2021 年 62 巻 1 号 p. 16-19
A heat flow switching device was developed using semiconductors characterized by very small lattice thermal conductivity. We selected Ag2Ch (Ch = S, Se) which possesses semiconducting electron transport properties and very small lattice thermal conductivity, and tried to control their electron thermal conductivity using bias voltage. The samples were prepared by means of self-propagating high-temperature synthesis under vacuum atmosphere, and mechanically rolled into ribbons of 10 µm in thickness. For making the capacitor-type device, amorphous Si and Mo were deposited on the rolled films using RF-sputtering. We compared thermal conductivity with and without bias voltage by means of the AC heating method. As a result, we succeeded in observing a 10% increase of heat flow in the capacitor type heat flow switching device.
This Paper was Originally Published in Japanese in J. Thermoelec. Soc. Jpn. 16 (2019) 73–76.
Thermal management is one of the key ideas for building a low carbon-emitting, low energy consuming, and sustainable society. Thermal management is a concept leading to such a society by using technologies such as heat insulation, heat storage, thermoelectric generation, heat transportation, etc. Among them, we have been studying the effective heat transportation technologies to control the direction and amount of heat flow for heat waste recovering.1–3)
Heat conduction in an isotropic solid material is generally described by Fourier’s law jQ = −κ∇T, where jQ, ∇T, and κ represent the heat flow density, the temperature gradient, and the thermal conductivity of a material, respectively. The magnitude of heat flow density in a given material is determined almost solely by temperature gradient |∇T| because it is difficult to modify the magnitude of thermal conductivity using the conditions applicable externally. Therefore, it is difficult to control heat flow by using a solid material.
Two alternative ways of controlling heat flow are naturally considered; one is mechanical using the difference between contact and non-contact, and the other uses a heat transfer fluid. But both systems have difficulties in controlling heat flow like maintenance, miniaturization, and operating temperature.
If we develop a heat flow switching device solely consisting of solid materials without having any mechanical moving parts, it would become one of the most important technologies of thermal management.
In 2006, Peyrard proposed a mechanism leading to thermal rectification for controlling the heat flow.4) To achieve direction dependence of heat flow, they used a composite in which two types of materials connected; one has increasing thermal conductivity with increasing temperature, while the other decreasing. The experiments done later proved the validity of the concept by showing the significant rectification of heat flow with the rectification ratio of 2 to 2.7.5–9) However, these numbers have not satisfied the requirements. Besides, thermal diodes have difficulty in practical usage, because the working temperature was determined by the temperature dependence of component materials and this characteristic prevents us from easily optimizing the function for each application. Therefore, the thermal diodes have not been practically used yet.
In this study, therefore, we propose a new, innovative concept of heat flow switching device working with an external field effect and usable in a wide temperature range. We also constructed a device and evaluated its performance to confirm the validity of our concept.
Electron concentration in semiconductors can be controlled by a bias voltage in devices such as MOSFETs. It is also widely known that the electron concentration in metals and degenerate semiconductors is altered by bias voltage in capacitors. This change in electron concentration should lead to a change in electron thermal conductivity, and therefore we propose a new concept leading to an effective switching in heat flow; controlling the electron thermal conductivity using bias voltage.
Unfortunately, however, it would be very difficult to observe the tiny change in electron thermal conductivity under the influence of large lattice thermal conductivity generally existing in semiconductors. Semiconductors that having extremely small lattice thermal conductivity are required for developing a heat flow switching device working with this mechanism.
In this study, therefore, we employed Ag2S1−xSex as the electrodes because these materials were reported to possess very small lattice thermal conductivity as low as 0.5 W m−1K−1 10) together with behaviors of electron transport properties of a semiconductor. The very small lattice thermal conductivity is presumably brought about by the anharmonic vibration of the lattice in association with the hopping of atoms between the naturally existing split-sites. Notably, since Ag2S has been reported to be ductile,11) ribbon samples are easily fabricated from bulk. A capacitor type device was made with the Ag2S1−xSex ribbon with an insulating amorphous Si.
Reagent grade raw materials of Ag (purity 99.9%), S (99.99%), and Se (99.9%) were used to prepare the starting powders for synthesizing Ag2S1−xSex. The raw powders were well mixed by the mortar and the pestle and pressed at 100 MPa into a cylindrical shaped pellet of 1 to 2 mm in thickness and 10 mm in diameter at room temperature. Ag2S1−xSex was synthesized by means of self-propagation high temperature synthesis method under a vacuum atmosphere. For homogenization, the samples were heated up to 1000°C and kept at 1000°C for five hours. After that, the samples were cooling down to 300°C with a cooling rate of 10°C/hour and furnace cooling to room temperature. The obtained ingots (x = 0.3, 0.4) were mechanically rolled into ribbons of 10 µm in thickness.
Some of the homogenized samples were crushed into powder and sintered into high density bulks at 350–650°C, 40 MPa, 10 minutes under vacuum atmosphere. The density of all the bulk samples was more than 95% of the theoretical value.
The phases involved in the samples were identified by conventional powder X-ray diffraction using BRUKER D8 Advance with Cu-Kα radiation (λ = 0.15418 nm). The electrical conductivity (σ) was measured by the conventional four-probe method. The thermal conductivity of the Ag2S1−xSex in bulk was measured at room temperature using a Laser flash method (LFA457, NETZSCH).
For making capacitor-type devices, amorphous Si and Mo of 500 nm in thickness were deposited on the rolled ribbons using a radio frequency (RF) magnetron sputtering system (ULVAC VTR-150M/SRF) at room temperature. The background pressure of the chamber was 2 Pa and high purity Ar was used as a sputtering gas. The multilayer was sputtered at the Si and Mo RF power of 80 W and distance between target and substrate was 60 mm.
We compared thermal conductivity with and without bias voltage by means of the AC heating method10) under a vacuum atmosphere of ∼1 Pa to decrease the effect of air convection. AC temperature waves with f = 30 mHz, Tmax ∼ 50°C were applied to the device. Time dependence of temperature was measured at various points on the device by an infrared camera (Optris PI 200). The experimental setup is schematically drawn in Fig. 1.
(a) Schematic drawing of the AC heating measurement with and without bias voltage. (b) Temperature distribution at the sample surface observed using IR camera.
The measured X-ray diffraction (XRD) patterns are plotted in Fig. 2. The XRD patterns of the samples x ≥ 0.8 show the structure of Ag2Se (Orthorhombic, Pearson sign: oP12)12) and, XRD patterns turn up to show the structure of Ag2S (monoclinic, mP12)13) at x ≤ 0.4. In any compositions, we did not find precipitation of secondary phase. Our result agreed with a reported pseudo-binary phase diagram of Ag2S–Ag2Se.14)
XRD patterns of Ag2S1−xSex (x = 0, 0.3, 0.4, 0.8, 1).
The samples which have the structure of Ag2S (x ≤ 0.4) could be mechanically rolled into ribbons of 10 µm in thickness. Figure 3 shows a photograph of a rolled Ag2S0.6Se0.4. We confirmed that the samples possessing the structure of Ag2S showed clear ductility in the same manner as pure Ag2S, while the samples which have the structure of Ag2Se (x ≥ 0.8) did not show any ductility. Notably, it is a newly found fact that not only Ag2S but also Ag2S1−xSex, which has a structure of Ag2S (x ≤ 0.4), is ductile.
A mechanical rolled, ribbon shaped sample of Ag2S1−xSex (x = 0.4). The thickness is ∼10 µm. The rolled samples prepared at x ≤ 0.4 possess good flexibility.
Figure 4 shows Se concentration dependence of electrical conductivity, thermal conductivity, electron thermal conductivity, and lattice thermal conductivity of Ag2S1−xSex at room temperature. We observed drastic increases both in the electrical conductivity and the thermal conductivity with increasing x from 0.4 to 0.8. The electron thermal conductivity was roughly estimated using the Wiedemann-Franz law (κel = L0σT: L0 = 2.44 × 10−8 W ΩK−2), and the lattice thermal conductivity was estimated by subtracting the electron contribution from the total thermal conductivity as κlat = κ − κel. The lattice contribution of thermal conductivity possesses less significant Se concentration dependence, while the electron contribution is mainly determined by the Se concentration. The electron thermal conductivity reached four times larger value from that of Ag2S with increasing Se concentration and the largest electron thermal conductivity is much higher than the magnitude of lattice thermal conductivity. Therefore, we strongly expect that their thermal conductivity is capable of being controlled by the variation of electron concentration using the bias voltage.
(a) Electrical conductivity, (b) thermal conductivity, (c) electron thermal conductivity, and (d) lattice thermal conductivity of bulk Ag2S1−xSex (x = 0, 0.3, 0.4, 0.8, 1).
Now we are ready to prepare the capacitor type thermal switching devices by using Ag2S1−xSex (x = 0.3, 0.4). The thermal diffusivity was measured for the prepared device using the AC heating method with and without bias voltage (Fig. 5).
Bias voltage dependence of normalized heat flow in the heat flow switching devices.
From the results in Fig. 5, it is expected that the heat flow would increase substantially in proportion to the bias voltage, provided that a voltage is applied in the forward bias direction. We confirmed that the heat flow indeed increased with increasing bias voltage presumably due to the increased electron thermal conductivity. As a result, we observed a 10% increase in heat flow with a small bias voltage of 3 V. In this study, the limitation of power source prevented us from applying more than 3 V to the device. Although the variation of thermal conductivity is not large but rather small, this preliminary result could lead to the development of high-performance, practical heat flow switching devices.
It would be important to note here that no significant change in heat flow was observed at negative bias. In this study, Ag2S1−xSex shows n-type characteristics regardless of the composition. Once some carriers are eliminated from the sample near the electrodes, that portion would become insulating and prevent the electrons/holes from moving towards the interface of the capacitor. This could explain the asymmetric bias voltage dependence of heat flow.
We consider that the portion of varied thermal conductivity would be much thinner than the thickness of ribbon samples and therefore makes the variation of heat flow less significant in the prepared device: the thickness of the accumulated electrons layer would be less than a few tens of nm, while the thickness of the ribbons is 10 µm. This means that the thickness of the portion of the ribbon that does not contribute to the variation of thermal conductivity is 1000 times larger than the layer in which the electron thermal conductivity changes.
It should be emphasized that, even under this negative condition, we observed a 10% change in the heat flow. It means that the variation in electron thermal conductivity in the portion where electron concentration changed is significantly enlarged. This consideration naturally led us to a confidence that a very large change in heat flow should be achieved by making the samples thinner and multi-layered.
If we could reduce the thickness of the semiconductor layers from 10 µm to several tens of nanometers, a change in thermal conductivity would exceed 10 times. It is also considered that the capacitance is capable of being increased by reducing the thickness of the insulating layer or using an insulator with a large dielectric constant. We are now conducting such device preparations and the results will be presented elsewhere.
The origin of the asymmetric change of the heat flow concerning the bias voltage is related to the type of carrier in semiconductors. Since the samples used in this study are an n-type semiconductor, when an electric field is applied in the direction of sucking up electrons, carriers are reduced only around the electrodes and become insulators, so that the electron concentration near the interface cannot be changed. To avoid this effect, it is necessary to develop capacitors consisting of p-type and n-type, or those consisting of a material having a pseudo gap at the Fermi level.
In the near future, the structure and physical properties including the semiconductor layer and insulator layer will be optimized for a huge variation in magnitude of heat flow at least exceeding 10 times. If once such a large change in heat flow was observed in a heat flow switching device, it should be widely used in a variety of applications together with a variety of heat sources.
In this study, we tried to fabricate a capacitor-type heat flow switching device that can control the magnitude of heat flow by applying a bias voltage to alter the electron concentration. The semiconducting Ag2S1−xSex (x = 0.3, 0.4) with extremely low lattice thermal conductivity was used in the device. By comparing thermal conductivity with and without bias voltage by means of the AC heating method, we confirmed the small but finite change in the heat flow with bias voltage. We succeeded in demonstrating that our newly proposed “thermal switch mechanism using variations in electronic thermal conductivity via bias voltage” works effectively.
