by the slow de-trapping of electrons from the perovskite layer. Once these electrons are de-trapped, they can transfer to graphene, which reduces the net electron population in the perovskite layer and the overall signal. The graphene-on-perovskite sample shows slower signal decay as compared to the perovskite-on-graphene sample. Currently, the reason behind the different decay times is unclear. Factors such as the difference in the interfacial band bending and the perovskite film morphology would prolong the electron trapping time within the perovskite layer. Nevertheless, an increase in the electron trapping time can indeed improve the photoresponsivity of the detector because the photoconductive gain increases with the carrier lifetime in the light-sensitizer.[50] As shown below, the grapheneon-perovskite detector indeed has a better photoresponsivity. Figure 5 compares the optoelectronic performance of the graphene-on-perovskite photodetectors with their perovskite-on-graphene counterparts reported in our previous work,[23, 24] which is fabricated by spin-coating perovskite on graphene. The previous results from perovskite-on-graphene photodetectors are used here as a benchmark to gauge the performance of the graphene-on-perovskite device. In these measurements, a continuous light source with a controllable intensity was used to excite the device. A source-drain voltage of 1 15
V was applied across the graphene channel. The change in the current (∆I) across the channel before (Idark) and after (Ilight) the light exposure was measured. Typical dynamical responses of the two types of the devices when the light was switched on and off are shown in Figure 5(a) with the vertical axis representing the percentage change in the source-drain current (∆𝐼𝐼/𝐼𝐼𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 ).
The rise/recovery times (when 70% of the signal is reached/recovered) for the graphene-on-
perovskite and the perovskite-on-graphene devices are 8.6 s/44.2 s and 2.5 s/65.0 s, respectively. The two devices show comparable response times under the exposure of a continuous light source. Note that this result seems to contradict the ultrafast measurement in Figure 4b (right panel), which shows significantly different decay times for the two samples. This apparent discrepancy can be explained by different trap states populated under different photoexcitation conditions. Impulsive (fs) excitation populates both shallow and deep traps depending on their availability. On the other hand, continuous light excitation allows the preferential accumulation of trapped electrons that have longer lifetimes (i.e. deeper traps) because the steady-state population increases with the electron trapping time. Another important figure of merit for photodetectors is the photoresponsivity, which is defined as 𝑅𝑅 = ∆𝐼𝐼/𝑃𝑃, where P is the power of the incident light. The photoresponsivity R for
the two devices as a function P are shown in Figure 5b. For both devices, R increases as P is decreased, which is a typical behavior observed in semiconductor-sensitized GFET photodetectors.[22, 50] This can be explained by the reduction of electron-hole recombination or other carrier-carrier annihilation processes as the light power, and hence the carrier concentration, is reduced. Over the whole light intensity range, the graphene-on-perovskite device shows better R values. We further note that the channel length of the graphene-on 16
perovskite detector (1 mm) is larger than that of the perovskite-on-graphene detector (0.3 mm), while other factors such as the channel width (2 mm), source-drain voltage (1 V), and perovskite thickness (~400 nm) are kept the same. Reducing the channel length of the graphene-on-perovskite detector, which already has a higher R, to match that of the perovskiteon-graphene detector can further boost its photocurrent and, hence, increase R proportionally.
Figure 5: (a) The dynamic responses of the graphene-on-perovskite (top) and the perovskiteon-graphene (bottom) devices. The data is collected with a source-drain bias of 1.0 V and a continuous light source with wavelength of 500 nm (b) The photoresponsivity as a function of the incident light power for the two photodetectors. 3. CONCLUSIONS In conclusion, we have developed a modified graphene dry transfer method by implementing high-vacuum cleaning steps and succeeded in transfer of continuous cm-sized graphene onto water-sensitive CH3NH3PbI3 thin films with a rough surface using this method. By using photoelectron spectroscopy, we show that the graphene surface and the graphene/perovskite interface do not contain significant amount of polymer residues and trapped solvent molecules. The transferred graphene is continuous, which allows us to use this method to fabricate mm-sized perovskite-sensitized GFET photodetectors. Characterization using time-resolved photoelectrical measurement confirms that charge injection efficiency across the graphene-on-perovskite interface is comparable to that across the optimized 17
perovskite-on-graphene interface. On the device level, we show that the graphene-onperovskite photodetector has a better photoresponsivity as compared to that of the state-of-theart perovskite-on-graphene photodetector[24] fabricated by coating perovskite onto graphene. This dry transfer method has several advantages. First, it provides a viable pathway to stack large-sheet graphene onto a variety of functional materials that may be damaged through exposure of solutions, mechanical deformation and excessive heat. The feasibility in arranging the sequence of the layers is important for constructing and designing multilayer devices. Second, it enables the dry graphene transfer on functional materials with rough surfaces and the formation of a clean interface between them to facilitate charge injection across the interface. Finally, the process is scalable since it can be readily implemented to roll-to-roll device production. Therefore, it will be useful for fabricating multilayer optoelectronic devices and photovoltaics. EXPERIMENTAL SECTION Sample Preparation. The glass substrates were sonicated in acetone and isopropyl alcohol (IPA) bath, followed with ozone-treatment for 15 min before perovskite deposition. The CH3NH3PbI3 perovskite film was prepared by a one-step spin-coating method. The lead iodide (Alfa Aesar, 99.9985%) and methylammonium iodide (Luminescence Technology, 99.5%) in a stoichiometric ratio were dissolved in dimethylformamide (DMF) (Sigma-Aldrich, 99.8%) with a concentration of 1 M. The solution was stirred at 70 oC overnight prior to the use. The perovskite film was deposited on the glass substrate by spin-coating at 3000 rpm for 30 s in a nitrogen glovebox. During the spin-coating, chlorobenzene (Macron), which acted as an antisolvent, was casted onto the film. Then, the perovskite film was annealed sequentially at 60 oC (5 minutes), 80 oC (5 minutes) and 100 oC (10 minutes) to remove the residual solvent. For the graphene transfer, chemical vapor deposition (CVD)-grown, monolayer graphene on Cu was purchased commercially from Graphene Supermarket. We utilized commercial smart-phone 18
screen protectors (Supershieldz) as the PET/silicone stamp. These screen protectors consist of a PET layer coated with the silicone adhesive on one side. They can be peeled off from surfaces easily without leaving visually noticeable residues. Film Characterization. The samples were characterized by various techniques. The optical absorption spectra were collected using a home-built setup consisting of a tungsten light source and a spectrometer. UPS was carried out in a home-built ultra-high vacuum (UHV) chamber, with a base pressure of < 1 × 10-10 Torr, equipped with a hemispherical energy analyzer (SPECS, Phoibos 100). The He-I emission line (21.2 eV) was used in this experiment. The sample was biased at -2 V during the UPS experiment. The XPS spectra were collected in a commercial XPS system (PHI VersaProbe II) equipped with a Kα (1486.6 eV) excitation source (base pressure ~ 1 × 10-10 Torr). Nano-second photoelectrical conductivity measurement. For the ultrafast conductivity measurement, a pair of Cu electrodes were deposited on the graphene using a shadow mask. The typical size of the graphene channel is 1 mm (length) × 2 mm (width). The perovskite was excited by a femtosecond (fs) laser beam. To produce the required fs pulses, an amplified Yb:KGW laser system (Light Conversion, PHAROS) was used to pump a noncollinear optical parametric amplifier (NOPA) (Light Conversion, ORPHEUSN-2H). The fs laser pulses produced by the NOPA had a wavelength centered at 700 nm. The pulse energy and pulse duration were ~ 200 nJ and ~ 20 fs, respectively. The full-width half maxima beam size at the sample was ~ 1.1 mm. During the experiment, the sample was mounted inside a high vacuum (~ 10-6 Torr) cryostat. The temporal change of the graphene channel resistance after the photoexcitation was measured by a 200 MHz oscilloscope that was synchronized with the laser. During the measurement, a constant source-drain bias voltage of 1 V was applied across the graphene channel and a fixed resistor which were connected in series. The circuit used in the experiment can be found in the supporting information and in Ref. [23, 45]. Device measurement. The performance of the graphene/perovskite photodetectors was characterized using a continuous light source (wavelength = 500 nm). The graphene channel current was measured with an electrochemical workstation (CH Instruments, CHI660D) at a source-drain bias voltage of 1 V. The light intensity was measured by a Newport optical power 19
meter. Before the measurement, the devices were kept in dark overnight. During the experiment, the devices were not exposed to other lights except the designated light source of 500 nm in wavelength. Photoresponsivity and the response time of the photodetector were extracted. ACKNOWLEDGEMENTS The authors like to acknowledge Ti Wang for his help in our early attempt to transfer graphene on perovskite. We also like to thank Prof. Liu and his students from Nankai University for sharing experimental details regarding their earlier works on the graphene transfer. This work was supported by US National Science Foundation grants DMR-1351716. The support by the University of Kansas General Research Fund allocation #2151080 is also acknowledged. J. W. acknowledges the support by US National Science Foundation DMR1337737, and DMR-1508494 and Army Research Office grant W911NF-16-1-0029. Y. H. acknowledges the support from Natural Science Foundation of China (Grant No. 61475017 and 61775011). CONFLICT OF INTEREST The authors declare no conflict of interest. TOC FIGURE
REFERENCES
[1] M. A. Green, A. Ho-Baillie, H. J. Snaith, Nat. Photon. 2014, 8, 506. [2] G. C. Xing, N. Mathews, S. Y. Sun, S. S. Lim, Y. M. Lam, M. Gratzel, S. Mhaisalkar, T. C. Sum, Science 2013, 342, 344. [3] S. D. Stranks, G. E. Eperon, G. Grancini, C. Menelaou, M. J. P. Alcocer, T. Leijtens, L. M. Herz, A. Petrozza, H. J. Snaith, Science 2013, 342, 341. [4] W. Y. Nie, H. H. Tsai, R. Asadpour, J. C. Blancon, A. J. Neukirch, G. Gupta, J. J. Crochet, M. Chhowalla, S. Tretiak, M. A. Alam, H. L. Wang, A. D. Mohite, Science 2015, 347, 522. [5] G. E. Eperon, S. D. Stranks, C. Menelaou, M. B. Johnston, L. M. Herz, H. J. Snaith, Energy Environ Sci 2014, 7, 982. [6] M. Saliba, T. Matsui, K. Domanski, J. Y. Seo, A. Ummadisingu, S. M. Zakeeruddin, J. P. CorreaBaena, W. R. Tress, A. Abate, A. Hagfeldt, M. Gratzel, Science 2016, 354, 206. [7] J. W. Lee, Y. T. Hsieh, N. De Marco, S. H. Bae, Q. F. Han, Y. Yang, J. Phys. Chem. Lett. 2017, 8,
20
1999. [8] S. Q. Dong, Y. S. Liu, Z. R. Hong, E. P. Yao, P. Y. Sun, L. Meng, Y. Z. Lin, J. S. Huang, G. Li, Y. Yang, Nano Lett. 2017, 17, 5140. [9] L. J. Zuo, C. Y. Chang, C. C. Chueh, S. H. Zhang, H. Y. Li, A. K. Y. Jen, H. Z. Chen, Energy Environ Sci 2015, 8, 1712. [10] K. Zhang, K. Gao, R. X. Xia, Z. H. Wu, C. Sun, J. M. Cao, L. Qian, W. Q. Li, S. Y. Liu, F. Huang, X. B. Peng, L. M. Ding, H. L. Yip, Y. Cao, Adv. Mater. 2016, 28, 4817. [11] G. Hong, Q. H. Wu, J. G. Ren, C. D. Wang, W. J. Zhang, S. T. Lee, Nano Today 2013, 8, 388. [12] C. Petridis, G. Kakavelakis, E. Kymakis, Energy Environ Sci 2018, 11, 1030. [13] F. Lang, M. A. Gluba, S. Albrecht, J. Rappich, L. Korte, B. Rech, N. H. Nickel, J. Phys. Chem. Lett. 2015, 6, 2745. [14] P. You, Z. Liu, Q. Tai, S. Liu, F. Yan, Adv. Mater. 2015, 27, 3632. [15] Y. P. Zhang, J. Y. Liu, Z. Y. Wang, Y. Z. Xue, Q. D. Ou, L. Polavarapu, J. L. Zheng, X. Qi, Q. L. Bao, Chem. Commun. 2016, 52, 13637. [16] W. S. Tseng, M. H. Jao, C. C. Hsu, J. S. Huang, C. I. Wu, N. C. Yeh, Nanoscale 2017, 9, 19227. [17] X. S. Li, Y. W. Zhu, W. W. Cai, M. Borysiak, B. Y. Han, D. Chen, R. D. Piner, L. Colombo, R. S. Ruoff, Nano Lett. 2009, 9, 4359. [18] J. D. Caldwell, T. J. Anderson, J. C. Culbertson, G. G. Jernigan, K. D. Hobart, F. J. Kub, M. J. Tadjer, J. L. Tedesco, J. K. Hite, M. A. Mastro, R. L. Myers-Ward, C. R. Eddy, P. M. Campbell, D. K. Gaskill, ACS Nano 2010, 4, 1108. [19] J. Kang, S. Hwang, J. H. Kim, M. H. Kim, J. Ryu, S. J. Seo, B. H. Hong, M. K. Kim, J. B. Choi, ACS Nano 2012, 6, 5360. [20] M. J. Allen, V. C. Tung, L. Gomez, Z. Xu, L. M. Chen, K. S. Nelson, C. W. Zhou, R. B. Kaner, Y. Yang, Adv. Mater. 2009, 21, 2098. [21] M. Spina, M. Lehmann, B. Nafradi, L. Bernard, E. Bonvin, R. Gaal, A. Magrez, L. Forro, E. Horvath, Small 2015, 11, 4824. [22] Y. Lee, J. Kwon, E. Hwang, C. H. Ra, W. J. Yoo, J. H. Ahn, J. H. Park, J. H. Cho, Adv. Mater. 2015, 27, 41. [23] L. Qin, L. P. Wu, B. Kattel, C. H. Li, Y. Zhang, Y. B. Hou, J. Wu, W. L. Chan, Adv. Funct. Mater. 2017, 27, 1704173. [24] L. P. Wu, L. Qin, Y. Zhang, M. Alamri, M. G. Gong, W. Zhang, D. Zhang, W. L. Chan, J. Z. Wu, Acs Appl Mater Inter 2018, 10, 12824. [25] J. Song, F. Y. Kam, R. Q. Png, W. L. Seah, J. M. Zhuo, G. K. Lim, P. K. H. Ho, L. L. Chua, Nat. Nanotechnol. 2013, 8, 356. [26] X. D. Chen, Z. B. Liu, C. Y. Zheng, F. Xing, X. Q. Yan, Y. S. Chen, J. G. Tian, Carbon 2013, 56, 271. [27] S. J. Kim, T. Choi, B. Lee, S. Lee, K. Choi, J. B. Park, J. M. Yoo, Y. S. Choi, J. Ryu, P. Kim, J. Hone, B. H. Hong, Nano Lett. 2015, 15, 3236. [28] C. R. Ma, Y. P. Gong, R. T. Lu, E. Brown, B. H. Ma, J. Li, J. Wu, Nanoscale 2015, 7, 18489. [29] R. T. Lu, J. W. Liu, H. F. Luo, V. Chikan, J. Z. Wu, Sci. Rep. 2016, 6, 19161. [30] K. S. Kim, Y. Zhao, H. Jang, S. Y. Lee, J. M. Kim, K. S. Kim, J. H. Ahn, P. Kim, J. Y. Choi, B. H. Hong, Nature 2009, 457, 706. [31] X. L. Liang, B. A. Sperling, I. Calizo, G. J. Cheng, C. A. Hacker, Q. Zhang, Y. Obeng, K. Yan, H. L. Peng, Q. L. Li, X. X. Zhu, H. Yuan, A. R. H. Walker, Z. F. Liu, L. M. Peng, C. A. Richter, ACS Nano
21
2011, 5, 9144. [32] Q. Liu, Y. P. Gong, T. Wang, W. L. Chan, J. Wu, Carbon 2016, 96, 203. [33] K. C. Kwon, K. S. Choi, S. Y. Kim, Adv. Funct. Mater. 2012, 22, 4724. [34] Q. H. Wu, G. Hong, T. W. Ng, S. T. Lee, Appl. Phys. Lett. 2012, 100, 161603. [35] P. Schulz, E. Edri, S. Kirmayer, G. Hodes, D. Cahen, A. Kahn, Energy Environ Sci 2014, 7, 1377. [36] Y. Wang, T. Gould, J. F. Dobson, H. M. Zhang, H. G. Yang, X. D. Yao, H. J. Zhao, PCCP 2014, 16, 1424. [37]V. N. Strocov, A. Charrier, J. M. Themlin, M. Rohlfing, R. Claessen, N. Barrett, J. Avila, J. Sanchez, M. C. Asensio, Phys. Rev. B 2001, 64, 075105. [38] S. Kim, J. Ihm, H. J. Choi, Y. W. Son, Phys. Rev. Lett. 2008, 100, 176802. [39] R. Lindblad, N. K. Jena, B. Philippe, J. Oscarsson, D. Q. Bi, A. Lindblad, S. Mandal, B. Pal, D. D. Sarma, O. Karis, H. Siegbahn, E. M. J. Johansson, M. Odelius, H. Rensmo, J. Phys. Chem. C 2015, 119, 1818. [40] A. Ermolieff, A. Chabli, F. Pierre, G. Rolland, D. Rouchon, C. Vannuffel, C. Vergnaud, J. Baylet, M. N. Semeria, Surf. Interface Anal. 2001, 31, 185. [41] A. Pirkle, J. Chan, A. Venugopal, D. Hinojos, C. W. Magnuson, S. McDonnell, L. Colombo, E. M. Vogel, R. S. Ruoff, R. M. Wallace, Appl. Phys. Lett. 2011, 99, 122108. [42] Y. Z. Li, X. M. Xu, C. G. Wang, C. C. Wang, F. Y. Xie, J. L. Yang, Y. L. Gao, Aip Adv 2015, 5, 097111. [43] J. Chastain, R. C. King, J. Moulder, Handbook of X-ray photoelectron spectroscopy: a reference book of standard spectra for identification and interpretation of XPS data, Physical Electronics Division, Perkin-Elmer Corporation Eden Prairie, Minnesota, 1992. [44] T. Wang, Q. F. Liu, C. Caraiani, Y. P. Zhang, J. Wu, W. L. Chan, Phys. Rev. Appl. 2015, 4, 014016. [45] B. Kattel, L. Qin, T. R. Kafle, W. L. Chan, J. Phys. Chem. Lett. 2018, 9, 1633. [46] C. Xie, F. Yan, Acs Appl Mater Inter 2017, 9, 1569. [47] G. Konstantatos, M. Badioli, L. Gaudreau, J. Osmond, M. Bernechea, F. P. G. de Arquer, F. Gatti, F. H. L. Koppens, Nat. Nanotechnol. 2012, 7, 363. [48] A. Senocrate, I. Moudrakovski, G. Y. Kim, T. Y. Yang, G. Gregori, M. Gratzel, J. Maier, Angew Chem Int Edit 2017, 56, 7755. [49] Y. Chen, H. T. Yi, X. Wu, R. Haroldson, Y. N. Gartstein, Y. I. Rodionov, K. S. Tikhonov, A. Zakhidov, X. Y. Zhu, V. Podzorov, Nat. Commun. 2016, 7, 12253. [50] M. G. Gong, Q. F. Liu, B. Cook, B. Kattel, T. Wang, W. L. Chan, D. Ewing, M. Casper, A. Stramel, J. Z. Wu, ACS Nano 2017, 11, 4114.
22
