Letter pubs.acs.org/NanoLett
Redesigning Photodetector Electrodes as an Optical Antenna Pengyu Fan,† Kevin C. Y. Huang,† Linyou Cao,†,‡ and Mark L. Brongersma*,† †
Geballe Laboratory for Advanced Materials, Stanford University, Stanford, California 94305, United States Department of Materials Science and Engineering, North Carolina State University, Raleigh, North Carolina 27607, United States
‡
ABSTRACT: At the nanoscale, semiconductor and metallic structures naturally exhibit strong, tunable optical resonances that can be utilized to enhance light-matter interaction and to dramatically increase the performance of chipscale photonic elements. Here, we demonstrate that the metallic leads used to extract current from a Ge nanowire (NW) photodetector can be redesigned to serve as optical antennas capable of concentrating light in the NW. The NW itself can also be made optically resonant and an overall performance optimization involves a careful tuning of both resonances. We show that such a procedure can result in broadband absorption enhancements of up to a factor 1.7 at a target wavelength of 660 nm and an ability to control the detector’s polarization-dependent response. The results of this study demonstrate the critical importance of performing a joint optimization of the electrical and optical properties of the metallic and semiconductor building blocks in optoelectronic devices with nanoscale components. KEYWORDS: Nanowire photodetector, optical antenna, antenna electrodes, coupled optical resonators, cascading optical resonances, plasmonics
P
taught us how a reduction in device footprint comes with the benefits of lower power consumption, a higher signal-to-noise, and an increased speed of operation. From an optical perspective, the benefit of having closely spaced electrodes made from lossy and reflective metals is less obvious as they may partially block or absorb an incident light beam. However, the narrow, finite-length slit that naturally forms between two closely spaced metallic electrodes can support strong standing wave resonances of metal−insulator−metal (MIM) or gap surface plasmon-polaritons (SPPs).20−22 When such a resonance is excited, one can capitalize on the high field intensities that build up in the slit to enhance light absorption in a semiconductor medium placed inside the slit.4 The high refractive index NW in the gap also exhibits strong optical resonances and with proper engineering it should be possible to cascade a semiconductor NW resonance with the metallic slit resonance. In the overall optimization one needs to be cognizant of the fact that both types of resonances are strongly dependent on the size and geometry of the structures as well as the state of polarization of the incident light. Below, we discuss our holistic design strategy for an ultracompact photodetector that is aimed at a joint optimization of the semiconductor medium and the electrical leads. We also analyze the performance of optimized devices with the Ge NWs parallel and orthogonal to the slit. We will first discuss the case where a Ge NW is placed inside and along the length of the slit formed between two closely spaced silver (Ag) pads. Figure 1c shows a cross-sectional
hotodetectors play a critical role in many optoelectronic applications, including optical interconnection 1 and imaging.2 Scaling of these devices has brought many advantages in terms of their operating speed, signal-to-noise ratio, and power consumption. At first sight, it appears that scaling of photodetectors below the diffraction limit would also result in the undesired inability to capture all incident photons from an optical beam or waveguide. To overcome this challenge, researchers have successfully employed the light concentrating properties of metallic nanostructures that support surface plasmon excitations.2−8 More recently, it has also been shown that the optical Mie9 resonances that naturally occur in highdielectric-constant nanostructures can be used to boost light absorption in photodetectors10,11 and solar cells.12−14 Most recently, the resonances in metallic and semiconductor nanostructures have been cascaded to even further boost device performance15−17 or to add entirely new functions such as invisibility.18 In this work, we demonstrate that the metallic leads to a semiconductor photodetector can be re-engineered to perform simultaneous charge extraction and optical antenna functions. We will illustrate this with metal−semiconductor− metal (MSM) photodetector consisting of a Ge nanowire (NW) placed between two metallic contacts. MSM photodetectors are known for their high speed and low-noise operation.19 Figure 1a,b illustrates the two most basic configurations in which a Ge nanowire can be contacted to realize a MSM photodetector. The NW (gray) can be positioned either perpendicular or parallel to the gap formed between two metallic contacts (orange). From an electronics viewpoint, it is clearly beneficial to bring the metallic electrodes closer together. Experience with scaling of electronic devices has © 2013 American Chemical Society
Received: September 22, 2012 Revised: December 19, 2012 Published: January 8, 2013 392
dx.doi.org/10.1021/nl303535s | Nano Lett. 2013, 13, 392−396
Nano Letters
Letter
Figure 1. Design of semiconductor nanowire (NW) photodetectors that exploit optical resonances in both the NW itself and the narrow gap formed between the metallic contacts. Metallic contacts in a metal−semiconductor−metal photodetector with a semiconductor NW can be redesigned to serve as optical antennas capable of enhancing the optical response. Two basic configurations are explored where the NW is either (a) perpendicular or (b) parallel to the gap formed between the metallic leads. The image also shows the direction of the electric field for the case that the wires are top-illuminated with transverse magnetic (TM, E-field along the wire axis) or transverse electric (TE, E-field normal to the wire axis) light. When the electric field is normal to the metal slit a strong plasmonic resonance can be excited in the slit. (c) SEM image of a 90 nm diameter Ge NW placed in a 150 nm wide slit formed between two 150 nm thick Ag pads. Scale bar is 200 nm. (d) Simulated enhancement of the absorption in a Ge NW in an Ag slit as a function of the metal electrode thickness. The enhancement was measured with respect to a bare Ge NW of the same diameter. The simulation was performed at an excitation wavelength of 650 nm and for TE polarization (electric field normal to the NW and slit). The dashed blue line represents an enhancement factor of 1 (i.e., no enhancement). (e,f) Distribution of magnetic field intensity Hz in the cross sectional plane of the device with a 150 or a 420 nm Ag thickness respectively. Scale bar is 200 nm.
scanning electron microscopy (SEM) image of this basic configuration. The geometrical parameters of this structure were chosen to optimize the light absorption at a (somewhat arbitrarily chosen) target wavelength of 660 nm. To maximize the light absorption at this wavelength, one needs to first select a Ge NW diameter that produces a strong optical resonance at this wavelength. For top-illuminated NWs, the resonant modes can be divided into transverse-magnetic (TM, with the electric field along the axis of the nanowire) and transverse-electric (TE, with the electric field normal to the axis of the nanowire) modes.10 In deciding whether to select a NW that features a TM or TE resonance, it is important to realize that a plasmonic mode in a slit can only be excited when the electric field is orthogonal to the slit (and thus also orthogonal to the wire in the slit). With the aim to ultimately cascade the NW and slit resonances, we choose a 90 nm diameter Ge NW that exhibits the lowest order TE resonance at the target wavelength of 650 nm. After having chosen a NW, we can turn our attention to the optimization of the metallic electrodes. Here, the goal is to effectively drive the standing wave SPP resonances in the slit that result from gap SPP reflections at the top (entrance) and bottom (exit) of the slit. The spectral locations of these resonances are effectively controlled by the length of the slit, that is, the thickness of the metallic electrodes. Figure 1d shows the simulated dependence of the light absorption in the Ge NW as a function of the thickness of the adjacent Ag pads and a pad spacing of 150 nm. It can be seen that the magnitude of the NW absorption oscillates with the Ag thickness and shows maxima near 150 and 420 nm. The highest peak is near 150 nm, where the NW absorption is increased by as much as 1.9 times over the bare Ge NW. This governed our choice of the metal thickness for the fabricated NW detector structure shown in Figure 1b. To gain physical insight into the nature of the
absorption enhancements, we also plot the distribution of the magnetic field intensity |Hz| in the device cross sections for two optimal thicknesses. For both the 150 nm (Figure 1e) and 420 nm (Figure 1f) thick Ag pads the Ge NW core features a single broad maximum in the |Hz| intensity, indicating the excitation of the lowest order TE resonance in the NW. Depending on the thickness, one (Figure 1e) or two (Figure 1f) antinodes in the |Hz| intensity could be identified inside the Ag slit. This is consistent with the first and second order standing wave gap SPP resonances supported in the Ag slits respectively. From these results, it can be concluded that a slit resonance between metallic electrodes can effectively be cascaded with a semiconductor NW resonance to boost the absorption of light beyond what is feasible in the NW by itself. Next, we aim to verify this point experimentally. Figure 2 shows the photocurrent spectra taken from a bare 90 nm diameter Ge NW and from the same 90 nm diameter NW, in a region where it is surrounded by two 150 nm thick Ag pads. The measurement geometry that allowed both measurements on the same NW is shown schematically in the inset to Figure 2. Figure 3a also shows a top view SEM image of the NW in the area where it is surrounded by the Ag pads. The spectra for both the bare wire and the wire in the slit show broad peaks around 660 nm, close to the fundamental TE resonance of the bare wire. To analyze the magnitude of the enhancement that results from the metal pads, we divided these photocurrent spectra (magenta dots). It was found that a peak enhancement of almost 1.7 can be reached at a wavelength that is very close to the target of wavelength of 660 nm by simultaneously harnessing the beneficial effects of slit and NW resonances. Metallic leads/pads can not only enhance the photocurrent from a Ge NW photodetector, they can also be used to modify its polarization response. This opportunity arises from the fact 393
dx.doi.org/10.1021/nl303535s | Nano Lett. 2013, 13, 392−396
Nano Letters
Letter
the Ge NW in the Ag slit the situation is reversed and it absorbs TE polarized light more strongly than TM polarized light (a 1.7 ratio is measured). This results from the fact that the standing SPP resonance in the slit can only effectively be driven with the electric field normal to the slit. It is thus clear that the polarization dependent response of a NW can be significantly altered by the presence of the metal pads. Next, we discuss the case in which a GeNW is electrically hooked up in an orientation normal to the metallic slit. This configuration allows for a convenient study of the effects of device scaling and the significant impact the spacing of electrical contacts can have on the optical device properties. Figure 4a,b shows a 100 nm diameter Ge NW that is electrically contacted by three 150 nm thick aluminum (Al) leads. Two of the leads form a narrow, 200 nm wide gap and the other pair forms a wider 400 nm gap. Through these gaps light from a free-space beam can reach the Ge NW to produce photocurrent. The use of three contacts allows one to fairly compare the photocurrent measured from wide and narrow lead-spacings on one and the same wire while avoiding wire-to-wire performance variations. Photocurrent maps on these devices were taken at a 650 nm illumination wavelength and for different polarizations. Figure 4c shows that a significantly larger photocurrent response is obtained from the NW region in the wide slit as compared to the narrow slit for TE illumination. Figure 4d shows how the photoresponse from both slit regions changes upon rotation of the polarization. Whereas the photoresponse from the wide slit increases only by factor 1.6, the response from the narrow slit by as much as 3.8. From Figure 4e, under TM-polarized illumination (electric field along the NW and normal to the slit), it can in fact be seen that the narrow slit produces more photocurrent than the wide slit, despite the fact that a shorter length of NW is exposed to the incident beam. These observations, as for the case of NW parallel to the metallic slit, are consistent with the existence of a strong plasmonic resonance in narrow metallic slits. Figure 4f shows the ratio between NW responses for TM to TE polarization could be enhanced from about 1.6 to almost 4 by placing the metallic leads closer together. A Ge NW detector with closely spaced leads would thus possess a higher sensitivity to the incident polarization. In this work we have demonstrated that the metallic leads to a nanoscale photodetector not only should be optimized from an electrical viewpoint, but also can greatly impact the optical properties. Through rational design, the photocurrent can be enhanced and the polarization response can be modified. This unique opportunity is naturally derived from the fact that nanoscale semiconductor and metallic building blocks exhibit strong optical resonances. In the past, the metallic leads where mainly designed from an electrical perspective to minimize charge extraction and RC delay times and to reduce electrical power consumption. Due to device scaling the building blocks making up a detector have become small compared to the wavelength of light and optically resonant. As a result, a simultaneous electronic and optical optimization needs to take place. Such an optimization is nontrivial as the optical modes of nanostructures extend well beyond their physical boundaries and strong near-field interaction and mode hybridization can occur. However, the pay-off can be large as high performance devices and new functionalities can be implemented in this fashion. By packing multiple electronic and optical functions in the same physical space, their integration density can be increased and the cost per device can be reduced. Whereas the
Figure 2. Spectral response of a Ge NW photodetector for which the NW lies along the length of the Ag slit under TE incidence (electric field polarized normal to the NW and slit). Green curve with hollow data points represents the photoresponse of a bare 90 nm diameter Ge NW as a function of the illumination wavelength. The green curve with solid data points represents the photoresponse of the same Ge NW, but now placed between in a 150 nm wide and 150 nm thick Ag pads. The incident light is polarized normal to the NW and slit (TE polarization). The spectral enhancement in the photoresponse for a Ge NW in the Ag slit over a bare Ge NW is plotted in magenta dots. The enhancement due to the presence of the Ag pads peaks around 660 nm at a similar spectral location to the photoresponse maxima. The inset is a schematic top-view illustration of the device and measurement geometry where the photocurrent is measured by focusing incident light on different locations of a Ge NW photodetector.
that the optical responses of both the NW and slit resonances are strongly polarization-dependent. We experimentally demonstrate this opportunity by presenting spatially resolved photocurrent maps of our Ge NW detector from the same area as is shown in the SEM image in Figure 3a. Figure 3b,c shows photocurrent maps taken using a 650 nm excitation wavelength for the cases that the electric field is polarized either normal or parallel to the slit. When the incident light is polarized normal to the slit and NW (Figure 3b), the largest photoresponse is obtained near the slit in the metal pad. This is consistent with the enhancement of Ge NW absorption due to the excitation of a plasmonic resonance in Ag slit. When the electric field direction is rotated by 90 degrees and aligned parallel to the slit and the NW (Figure 3c), the photoresponse is reversed. For this polarization, the photocurrent obtained from bare NW regions is larger. For this polarization, the excitation of the plasmonic mode in the slit is symmetry forbidden and less light is able to reach the NW (i.e., more is reflected back to the source). Figure 3d shows the polarization-dependence of the photocurrent response obtained for the bare Ge NW and the Ge NW in the slit. It can be seen that for these geometries the polarization direction that provides the strongest response is rotated by 90°. The bare Ge NW absorbs more strongly under TM than TE incidence (a 1.5 ratio is measured) due to the presence of a degenerate TM resonance of a higher order supported by the NW at the same wavelength.10 However, for 394
dx.doi.org/10.1021/nl303535s | Nano Lett. 2013, 13, 392−396
Nano Letters
Letter
Figure 3. Comparison of polarization dependent photocurrent response for a bare Ge NW and a Ge NW in a Ag slit. (a) Scanning electron microscopy image of the device that is also shown in the inset of Figure 2. It features a 90 nm diameter Ge NW located inside a slit defined by two Ag pads. Scale bar is 500 nm. (b,c) Photocurrent maps taken from the NW in the region where is flanked by the two Ag pads. The magnitude of the photocurrent is signified by the color of each point in the image. The device geometry is overlapped onto the photocurrent image with black lines. Scale bar is 500 nm. (d) Measured polarization-dependence of the photocurrent response for the bare NW region and NW in the Ag slit. Both curves are normalized to their own minimum.
Figure 4. Photocurrent response of a Ge NW oriented perpendicular to the slit formed by Al electrodes. (a) Schematic of the device geometry featuring a single Ge NW contacted by three Al electrodes. The two pairs of electrodes define a narrow and a wide gap. (b) SEM image of the fabricated device with a 100 nm diameter Ge NW crossing 200 nm wide and 400 nm wide slits between Al electrodes. Scale bar is 500 nm. (c) Photocurrent map of the device for the TE illumination (electric field normal to the NW and along the slit) with the polarization of the electric field as indicated by the orange arrow. The intensity of the photocurrent is signified by the color of each point. The device geometry is overlapped onto the photocurrent image as dark lines. A bright spot can be identified when the light shines on the NW part in wide slit and little signal can be detected from the narrow slit. Scale bar is 500 nm. (d) Photocurrent maps showing the polarization dependence of the optical response of the NW in the narrow and wide slits. The variation of polarization angle between each panel is 15°, as indicated by direction of the orange arrow. The photocurrent response from the NW in Al slit increases more rapidly than for the wide slit. (e) Photocurrent map for TM polarization with the electric field along the NW and normal to the slits. (f) Comparison of measured polarization dependent response for the NW in narrow and wide slit regions. The photocurrent signals were normalized to their own minimum reached for TE polarization. 395
dx.doi.org/10.1021/nl303535s | Nano Lett. 2013, 13, 392−396
Nano Letters
Letter
(3) Ishi, T.; Fujikata, J.; Makita, K.; Baba, T.; Ohashi, K. Si nanophotodiode with a surface plasmon antenna. Jpn. J. Appl. Phys., Part 2. 2005, 44, L364−L366, DOI: 10.1143/jjap.44.l364. (4) White, J. S.; et al. Extraordinary optical absorption through subwavelength slits. Opt. Lett. 2009, 34, 686−688. (5) Tang, L.; et al. Nanometre-scale germanium photodetector enhanced by a near-infrared dipole antenna. Nat. Photonics 2008, 2, 226−229. (6) Ditlbacher, H. Organic diodes as monolithically integrated surface plasmon polariton detectors. Appl. Phys. Lett. 2006, DOI: 10.1063/ 1.2362975. (7) Neutens, P.; Van Dorpe, P.; De Vlaminck, I.; Lagae, L.; Borghs, G. Electrical detection of confined gap plasmons in metal-insulatormetal waveguides. Nat. Photonics 2009, 3, 283−286. (8) Schuller, J. A.; et al. Plasmonics for extreme light concentration and manipulation. Nat. Mater. 2010, 9, 193−204. (9) Mie, G. Articles on the optical characteristics of turbid tubes, especially colloidal metal solutions. Ann. Phys.-Berlin 1908, 25, 377− 445. (10) Cao, L. Engineering light absorption in semiconductor nanowire devices. Nat. Mater. 2009, 8, 643−647 http://www.nature.com/ nmat/journal/v8/n8/suppinfo/nmat2477_S1.html. (11) Cao, L.; Park, J.-S.; Fan, P.; Clemens, B.; Brongersma, M. L. Resonant Germanium Nanoantenna Photodetectors. Nano Lett. 2010, 10, 1229−1233. (12) Cao, L.; et al. Semiconductor Nanowire Optical Antenna Solar Absorbers. Nano Lett. 2010, 10, 439−445. (13) Grandidier, J.; Callahan, D. M.; Munday, J. N.; Atwater, H. A. Light Absorption Enhancement in Thin-Film Solar Cells Using Whispering Gallery Modes in Dielectric Nanospheres. Adv. Mater. 2011, 23, 1272−1276. (14) Spinelli, P.; Verschuuren, M. A.; Polman, A. Broadband omnidirectional antireflection coating based on subwavelength surface Mie resonators. Nat. Commun. 2012, 3, 692 http://www.nature.com/ ncomms/journal/v3/n2/suppinfo/ncomms1691_S1.html. (15) Colombo, C.; Krogstrup, P.; Nygard, J.; Brongersma, M. L.; Morral, A. F. I. Engineering light absorption in single-nanowire solar cells with metal nanoparticles. New J. Phys. 2011, DOI: 10.1088/13672630/13/12/123026. (16) Brittman, S.; Gao, H.; Garnett, E. C.; Yang, P. Absorption of Light in a Single-Nanowire Silicon Solar Cell Decorated with an Octahedral Silver Nanocrystal. Nano Lett. 2011, 11, 5189−5195. (17) Hyun, J. K.; Lauhon, L. J. Spatially Resolved Plasmonically Enhanced Photocurrent from Au Nanoparticles on a Si Nanowire. Nano Lett. 2011, 11, 2731−2734, DOI: 10.1021/nl201021k. (18) Fan, P. An invisible metal-semiconductor photodetector. Nat. Photonics 2012, 6, 380−385 http://www.nature.com/nphoton/ journal/v6/n6/abs/nphoton.2012.108.html#supplementaryinformation. (19) Chi, On, C.; Okyay, A. K.; Saraswat, K. C. Effective dark current suppression with asymmetric MSM photodetectors in Group IV semiconductors. IEEE Photonics Technol. Lett. 2003, 15, 1585−1587. (20) Miyazaki, H. T.; Kurokawa, Y. Squeezing Visible Light Waves into a 3-nm-Thick and 55-nm-Long Plasmon Cavity. Phys. Rev. Lett. 2006, 96, 097401. (21) Bozhevolnyi, S. I.; Søndergaard, T. General properties of slowplasmon resonant nanostructures: nano-antennas and resonators. Opt. Express 2007, 15, 10869−10877. (22) Chandran, A.; Barnard, E. S.; White, J. S.; Brongersma, M. L. Metal-dielectric-metal surface plasmon-polariton resonators. Phys. Rev. B 2012, 85, 085416. (23) Dufaux, T.; Burghard, M.; Kern, K. Efficient Charge Extraction out of Nanoscale Schottky Contacts to CdS Nanowires. Nano Lett 2012, 12, 2705−2709.
current demonstration is limited to a photodetector, NW solar cells and multiple other nanoscale optoelectronic devices including optical sources, modulators, optical routers, wavelength splitters etc. can benefit from applying similar design concepts. Methods. Germanium NW were grown using a goldcatalyzed chemical vapor deposition process with gold nanoparticles with diameters in the range from 90 to 100 nm. The NWs were dispersed onto a silicon substrate with a 300 nm thick thermally grown oxide. Silver or aluminum contacts were then deposited onto the nanowires with standard electron beam lithography, metal evaporation, and lift-off to form metal−semiconductor−metal photodetectors. Ohmic contacts are formed to ensure photocurrent is generated uniformly along the NW for fair comparison of optical effects of adding metallic slits.23 A 2 nm thick film of germanium was first evaporated as seeding layer before the silver deposition to obtain smoother films. For the NWs placed parallel to the silver slit, a 30 nm thick conformal layer of aluminum oxide (Al2O3) was deposited via atomic layer deposition (Cambridge NanoTech) on top of the detector to prevent shorting the device. Photocurrent measurements were carried out with a supercontinuum laser source with an acousto-optic tunable filter (Fianium) to select a specific target wavelength. Incident light was focused on to the sample with a diffraction limited spot by a 50× objective (Mitutoyo). The sample was placed on a three-axis piezo stage (Physik Instrumente) to allow accurate focusing and spatial control over the location of the excitation spot to determine photocurrent generated as a function of illumination spot position. To increase the signal-to-noise ratio, the incident beam was chopped and the photocurrent signal was measured via a source meter (Keithley) connected to a lock-in amplifier (Stanford Research Systems).
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS Authors thank R. Pala, F. Afshinmanesh, and E. Barnard for useful discussions and help with photocurrent measurements. This work is supported by the Interconnect Focus Center, one of six research centers funded under the Focus Center Research Program (FCRP), a Semiconductor Research Corporation entity. P.F. would also like to acknowledge support from Stanford Graduate Fellowship. P.F. and M.L.B. conceived of the experiments. P.F. performed device fabrication and characterization. P.F. and K.C.Y.H. performed numerical simulations. L.C. performed Ge nanowire synthesis. P.F. and M.L.B. analyzed the data and wrote the first draft of the manuscript. All authors discussed and contributed to the final version of the manuscript.
■
REFERENCES
(1) Miller, D. Device Requirements for Optical Interconnects to Silicon Chips. Proc. IEEE 2009, 97, 1166−1185. (2) Laux, E.; Genet, C.; Skauli, T.; Ebbesen, T. W. Plasmonic photon sorters for spectral and polarimetric imaging. Nat. Photonics 2008, 2, 161−164. 396
dx.doi.org/10.1021/nl303535s | Nano Lett. 2013, 13, 392−396
