Microcavity-Integrated Carbon Nanotube Photodetectors Shuang Liang,† Ze Ma,† Gongtao Wu,† Nan Wei,† Le Huang,† Huixin Huang,† Huaping Liu,§,∥ Sheng Wang,*,† and Lian-Mao Peng*,†,‡ †
Key Laboratory for the Physics and Chemistry of Nanodevices, Department of Electronics, and ‡Academy for Advanced Interdisciplinary Studies, Peking University, Beijing 100871, China § Beijing National Laboratory for Condensed Matter Physics, Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China ∥ Collaborative Innovation Center of Quantum Matter, Beijing 100190, China S Supporting Information *
ABSTRACT: Carbon nanotubes (CNTs) are considered to be highly promising nanomaterials for multiwavelength, room-temperature infrared detection applications. Here, we demonstrate a single-tube diode photodetector monolithically integrated with a Fabry−Pérot microcavity. A ∼6-fold enhanced optical absorption can be achieved, because of the confined effect of the designed optical mode. Furthermore, taking advantage of Van-Hove-singularity band structures in CNTs, we open the possibility of developing chirality-specific (n,m) CNT-film-based signal detectors. Utilizing a concept of the “resonance and off-resonance” cavity, we achieved cavityintegrated chirality-sorted CNT-film detectors working at zero bias and resonance-allowed mode, for specific target signal detection. The detectors exhibited a higher suppression ratio until a power density of 0.07 W cm−2 and photocurrent of 5 pA, and the spectral full width at half-maximum is ∼33 nm at a signal wavelength of 1200 nm. Further, with multiple array detectors aiming at different target signals integrated on a chip, a multiwavelength signal detector system can be expected to have applications in the fields of monitoring, biosensing, color imaging, signal capture, and on-chip or space information transfers. The approach can also bring other nanomaterials into on-chip or information optoelectronics, regardless of the available doping polarity. KEYWORDS: photodetector, carbon nanotube, microcavity, Schottky barrier, photocurrent − e−αd for α ≈ 2 × 105 cm−1 and d ≈ 1.5 nm). Besides, for signal sensing or information imaging applications, which involve the detection of a specific target signal among a broad background of signals or noise, the photoresponse of CNT aggregations (chirality-mixed arrays or films) usually covers a wide near-infrared scope with various characteristic peaks induced by (n,m) chirality-dependent Van-Hove-singularity (VHs) transitions,8 leading to random responses without any targets. That is to say, a specific response to a target signal requires a higher suppression ratio (SR). Despite all this, CNTs still exhibit some strong advantages for applications in infrared detectors in comparison with other candidate materials, e.g., the naturally nanoscale size, simplified growth and device fabrication, chirality-dependent direct band structures that cover important communication wavebands, and room-temperature operation due to their moderate band-gap energies (≫26 meV).8−12 The very small size and good extendibility also
D
riven by emerging semiconductor nanomaterials and modern optoelectronics technology, the amount of scientific research on photodetectors is constantly increasing due to their extensive and significant applications in military and civil use, such as night vision, imaging, sensing, and even information communication systems.1−4 Carbon nanotubes (CNTs) from rolled graphene sheets are hollow cylinders with typical diameters (d) of ∼0.5−2 nm, which have shown significant potential for use in next-generation high-performance nanoelectronics.5 However, for CNT-based near- or shortwave-infrared optoelectronics, the existing developments for enhanced absorption or signal detection are insufficient. At technologically important wavelengths (e.g., ∼1200−1700 nm), although the optical absorption coefficient (α) is nominally 1 or 2 orders of magnitude larger than that of conventional semiconductor materials (such as Ge, HgCdTe, and InxGa1−xAs),2,6,7 the effective light absorption is indeed impeded due to the smaller cross-section diameter of a single tube and the shorter interaction distance with passing photons (an absorption of ∼4% of the incident photons can be simply estimated for a single layer of dense CNTs, with the formula 1 © 2016 American Chemical Society
Received: May 1, 2016 Accepted: July 5, 2016 Published: July 5, 2016 6963
DOI: 10.1021/acsnano.6b02898 ACS Nano 2016, 10, 6963−6971
Article
www.acsnano.org
Article
ACS Nano provide great prospects for their flexible integrations into other photonic or hybrid optoelectronic systems.9−12 Unlike conventional doping processes used in bulk semiconductors, a built-in electrostatic field can be easily established inside a single-CNT channel region through contact engineering at the source− drain ends.13 To further increase the signal outputs, the hybrid systems that enhance light absorption or exciton dissociation in the interfaces can also be considered.14−16 Compared with the fast progress in the fields of two-dimensional (2-D) materials (e.g., graphene on waveguides or with plasmonics nanostructures),17−19 some concepts for enriching CNT photodetectors should be further investigated and developed. As a complementary part of highly integrated CNT-based electronics,5 optoelectronics will introduce more functional integrations on a chip, and the search for strong photoelectric conversion or signal detection recognition is very important for the potential development of large-scale 2-D array detectors, on-chip (opto)electronic unity, signal capture, imaging or sensing, and even information communication systems. The naturally small dimensions of CNTs also ensure its flexible insertion into other photonic systems.9−16 Microcavityintegrated optoelectronics evolved from conventional III−V or IV semiconductors, for the expected optical gains and selectable mode outputs.20 Until recently, an extension of this concept into low-dimensional or few-layer nanomaterial families has been promoted and pursued. For nanoscale materials (including the CNT), optimizing the design of optoelectronic devices with a microcavity or other optical mode couplings will have obvious significance, because of the specific wavelength response and enhanced performance, especially for future onchip information (opto)electronics and nanophotonic circuits.21−27 However, there have yet to be any reports on making and analyzing cavity-integrated CNT photodetectors for enhanced photoelectric conversion or high-performance signal detection. Besides, although many emerging nanomaterials and nanofabrication methods developed for photodetectors have been reported, research aiming at recognizing specific signals is still lacking. In this study, we realized the monolithic integration between a single-tube diode detector and a metal-end-face Fabry-Pérot (F−P) cavity, and improvements in photocurrent (PC) or signal recognition (the distinct response to a specific signal) have been confirmed, respectively. We show the importance of a high SR for high-performance target signal detection. With a deepening of the research, the chirality-sorted CNT films were integrated into a designed cavity, consisting of resonant and off-resonant parts. In the resonant part, the chirality-specific VHs-defined absorption energy of the CNTs overlaps with the designed cavity-allowed optical mode, whereas the off-resonant part obstructs all incident excitation and detunes the resonance mode, forming an artificially redistributed optical field within the optical center of the cavity and leading to asymmetric excitation around the Schottky contacts. Overall, in this study, the mechanism related with cavity-integrated devices will be first elaborated using a single-CNT system, and ∼6-fold enhanced photoelectric conversion efficiency (compared with the result from the free-space device) has been achieved. Furthermore, a cavity configuration for the signal detection has been demonstrated. With the advantages of better signal detection recognition, zero bias operation mode, potentially better dynamic response, and the applicability to other optoelectronic nanomaterials, this concept will enrich information optoelectronics.
RESULTS AND DISCUSSION The typical device architecture and an illustration of the standing-wave field distribution are shown schematically in Figure 1a and b. A Si wafer was used as the supporting
Figure 1. (a) Schematic representation of a single-tube diode-type photodetector integrated with a planar λ/2 F−P microcavity. The Ag and Au metal layers are the bottom and the top mirrors, respectively. The source and drain electrodes are Pd (25 nm) and Sc (25 nm), respectively. (b) The λ/2 photonic mode profile, showing a geometrical overlap between the single tube and the center of the optical mode. The single CNT is magnified to illustrate the detailed device configuration. (c) Typical I−V curve showing the diode-type rectification behavior (labeled by navy symbols) and the corresponding fitting curve (red). Inset: Symmetric transfer curve of the diode measured at Vds = 0.1 V.
structure, and the Ag and Au layers are bottom and top mirrors, respectively. Instead of the quarter-wave-stack configuration (in which multilayers of many pairs of alternating refractive index are used to form Bragg mirrors and each region has an optical thickness of λ/4 with a high interfacial smoothness), metal-endface mirrors were adopted in the present wavebands, and a ∼99% reflectivity of the Ag mirror can meet the actual demand (see Figure S1 in the Supporting Information). The standard boundary conditions for the reflection of light suggest that the optical field is nearly zero inside the metal mirrors (exponential decay); thus, the modal extent in a metal cavity can be very small. That is to say, a high optical confinement can be achieved through relatively thin metal mirrors, especially for wavelengths farther into the infrared. The compact cavity configuration, small modal extent, and simplified deposition fabrication provide large flexibility for better optical confinements in nanoscale systems. Integrated with the cavity, the single-CNT and chirality-sorted CNT-film devices were used to fulfill the respective goals. For the single-tube diode-type detector, the CNTs were grown by chemical vapor deposition (CVD) at 950 °C using a CuCl2 solution catalyst and were transferred onto the HfO2 substrate by a poly(methyl methacrylate) (PMMA)mediated method, following HF acid etching.28 The semiconducting CNTs were electrically selected by a high on−off ratio (>103), which was measured using the back-gated field6964
DOI: 10.1021/acsnano.6b02898 ACS Nano 2016, 10, 6963−6971
Article
ACS Nano
Figure 2. (a) Dependence of the short-circuit current (Isc) and the open-circuit voltage (Voc) on the excitation power density, showing the linearly varying Isc and the logarithmically varying Voc. (b) Dependence of Voc on Isc/Is, derived from the diode rectification equation. Inset: Light-polarization-dependent Isc, which follows a cos2 θ trend (Imax/Imin = 3.5). (c) Equivalent circuit for the photoresponse measurement. (d) Lch scaling from 0.4 to 8 μm performed on the same single CNT; the PC was obtained with a power density of ∼104 W cm−2 and a 3-μmdiameter excitation spot. The data were normalized by the maximum PC from the 1-μm-channel device. Region I represents the short-channel device (Lch < 2 μm). For the long-channel device (Lch > 3 μm), an obvious position-dependent response can be observed. Region II represents the excitation location on the contact, and region III represents the excitation location in the center of the channel region.
effect modulation. The built-in p−n electrostatic potential fields for separating electron−hole pairs (or excitons) were created with asymmetric source/drain contacts (Pd [25 nm]/Sc [25 nm]). With further integration of the device with the designed cavity, the enhanced performances are analyzed. As further research, the chirality-sorted semiconducting CNT films were adopted as the active materials, which were grown by highpressure catalytic carbon monoxide (HiPco), separated by gel chromatography, and deposited on the insulating Al2O3 layer by the dip-coating method, with an average length of ∼1.2 μm. With the chirality-sorted film and symmetric Ti/Au (0.5 nm/25 nm) contacts, a device configuration of the resonance and offresonance cavity will be introduced and discussed in the following. For all devices, reactive ion etching (RIE) was used to remove redundant CNTs (for details, see Methods and Figures S2, S4 in the Supporting Information). Until now, the response mechanisms of CNT-based photodetectors have been classified into certain categories, namely, photoconductive,29 photogating,30 and bolometric models.31 To investigate the optoelectronic characteristics of the present detectors in free space (without the top Au mirror), the small source−drain bias Vds (−0.5−0.5 V) and gate voltage Vg = 0 V (off-state) were adopted. Figure 1c displays the I−V curve obtained using Pd/Sc contacts, channel length Lch = 2 μm, and contact length Lc = 1 μm. The obvious rectification behaviors revealed as the logarithm and symmetrical transfer characteristics versus gate bias (V-shaped conductance) can be observed (Figure 1c). This implies favorable band-gap alignments; that is to say, the Pd and the Sc electrodes can align with the valence and conduction bands of the CNTs, respectively.13 Using the fitting curve (red) obtained from the
⎡ ⎤ q(V − IR s) equation I = Is⎢exp − 1⎥, the reverse saturation kBT ⎣ ⎦ current is determined as Is ≈ 1.1 × 10−11 A and the series resistance as Rs ≈ 142 kΩ (kBT ≈ 26 meV). Upon optical excitation, the built-in electrostatic field is expected to separate electron−hole pairs, producing a short-circuit current (Isc) and an open-circuit voltage (Voc). For the photoresponse measurement, the ∼3-μm-diameter excitation spot (at an excitation wavelength λe of 1400 nm) was focused on the devices. Figure 2a shows that with increased power density (normalized by a factor of 2 due to the reflection of the Ag mirror), a linearly varying short-circuit current (Isc) and a logarithmically varying open-circuit voltage (Voc) can be obtained. The logarithmic dependence of Voc can be explained by the diode-type
(
rectification relation32 Voc ∝
)
(
I kBT ln Isc q s
)
+ 1 , confirming the
photovoltaic operation mechanism (Figure 2b). The low photoresponsivity of ∼2.3 μA/W can be mainly attributed to the smaller d (0.5−2 nm) of the single tube and the excitation condition (e.g., λe or tube chirality [n,m]). The inset shows the light-polarization-dependent Isc fitted by the cos2 θ curve (θ is the polarization angle of the incident light with reference to the tube axis), reflecting a ratio of ∼1:3.5, which is usually defined by the confined electronic states in the one-dimensional CNT. Metal-contacts-induced charge transfers in nanomaterials can play an important role in defining the optoelectronic properties of the nanodevices.33−37 Thus, the Lch scaling of devices from ∼0.4 to 8 μm was performed on the same single CNT. The PC from the localized excitation (on the contacts and center regions of channels) was investigated using a lock-in system with a 3-μm-diameter excitation spot, and the equivalent circuit 6965
DOI: 10.1021/acsnano.6b02898 ACS Nano 2016, 10, 6963−6971
Article
ACS Nano
Figure 3. (a) Photoresponse of the single-tube device in free space (black dashed line) and integrated with the cavity (red line). The extra peak at 1400 nm confirms the ∼6-fold-enhanced photoelectric conversion efficiency. Inset: Simulated E-field amplitude at the TM excitation mode, confirming the cavity-induced enhancement effect. (b) Photoresponse of integrated devices with different TAu, showing the improved SR obtained with a 30-nm-thick top Au mirror (purple). The gray-shadowed frame indicates the cavity-related characteristic responses. The dotted frame shows the response peak from the CNT, which affects the target signal recognition. (c) Basically linear dependence of the response of the single-CNT diode detector on the power density, until a power density of 3 W cm−2 and a PC of 3.5 pA.
to 14 nA). Obviously, compared with the intrinsic response of the device in free space (black dashed line), the peak profile (including the position, fwhm, and intensity) confirms the cavity-related confinement effect. A vivid interpretation of the result is that photons with a specific energy decided by the resonance mode will be allowed into and stay within the designed cavity and pass multiple times through active materials at the optical center. The processes spatially enhance the interaction between photons and CNTs and also increase energy conversion efficiency within a designed waveband. The simulated result of E-field distribution in the inset of Figure 3a (with the TM polarization and complex refractive index nCNT ≈ 3.0 + 1.5i) depicts a ∼2.5-fold enhancement as compared with results from free-space devices, which can be attributed to the resonant input and trapping processes of incident photons. For the TE polarization, see Figure S3b in the Supporting Information. The spectral profile of the enhanced peak is affected by the response of CNTs and the cavity confinement effect, showing a slight asymmetry toward the high-energy (short-wavelength) side. With changed incident polarizations of the light, we confirmed that for the present configuration the enhancement is isotropic (that is to say, the polarization absorption of the intrinsic single CNT can be retained). For further adjustment of TAu, the enhancement effect will be reduced. In fact, for a certain CNT active layer (αT, where α is the absorption coefficient and T is the thickness of the active materials), an appropriate resonant feedback mechanism arising from the counter mirrors is required to enhance the energy distribution within the cavity, and a smaller or larger Au reflectivity (RAu) will lead to a very loss or a fully reflective cavity, thus affecting the enhancement ratio. However, the optical design for the high mode density confined within thinner active layers or the heterointegrations should be further pursued to enhance the interactions between photons and nanomaterials. The recognition or selection of a specific signal from the information channels (or among broad background signals) is also very important, for actual applications in information fields, such as monitoring, biosensing, color imaging, information transfers and recognition, and (opto)electronic unity on a chip.19,38−40 Conventionally, photodetector performances suffer from two very fundamental limitations. One is the internal noise (such as shot, flicker, and thermal noises), which is restricted by defects in materials, thermal fluctuations, or device configurations. While the noises are suppressed with
is displayed in Figure 2c. As illustrated in Figure 2d, the normalized data indicate that the shorter Lch values (Lch < 2 μm, region I) are more conducive to the larger response, which can be attributed to the high-efficiency carrier extraction achieved under a sharper voltage drop. For devices with Lch < 1 μm, a slightly reduced PC can be observed, which can be mainly attributed to the smaller Lch (that is, a smaller active area) or the light scattering by the metal electrodes. For longchannel devices (e.g., Lch = 5 μm), an excitation-positiondependent effect can be clearly observed. When the excitation is focused in the vicinity of the Pd or Sc contact (region II), a relatively strong PC can be obtained (∼70−80% of the maximum); however, when the excitation occurs in the central regions of the channels (region III), the response is reduced to ∼40%. This difference can become more obvious with increased Lch, which is associated with the high-efficiency separation and transport of electrons and holes. During the measurement, the unidirectional PC achieved at zero external bias can be kept, regardless of the excitation position. The varied light absorption area (excess carrier density distribution) due to Lch scaling, the potential defects, or carrier traps in the CNT can affect the dependence of PC on Lch. The optimal Lch should be 1−2 μm to avoid numerous recombinations within analogous neutral regions owing to the greater length of the channels. The parameters Lch = 2 μm and Lc = 1 μm are adopted in the following studies. Let us now turn to the performances of the cavity-integrated devices. It is well known that a microcavity can create an optical confinement effect, leading to the in-depth reflection dip corresponding to the optically allowed mode mainly defined by the cavity length.20−23 Ideally, for wavelength-scanning excitation, if the resonance occurs at a mode wavelength λc, the light will be trapped within the cavity, achieving a lower reflection; otherwise, the cavity can be regarded as a high reflector and rejects the incident light (see Figure S1). Overall, the allowed entrance of more photons into a designed cavity, the prolonged interaction time between photons and active materials, and the designed resonance mode are expected to result in improvements in device performances (including enhanced absorption and signal detection recognition). The red line in Figure 3a corresponds to the measured spectral response from the cavity-integrated device, with an Au thickness (TAu) of ∼7−8 nm. The result exhibits an extra enhanced peak at λc ≈ 1400 nm with a full width at half-maximum (fwhm) of ∼140 nm, and a ∼6-fold enhanced PC can be observed (from 2.3 nA 6966
DOI: 10.1021/acsnano.6b02898 ACS Nano 2016, 10, 6963−6971
Article
ACS Nano
Figure 4. (a) Schematic of the device based on the “resonance and off-resonance” cavity configuration. The chirality-sorted (8,3) and (8,4) films were used as active materials. (b) Tilted SEM image of a real array device, with its different regions labeled. The resonant region has a mode (signal) wavelength of ∼1200 nm that overlaps with the resonant absorption of the (8,4) CNT films. (c) Simulated optical field distribution within the cavity-integrated film device, demonstrating the redistribution of the optical fields owing to the lateral detuning effect by the nonresonant parts. Depending on the Schottky barriers and localized excitation, the device can operate at zero bias. The green dashed line outlines the optically active region, and the orange dotted line indicates the Schottky-barrier-induced voltage drop that separates electron−hole pairs. The Ti/Au contacts are colored to depict their different roles in obtaining the unidirectional PC.
CNTs would severely affect signal recognition.8 Thus, an optimization strategy, adopting the chirality-specific (n,m) CNTs with the VHs transition energy that overlaps with the optically allowed mode of the designed cavity, should be adopted. After finishing the above detectors based on the single-tube system, the chirality-specific (n,m) CNTs films were integrated with a designed cavity for the specific target detection under optical-mode-allowed resonance excitation. The advantages of this concept consist in a concentrated and consistent photoresponse from the same chirality CNTs, enhanced signal recognition ability, a narrower fwhm, and the rejection of interferences from other CNTs. In the future, the incorporation of multiple cavity-integrated array layouts aiming at different targets on a chip with a stronger SR can considerably promote the use of CNT-film-based detectors in broadband information sensing applications. Although an energy overlap is required between the intrinsic-VHs-defined absorption of the chiralityspecific (n,m) CNTs and the allowed mode of the designed cavity, some remaining difficulties must be addressed. For example, the fabrication of high-quality and stable CNT-filmbased homogeneous diodes is still largely unsatisfactory, and the growth of chirality-specific (n,m) CNT arrays is also uncontrollable. These obstacles make it impossible (or very difficult) to directly fabricate homogeneous diode detectors with a process similar to that described above for single-tube diodes. To overcome this difficulty, we propose the concept of a “resonance and off-resonance” cavity, utilizing chirality-sorted (8,3) and (8,4) semiconducting CNT films. In this configuration (Figure 4a), symmetric electrodes (0.5 nm Ti/ 25 nm Au) were adopted to make electrical contacts with ∼10-
some techniques, such as refrigeration or modest band-gap engineering, external factors may become sufficiently prominent to disturb the detection processes. In principle, the detector cannot show a distinct response to the weaker target signal, because the surrounding radiation (or nontarget signals) and noise can also randomly excite electron−hole pairs in the active materials and trigger the response. Therefore, for a clear recognition, a higher SR (SR = Sc/Snc, where Sc is the target response at the signal wavelength λc and Snc is the response to the noise or other undesired signals at the wavelength λnc) is expected to highlight the target response. With TAu = 30 nm (Figure 3b), for which a high quality factor (Q-factor) can be obtained, the SR can be significantly improved and the fwhm can be reduced in comparison with the results of the cavity with TAu = 8 nm. For example, S1400/S1760 ≈ 4.9 for TAu = 8 nm, and the ratio can be increased to ∼50 for TAu = 30 nm. For other wavelengths (e.g., λnc = 1716 nm), the SR can surpass 2 orders of magnitude (>200). Note that in the CNT the VHs-like energy band structure can lead to a strong photoresponse peak from the resonance excitation, with ∼5−10-fold enhanced intensity; thus, if λc resonates with the band energy of the single CNT, the SR is expected to approach 3 orders of magnitude. The selection of the 30-nm-thick Au mirror is a compromise between a high Q-factor and sufficient incident energy to enhance the interactions between the target photons and active materials. In the single-tube configuration, a clear response can be confirmed until the power density of ∼3 W cm−2 (Figure 3c). To further improve the ability of signal detection, the CNT films or arrays can be considered. However, for CNT aggregates (chirality-mixed films or arrays), discrete and random transition peaks from various types of uncertain 6967
DOI: 10.1021/acsnano.6b02898 ACS Nano 2016, 10, 6963−6971
Article
ACS Nano
Figure 5. (a) Top panel: Spectral response from (8,3) and (8,4) CNTs films in free space, with local excitation on the one contact. Bottom panel: Spectral response from cavity-integrated devices under global illumination, with a fwhm of ∼33 nm. The target signal (or mode wavelength) is located at 1200 nm. (b) Spatial PC mapping over the cavity-integrated array devices with a Pe of 1 mW; the resonant region is labeled by the gray dashed frame. The maximum PC of ∼90 nA occurs at the center of the optically active region due to the 3-μm-diameter excitation spot. The resonant and nonresonant regions lead to the obviously different PC distributions. (c) Dynamic response of integrated devices, showing the rise time τr and the fall time τf (∼45 μs) that are restricted by the available maximum f m (4k Hz). (d) Basically linear dependence of the PC on the excitation power until a power density of 0.07 W cm−2 and a PC of ∼5 pA. Inset: Invariant response with increasing f m up to 4k Hz, implying the possible broad frequency bandwidth.
of the intensity of the local optical fields compared with that of the incident light. Depending on the Schottky-barrier-induced voltage drop, the electron−hole pairs created by resonant excitation in subdevice regions will be separated and gained by opposite electrodes, forming a stable, unidirectional, and wavelength-defined PC at zero bias (Figure 4c). Unlike in the waveguide-integrated graphene detector,41 the detection here can actively aim at a specific target signal, regardless of the incidence polarizations of the excitation light. Besides, the device fabrication process is compatible with complementary metal-oxide-semiconductor (CMOS) techniques. The concept enriches cavity-related optoelectronic devices and can also be extended to other few-layer nanomaterials. This operation avoids the potential damages of the space dielectric layers caused by a large bias in usual hybrid phototransistors and also prevents relatively slow response speeds (in phototransistor detectors) that are induced by hybrid barriers within channels and provides a higher on-chip compatibility with other functions owing to the common Schottky barriers. Figure 5a shows the spectral response of the cavity-integrated array device, and the free-space response from the chiralitysorted (8,3) and (8,4) semiconducting CNT films is shown in the top panel as a comparison (also see Figure S4a,c). In the bottom panel, a single response peak from the cavity-integrated device at 1200 nm can be observed, with a narrowed fwhm of ∼33 nm. For λnc > 1300 nm, Snc can quickly approach zero (thus, the SR is infinite owing to the suppressed absorption near the band tails of the [8,4] CNTs). For λnc < 1100 nm, the relatively low SR (e.g., S1200/S1000 ≈ 50) results from the existence of the (8,3) CNTs, which produce strong band transitions in the range 900−1050 nm. Without the interference, the SR can be expected to approach 3 orders of
nm-thick CNT films, with Lch = 2 μm and Lc = 1 μm. The bottom-half space of the cavity contains the stacks of Ag (120 nm)/SiO2 (100 nm)/Al2O3 (40 nm). In the top-half space, the stacking of HfO2 (140 nm)/Au (30 nm) layers is designed as the resonant part and forms the optically active regions, while the stacking of HfO2 (90 nm)/Au (80 nm) layers is designed as the nonresonant part and obstructs incident excitations. The two parts divide Lch equally to further redistribute optical fields at the optical center (for detailed fabrications and materials, see Figures S2 and S4). For the present film device, we did not consider the photothermoelectric (PTE) effect as the dominant photoresponse mechanism, because of the obvious diffusionand recombination-related response behaviors (Figure S5). Figure 4b shows a tilted scanning electron microscope (SEM) imaging of the corresponding array device, with its different regions labeled. In this case, the incoming photons with a specific energy (∼1.03 eV/1200 nm) defined by the resonant (optically active) parts (outlined by the white frame) can be well allowed into the cavity and resonantly excite the (8,4) CNTs around the designed Schottky contacts within the confined regions. In contrast, the nonresonant parts (labeled by the yellow color) will disable the excitation due to tremendous optical losses. The further lateral confinement of the optical fields is enabled via the detuning (nonresonance) effect and by the designed metal blocks, which create a very small modal extent and bring great flexibility in the nanoscale design. Thus, a specific resonance-allowed localized mode is created for the target signal detection, and the E-field amplitude distribution (at the TM polarization) is illustrated in Figure 4c (for the TE polarization, see Figure S3c). Unlike the usual (conventional) F−P cavity, the further horizontal confinement within the cavity (with TAu = 30 nm) also induces a >4-fold enhancement 6968
DOI: 10.1021/acsnano.6b02898 ACS Nano 2016, 10, 6963−6971
Article
ACS Nano
approach also broadens the potential for the application of other nanomaterials in information photonics, regardless of their doping polarity. The concept offers a promising and practical pathway for the development of stable, large-scale, multiwavelength signal capture, recognition, and sensing systems.
magnitude due to the reduced (or nonresonant) absorptions from the (8,4) CNTs. The spatial PC mapping (obtained with a 3-μm-diameter excitation spot and an excitation power of 1 mW) over the region of a corresponding unit is displayed in Figure 5b, showing a peak value of ∼90 nA that corresponds to the resonant part of the cavity (indicated by the dotted frame). The result confirms the response distribution and potentially close assembly into a 2-D array detector system, which is consistent with the result of optical simulation (Figure S3d). The dynamic response exhibits a short response time (τ), and the rise time τr ≈ the fall time τf ≈ 45 μs as the 10−90% rule; however, the observation is restricted by the used chopper (Figure 5c). In fact, for the present cavity, the prolonged photon lifetime (Lp) within the cavity is expected to be dozens of femtoseconds (see Figure S1); thus, the integration cannot affect the response speed of devices. Additionally, the ultrafast carrier transport driven by the Schottky barriers through channels of several micrometers in length also provides the foundation for a fast dynamic response in high-speed information detection and video imaging applications,42 with a better dynamic response than that of usual phototransistor detectors. Figure 5d shows the basically linear dependence of the PC on the excitation power density, which indicates favorable conditions for array integration. A single response peak can be obtained until 0.07 W cm−2 with ∼5 pA, and further coupling with CNT-based light-emitting devices can be expected to achieve nanophotonic information circuits. The inset implies the invariable response as increasing f m up to 4 kHz, and a better dynamic performance (including broadfrequency bandwidth and shorter τ) can therefore be expected. For multiwavelength detectors on a chip, the chip should be divided into different regions for detection of different wavelengths. In each region, one kind of cavity (for one kind of wavelength detection) will be fabricated, with the corresponding single-chirality CNT film. Moreover, some common parameters (including the bottom and top mirrors, insulating Al2O3 layer, and deposition of the electrodes and optical structures in active regions) can be designed and simultaneously fabricated to simplify the processes. Furthermore, in the construction of the resonance and off-resonance cavity, the large difference of absorption in different parts of the device under uniform illumination indeed can be created. Thus, we also expect that the design may be considered in PTE-type photodetectors. However, more improved possibilities can be considered through Lch/Lc scaling, using an absolute singlechirality film, and localized plasmonic enhancement within the resonant regions. The present concept enriches optoelectronic devices and also provides a broad possibility for other unipolar nanomaterials in applications of signal sensing and analysis and imaging, on-chip optoelectronic unity, and information interconnects.
METHODS Device Fabrication. The Si wafer with a 300-nm-thick thermal oxide SiO2 was used as a supporting structure. The nanofabrication of devices was performed with a Raith 150 EBL system, followed by thinlayer deposition and lift-off processes. The Ag, SiO2, and Au layers and also the Pd/Sc electrical contacts were deposited with an e-beam evaporator. The HfO2 and Al2O3 layers were grown by atomic layer deposition at 90 °C. For the cavity-integrated single-CNT device, from bottom to top, the Ag layer (120 nm) and SiO2 layer (145 nm) were deposited in order. Subsequently, an insulating HfO2 layer (40 nm) was grown as the device substrate. The oxide layers, SiO2 and HfO2, were the lower-half space of the cavity. A spin-coating (∼185-nmthick) PMMA layer with a low surface roughness (
