Thermal Device Design for a Carbon Nanotube Terahertz Camera.

Terahertz (THz) wave detectors are increasingly expected to serve as key components of powerful nondestructive and noncontact inspection tools in a large variety of fields. In contrast to conventional THz detectors based on rigid solid materials, we previously developed an uncooled and bendable THz camera based on the THz-induced photothermoelectric effect of carbon nanotube (CNT) array devices and demonstrated omnidirectional THz imaging of three-dimensional curved samples. Although this development opened a pathway to flexible THz electronics, the physical parameters that determine the performance of the CNT THz camera have not been fully investigated. As a result, the thermal device design has not been optimized in terms of the camera sensitivity and spatial resolution. In this work, we studied the underlying mechanism of the THz-induced photothermoelectric effect of the CNT camera and found physical factors related to the detector performance. Through simulation and experiments, we observed that the detection sensitivity and response time strongly depend on the CNT channel width and film thickness. We further identified that the irradiated wave penetration into the CNT film through the electrode materials deteriorates the detection area, which is directly linked to the camera spatial resolution. By utilizing the improved CNT device design fabricated based on these findings, we eliminated undesired signals generated via thermal diffusion and THz wave penetration and achieved higher-sensitivity THz detection and higher imaging resolution compared to our previously reported THz camera. The presented technologies are expected to contribute to future flexible THz imaging applications and will also be applicable to other types of photothermoelectric devices.

Introduction

Visualization is one of the most powerful measurement schemes and is applicable to a large variety of fields, including medical care, security screening, and materials and biological research.[1−6] Along with ultrasound scan,[7,8] infrared spectroscopy,[9,10] X-ray examination,[11,12] and so forth, the terahertz (THz) imaging technology has attracted increasingly considerable attention because of its ability to penetrate nonpolarized objects and because of the fingerprint spectra of many materials and molecules lying in this frequency region.[13−15] As a nondestructive and noncontact inspection tool, THz imaging can thus be powerfully employed for the characterization of organic/inorganic materials, pharmaceutical quality control, and agricultural, medical, and biological examinations. To practically utilize THz imaging techniques for industrial and medical applications, high-speed (real-time) visualization of objects is inevitably required. This issue can be resolved by arraying multiple THz detectors in a one- or two-dimensional configuration. In contrast to technically mature frequency regions, such as visible light and microwave, the development of cameras in the THz region remains a challenge; however, several types of THz camera devices have been reported.[16,17] In particular, one difficulty of existing THz cameras is their substantially rigid unbendable structure, which primarily limited their use to flat samples; however, surfaces of most real samples have three-dimensional curvatures. This restricts the adaptable range of THz imaging measurements.

The high thermoelectric properties of carbon nanotube (CNT) films are expected to lead to high-performance thermal THz detectors because of their large Seebeck coefficient, high electrical conductivity, mechanical strength, and high absorption ratios over a broad frequency range from the subTHz to ultraviolet regions. By utilizing macroscopically bendable CNT films,[18−21] we recently developed flexible THz cameras based on multiarray CNT devices and demonstrated 360° multiview THz imaging of bent samples, such as syringes and PET bottles.[22] This unique THz camera has thus eliminated the need for bulky components required for three-dimensional surface imaging, such as THz tomography techniques. Although our technology potentially expands the applications of THz imaging, the physical parameters that govern the detection mechanism have not been clarified in detail, and the sensitivity and the spatial resolution of the CNT THz camera can be much improved in terms of the thermal device design.

In this work, through simulations and experiments, we examined the underlying mechanism behind the THz-induced photothermoelectric effect and determined parameters that primarily govern the detector performance, such as the sensitivity and response speed. We also observed that the irradiated wave penetration through electrode materials affects the spatial resolution of the CNT THz camera and accordingly suggested an improved device design for sufficiently suppressing wave penetration. On the basis of all these findings, we fabricated a refined structure of the CNT THz camera and enhanced the sensitivity and spatial resolution. The presented thermal design is a general feature of the photothermoelectric properties of the CNT film and, along with the THz camera, can be utilized in a variety of CNT-based photothermoelectric applications, such as optically activated energy harvesters.

Results and Discussion

We employed CNT films formed through a filtration process (Figure a). A CNT solution was deposited on a cellulose membrane filter and was vacuumed until the CNT solution was fully filtered (typically over several hours). Then, the film was dried, and the CNT film was picked up by using tweezers. As shown in Figure a,b, the CNT film can be easily bent and almost totally absorbs the THz wave over 99%. The mechanical strength and high absorption from subTHz to several THz frequency region indicates that the CNT films are preferable for a broadband and flexible THz detector. Figure a schematically provides a general and simple explanation of the photothermoelectric effect. When the THz wave is irradiated onto the CNT detector, the THz absorption by the CNT film causes local heating. Along with the resulting temperature gradient, the carriers diffuse in the CNT film, and the voltage, as a THz-detected signal, is generated, which is described by the following equationHere, V is the photoinduced voltage, S is the Seebeck coefficient of the material, and ΔT is the temperature gradient of the material. On the basis of the detection mechanism of the photothermoelectric effect, THz illumination caused a voltage change (Figure b), and the THz-induced voltage reached a maximum at the two interfaces (Figure c). Note that the polarity of the THz signal was opposite between the source- and drain-electrode sides, which is characteristic of the photothermoelectric effect. Although not shown here, we experimentally confirmed that the CNT device responded to broadband THz waves from the subTHz to 39 THz frequency region.[22]

Figure 1

(a) Schematic illustration of the filtration process. A monodispersed CNT solution was deposited on a cellulose membrane filter and was vacuumed until the CNT solution was fully filtered (typically during several hours). The film was then dried and finally the CNT film was picked up using tweezers. (b) THz absorption spectrum taken by THz time-domain spectroscopy.

Figure 2

(a) Schematic illustration of the general mechanism of the photothermoelectric effect. Along with thermal gradient generated by THz irradiation, the carriers diffused and the voltage (THz response) was generated. (b) THz response map taken by scanning the THz laser spot over the CNT device. According to the expression of the photothermoelectric effect, as shown in (a), the highest THz responses were generated at the interfaces between the CNT film and the source/drain electrodes, and the polarities of the THz response signal were opposite between the source and drain electrodes. (c) Current vs voltage with and without THz irradiation at 1.4 THz.

To gain a deeper understanding of the photothermoelectric effect, we elaborated eq according to the actual model by considering the Seebeck coefficient of the metal–CNT composite material Stotal (Figure a). When the heat flows parallel through the metal and CNT, as shown in Figure a, the Seebeck coefficient of the metal–CNT composite material, Stotal, is expressed aswhere σmetal and σCNT are the electrical conductivities and tmetal and tCNT are the thicknesses of the metal and CNT film, respectively. Equation indicates that the value of Stotal is determined by the ratio of the CNT thickness and the conductivity of the electrode metal and the CNT film when the THz-induced heat flows parallel through the metal and the CNT film. From Figure a, we obtained the following equationwhere SCNT and Stotal are the Seebeck coefficients of the CNT film and the metal–CNT composite material, respectively and TMax, TA, and TB are the temperatures at the interface of the metal and CNT film, at the edge of the metal, and at the CNT film, respectively. When Smetal < Stotal < SCNT, an effective approach to achieve a high photoinduced voltage V (detection sensitivity) is to increase TA and to decrease TB. We previously reported a sensitivity enhancement through this approach by optimizing the thicknesses of the electrode metals and employing a high thermal conductivity metal for higher TA and a low thermal conductivity metal for lower TB.[22]

Figure 3

Thermal analysis of the CNT THz detector through simulations and experiments. (a) Actual model by considering the Seebeck coefficient of the metal–CNT composite material. (b) Schematic image of the thermal conduction model and the simulation results of the temperature distribution in the CNT device for a channel width of 1 mm and film thickness of 2 μm. (c) Transient response of the CNT THz detector. The solid and dashed lines indicate the experimental results and the fitting curve given by V/Vmax = (1 – exp(−t/τ)), respectively. (d) Simulation results obtained by the model shown in (b). THz-induced temperature rise ΔT as functions of the channel width and film thickness and time constant vs the film thickness. (e) Experimental results and their comparison with the simulation results of (d). Good agreement between the simulation and experimental results were observed.

An alternative approach is to increase TMax. Here, we examined this approach (a TMax increase) through simulations based on the thermal conduction model of the CNT device and experimental verification. Figure d shows the simulated temperature gradients for different channel widths and thicknesses of the CNT film, which were calculated through a steady-state thermal analysis. The simulation results reveal that a narrower channel and a thinner film result in a larger temperature gradient and that a thinner film leads to shorter time constant (THz response time). These behaviors originate from the thermal localization via the high thermal resistance of the CNT film, which is more effective for such a narrow and thin CNT films. Note that the above thermal simulation was carried out under the condition of high THz absorption rate of the CNT film. We experimentally observed that the CNT film is required to be thicker than several-tens nanometer so that the THz absorption rate exceeds 90%. We also note that the thickness of the CNT should be larger than 1 μm to retain its free-standing shape. Figures S2 and S3 of the Supporting Information clearly indicate that the TMax value and time constant are strongly governed by the cross-sectional area of the CNT film, supporting the above interpretation. Figure e displays the experimental results describing the CNT size dependence of the THz response, showing features that well agree with the simulation results. On the basis of the strong relationship between the cross-sectional area of the CNT film and the THz response, we enhanced the THz-induced temperature gradient in the CNT film. We estimated the noise equivalent power (NEP) of the improved CNT detector, which was calculated by the noise voltage Vnoise and the sensitivity Vsens as followswhere kB is the Boltzmann constant, T is the temperature, R is the resistance, Δf is the frequency bandwidth, and Peffect is an effective power of the THz wave at the detector. The measured values are Vnoise = 2 nV Hz–1/2, Peffect = 58 μW, and S × ΔT = 6.65 mV under vacuum, respectively (Figure ). According to the above values and eq , the NEP was estimated to be 17 pW Hz–1/2, which is of a similar or superior level compared to competing room-temperature thermal detectors.[23]

Figure 4

(a) Noise voltage spectrum of the CNT THz detector. The measured noise voltage was reduced to 2 nV Hz–1/2, which well agrees with the theoretical value of the thermal noise limit of 2.75 nV Hz–1/2, given by, where kB is the Boltzmann constant, Δf is the frequency bandwidth, T is the temperature, and R is the resistance. (b) THz response of the optimized-thermal-design CNT THz detector based on Figure , where the channel width is 300 μm, the film thickness is 2 μm, and the effective irradiation power is 58 μW at 39 THz. The obtained THz response was 6.65 mV under vacuum.

We then studied a design for a THz detector array to enhance the performance of the THz camera. For high-resolution THz camera imaging, miniaturization of a single-element detector is indispensable. In this sense, a narrow CNT channel width, as shown in Figure , is also favorable for a high camera resolution. On the other hand, regarding the effect of the channel length, we previously observed an issue where, as the channel length was reduced, the positive and negative THz signals between the source and drain electrodes were canceled. To resolve this issue, as shown in Figure S4 of the Supporting Information, we previously fabricated the detector array with different electrode materials (Ti and Au) and avoided the cancelation of the THz responses generated at the interfaces of the source and drain electrodes. Although we demonstrated multiview THz imaging with the flexible THz camera based on this design, the spatial resolution of the acquired images (1 mm) were much lower than the estimated detection area (0.5 mm). Additionally, as shown in Figure a, we found that the spatial resolution was different between the source- and drain-electrode sides, where the detection area at the interface with the Au electrode (0.4 mm) was 2.5 times narrower than that with the Ti electrode (0.95 mm). This led to a low spatial resolution of the THz camera.

Figure 5

Effect of the irradiated wave penetration through the electrode metals. (a) Spatial resolution comparison between the Au and Ti electrode sides. A sharp tail appeared on the Au electrode side (HWHM = 0.4 mm), which is 2.5 times narrower than that on the Ti electrode side (HWHM = 0.95 mm). (b) Schematic illustration of the actual detection area. The detection area was expanded because of the irradiated wave penetration into the CNT film through the electrode metals, consequently resulting in an ideal detection area and undesired detection area. (c) HWHMs of the THz laser spot scanning for the detectors with electrode thicknesses of 5, 40 and 100 nm. A thicker electrode and higher frequency of the illumination wave led to a sharper peak. (d) Undesired detection area and electric field intensity of the THz wave vs the electrode thickness. The symbols are the experimental results plotted from Table , and the dashed lines denote the electric field intensity of the THz wave penetrated into the CNT film through the electrode material, which was calculated by eq .

Table 1

Undesired Detection Area and Spatial Resolution as a Function of the Electrode Thickness for Frequencies of 29 and 1.4 THz

frequency [THz]	laser spot size [μm]	Au thickness [nm]	HWHM [μm]	undesired detection area [μm]	channel length [μm]	spatial resolution [μm]
29	500	5	1200	950	280	1230
		10	800	550		830
		20	450	200		480
		40	310	60		340
		100	260	10		290
		200	280	30		310
1.4	1500	5	1300	550		830
		10	1700	950		1230
		20	1300	550		830
		40	1200	450		730
		100	760	10		290
		200	800	50		330

To elucidate these phenomena, we fabricated detectors with Au electrodes with different thicknesses ranging from 5 to 200 nm and compared the sensing areas of these detectors. Figure and Table represent the results of the line profiles of the THz signals and half widths at half-maximum (HWHMs) for different electrode thicknesses upon scanning the spots of THz waves with 1.4 and 29 THz frequencies across the channel-length direction of the CNT detectors. The HWHM corresponds to the spatial resolution of the arrayed detectors. The results reveal that the use of a thicker electrode produced a sharper profile and that the HWHM narrowed with an increasing frequency of the illuminated wave.

Generally, the spatial resolution of a detector is chiefly determined by the area receiving the incident waves, in this case, the channel size of the CNT film. This, however, is not true of the present situation, as the area size of the CNT channel used here is the same for all the measured devices. Because the HWHM depended on the Au film thickness and the frequency of the irradiated wave, the incident wave could have penetrated into the CNT film through the electrode material, thereby enlarging the sensing area. According to Maxwell’s equations, when an electromagnetic wave travels inside a material with attenuation, the electric field intensity E in the material and the penetration depth d are described aswhere ω is the angular frequency of the electromagnetic wave, σ is the conductivity of the material, and μ is the permeability of the material. Considering the penetration effect, the actual detection area can be expressed by summing the ideal detection area (the channel area) and the undesired detection area (the wave penetration area), as shown in Figure b. Figure d plots the undesired detection area and the electric field intensity of the THz wave as a function of the Au thickness for 1.4 and 29 THz frequency wave irradiation. Here, the symbols denote the experimental results of the undesired detection area, and the dashed lines correspond to the electric field intensity of the irradiated wave penetrated into the CNT film through the electrode materials, as calculated using eq . The general trend of the experimental results well fits the calculated curves, indicating that the THz wave penetrates into the electrode metals. Therefore, the additional THz photothermoelectric voltage occurs in the vertical CNT–electrode interface, which unintentionally expands the actual detection area.

To the best of our knowledge, this phenomenon has not been noted or discussed in the photothermoelectric effect and device research, which we believe is partly because the majority of research on the photothermoelectric effect has been long focused on the near-infrared and visible light regions (over 100 THz).[24−27] In these regions, the penetration depth is estimated to be at most a few nanometers and can therefore be ignored in most cases (Table ). We can thus conclude that carefully choosing a metal and electrode thicknesses is of much importance, especially in the THz region, to prevent wave penetration through the electrodes. The obtained resolution of our THz detector was about 290 μm for the 1.4 THz irradiation as shown in Table , whose value was comparable to the wavelength of the 1.4 THz wave, 214 μm. This means that our THz detector almost reached the diffraction-limited resolution. We note that the current channel length was optimized for the 1.4 THz irradiation and therefore expect that when a shorter channel is used, spatial resolution for the 29 THz irradiation should also reach the order of the wavelength.

Table 2

List of the Penetration Depths for Various Materials as a Function of the Irradiation Frequency

	penetration depth [nm]
Material	irradiation frequency [THz]
	0.1	1	10	100
Ni	41	13	4	1
Cu	207	65	20	6
Au	237	75	23	7
Al	267	84	26	8
Pt	519	164	51	16
Ti	1156	365	115	36

By incorporating all the findings and techniques clarified in this work, we fabricated a higher-sensitivity and higher-resolution CNT THz camera than our earlier type of camera. Figure a displays the photographic and schematic images of the fabricated THz camera. A CNT film with a 30 μm thickness and 500 μm channel width was used, and 10 nm thick Ti and 200 nm thick Au were deposited as source and drain electrodes, respectively. A TiO2 film was then formed through surface oxidation of the Ti film as the thermal isolation layer to avoid thermal diffusion to the adjacent metal layer. Finally, a THz-reflection layer with a 100 nm thick Au film was evaporated to suppress the THz wave penetration into the CNT film. (Note that the calculated penetration depth for a 1 THz wave into the Au film is 75 nm.) In order to acquire clear two-dimensional THz images, the resolution along the channel-width direction has to be improved together with the resolution along the channel-length direction, as discussed above. To remove unnecessary signals generated through the interference between the neighboring detectors, we also thermally isolated the detector elements by cutting the CNT film in between the two detection areas (Figure S6 of the Supporting Information). As shown in the full width at half-maximum (FWHM) results in Figure S6, the detection area along the channel-width direction was reduced by almost 85%.

Figure 6

(a) Photographic image of the fabricated THz camera and device structure of the single element. The source electrode was composed of the thermal conduction layer (Ti), thermal isolation layer (TiO2), and THz-reflection layer (Au). (b) THz images of the metal clip acquired by the previous THz scanner and the refined THz camera using the high-resolution technique in (a). As shown in the line scan data (lower figures), the undesired detection area was reduced by a factor of 70.

Figure b compares the imaging performances of our previous THz camera and the present high-resolution THz camera. The upper panels are the THz images of a metal clip with a diameter of 1.28 mm, and the lower plots are line scan data taken from a part of the upper images. As shown in Figure b, a clearer image was obtained by suppressing the THz wave penetration through the electrode materials and consequently eliminating undesired signals. By subtracting the sum of the channel area and the diameter of the object (1.28 mm) from the FWHM value, we estimated that the undesired detection areas of the two cameras were 1.5 mm for the previous type and 20 μm for the present type, indicating improved camera resolution. The calculated values of the penetration depth in Table suggest that the use of Ni as an electrode material would more strongly suppress wave penetration and the resulting undesired detection area.

These THz images were taken with the acquisition time of 3 h simply because of a limited number of signal readout being used. We expect that it should be improved up to 13 ms that corresponds to the fundamental response time of our THz detector (Figure c) by integrating a large detector array with a multichannel readout and processing system.

Conclusions

In summary, we investigated the underlying mechanism of a photothermoelectric THz CNT camera and improved the thermal device design by optimizing the physical factors influencing the sensitivity, response time, and spatial resolution. The simulated and experimental results reveal that a smaller cross-sectional area of the CNT channel results in higher sensitivity and a faster response speed because of thermal localization via a high thermal resistance. In addition, we found that the actual detection area was strongly governed by the incident wave penetration into the CNT film through the metallic electrode materials, leading to a degradation of the spatial resolution of the multiple-array THz camera. The implementation of a blocking layer for the THz penetration and a thermal isolation layer allowed us to sufficiently suppress this effect and to miniaturize the actual detection area. By utilizing all the technologies based on these findings, we enhanced the imaging performance of the CNT THz camera and obtained a much clearer THz image compared to the previous type of CNT THz camera. In addition to the THz camera, the presented thermal scheme will also be applicable to other thermal-type photodetectors,[28,29] optically activated portable batteries,[30] and so forth and suggests a huge potential for further optimization by engineering the thermal structures.

Materials and Methods

Device Fabrication Process

We employed a flexible and inch-size CNT film as the material for the bendable THz camera. The original CNT film was produced through a filtering process of a CNT solution with a membrane (Figure a).[31] The film structure, such as scale, thickness, and density, were controlled by adjusting the concentration of the CNT solution. The typical structure parameters of the CNT film were as follows: scale (A3, 30 × 42 cm), thickness (2–150 μm), density (0.4 g/cm3), basis weight (0.2 g/m2), porosity (63%), and specific surface area (740 m2/g). The Seebeck coefficient of the CNT film was measured to be 55 μV/K. We cut a CNT film into the appropriate size for a single-element THz detector. The Ti and Au source and drain electrodes, separately, were evaporated through a hard mask without a photoresist to prevent chemical contamination. Each THz detector was placed onto a chip carrier and a universal board, and readout wires were connected by silver paste.

Measurement System

The results of the THz response of the CNT THz detector were obtained using three THz sources that cover a wide range of frequencies: a quantum cascade laser (λ = 7.7 μm), a CO2 gas laser (λ = 10.2 μm), a THz laser pumped by a CO2 laser (λ = 214.6 μm). The THz radiation from these sources were guided and concentrated to the detection area by lenses or fibers. The THz responses (generated THz-induced voltages) were directly recorded with a digital multimeter. For the low-response signals, a current amplifier was used. For the THz response mapping (Figures b, 5a,c, and S6) and the THz imaging measurement (Figure b), we used two-axis translation stages and their corresponding controllers. We used a germanium substrate as the opaque object because of the high transmission through this object at a frequency of 29 THz. We employed a lock-in amplifier to measure the noise density spectrum of the THz detector (Figure S2) with a bandwidth of 1 Hz (or 0.3 Hz) and a smoothing time constant of 1 s (or 10 s). The NEP was estimated using the equation NEP [W Hz–1/2] = noise voltage density [V Hz–1/2]/sensitivity [V/W]. The sensitivity [V/W] was derived by dividing the voltage response [V] by the irradiation power at the detection area [W]. The THz absorption spectra of the CNT film (Figure b) were taken with a THz-time domain spectroscopy (THz-TDS) transmission measurement at room temperature with a data averaging of 1028 times, a frequency bandwidth of 0.5–7 THz, a time resolution of 2 fs, a frequency resolution of 3.8 GHz, a scan range of 262 ps, a throughput of 16 ms/scan, and a frequency accuracy of 10 GHz (at 1.41 THz). The THz-TDS system was placed inside a glovebox, which was filled by dry air to avoid THz absorption by water vapor.

Thermal Conduction Analysis of the CNT Device

To simulate the device shape dependence of the thermal conduction (Figures , S2, and S3), we performed a steady-state thermal analysis and a transient thermal analysis using the ANSYS software package. The series of simulation results were calculated under the conditions of the thermal conductivity of the CNT film: 10 W/m K for the X–Y plane and 0.1 W/m K for the Z axis;[32] the thermal conductivity of the electrode metal (Au): 315 W/m K; the heat transfer coefficient of air: 10 W/m2 K, and under calm conditions at 300 K. The simulation model that we employed is depicted in Figures S2a and S3a. Assuming that the power of incident THz wave was totally absorbed by the CNT film, as experimentally shown in Figure b, and was converted into the heat, we added the heat on the surface of the CNT film. We considered the situation in which the CNT film retained its free-standing shape without the substrate and was exposed to the atmosphere. The outside (air) temperature was set to be 22 °C. The temperature distribution was calculated by solving the following heat equation given bywhere ρ is the density, C is the specific heat capacity, k is the thermal conductivity, and Q is the total amount of heat.