Improved photoresponse with enhanced photoelectric contribution in fully suspended graphene photodetectors

Graphene's unique optoelectronic properties are promising to realize photodetectors with ultrafast photoresponse over a wide spectral range from far-infrared to ultraviolet radiation. The underlying mechanism of the photoresponse has been a particular focus of recent work and was found to be either photoelectric or photo-thermoelectric in nature and enhanced by hot carrier effects. Graphene supported by a substrate was found to be dominated by the photo-thermoelectric effect, which is known to be an order of magnitude slower than the photoelectric effect. Here we demonstrate fully-suspended chemical vapor deposition grown graphene microribbon arrays that are dominated by the faster photoelectric effect. Substrate removal was found to enhance the photoresponse by four-fold compared to substrate-supported microribbons. Furthermore, we show that the light-current input/output curves give valuable information about the underlying photophysical process responsible for the generated photocurrent. These findings are promising towards wafer-scale fabrication of graphene photodetectors approaching THz cut-off frequencies.


Transfer of LPCVD-grown graphene and optical transmission measurement
To transfer the graphene layer onto a substrate, the copper layer was etched away in a copper etchant bath using a commercially available etchant solution (Transene 49-1) as shown in Fig.   S1a. After copper was removed, the graphene was transferred into a distilled (DI) water bath ( Fig. S1b) to remove any trace of the etchant. After three DI water bath transfers, one part of the clean graphene layer was transferred onto a Si substrate with 300 nm SiO 2 for photodetector device fabrication and another part onto a clean glass slide (Fig. S1c) for optical transmission measurements. To confirm the quality of LPCVD-grown graphene over a large-area after the transfer process, optical transmission measurements were carried out (Fig, S1d). For the measurements, white light was focused with a microscope objective onto the graphene/glass sample and the transmitted light was collected with another microscope objective and sent into a spectrometer. The spectrum was normalized to the transmission of the bare glass substrate. The normalized data was smoothed by adjacent averaging using 500 points. An average of ~97% universal transmission through various locations on the graphene sample shows good, consistent quality of the monolayer CVD graphene. Figure S1: Optical transmission measurements through graphene on a glass substrate. a,b,c. CVD-grown graphene transfer onto a glass slide after processing through a copper etchant and two DI water baths. d. Transmission measurements on the glass slide.

Raman spectrum of graphene
Graphene was transferred onto a 300 nm SiO 2 substrate to evaluate its quality using Raman spectroscopy. Raman measurements were taken with a 523 nm laser focused to a spot diameter of ~2 m (FWHM) onto the graphene layer using a 100x objective. The signal was accumulated for 5 seconds and summed over five consecutive spectra. Figure S2 shows the characteristic graphene-Raman peaks G and G' at 1570 cm -1 and 2650 cm -1 , respectively. A G' peak width

Fabrication of the graphene devices
Graphene ribbons were fabricated to be 2 m wide and 5 m long between the metal contacts.
To etch the graphene ribbon, target graphene areas were protected using a nickel layer which was patterned and deposited using a JEOL6300 e-beam lithography system and a Kurt-Lesker e-beam evaporator at Brookhaven National Laboratory. After the graphene etching using the Trion Tech RIE system, the nickel layer was etched away using standard nickel etchant. We chose to use a nickel mask for graphene etching since it does not contaminate graphene ribbons unlike other resist materials such as PMMA as shown in recent study by Kumar et al 43 Figure S3a shows the device geometry with each graphene ribbon contacted using individual pairs of source and drain electrodes. Each electrode was then wire-bonded to individual pins and a bias was applied separately for each source-drain pair. This avoided cross-talk between neighboring devices, especially during the backgated photocurrent measurements. Figure S3 The well-defined under-etching ensures that the suspended graphene ribbon is well isolated from the p-doped silicon substrate by an approximately 300 nm air-gap. The CPD process was carried out using high purity ethanol as a transfer chemical and liquid CO 2 as a sample drying chemical.

Fabrication of the suspended devices
During the CPD process, the chamber was pressurized and heated above the critical point of CO 2 and then slowly de-pressurized such that the drying of the sample occurred without any phasechange of CO 2 avoiding any damage to the suspended graphene microribbons. The rate of depressurization of the chamber was carefully controlled to avoid condensation of any trapped impurities on the surface of the device.

Current and laser annealing of graphene ribbons
The as-fabricated graphene devices in this study typically show heavy p-doping of the graphene layer. Incident laser excitation with current flowing through the graphene ribbon has been shown to remove the p-dopants from the graphene layer, shifting the Fermi-level towards the intrinsic Dirac point 45 . Therefore, laser-current annealing was done on as-fabricated graphene devices before the photocurrent measurements to avoid unwanted shift in the photocurrent values during the measurements due to the shift in the Dirac point. Figure S3 shows a significant shift in the Dirac point after laser-current annealing. This indicates effective removal of dopants from the surface of the graphene ribbon and leading to enhanced mobility and reduced hysteresis 43 1) where, η is the quantum efficiency of the photodetector layer, P opt is the incident power, A Laser is the area under the laser excitation spot, D is the layer thickness and hν is the energy of the incident photon. For steady state, the photocarrier density is given as: where τ is an average lifetime of the photo-generated carriers. Combining the supplementary equations SE (1) and SE (2), the steady state photocarrier density can be evaluated as: The photocurrent generated in the photodetection layer can then be expressed using the following equation: where σ p is conductivity of the photodetector layer due to the photo-generated carriers, ξ is an applied electric field across the layer and W is width of the layer. Thus, the photocurrent measured at the source and drain contacts with a carrier separation efficiency β is: where µ n and µ h are the electron and the hole mobility respectively and q is the electron charge. The thermoelectric current originates from carrier diffusion when a thermal gradient is generated between the photoexcited electrons at the excitation spot and the graphene-metal interface. The thermoelectric contribution can be mathematically expressed as: SE (7) where S(T) is the thermoelectric power which can be obtained using the Mott formula 47 . S(T) is directly proportional to the hot electron temperature term T e such that . The term is proportional to the hot electron temperature difference between the excitation area and near the contacts. The portion of the incident energy received by the hot electrons is proportional to the incident power which results in . The term is the heat capacity of the electrons which is temperature dependent 48 such that . Therefore, the electron temperature is related to the incident power as: or . Thus from the supplementary equation SE (7), the relationship between the thermoelectric current and the incident power can be expressed as: .