1. INTRODUCTION

A surface plasmon (SP) constitutes electron oscillation or wave propagation along the surface of a conductor (usually a metal), which is generated from the interaction of light and electrons [1–3]. An SP possesses several optoelectronic characteristics, such as collecting incident light to increase the photon density of states, generating a strong electric near field [4–6], guiding the incident photon into a surface polariton to propagate the energy on the surface of metal [7], or modulating the refractive index to adjust the phase shift of incident light [8,9]. Therefore, SPs have attracted significant attention recently and have been widely utilized in physics [10–12], electrical engineering [13–16], biology [17–19], chemistry [20], and novel medical therapies [21,22]. SPs can be divided into two modes, surface plasmon polaritons (SPPs) and localized surface plasmon resonance (LSPR). SPPs are incident photons being coupled into a propagating wave on the surface of metal [23]. LSPR constitutes incident photons being trapped and generating a strong electric near field on the surface of metal [4,5]. In comparison to SPPs, the SP from LSPR cannot propagate along the surface of nanostructures. The incident energy would be localized in a specific region or boundary on the metallic nanostructure (light trapping effect) [24,25], which can accumulate the energy and form a more powerful electric near field than an SPP [26]. However, the induced SP is limited. Since both modes can exist only as the momentum between the incident photon and the electron on the surface of metal is matched, the wavelength and polarization of incident light have to correspond highly to the scale, period, and shape of the metal structures in order to couple light into SPs (SPP or LSPR) [27–31]. Thus, the corresponding SP is mainly in a narrow wavelength range only. Knight et al. proposed an active optical antenna device that allows each single length metallic nanorod to correspond to a single wavelength to excite LSPR [4]. Lin et al. proposed a deep-trench/thin-metal active antenna, which claimed to cover a broadband absorption [5], but the high-intensity spectrum region of LSPR is actually only 200 nm. Recently, Wang et al. proposed a broadband achromatic optical metasurface device [8], which shows the phase shift from LSPR caused by various metallic patterns and sizes. Nevertheless, the effective conversion spectrum region for each single pattern is approximately 500 nm. In addition, other LSPR structures usually include anisotropic metallic nano-islands [29,32], nano-particles [33], gratings [30,34], nano-antennas [27,31], and geometric nano-patterns [24,35]. However, the enhanced spectral range remains confined to hundreds of nanometers due to the strong correlation between wavelength and structure scale.

Here, we reveal that broadband-induced LSPR exists in a certain three-dimensional (3D) metallic structure. The LSPR is possibly induced in a very broad spectral region of 300–2700 nm from the experiment, while the theoretical simulation further predicts that ultra-broadband LSPR could extend to wavelengths beyond 4000 nm. The absorption of a copper layer with a nano-scale thickness can be even over 80% in the spectral region of 300–2700 nm. Moreover, the polarization-sensitive issue, which is common in the general LSPR, could be avoided through two-dimensional (2D) symmetric metallic structures. Furthermore, the photoresponse of photodetectors can be effectively optimized by the LSPR in the ultra-broadband spectral region. In other words, the application of LSPR can be very easy, universal, and effective. It is also not necessary to use specific nano-scale linewidths or patterns for various wavelengths to excite LSPR through a focused ion beam or electron-beam lithography. This structure can be realized on silicon substrates through a CMOS-compatible process using regular apparatuses and materials, so it has a high potential of integration with Si-based electronics on a single chip. In this research, a Si-based Schottky [36,37] photodetector using the above broadband LSPR can effectively enhance the photoresponse by nearly 40 times in the spectral range from 1150 to 2700 nm, limited by our measurement instrument.

The response of a photodevice is improved due to incident photons being trapped by the LSPR-induced structure. However, the photocurrent is not directly generated by LSPR. In the past, Schottky photodiodes were one of the devices used to transfer incident photons into hot carriers or electron–hole pairs [29,38–41]. Although it can detect photons below the bandgap of a semiconductor, the detection spectrum is still limited by the Schottky barrier height. As shown in Fig. 1A, the carriers (electrons or holes) in metal are excited by incident light and become hot carriers. Afterward, the hot carriers will pass through the Schottky barrier and be separated by the built-in field of a Schottky junction to generate a photocurrent. By this internal photoemission absorption mechanism (IPA) [37], photocurrent can be generated only when the energy of photons is greater than the Schottky barrier height. However, our study demonstrates that such a limitation can be overcome. When the energy of incident photons is less than the Schottky barrier (wavelength 2.3–2.7 μm), the plasmon decay excited by the incident light propagates into the metal–semiconductor junction. This intense electronic near field excites a large number of hot carriers. During the early stage of thermal equilibrium, the hot carriers of high density collide with each other, and some of them obtain sufficient energy to pass through the Schottky barrier. The aforementioned mechanism can surmount the detection ability limit defined by the Schottky barrier height of previous devices. This not only enhances the photoresponsivity performance of the Schottky photodetector in the infrared [4,5,42,43], but also extends the detection spectrum to the mid-infrared range for Si-based photodetectors.

 

Fig. 1. Silicon-based Schottky photodetector with IPAS. (A) Internal photoemission absorption, the mechanism of the Schottky photodetector to generate a photocurrent. The detection ability is defined by the Schottky barrier height. (B) Schematic of the ultra-broadband detection photodiode with IPAS. Color-coded layers: orange, metallic nano-film; gray, p-Si based IPAS substrate. (C) Cross section of the IPAS. The linewidth inside of the IPAS cavity is L(H); gradient linewidth $\mathit{L}(\mathit{H})=\mathit{H}\times 2\cot (54.74°)$. (111) and (100) are the crystal planes of a single crystal silicon substrate. (D) Top view of the IPAS. The dimensions of the 2D symmetric periodic array and geometry unit are shown.

Download Full Size | PDF

To achieve the LSPR induced and photoresponsivity optimized by a structure in an ultra-broadband spectrum region, in this study, a device based on a novel concept (Fig. 1B) that combines a Schottky photodetector and a 3D inverted pyramid array structure (IPAS) was investigated. The cross section of the IPAS is shown in Fig. 1C. The height of the inverted pyramid cavity is represented as $H$, and the linewidth between both metallic sidewalls of the IPAS cavity is represented as L(H). As H increases, the inside horizontal linewidth of the IPAS cavity increases $[\mathit{L}(\mathit{H})=\mathit{H}\times 2\cot (54.74°)]$. Here, although the period of this metallic IPAS is constant, each structure unit has gradient metallic linewidth. Figure 1D is the top view of the IPAS, which conforms to the 2D symmetric periodic array and geometry unit. According to the near-field simulation of the interaction of the electromagnetic wave and IPAS, this structure can actually induce ultra-broadband and polarization-insensitive LSPR. Furthermore, the IPAS exhibits excellent 3D cavity effects in the cavity of the inverted pyramid, which can collect incident light and induce strong LSPR. Moreover, the plasmon decay generates a great number of hot carriers at the Schottky junction, which then improves conversion efficiency. The IPAS with extremely strong and broadband-induced LSPR is used to overcome the limit and advance the performance of the photodetector. The detection range in the measurement covers from the visible to mid-infrared (300–2700 nm). The ideal detection spectrum, according to the theoretical estimation and simulation, would be even broader.

2. MATERIALS AND METHODS

A. Device Fabrication

$\mathrm{P}$-type double-side polished Si wafers were used to fabricate the IPAS. The doping level was $\sim {10}^{15}{\mathrm{cm}}^{-3}$, and the thickness was approximately 400 μm. The piranha and buffer oxide etching (BOE) process was utilized to remove organic contaminants (such as dust particles, grease, or silica gel) on the wafers. On both sides of the Si wafer, 500 nm ${\mathrm{SiO}}_2$ was deposited (Fig. 2A) by plasma-enhanced chemical vapor deposition (PECVD). Through the photolithography process, the period of the IPAS (Fig. 2B) was defined, and 40 nm Cr was deposited by a thermal evaporation system (TES) on the patterned photoresist (Fig. 2C). The samples were dipped into acetone with ultrasonic agitation to lift off the photoresist (Fig. 2D), and were etched by reactive-ion etching (RIE) to finish the etching mask during the wet etching process (Fig. 2E). Afterward, the samples were dipped into KOH solution to form the shape of inverted pyramids by the anisotropic wet etching process (Fig. 2F). Finally, through the BOE, the residual ${\mathrm{SiO}}_2$ and Cr mask were removed (Fig. 2G) to finish the IPAS fabrication process.

 

Fig. 2. Fabrication process of Si-based IPAS substrate. (A) Si substrate with 500 nm thick ${\mathrm{SiO}}_2$. (B) Photolithography. (C) 40 nm thick Cr is deposited. (D) Photoresist liftoff. (E) RIE. (F) KOH anisotropic wet etching. (G) Residual ${\mathrm{SiO}}_2$ and Cr were removed. Color-coded layers: yellow, Cr; orange, photoresist; blue, ${\mathrm{SiO}}_2$; gray, p-Si.

Download Full Size | PDF

Organic contaminants were removed from the Si-based IPAS substrate again by the piranha and BOE process, thereby ensuring good contact between the metal and Si. Next, a 100 nm Pt Ohmic contact electrode was deposited on the polished side of the IPAS substrate by an electric evaporation system (E-gun). Finally, a 13 nm Cu nano-film was deposited by E-gun to cover the IPAS, which has a gradient metallic linewidth and Schottky junction, providing a device possessing an IPAS-type device. The planar-type device was fabricated by the same process, and the substrate was changed from a Si-based IPAS to a double polished p-Si substrate.

B. Simulation and Measurement

The simulations of plasmon behaviors were investigated using the 3D finite element method (3D-FEM) by COMSOL Multiphysics (COMSOL, Inc., Massachusetts, USA). The absorption spectrum of devices was measured by a spectrometer (JASCO V-770, JASCO Cooperation, Tokyo, Japan; measured wavelength region of 300–2700 nm) with an integral sphere (ISN-923, JASCO Cooperation), and the absorption (A) was defined as 100% deducted from the measured reflection (R; %) and transmission (T; %) of the device (A=1-R-T). A blackbody light source (tungsten lamp with filter and grating; wavelength of 1150–2700 nm) and a tunable IR-laser (Thorlabs ITC4005QCL; operating power of 1.2–5.8 mW) were used to measure the photoresponse of the devices, and the current-voltage (IV) curves were measured by a digital source meter (Keithley 2400, Tektronix, Ohio, USA). The photoresponse was defined as the difference between the photocurrent and the dark current.

3. RESULTS AND DISCUSSION

A. Design and Simulations

Previously [27–35], the linewidth and period of the conventional LSPR-induced structures are fixed. The broad bandwidth resonance and absorption then cannot be achieved. Our study shows that such limitation can be overcome by utilizing a gradient metallic linewidth. In addition, as the 2D symmetric geometry is encompassed, the polarization sensitivity can be avoided. As a result, the IPAS structure used here can satisfy the above requirements.

To theoretically simulate the reaction of plasmon behaviors between incident light and IPAS, the 3D-FEM provided by COMSOL Multiphysics is utilized. Whether IPAS with a gradient metallic linewidth can arouse LSPR in a broad spectrum region can be confirmed through the simulation. The details are described as follows. Figure 3A shows the structure set in the simulation. An IPAS with a 4 μm period is designed, and a 30 nm metal nano-film is set to cover the IPAS. Then, the structure is illuminated by a plane wave at the wavelength of 1000 nm propagating through the $-z$ direction. In the beginning, the three most common materials used in the LSPR-induced structure, Cu, Ag, and Au, are chosen for the discussion of the induced plasmon intensity. The simulated absolute ratios of amplitude of electromagnetic waves are presented in Figs. 3B and 3C. The incident light wave is trapped inside the inverted pyramid cavity and generates a strong localized electric field on the surface of the metallic thin film, as shown by the red spot at the bottom of the cavity. The IPAS based on these three metals can induce strong LSPR under 1000 nm incident radiation, and the LSPR intensities are very much similar. The strongest amplitude of the LSPR is about 10.167 times of that of the incident wave. However, because the Schottky barrier height formed by Cu/p-Si is approximately 0.52 eV, which is lower than the junction formed by Ag and Au, the smaller barrier should have better performance for mid- or far-infrared detection. Hence, in the following research, Cu is chosen as the Schottky contact metal for the Schottky photodetector.

 

Fig. 3. Plasmon-induced effect of IPAS with Cu, Au, and Ag nano-films. 3D-FEM COMSOL simulations are used to investigate the effects of structures with various metal nano-films, and the wavelength of incident radiation is set to be 1000 nm, both X-polarized and Y-polarized separately. (A) The formation of lateral modes around the IPAS. Material, from top to bottom, air, 30 nm thick nano-film, and 4 μm period Si-based IPAS. (B) X-polarized and (C) Y-polarized E-field incident wave. The incident waves are trapped and then induce LSPR. The IPASs based on these three metals exhibit similar intensity distributions inside the cavity. Metal from left to right: Cu, Au, and Ag. The dimensions of the configuration for simulation are expressed in the form of vertical and horizontal coordinates, with zero meaning the vertical and horizontal center of the configuration for simulation, in the top of (B) and the right of both (B) and (C).

Download Full Size | PDF

Figures 4A and 4B are the simulated absolute amplitude ratio of incident waves and induced SP of Si-based IPAS with Cu nano-film (Cu-IPAS), and the wavelengths include 1000, 2000, 3000, and 4000 nm. Here, the significantly localized electric fields (red spot) are generated by incident light under every wavelength, and the localized electric field is distributed along the surface of the metallic sidewall and across the cavity. The Cu-IPAS shows a superior photon-trapping effect in the cavity under every wavelength. This demonstrates that the Cu-IPAS provides a great 3D optical cavity effect. As the wavelength increases from 1000 to 4000 nm, the resonance region shifts from the bottom to the top of the IPAS. When the incident wavelength is 1000 nm, the shorter wavelength requires a relatively shorter metallic linewidth to induce LSPR. Therefore, the incident light matches the adaptive nano-linewidth on the bottom of the Cu-IPAS. When the incident wavelength is in the range of 2000–4000 nm, the longer wavelength necessitates a relatively longer metallic linewidth to induce LSPR, and thus the corresponding linewidths are on the top of the Cu-IPAS. In addition, the strength of induced LSPR decreases, while the wavelength of the incident wave increases. The times of the strongest amplitude of LSPR to the amplitude of incident waves with wavelength from 1000 to 4000 nm are about 10.167, 5.619, 3.257, and 2.036. This is because the resonance regions increase, while the wavelength of incident waves increases. The energy in a narrow region is much more concentrated than that in a wide region. Consequently, the Cu-IPAS can induce LSPR in an extreme broadband infrared spectrum region by the gradient metallic linewidth, and the wavelength region from 1000 to 4000 nm can find the corresponding resonance linewidth to excite LSPR, as shown in Fig. 4C. This structure effectively overcomes the shortcomings of previous LSPR-induced structures, which can induce LSPR only in a narrow wavelength region. Furthermore, since the Cu-IPAS possesses the characteristics of both a 2D symmetric periodic array and structure units, the device with the IPAS is polarization insensitive to the incident light. On the other hand, those phenomena seem to be similar with SPP-induced nano-focusing. However, here, the induced SP occurs only in the space between each plane of the IPAS. The SP does not appear in the apex of the IPAS. This situation is not like the SPP-induced nano-focusing [44]. Also, according to the definition of nano-focusing provided by Gramotnev and Bozhevolnyi [44], metal nano-film-covered-IPAS-induced SP is not a type of SPP-induced nano-focusing. In detail, metal nano-film-covered IPASs confine optical fields only inside the IPAS. The induced SP does not have transport phenomena. Also, the induced SP is polarization insensitive. In addition, the metal nano-film-covered IPAS does not contain plasmonic waveguides, which are used for transporting SPPs. Therefore, the metal nano-film-covered-IPAS-induced electromagnetic phenomenon is not a type of SPP-induced nano-focusing.

 

Fig. 4. Plasmon behaviors of the IPAS when illuminated by various wavelengths and polarization of waves. The incident wavelengths from left to right: 1000, 2000, 3000, and 4000 nm, which are used to evaluate the plasmon behaviors of the Cu-IPAS. (A) X-polarized E-field incident wave. (B) Y-polarized E-field incident wave. The LSPR-induced linewidth and photo-trapping region would be different under various wavelengths. The dimensions of the configuration for simulation are expressed in the form of vertical and horizontal coordinates, with zero meaning the vertical and horizontal center of the configuration for simulation, in the top of (A) and the right of both (A) and (B). (C) The resonance region (yellow square) of the Cu-IPAS shifts from center to peripheral as the incident wavelength increases from 1000 to 4000 nm. (D) The estimated absorption of the Cu-IPAS, showing that the intensities of reflection, transmission, and absorption are fairly smooth in the simulated spectrum.

Download Full Size | PDF

The simulations of Figs. 4A and 4B are used to estimate the absorption for the Cu-IPAS, as shown in Fig. 4D. In the region from 1000 to 4000 nm, the curve of reflection is smooth. Because the thickness of the metal nano-film is set as 30 nm, the transmission is nearly zero. Specifically, the transmission is $1.7\times {10}^{-4}$. In contrast, the reflection is $\sim 0.2$, and the absorption of the Cu-IPAS is $\sim 0.8$. As shown in the simulation of plasmon behaviors, the induced region of LSPR for the Cu-IPAS is different under various wavelengths, and the intensity of LSPR decreases as the wavelength increases, but the resonance regions expand in the cavity. Thus, the Cu-IPAS can arouse LSPR with similar energy in the broadband spectrum region. This leads to the Cu-IPAS being different from previous structures [27–35], and the estimated spectrum of the Cu-IPAS does not display any characteristic absorption peak but performs broad bandwidth absorption. According to the result of the estimated spectrum, the absorption ability of the Cu-IPAS does not seriously relate to wavelengths in the range of 1000–4000 nm, which is under the condition of the IPAS with 4 μm period. If the period of the IPAS increases, the absorption spectrum should extend to an even broader range.

B. Experiments and Results

Based on the above simulation, we further manufactured a large area ($2.5\mathrm{cm}\times 2.5\mathrm{cm}$) IPAS with a 4 μm period on a $\mathrm{p}$-type Si wafer (as shown in Fig. 5A) by the usual photolithography and KOH anisotropic wet etching process [10,45]. Figure 5B is the top view of the IPAS taken by a scanning electron microscope (SEM), which displays an array with squares being closely arranged. The array and structure units of the IPAS both have the characteristic of 2D symmetry. According to the SEM cross section of the IPAS (Fig. 5C), the linewidth between the sidewall of the IPAS cavities is gradient, ranging from 10 nm to 3.78 μm. Furthermore, depending on the 3D structure of the IPAS, the active area is extended to 1.74 times the planar-type device (without IPAS), and the larger active area of the device has the potential for higher photoresponse. In this research, the p-Si-based IPAS has 13 nm thick Cu (Schottky contact) deposited on the top and Pt (Ohmic contact) on the backside. A planar-type device was also fabricated and deposited with the same metals for comparison.

 

Fig. 5. Morphology of the inverted pyramid array nanostructure. (A) The area size of the IPAS substrate is $2.5\mathrm{cm}\times 2.5\mathrm{cm}$, and the clear interference fringes can be observed by the naked eye on the substrate. (B) Top view and (C) cross-section SEM images of IPAS on Si substrates. The linewidth range of the 4 μm period IPAS is embraced: $3.78\mathrm{\mu m}\ge \mathit{L}(\mathit{H})\ge 10\mathrm{n}\mathrm{m}$.

Download Full Size | PDF

To estimate the enhancement of performance from metallic IPAS for the Schottky photodetector, the planar-type and IPAS-type devices were both measured. Figures 6A–6C present the reflection, transmission, and absorption spectra of the two types of devices and the p-Si substrate for comparison. The bandgap of Si is approximately 1.12 eV, and thus the absorption cut-off wavelength of Si (${\lambda }_{\mathrm{c}-\mathrm{Si}}$) is approximately 1107 nm. The Schottky barrier height of Cu/p-Si is approximately 0.52 eV, and thus the corresponding absorption cut-off wavelength of the Schottky junction (${\lambda }_{\mathrm{c}-\mathrm{Sch}}$) is 2380 nm. The black line in Fig. 6C is the absorption of double polished p-Si wafers, which is approximately 60%–70% in the spectrum region shorter than ${\lambda }_{\mathrm{c}-\mathrm{Si}}$. In the region near ${\lambda }_{\mathrm{c}-\mathrm{Si}}$, the value declines sharply. In the wavelength region longer than ${\lambda }_{\mathrm{c}-\mathrm{Si}}$, the value is 0 due to the photon with energy lower than the Si bandgap, making it difficult for Si to absorb. Through the absorption of the planar-type device (Fig. 6C, blue line), because the Cu nano-film increases the reflection of the device, the absorption of the planar-type device is lower than that of the Si wafer by 10%–20% for wavelengths shorter than ${\lambda }_{\mathrm{c}-\mathrm{Si}}$. In the region of ${\lambda }_{\mathrm{c}-\mathrm{Si}}$ to ${\lambda }_{\mathrm{c}-\mathrm{Sch}}$, some of the photons would be absorbed by the metal and then the Schottky junction, and so the absorption of the planar-type device is augmented to 40%. In the region after ${\lambda }_{\mathrm{c}-\mathrm{Sch}}$, the absorption is still maintained at 40%. The deposited Cu nano-film is 13 nm thick, and forms a semicontinuous film on the Si wafer (Fig. 7), which can optimize the absorption of the device in the long wavelength region [46]. However, for photodetection, the absorption of the planar-type device remains insufficient. Here, the IPAS-type device has the superiority of broadband-induced LSPR, 3D optical cavity effects, and a larger active area, and thus the absorption of the IPAS-type device is higher than 80% in the region from the visible to mid-infrared (450–2700 nm). This is highly similar to the simulation results. Therefore, we expect that the IPAS-type devices would have high potential to be applied in ultra-broadband detection photodevices.

 

Fig. 6. (A) Transmission, (B) reflection, and (C) absorption (A = 1−R−T) spectra for p-Si wafer (black line), planar-type (blue line), and IPAS-type (red line) devices.

Download Full Size | PDF

 

Fig. 7. SEM morphology of a Cu semicontinuous nano-film. (A) Cross-section of Cu nano-film. The thickness is about 13 nm. (B) Top view of Cu nano-film. It can obviously be seen that the morphology of Cu nano-film is not flat. Instead, it has many nano-islands.

Download Full Size | PDF

Figure 8A presents the dark IV-measurement result of the planar-type and IPAS-type devices. Both devices exhibit the IV characteristics of the standard Schottky diode [5,42], which has a turn-on voltage lower than the typical PN junction diode in the forward bias region. The turn-on voltage is approximately 0.1 V, and the reverse current is small. In addition, because the dark IV characteristics of the planar-type and IPAS-type devices are similar, the band diagram of the Schottky diode is not influenced by the IPAS.

 

Fig. 8. (A) Dark IV curve of both IPAS-type and planar-type devices. (B) The photoresponse of both IPAS-type and planar-type devices measured at different applied voltages and various incident wavelengths.

Download Full Size | PDF

Both the planar-type and IPAS-type devices are used to measure the photoresponse under the infrared (1150–2700 nm), and the devices are operated at 0 bias and $-5\mathrm{mV}$ to measure the dark current and photocurrent [38]. The photoresponse is defined as the difference between the dark current and photocurrent, as shown in Fig. 8B. For the planar-type device, in the condition of operating at 0 bias and $-5\mathrm{mV}$, the photoresponse decreases as the incident wavelength increases. These response tendencies approximately match the quantum transmission probability curves [42]. Since the photon energy decreases as the incident wavelength increases, the response in the short wavelength is more significant than that in the long wavelength spectrum region. Furthermore, as $-5\mathrm{mV}$ is applied to the device, the response improves by 3–10 times. The detection cut-off wavelength of the planar-type device is approximately 2350 nm, converting into a Schottky barrier height of nearly 0.53 eV, which is very close to the theoretical value of 0.52 eV of the Cu/p-Si Schottky junction. This result proves that the Schottky photodetector formed by Cu/p-Si can actually detect photons with energies lower than the Si bandgap. However, the mechanism to generate the photocurrent of the planar-type device is only IPA, which lacks other enhanced mechanisms to reach high photoresponsivity. In particular, very few photogenerated electrons could surpass the Schottky barrier height, thus limiting its range of spectral response. According to the measurement results, when the device is operated at $-5\mathrm{mV}$ and illuminated by a 1550 nm laser, the best photoresponse of the planar-type device is just 187 nA/mW, as shown in Fig. 9.

 

Fig. 9. Photoresponse of planar-type devices. A 1550 nm IR-laser with 2 mW is used to measure the photocurrent of the device. (A) The IV curves of the planar-type device. Both the dark current and photocurrent are shown with stable performance. (B) The detailed IV-curve for the range across $-5\mathrm{mV}$. There is an obvious difference between dark current and photocurrent.

Download Full Size | PDF

In comparison, the measurement of the IPAS-type device displays high photoresponse in the ultra-broadband spectrum region (Fig. 8B). In the condition of operating at 0 bias and $-5\mathrm{mV}$, the photoresponses are higher than 300 and 3700 nA, respectively. This value is nearly 40 times higher than the planar-type device. The photoresponse of the IPAS-type device decreases as the wavelength increases, but the decrease is not as rapid as the planar-type device, and the detection cut-off wavelength of the IPAS-type device is not obtained in the measured spectrum. It is expected to be beyond the measurement limit of our instrument. Because the IPAS-type device possesses the superiority of the LSPR-induced structure and 3D optical cavity effects, it can effectively trap the photon energy at the Schottky junction to provide a strong electric field within the active region and generate a large number of hot carriers to improve photoresponse performance. According to the IPAS with a gradient metallic linewidth, the simulation results demonstrate that LSPR can be excited in the spectral region from 1000 to 4000 nm. Consequently, the photoresponse of the device can be optimized not only at a specific wavelength, but also in an extreme broadband spectrum region. Moreover, according to the measured results, the IPAS-type device has the capability to detect photons with energies well below the Schottky barrier (0.53 eV: 2350 nm). As shown in Fig. 10, the IPAS can induce excellent LSPR effects. Initially (step 1), the hot carriers excited by the photons with weak energy cannot pass through the Schottky barrier, but oscillate in the metal due to LSPR. Next (step 2), a great number of hot carriers are continuously excited in the metal by photons and are accumulated in the metal by step 1. Finally (step 3), the high density of the hot carriers is held, they collide with each other, and some of them obtain sufficient energy to pass through the Schottky barrier to turn into photocurrent. Thus, the IPAS-based device does not merely elevate the photoresponse under photon energies below the Si bandgap, but can also detect mid-infrared with energies lower than the Schottky barrier height.

 

Fig. 10. Mechanism of the IPAS-type devices to generate photocurrent when the incident wavelength is longer than ${\lambda }_{\mathrm{c}-\mathrm{Sch}}$. The red curves indicate the energy profile of carriers from the beginning of photoexcitation to the stage of some carriers with energy exceeding the Schottky barrier height. The blue curve represents the energy profile of the post-excited hot carriers. In other words, in step 3, the pre-excited hot carriers held on the Cu-IPAS due to LSPR collision with the post-excited hot carriers, and following, the energies of hot carriers are redistributed. Therefore, some of hot carriers will possess energy larger than the Schottky barrier height and transport to the outside of the device (top tail of red energy profile in step 3).

Download Full Size | PDF

To quantitatively obtain the photosensitivity of the IPAS-type device for input power of the infrared, the IPAS-type device is used to measure the photocurrent under a 1550 nm IR laser with various input powers, and the power range is 1.2–5.8 mW. The measured photoresponse is shown in Fig. 11. When the IPAS-type device is operated at $-5\mathrm{mV}$, the photoresponse is approximately proportional to the input laser power. The photoresponse result provides a high degree of linearity ($R^2=0.986$), and the photoresponsivity is approximately 1343 nA/mW. The good linearity demonstrates that the IPAS-type device is excellent to define infrared power.

 

Fig. 11. Photosensitivity measurement of the IPAS-type device. A 1550 nm IR-laser with tunable power is used to measure the photoresponse of the device under various input powers. For the photoresponsivity measured at $-5\mathrm{mV}$, the positive correlation coefficient ($R^2$) is approximately 0.986.

Download Full Size | PDF

4. CONCLUSION

In this paper, we successfully create a Si-based Schottky photodetector by Cu/p-Si junction, and an LSPR-induced structure is used to augment the photoresponse performance. The IPAS designed in this article possesses a 1D gradient metallic linewidth, a 2D symmetric periodic array, and 3D optical cavity effects, which can effectively optimize the capability of photon trapping and photocurrent generation. According to the simulation results, in the spectral range of 1000–4000 nm, the corresponding metallic linewidth to induce LSPR can be found in the Cu-IPAS, which has the capacity to induce robust LSPR in a superlative broadband spectrum region. The experimental absorption of the IPAS-type device is completely higher than 80% in the extreme broadband spectrum region (450–2700 nm). The IPAS overcomes the deficiency of the anterior structure, which can merely induce LSPR in a narrow or specific wavelength range. A noteworthy point of the IPAS is that, through the excellent LSPR-induced effects, it can not only improve the photoresponse under photon energies below Si bandgap, but also detect mid-infrared with energies lower than the Schottky barrier height. Depending on the photoresponse with varied wavelengths, the IPAS-type device maintains excellent value in the measured wave range of 1150–2700 nm. When 0 V and $-5\mathrm{mV}$ are applied to the IPAS-type device, the photoresponses are respectively higher than 300 and 3700 nA, and the value does not decrease abruptly as the wavelength increases. In the condition of applying $-5\mathrm{mV}$ and illuminating with a 1550 nm IR laser, the photoresponsivity can even reach 1343 nA/mW. In brief, the IPAS-based devices possess several superiorities: ultra-broadband powerful absorption, polarization insensitivity, and high photoresponsivity. The IPAS is manufactured by CMOS and a wet etching process, which involves regular instruments and materials, and a mature fabrication process. Thus, these IPAS-based devices have high potential to be integrated into Si-based chips and to develop Si-based visible/infrared photonics or infrared thermal imagers.