Engineering Graphene Phototransistors for High Dynamic Range Applications

Phototransistors are light-sensitive devices featuring a high dynamic range, low-light detection, and mechanisms to adapt to different ambient light conditions. These features are of interest for bioinspired applications such as artificial and restored vision. In this work, we report on a graphene-based phototransistor exploiting the photogating effect that features picowatt- to microwatt-level photodetection, a dynamic range covering six orders of magnitude from 7 to 107 lux, and a responsivity of up to 4.7 × 103 A/W. The proposed device offers the highest dynamic range and lowest optical power detected compared to the state of the art in interfacial photogating and further operates air stably. These results have been achieved by a combination of multiple developments. For example, by optimizing the geometry of our devices with respect to the graphene channel aspect ratio and by introducing a semitransparent top-gate electrode, we report a factor 20–30 improvement in responsivity over unoptimized reference devices. Furthermore, we use a built-in dynamic range compression based on a partial logarithmic optical power dependence in combination with control of responsivity. These features enable adaptation to changing lighting conditions and support high dynamic range operation, similar to what is known in human visual perception. The enhanced performance of our devices therefore holds potential for bioinspired applications, such as retinal implants.


Figure S1. Raman characterization of graphene
(a) Raman spectrum of graphene as used for the proposed device after the transfer process (layer stack consisting of silicon, 20 nm Al 2 O 3 and graphene) recorded using a 471 nm laser.(b) A total of 100 Raman spectra as shown in (a) have been measured on an area of 100×100 µm 2 .Taking into consideration the position of graphene's G and 2D peak for each measured spectrum (obtained by Lorentzian fitting), one can assemble a two-dimensional plot, where each point corresponds to a Raman measurement on a different location of the graphene sample.This information can be translated into information on graphene's doping and strain levels. 1 The data shows that the transferred graphene has very low strain/stress, and a moderate amount of doping.Based on our observations, it is anticipated that the introduction of a passivation layer results in a reduction in both strain and doping levels.

Figure S3. Measurement setup for photodetector characterization
The figure shows the measurement setup as used for the characterization of photogating devices.
A 532 nm laser is attenuated with help of neutral density (ND) filters and modulated with an optical chopper at a frequency given by the reference signal of the lock-in amplifier (LIA).The laser beam is directed to the objective of the microscope (5× or 10× magnification).For alignment purposes a white LED is used to illuminate the sample.The objective focuses the incoming light onto the device under test, which is probed with help of DC-needles.The device is connected in series with a source measure unit (SMU) and a LIA.The SMU is used to apply the source-drain bias, the topgate voltage (if applicable) as well as the bottom-gate voltage.For a p-type silicon substrate negative bottom-gate voltages must be in order to achieve depletion mode biasing.

Figure S6. Time trace of the photocurrent
The figure shows normalized photocurrent time trace measurements of the proposed device with a modulated 532 nm laser.The measurements were performed with a source-drain voltage of 0.1 V.Under a bottom-gate voltage of -1 V (blue curve), the silicon substrate is biased into depletion mode resulting in a prolonged lifetime of minority charge carriers in silicon.This is clearly visible by considering the slower recovery of the photocurrent signal indicated with a black arrow as compared to the red curve with a bottom-gate voltage of 0 V (red curve).finger structure as used for the measurement in Fig. 6 of the main text ( f of 1200 nm and a finger width  f of 800 nm) at a laser wavelength of 532 nm in dependence of the polarization angle (0°: E-field perpendicular to the finger axis).The reflection is only weakly polarization dependent.

Figure S9. Spectral responsivity at different optical powers
Responsivity in dependence of the wavelength and incident optical power for an enhanced device with a semi-transparent top-gate and interdigitated fingers ( f = 800 nm,  f = 1200 nm).The laser spot size was comparable for all wavelengths.The measurement has been limited by the calibration measurements of the used neutral density filters at the lower end of optical powers.At higher powers the device exhibits a faster response.This can be attributed to the fact that under depletion biasing (as is the case under -1 V  BG ) the increasing power levels lead to an increased amount of photogenerated charge carriers, which reduces the lifetime of the collected minority charges at the Si/Al 2 O 3 interface.This reduction in lifetime leads to a faster photoresponse (b) Frequency sweep measurement of the same device under different bottom-gate voltage conditions.
Under -1 V, with which silicon is biased into the depletion mode, the response time is longer and thus the photoresponse slower as compared to the case for 0 V bottom-gate voltage.

Figure S13. Stability measurement of the presented photodetectors
Measurement of the source-drain current and the photocurrent for a time span of more than 10 hours for an enhanced device with a semi-transparent top-gate and interdigitated fingers ( f = 800 nm,  f = 1200 nm).These stability measurements suggest that the device operates stably under ambient conditions in air and no performance degradation has been observed.

Table S1. Table for comparing various photogating devices
The table shows various photodetectors utilizing the photogating effect.This table contains recent relevant works, but it is not complete. ph is the photoresponsivity,  min the minimum measured power,  r the slowest time component of the photoresponse (rise or fall time) and  the wavelength at which the values were measured.CVD stands for chemical vapor deposition, ME for mechanical exfoliation, SOI for silicon-on-insulator and QD for quantum dots.We differentiate between three photogating mechanisms, namely the interfacial, the deep-depletion graphene-oxidesemiconductor (D 2 GOS) and the direct photogating mechanisms.

Figure S5 .
Figure S5.Comparison of n-and p-type silicon

Figure S7 .
Figure S7.Resistance of devices with and without interdigitated finger structures

Figure S8 .
Figure S8.Reflection simulations of interdigitated finger structures

Figure S10 .
Figure S10.Comparison to demonstrations in literature

Figure S12 .
Figure S12.Power and bias dependence of frequency measurements