- Photonics Research
- Vol. 10, Issue 8, 1996 (2022)
Abstract
1. INTRODUCTION
Charge transport through a metal–insulator–semiconductor (MIS) heterogeneous junction holds the key to the performance of many classes of electronic devices, for example, MIS field-effect transistors and MIS capacitors [1]. In recent years, the MIS configuration has attracted great attention in the research community due to its novel applications in optical devices, in particular, nanoscale plasmonic devices such as hybrid plasmonic waveguides, hybrid plasmonic lasers [2–4], amplifiers [5], and nonlinear light conversion devices [6,7], in which the optical modes are tightly confined into a deep subwavelength scale of the interfaces of MIS, promoting strong light–matter interaction at the interfaces and bridging the size gap from a few nanometers of electronic components to a few micrometers of optical components.
For MIS based capacitors, the thickness of the insulator is usually larger than 5 nm to prevent any leakage current. For the application of ideal Schottky diodes, the insulator thickness is considered less than 1 nm. In photodetector applications, the insulator thickness is kept below 2 nm to allow an efficient carrier transport. In hybrid plasmonic structures, the thickness of the insulator is usually from 3 nm to 6 nm, allowing a strong electric field enhancement and a much reduced ohmic loss in the devices [8]. The interesting phenomenon of loss compensation in hybrid plasmonic waveguides at very low pump intensities was reported [5]. This demonstration opens up possibilities to significantly reduce the pump energy required to achieve gain-assisted signal modulation in plasmonic circuits. However, the mechanism is yet to be clarified. While the electric field enhancement and confinement have been discussed intensively in MIS based plasmonic structures, little is known about charge transport across the MIS heterojunction upon radiation of light, the redistribution of photo-generated carriers at the interfaces, their impacts on the optical gain/absorption of semiconductor material, and the potential application of MIS heterojunctions as photodetectors, particularly for oxide thickness ranging from 3 nm to 5 nm. Answers to above questions will also shed light on optically and electrically driven plasmonic signal modulation, which plays a pivotal role in integrated plasmonic circuits.
In this paper, we investigate the photocurrent transport through single CdSe microbelt interfaces with thickness varying from 3 nm to 5 nm. Peak photocurrent from a few nanoamperes to a few microamperes at peak incident laser intensity of with single photon energy exceeding the bandgap energy of CdSe is detected at zero bias for thickness , which is demonstrated through finite element simulation to be sensitively dependent on the positions of the conduction band edge of the ultrathin oxide layer. We show that the requirement for the aligned Fermi level across the heterojunction leads to a charge redistribution across the heterojunction and high optical gain in the CdSe region at low pump intensity. As the oxide thickness increases to 5 nm, no photocurrent is obtained at zero bias upon exposure to light. Instead, we observe an abrupt photocurrent increase at a moderate bias voltage of to , which is attributed to the Coulomb blockade effect. The current gain compared to dark current can exceed at a bias voltage of with a photoresponsivity of 0.03 A/W, which is of great importance for photodetection applications where low dark currents are highly desirable [9,10].
Sign up for Photonics Research TOC. Get the latest issue of Photonics Research delivered right to you!Sign up now
2. SAMPLE FABRICATION AND DARK CURRENT–VOLTAGE CHARACTERIZATION
To fabricate the heterojunction, CdSe MBs grown by chemical vapor deposition are deposited on the glass coverslip first. The Au electrode is then deposited on one side of the chosen CdSe MB, followed by ultrathin layer growth via atomic layer deposition. Finally, the Ag electrode is formed on the other side of the CdSe MB (see Appendix A for details). One example of the final devices is given as an inset of Fig. 1(a). Before the photocurrent measurements are conducted, current–voltage (I-V) characteristics of the devices are obtained first without light. Figure 1(a) shows the typical dark I-V characteristics obtained on three devices with thickness . As the current through individual CdSe MB is extremely small (tens of picoamperes), a lock-in technique is used to obtain the current at ambient conditions and room temperature. In the measurements, the Ag electrode is kept at zero bias, while the voltage applied to the Au electrode varies (see Appendix A for details). It is clear from Fig. 1(a) that the absolute current is significantly larger for the negative bias voltage, behaving as a diode where the forward direction corresponds to the negative voltage. The absolute value of the current varies from one device to another, sensitively dependent on the thickness of CdSe MB, the contact area with the Au and Ag electrodes, and the lateral distance between the Au and Ag electrodes (see Appendix B). Out of 10 samples we measured with thickness , 70% of them showed measurable dark current. For devices with thickness , no dark current is detected within the bias range of to 9 V, meaning that the current is below the detection limit of our setup ().
Figure 1.Current–voltage characteristics across
The I-V characteristics of the devices observed experimentally can be reproduced using the finite element simulation package COMSOL Multiphysics 5.5 semiconductor module. The simulated I-V curves on a 500 nm thick CdSe MB with an thickness of 3 nm are plotted in Fig. 1(b). In the simulations, in addition to the drift-diffusion equations [1], tunneling of electrons through the layer and Au-CdSe Schottky barrier [11] is also included to properly reproduce the experimental I-V characteristics (see Appendices A and B for details). Here the work functions of Ag and Au are set at 4.6 eV [12] and 5.1 to 5.13 eV [13] while the conduction band and valence band of the CdSe are set to 4.5 eV and 6.23 eV [14], respectively. The band positions and bandgap of the ultrathin layer, on the other hand, have been reported with very different values in the literature, with the conduction band edge varying from 1.27 eV [15] to 3.8–4.7 eV [16,17], sensitively dependent on the thickness of the oxide, the growth temperature, and fabrication methods. Our simulations show that a small change in the conduction band position of the function of Au can lead to a large change in the tunneling current at the positive/negative voltage, as illustrated in Fig. 1(b).
One important result from the simulation is that even at zero bias, the electron density within the CdSe is much larger than the hole density within the CdSe due to the presence of Ag (see Appendix D for details). Since the conduction band edge of CdSe is very close to the work function of Ag, if a common Fermi level is assumed across the heterojunction at equilibrium, extra electrons are transferred from the Ag electrode to the CdSe region. This means that even at zero bias, the CdSe MB is negatively charged and the highest electron density within CdSe is located at the region closest to the Ag electrode. The electron density across the entire heterojunction at zero bias with electron affinity at 3.74 eV and Au work function at 5.13 eV is given in the inset of Fig. 1(b). This redistribution of free charges across the heterojunction has a significant impact on I-V characteristics.
3. PHOTOCURRENT CHARACTERISTICS FOR
Upon exposure to light, the photocurrent through the MIS junction is observed for oxide thickness of 3 nm, as shown in Fig. 2. To measure the photocurrent, 532 nm pulsed laser light (Spark Antares laser, 80 MHz, pulse width 5–6 ps) is focused onto the sample by a 50× objective via an inverted Olympus microscope, as shown in Fig. 2(a). A wide field optical image of sample 1 with light focused on the CdSe heterojunction is displayed in Fig. 2(b). The generated photocurrent is also detected by a lock-in technique (see Appendix A for details). The calibrated peak currents as a function of bias voltage on the Au electrode for sample 1 and sample 4 are presented in Figs. 2(c) and 2(d), respectively. In both cases, we observed non-zero photocurrent at zero bias voltage. However, the polarity of photocurrent at zero bias varies from one sample to another. Among all samples, 25% exhibit positive current, 58% have negative current, and 17% have zero current at zero bias. We attribute the change in the polarity of zero bias photocurrent to the variation in the conduction band edge position of the ultrathin . As described in Figs. 2(e) and 2(f), the magnitude of photocurrent increases with the incident laser power even at zero bias. The simulated photocurrent curves, by the COMSOL semiconductor module in combination with the wave optics module, with the conduction band/Au work function at 3.74 eV/5.13 eV and 3.77 eV/5.1 eV are given in Figs. 2(g) and 2(h), respectively, and yield good agreement with experimental observations (see Appendix B for details). The photocurrent at zero bias is sensitively related to the local conduction band edge of , which can be related to the immobilized charges in the film [18,19] and the local work function of Au.
Figure 2.Photocurrent across
4. IMPACT OF CARRIER REDISTRIBUTION ON OPTICAL ABSORPTION IN
The simulation model we developed to describe the photocurrents can also be used to evaluate the optical absorption properties of CdSe in this heterojunction. As the net optical absorption coefficient can be expressed as [20], where is the maximum absorption coefficient at photon energy of , and and are the electron occupancy factors in the valence band and conduction band at the given photon energy, respectively; and follow Fermi–Dirac distribution (see Appendix B for details) and can be obtained directly from the simulation model. The net optical absorption loss is therefore directly related to the difference in electron occupancy factors in the two bands. The band alignment across the heterojunction can directly influence the electron occupancy factors within the semiconductor. Figure 3(a) gives a diagram of the cross section used in the following simulations. The thicknesses of and CdSe are set as 3 nm and 200 nm, respectively. Figures 3(b) and 3(c) show energy diagrams at zero bias for the heterojunction and a model system, respectively. The model system assumes the metal work function positioned at the middle of the bandgap of CdSe as well as at that of the insulator. As discussed previously, the heterojunction allows the carriers to be redistributed to maintain the same effective Fermi level across the heterojunction. Figure 3(d) demonstrates the electron density distribution along the red line highlighted in Fig. 3(a) in the configuration [Fig. 3(b)] upon excitation of light at 532 nm as a function of position and incident power. Figure 3(f) gives the corresponding at the emission wavelength of 716 nm. As a comparison, we give the electron density distribution [Fig. 3(e)] and [Fig. 3(g)] of the model system [Fig. 3(c)]. The hole density, and distributions are given in Appendix E.
Figure 3.Impact of carrier redistribution on optical absorption loss in
It is clear from the plots that for the heterojunction, the electron density of CdSe is strongly enhanced in the region close to Ag due to the accumulation of electrons, which leads to a significantly decreased value. At the pump power of 1 W, at in Fig. 3(f), meaning that the absorption coefficient decreases to only 10% of the original value, while at in the model system without the transfer of carriers from the metal [Fig. 3(g)]. It is worth noting that the input power required to reach the onset of absorption loss reduction, for example, , occurs at input power of for , which is four orders of magnitude smaller compared to that needed for the model system (occurring at input power of 0.2 W). Our analysis reveals the mechanisms for the pump intensity reduction observed in gain-assisted loss compensation in hybrid plasmonic waveguides compared to their photonic counterparts [5]. These results suggest that in gain-assisted hybrid plasmonic signal propagation, the presence of metal not only allows light to be confined into a deep subwavelength volume, but the redistribution of carriers across the heterojunction also allows significant modulation to the output signal with only a fraction of pump energy as required to reach the same level of signal modulation in its photonic counterpart (more details can be found in Appendix F). This unique property can be very useful for all-optical signal modulation and computation [21,22].
5. ABRUPT PHOTOCURRENT INCREASE FOR
There is, however, a limitation on how thick the is allowed to be for the detection of photocurrent at zero bias. Once the thickness of increases to and above 4 nm, not only an energy gap is developed in the photocurrent I-V characteristics, but also a step-like feature is observed in the photocurrent as illustrated in Fig. 4(a). It is more intuitive to convert the peak current into the unit of electron number per 5 ps, as indicated by the right vertical axis of Fig. 4(a). The three curves are obtained from two different experimental setups (see Appendix A). The bottom curve is obtained with the same setup as used for Fig. 2 while the top two curves are obtained using a supercontinuum laser (repetition rate 40 MHz, pulse width), after passing through a bandpass filter. The curves for sample 2 and sample 6 are offset for clarity. The dashed lines indicate the zero current lines for each curve. From the curves shown in Fig. 4(a), we can see that only a few electrons transport through the heterojunction from each light pulse. This repeatable step-like negative current onset occurs between and 0.5 V, too small to be related to the avalanche effect [23]. We speculate that the sharp increase in photocurrent is caused by the Coulomb blockade effect commonly observed in single electron tunneling events through a double-barrier heterojunction [19,24–26]. In the current case, the tunneling most likely happens through a localized state within the film [16]. It is difficult to predict the Coulomb blockade energy, the energy required for the electron to overcome to tunnel into the localized state within the layer, from the current simulation model. However, we can estimate the value to be less than half of the energy gap observed in Fig. 4(a) and larger than the thermal energy of , therefore between 0.03 eV and 0.75 eV. Figure 4(b) shows the photocurrent as a function of bias voltage on Au for different incident laser powers for sample 6. As shown in the inset of Fig. 4(b), the generated photocurrent after the threshold voltage linearly depends on the incident laser power. The photocurrent also depends on the wavelength of excitation light and is substantially quenched upon exposure to light with a photon energy below the bandgap of CdSe, as shown in Appendix G.
Figure 4.Photocurrent across
In samples with of about 5 nm thickness, we have observed the onset voltage spanning from to (see Appendix H). These results suggest that once the thickness goes beyond 5 nm, carrier redistribution across the MIS junction may be ignored at zero bias. As mentioned previously, the dark current of devices with thickness of 5 nm is lower than the detection limit of our setup () in the bias range of to . Nevertheless, the peak photocurrent detected from these devices can exceed 70 μA at with a modest peak incident power of , giving a current gain larger than and a photoresponsivity of 0.03 A/W [27]. For bias voltage magnitude larger than 2 V, the avalanche effect caused by impact ionization may also contribute to the giant current gain [23,28]. With optimization of CdSe thickness and excitation method to maximize the absorption of light [27,29], the photoresponsivity value of this type of device can be further improved. The very low dark current is in sharp contrast with the high photocurrent detected upon exposure to light. The low operation voltage also makes MIS based photodetection very compatible with on-chip applications and can be easily incorporated in an integrated optoelectrical platform.
6. CONCLUSION
In conclusion, we have studied the photocurrent generation and transportation through the heterojunction with thickness varying from 3 nm to 5 nm. The direct observation of nonzero photocurrents at zero bias voltage in the oxide thickness range of suggests a charge redistribution across the heterojunction when CdSe is optically excited, and this charge redistribution leads to a substantially reduced optical absorption coefficient upon excitation of the hybrid plasmonic mode supported by the heterojunction, making gain-assisted plasmonic signal modulation possible at low pump intensity. As the thickness of the oxide increases to , the Coulomb blockade effect is observed in the photocurrents, corresponding to a sharp onset in the I-V characteristics. The extremely low dark current and large photocurrent detected upon exposure to light at relatively low voltage make this configuration a promising candidate in photodetection applications where low dark current and low bias voltages are of paramount importance.
Acknowledgment
Acknowledgment. We thank the Laboratory of Microfabrication in the Institute of Physics, Chinese Academy of Sciences, for experimental support. N. Liu acknowledges the support from the Irish Research Council “New Foundations” Programme and Science Foundation Ireland Career Development award. X. H. Yan, L. Gao, and H. Wei acknowledge the support from the National Natural Science Foundation of China, and the Strategic Priority Research Program of Chinese Academy of Sciences.
APPENDIX A: METHODS
The CdSe MBs are grown by a chemical vapor deposition method. In short, an alumina boat containing CdSe powder is placed in the middle of a quartz tube furnace (single zone, Elite Thermal Systems). A Si substrate coated with a thin layer of Au () is then positioned at the downstream side of the tube. The furnace is first pumped down by a mechanical pump for 30 min. Ar gas is then introduced in the tube with a flow rate of 50 sccm (standard cubic centimeters per minute) while the pump is kept on. After the system is stabilized for 30 min, the furnace is heated up to 690°C and kept at this temperature for another 30 min for CdSe nanobelt growth. During growth, the Si sample is kept at a temperature of 500°C–600°C.
To fabricate the device for photocurrent measurements, an Au pad along with alignment marks is first defined on a glass coverslip by photolithography and deposited by thermal evaporation. CdSe MBs are then transferred onto the coverslip. An Au stripe (thickness 50 nm) connecting the Au pad and one side of the CdSe MB is defined by E-beam lithography and deposited by thermal evaporation. An ultrathin layer of is then deposited on the sample by atomic layer deposition. The thickness of the is measured by an ellipsometer or surface profilometer. Last, an Ag electrode (thickness 60 nm) is defined on the other side of the CdSe MB by E-beam lithography and deposited by thermal evaporation.
The applied voltage is supplied by a square wave function generator, which is applied to the Au electrode of the sample. The current is then fed into the current port of a lock-in amplifier (Signal Recovery 7265) from the Ag electrode. The current is calibrated against a resistor of known value.
To measure the photocurrent, pulsed lasers (Spark Antares laser at 532 nm, 80 MHz, pulse width 5–6 ps or NKT SC-400 Supercontinuum laser, 40 MHz, pulse width 76–90 ps, 450 nm to 2.5 μm) are used. To extract the small photo-induced current, a lock-in technique is used. The reference frequency of the lock-in amplifier is taken from an optical chopper, which is used to modulate the on-and-off of the light focused on the sample at a low frequency (). The modulated light is directed into an inverted microscope (Olympus) and then focused onto the sample by a long working distance objective (Olympus, NA 0.55). A homemade LabVIEW program is used to control the voltage output/input from a National Instruments multifunction data acquisition (DAQ) device (NI USB-6212). The output voltage is applied to the Au electrode of the sample. The photocurrent is directly fed into the current port of a lock-in amplifier (Signal Recovery 7265) from the Ag electrode. The output voltage from the lock-in amplifier is then read back by the same LabVIEW program via the DAQ device.
The simulation on the I-V characteristic of the devices without and with exposure to light is carried out by the COMSOL Multiphysics simulation package 5.5 using the semiconductor module coupled with the wave optics module. Stationary study is used to calculate the I-V behavior without light and stationary-frequency study is used to calculate the photocurrent upon exposure to light (see Appendix
APPENDIX B: COMSOL SIMULATIONS ON DARK CURRENT, PHOTOCURRENT, AND LOSS COMPENSATION
The semiconductor module in COMSOL 5.5 solves for the drift-diffusion equations of current density in the defined region:
The solution also satisfies current continuity equations:
The Wentzel–Kramers–Brillouin (WKB) tunneling model is used to allow tunneling current across the heterojunction to be calculated. Following Ref. [
For direct bandgap semiconductors, the net generation rate due to stimulated emission can be evaluated as [
The geometry we used to simulate the (3 nm)-CdSe (200 nm)-Au heterojunction is given in Fig.
Figure 5.Geometry of
APPENDIX C: DARK CURRENT-VOLTAGE CHARACTERISTICS OF MORE SAMPLES
Figure
APPENDIX D: HOLE DENSITY DISTRIBUTION AT ZERO BIAS WITHOUT EXPOSURE TO LIGHT
The hole density distribution at zero bias across the (3 nm)-CdSe (500 nm)-Au heterostructure is given in Fig.
APPENDIX E: HOLE DENSITY, fv, AND fc DISTRIBUTIONS RELATED TO FIG.?3
The distributions of hole density, , and at zero bias as a function of input power are shown in Fig.
APPENDIX F: IMPACT OF CHARGE REDISTRIBUTION ACROSS Ag-Al2O3-CdSe ON GAIN-ASSISTED HYBRID PLASMONIC MODE PROPAGATION
As discussed in previous work [
As discussed in the main text, we can assume the absorption coefficient of CdSe as , where is the absorption coefficient without the pump light. As shown in Fig.
We can approximate the output intensity of light through a waveguide as where is the initial intensity of light at the start of the waveguide and the length of the waveguide. Without pump light, . With pump light, . So, and . Here is the distance on the waveguide that the pump light is on. If can be ignored when the intensity of the stimulated emission is much larger than that of the spontaneous emission, can be further expressed as
For a hybrid plasmonic waveguide, is usually much larger than that of the photonic waveguides. For the two waveguides shown in Fig.
Figure 6.Variation of dark current density across the
Figure 7.Dark
Figure 8.Hole density distribution (
Figure 9.Hole density,
Figure 10.Loss compensation measurements on the fundamental hybrid plasmonic mode supported on CdSe nanobelt-
Figure 11.(a) Electric energy density distribution within
APPENDIX G: SPECTRAL DEPENDENCE OF THE PHOTOCURRENT
The photocurrent of sample 6 was measured at three different wavelengths, as shown in Fig.
Figure 12.Photocurrent measurements on sample 6 at three different wavelengths. (a) Typical photocurrent versus bias voltage at different incident laser wavelengths. The curve obtained from 800 nm excitation is multiplied by a factor of 20 for clarity. (b) Dependence of the plateaued current on the input power. The error bars indicate the range of current variation in the gradual linear increase regions.
APPENDIX H: PHOTOCURRENT I-V CHARACTERISTICS OF Ag-Al2O3 (5?nm)-CdSe MB-Au
Figure
Figure 13.Photocurrent measurement on four samples of
References
[1] S. M. Sze, K. K. Ng. Physics of Semiconductor Devices(2007).
[2] R. F. Oulton, V. J. Sorger, D. A. Genov, D. F. P. Pile, X. Zhang. A hybrid plasmonic waveguide for sub-wavelength confinement and long-range propagation. Nat. Photonics, 2, 496-500(2008).
[3] N. Liu, A. Gocalinska, J. Justice, F. Gity, I. Povey, B. McCarthy, M. Pemble, E. Pelucchi, H. Wei, C. Silien, H. Xu, B. Corbett. Lithographically defined, room temperature low threshold subwavelength red-emitting hybrid plasmonic lasers. Nano Lett., 16, 7822-7828(2016).
[4] R. F. Oulton, V. J. Sorger, T. Zentgraf, R.-M. Ma, C. Gladden, L. Dai, G. Bartal, X. Zhang. Plasmon lasers at deep subwavelength scale. Nature, 461, 629-632(2009).
[5] N. Liu, H. Wei, J. Li, Z. X. Wang, X. R. Tian, A. L. Pan, H. X. Xu. Plasmonic amplification with ultra-high optical gain at room temperature. Sci. Rep., 3, 1967(2013).
[6] Z. Li, B. Corbett, A. Gocalinska, E. Pelucchi, W. Chen, K. M. Ryan, P. Khan, C. Silien, H. Xu, N. Liu. Direct visualization of phase-matched efficient second harmonic and broadband sum frequency generation in hybrid plasmonic nanostructures. Light Sci. Appl., 9, 180(2020).
[7] J. Shi, Y. Li, M. Kang, X. He, N. J. Halas, P. Nordlander, S. Zhang, H. Xu. Efficient second harmonic generation in a hybrid plasmonic waveguide by mode interactions. Nano Lett., 19, 3838-3845(2019).
[8] H. Wei, D. Pan, S. Zhang, Z. Li, Q. Li, N. Liu, W. Wang, H. Xu. Plasmon waveguiding in nanowires. Chem. Rev., 118, 2882-2926(2018).
[9] R. Maiti, C. Patil, M. A. S. R. Saadi, T. Xie, J. G. Azadani, B. Uluutku, R. Amin, A. F. Briggs, M. Miscuglio, D. Van Thourhout, S. D. Solares, T. Low, R. Agarwal, S. R. Bank, V. J. Sorger. Strain-engineered high-responsivity MoTe2 photodetector for silicon photonic integrated circuits. Nat. Photonics, 14, 578-584(2020).
[10] A. H. Jones, S. D. March, S. R. Bank, J. C. Campbell. Low-noise high-temperature AlInAsSb/GaSb avalanche photodiodes for 2-μm applications. Nat. Photonics, 14, 559-563(2020).
[11] . Semiconductor Module User’s Guide, COMSOL 5.4(2018).
[12] A. W. Dweydari, C. H. B. Mee. Work function measurements on (100) and (110) surfaces of silver. Phys. Status Solidi A, 27, 223-230(1975).
[13] D. R. Lide. CRC Handbook of Chemistry and Physics(2008).
[14] F. Buch, A. L. Fahrenbruch, R. H. Bube. Photovoltaic properties of five II-VI heterojunctions. J. Appl. Phys., 48, 1596-1602(1977).
[15] M. Kaltenbrunner, P. Stadler, R. Schwödiauer, A. W. Hassel, N. S. Sariciftci, S. Bauer. Anodized aluminum oxide thin films for room-temperature-processed, flexible, low-voltage organic non-volatile memory elements with excellent charge retention. Adv. Mater., 23, 4892-4896(2011).
[16] W. H. Rippard, A. C. Perrella, F. J. Albert, R. A. Buhrman. Ultrathin aluminum oxide tunnel barriers. Phys. Rev. Lett., 88, 046805(2002).
[17] F. Wang, N. A. Melosh. Plasmonic energy collection through hot carrier extraction. Nano Lett., 11, 5426-5430(2011).
[18] X. H. Qiu, G. V. Nazin, W. Ho. Vibrationally resolved fluorescence excited with submolecular precision. Science, 299, 542-546(2003).
[19] N. Liu. Atomic scale understanding of nanostructures in a double barrier tunneling junction: scanning tunneling microscopy of alkali doped buckminsterfullerenes on partially oxidized NiAl(110)(2005).
[20] O. Svelto. Principles of Lasers(2010).
[21] J. Feldmann, M. Stegmaier, N. Gruhler, C. Ríos, H. Bhaskaran, C. D. Wright, W. H. P. Pernice. Calculating with light using a chip-scale all-optical abacus. Nat. Commun., 8, 1256(2017).
[22] J. Feldmann, N. Youngblood, M. Karpov, H. Gehring, X. Li, M. Stappers, M. Le Gallo, X. Fu, A. Lukashchuk, A. S. Raja, J. Liu, C. D. Wright, A. Sebastian, T. J. Kippenberg, W. H. P. Pernice, H. Bhaskaran. Parallel convolutional processing using an integrated photonic tensor core. Nature, 589, 52-58(2021).
[23] O. Hayden, R. Agarwal, C. M. Lieber. Nanoscale avalanche photodiodes for highly sensitive and spatially resolved photon detection. Nat. Mater., 5, 352-356(2006).
[24] B. Su, V. J. Goldman, J. E. Cunningham. Observation of single-electron charging in double-barrier heterostructures. Science, 255, 313-315(1992).
[25] J. B. Barner, S. T. Ruggiero. Observation of the incremental charging of Ag particles by single electrons. Phys. Rev. Lett., 59, 807-810(1987).
[26] M. A. Reed, C. Zhou, C. J. Muller, T. P. Burgin, J. M. Tour. Conductance of a molecular junction. Science, 278, 252-254(1997).
[27] R. Gherabli, S. R. K. C. Indukuri, R. Zektzer, C. Frydendahl, U. Levy. MoSe2/WS2 heterojunction photodiode integrated with a silicon nitride waveguide for visible light detection with high responsivity(2021).
[28] Y.-J. Doh, K. N. Maher, L. Ouyang, C. L. Yu, H. Park, J. Park. Electrically driven light emission from individual CdSe nanowires. Nano Lett., 8, 4552-4556(2008).
[29] C. Frydendahl, M. Grajower, J. Bar-David, R. Zektzer, N. Mazurski, J. Shappir, U. Levy. Giant enhancement of silicon plasmonic shortwave infrared photodetection using nanoscale self-organized metallic films. Optica, 7, 371-379(2020).
[30] K. Yang, J. R. East, G. I. Haddad. Numerical modeling of abrupt heterojunctions using a thermionic-field emission boundary condition. Solid State Electron., 36, 321-330(1993).
Set citation alerts for the article
Please enter your email address