Electrohydrodynamic Casting Bismuth Telluride Microparticle-Loaded Carbon Nanofiber Composite Material With Multiple Sensing Functions

Bismuth-based metallic alloys are semiconducting materials. For example, bismuth telluride has a narrow energy band of E_g = 0.19 eV [1]. Another important bismuth alloy, Bi₂Se₃, shows the band gap around 0.35 eV [2]. They are suitable for thermoelectric energy conversion. They also find applications for infrared (IR) radiation monitoring and temperature sensing. Bismuth telluride may be embedded into an insulating material, for example, alumina, to form a composite material. Such a composite material offers the enhanced thermoelectric figure-of-merit [3,4]. It is expected that the electrical conductivity and Seebeck coefficient of the composite material could be dramatically increased, while the thermal conductivity could be kept as low as possible by selecting a proper insulating matrix and adjusting the content of the loaded bismuth alloy filler. This approach could provide an effective way for improving the thermoelectric performance of materials. An ultrahigh molecular weight polyethylene-based composite material with separated networks of carbon nanotube/Bi₂Te₃ hybrids was made [5]. However, the Seebeck coefficient of the polyethylene composite is only about 29 μV/K which is much lower than that of the pure Bi-based thermoelectric material. Electrodeposition of polyaniline and Bi₂Te₃ simultaneously was conducted. It is found that the thermoelectric power of the composite material in the temperature range from 380 K to 420 K is higher than that of the pure polyaniline [6].

In order to increase the phonon scattering, adding nanoscale entities into the bismuth alloy was performed. The dispersion of SiC nanoparticles in Bi₂Te₃ reduced the thermal conductivity. But it did not significantly increase the electrical resistance of the alloy. This helped an 18% increase in the figure of merit [7]. The addition of carbon nanotube also enhanced the phonon scattering in Bi₂Te₃ [8]. The increased phonon scattering is also anticipated in some quantum-confined structures made by the epitaxial growth [9], hydrothermal synthesis [10,11], electrodeposition [12], and the combined chemical and thermal processing approach [13]. The mechanism for the increased phonon scattering may be due to the interface reflection effect as discussed in Ref. [14] for the Al₂O₃/Bi₂Te₃, and in Ref. [15] for silicon films.

Some major progress has been made in manufacturing Bi-based alloys and the Bi₂Te₃-containing composite materials. Bi-based alloys are typically manufactured through the spark plasma sintering [16,17], mechanical alloying [18], electrodeposition from ionic liquids [19], organic-assisted growth [20], and interface reaction [21]. For the composite materials, several major methods have been used including chemical reaction approach [22], mechanical alloying [23], plasma sintering [24], electrochemical synthesis [25,26], and aerosol deposition [27]. Recently, carbon-based or carbon and bismuth-based alloy mixture composite materials have caught much attention for thermoelectric applications [8,16,19,28–30]. However, due to the low figure of merit, the thermoelectric performance of the composites still remains to be improved.

This work focuses on making a new composite material consisting of carbon fibers and bismuth-based alloy particles. Specifically, it deals with manufacturing a bismuth telluride microparticle-loaded carbon nanofiber (CNF) composite material thorough an innovative approach, electrohydrodynamic casting. First, the Bi₂Te₃ microparticles were dispersed into a polyacrylonitrile (PAN) polymer nanofiber mat through the electrical force-assisted processing. Then, an intermediate temperature stabilization treatment on the PAN followed by a high temperature annealing was conducted to obtain the carbon fiber composite material.

The energy conversion and sensing multiple functions of the composite material were briefly shown. It is demonstrated that the patterned fiber mats can be manufactured through the electrohydrodynamic casting process. Characterization of the composite material was also conducted, which includes examining the structure of the material and testing its energy conversion performances. Specifically, scanning electron microscopic (SEM) analysis was performed to reveal the morphology and the composition of the composite material. The energy conversion functions in view of converting heat and photon energy into electricity were revealed. Hyperthermia behavior of converting electromagnetic wave energy to heat was investigated as well. The composite material shows thermoelectric, hyperthermia, and photovoltaic effects.

The solvent, dimethylformamide (DMF), was purchased from Alfa Aesar, Ward Hill, MA. The Bi₂Te₃ powder was purchased from Sigma Aldrich, St. Louis, MO. The PAN polymer was supplied by Scientific Polymer, Inc., Ontario, NY. The polymer solution was prepared by adding approximately 10% in weight of PAN powder into the DMF solvent. For a typical experiment, 1 g PAN powder was added into 9 ml DMF solvent and stirred for an hour to allow the PAN powder to dissolve into the solvent. Then 1 g Bi₂Te₃ powder with the nominal size less than 5 μm was added into the PAN-DMF solution.

The solution containing PAN polymer, Bi₂Te₃ powder, and DMF solvent was filled into a 10 ml plastic syringe with a gauge 20 stainless steel needle. The syringe was set on a precision syringe pump made by Chemyx, Inc., Stafford, TX. The pump can precisely control the injection flow rate of the mixture solution. The distance between the tip of the needle for injection and the receiving target was kept at about 150 mm.

A constant flow rate of the solution injected by the syringe pump was set as 0.1 ml/min. Under ambient temperature, pressure and humidity conditions, a DC voltage of 15 kV was applied at the tip of the needle to electrify the solution. This electrified state allowed the mixture to overcome the surface tension of the Bi₂Te₃/PAN/DMF solution.

The cast fibers were collected on the surface of a tissue paper which was attached to a grounded metal collector. The potential difference between the tip of the needle and ground collector led the charged jet to cast fibers onto the tissue paper continuously. The tissue paper has the advantage of being easily separated from the cast nanofibers. On the tissue paper, the orientation of cast Bi₂Te₃/PAN fiber was random. The collector was set on an x–y–z table, which can translate in x-, y-, and z-directions. From this processing step, the product collected is a composite mat containing Bi–Te particles connected by the PAN polymer fibers. The electrohydrodynamic manufacturing setup is schematically shown in Fig. 1(a).

In order to form electrically conductive networks around the Bi–Te particles, the PAN nanofibers were partially carbonized through the heat treatment using the facility as schematically shown in Fig. 1(b). The heat treatment can be divided into two steps. First, the composite mat was setting in the quartz tube and heated up to 250 °C in ambient atmosphere. After being kept at 250 °C for 1 h, the composite fiber mat was slowly heated up to 500 °C. Hydrogen was inducted into the chamber during the heating from 250 °C to 500 °C. This prevents the Bi–Te being oxidized. The PAN fiber started partial carbonizing above 300 °C. After the mat was heat treated at 500 °C for 1 h, it was cooled down naturally with the furnace to the room temperature of 23 °C.

The morphology of the composite material was examined using a JEOL JSM-6010PLUS/LA SEM. The acceleration voltages used were 15 kV and 20 kV depending on the magnifications of SEM images taken. The composition of the Bi–Te/C composite fiber material was measured by the energy dispersive X-ray spectroscopy (EDS).

To characterize the energy conversion behavior, thermoelectric and photovoltaic properties of the composite fiber were tested by measuring the open circuit voltage of the composite material under the action of heat waves generated by a heat gun and the illumination of a visible light source, respectively. An electrochemical analyzer, mode CHI 440C, made by CH Instruments, Inc., Austin, TX, was used to measure the time-dependent voltage signals of the fiber mat under phonon and photon energy excitations.

The fiber mat sample was wrapped at the two ends with metal strips (made of aluminum foil) as the electric collecting paths. Crocodile clips (electrodes) were attached to each end. The metal strips were also used as electrical conductors to link the jumpers (electrodes) to the composite mat. The voltage versus time data were recorded and processed using a matlab program.

The hyperthermia test measures the heat generation when the composite material was under exposure to electromagnetic waves. The hyperthermia response of the composite fiber mat to microwave was tested. The sample was heated in a microwave for seven different time frames. Prior to placing the sample in the microwave, the temperature throughout the surface of the sample was measured using an IR thermometer. After recording the temperature of the unheated sample, it was placed into a 900 W microwave. The specimen was heated for 0, 5, 10, 15, 20, 25, and 30 s, respectively, in each testing cycle. For each run, five temperature data points from different locations at the surface were obtained immediately after the sample was taken out from the microwave.

The microstructure and the composite information of the prepared composite mat were obtained through the SEM analysis. The results of morphology observed by the SEM and the composition profile of the composite material are shown in Fig. 2. Figure 2(a), an SEM image of the composite fiber mat at a relatively low magnification, provides a global view of two important microstructural constituents, the partially carbonized fibers, and the Bi–Te alloy powders. The microparticles are dispersed at different layers in the mat. At a slightly higher magnification as shown in Fig. 2(b), it can be seen that the smaller sized Bi–Te alloy particles are connected by the nanofibers. Each particle is connected by a single strand of nanofiber. The bigger-sized particles are connected by multiple nanofibers as also shown in Fig. 2(b). At a further higher magnification as shown by the micrograph of Fig. 2(c), the size of the nanofibers around a few hundred nanometers can be determined. The biggest size of the Bi–Te particles is around 5 μm. Figure 2(d), the SEM image with an even higher magnification, shows more details of the Bi–Te particles distributed among the interlaced carbon fibers. The fibers are randomly oriented and form the mat through the stacking of different layers.

In order to confirm the elemental composition of the composite material, an EDS generated by the area mapping technique is shown in Fig. 2(e). It is clearly shown that carbon, bismuth, and tellurium are the major elements. Oxygen signal is from the incomplete carbonization of the PAN polymer. It is believed that the PAN polymer underwent cyclization and oxidation when it was heated in the temperature range from 200 °C to 300 °C in air [31–33]. In this work, the stabilizing temperature used was 250 °C. The PAN molecules were cyclized and transformed into a nonmeltable ladder structure as demonstrated in Refs. [34–36]. Some oxygen in the functional groups such as =O and –OH are encapsulated into the backbone [37,38]. The quantitative results from the elemental analysis are shown in Table 1. From the quantitative results in Table 1, it can be seen that the atomic ratio of Bi–Te is about 3:4. Therefore, instead of the original Bi₂Te₃, the new nonstoichiometric compound, Bi₃Te₄, was obtained after the high temperature heat treatment in hydrogen.

To explain why the Bi₂Te₃ turns into Bi₃Te₄, the electronegativity of typical elements are given. Bi and Te has a closed electronegativity value of 2. But O has a different value of 3.5 and N has a value higher than 3. From the electronegativity difference, both Bi and Te have a tendency to donate electrons to the N and O elements in the PAN polymer. N and O become the electron receptors. But Te has much stronger affinity to electron acceptors than Bi from the coordinate bonding point of view. Consequently, Te is attached to N and O containing ligands to form complex groups. With the partial carbonizing of the PAN at high temperature, the Te element goes into the space between carbon layers. Therefore, the initial Bi₂Te₃ compound lost part of the Te to form a Bi-rich compound: Bi₃Te₄.

It must be noted that the EDS analysis is a half-quantitative method. Ideally, a second characterization method, such as X-ray powder diffraction (XRD) may be used to identify the crystalline structure and determine the phase amount of the material. However, the focus of this work is on manufacturing the nanofiber composite. The structure assessment is preliminary, and the information is mainly in the qualitative sense. However, the half-quantitative results did give the information of the change in compositions caused by both the organic–inorganic complex formation and the heat treatment. In view of the accuracy of EDS and XRD on quantitative elemental analysis, earlier studies showed that the difference in the results between the two techniques is less than 2% based on the Bi–Te atomic ratio measurements for the bulk alloys, particles, and/or thin films [39–49].

Another issue here is that XRD typically works perfect for crystalline materials analysis. But in this work, some Te has been already intercalated into polymeric or glassy carbon. This could generate some error in determining the Bi to Te ratio by the XRD approach. Earlier studies showed that the measured results could also be affected by the addition of other compositions. For example, the graphene addition changes the height of the diffraction peaks due to the addition of graphene sheets in Bi₂Te₃ [50]. MoS₂ nano-inclusions lead to incoherent interfaces between different phases. The strain built in Bi–Te alloy causes its diffraction peaks shifting [51]. Therefore, the XRD method could also introduce some error in the quantitative elemental analysis of the carbon nanofiber/Bi–Te alloy composite material.

Therefore, the estimated value of S for this material is about −3.937 mV/K. This result provided us some important information. First, the composite material shows a general n-type behavior, because the voltage dropped to a negative value when the specimen was heated. Second, the best carbon-based thermoelectric composite materials—the stacked graphene sheets can only reach a Seebeck coefficient value of −90 μV/K [30]. The results from this study showed that the Bi₂Te₃/CNF composite mat possesses a much higher absolute Seebeck coefficient value than the most existing carbon materials and Bi–Te alloys including the Bi₂Te₃ coating [52].

Based on this strong thermoelectric behavior of the Bi₂Te₃/CNF composite material, we used the sample to monitor the warm air flow due to exhaling and inhaling of human. The results are shown in Fig. 4. First, a relatively short exhaling phase (5 s) was followed by a longer inhaling stage (15 s) in a complete cycle, and the test results were plotted in Fig. 4(a). Obviously, the warm air from the exhaling stage caused the negative voltage generation. We monitored the surface temperature of the hot end of the specimen using an INF165 infrared thermometer manufactured by UEi Test Instruments, Beaverton, OR. The hot end surface temperature of the specimen was around 31 °C. And the cold end surface temperature was about 26 °C. From the voltage difference in the exhaling and inhaling stages, we found that the Seebeck coefficient is approximately 4 mV/K. This value is consistent with the result as calculated from the data presented in Fig. 3(b).

Second, we presented the test results related to a short exhaling phase (5 s) followed by a short inhaling stage (5 s) in a complete cycle, and the test results were plotted in Fig. 4(b). It is found that due to the short recovery time for the temperature in the inhaling cycle, the voltage change cannot reach so high as that shown in Fig. 4(a). The data demonstrated a general trend of decreasing in the open circuit voltage.

Third, we monitored the breathing pattern of a patient with the coughing symptom. The results were shown in Fig. 4(c). Evidently, the breathe-in signals were momentarily strong, because the air rushed out due to the coughing as shown by the sharp drop of the voltage. Because of the reflexing actions, the signals in the period right after the breathe-in, and the following breath-out cycle revealed the irregularity as an evident by the zig-zag patterns marked by the green circles in Fig. 4(c). The implication of the test results in Fig. 4 is that a cheap sensor may be built based on the highly sensitive thermoelectric responses of the Bi₂Te₃/CNF composite material.

The photoelectric energy conversion property of the fiber Bi₂Te₃/CNF composite material was tested using two different approaches, as schematically illustrated in Fig. 5. In Fig. 5(a), the experimental setup for measuring the time-dependent voltage generated by the composite fiber specimen under a 120 W florescent light is shown. The tubular light source was set 2 m away from the specimen so that the light beams shining on the specimen were parallel when they passed through the open window in the translating steel plate. In Fig. 5(b), a high speed rotating fidget spinner was used to test the light sensitivity of the composite material and determine the deceleration of the fidget spinner. The same light source was used for the two types of tests. The experiments were done with the intend to illustrating the feasibility of the composite material being used as optoelectronic sensors for the multiple function applications such as the light sensing, acceleration, and deceleration detections.

The voltage data shown in Fig. 6 were obtained when the sample was initially covered from the illumination of the fluorescent light located 2 m away from the sample. In the first cycle, the voltage increase is about +0.13 V. This is not the same response as observed by the thermal wave excitation. Under the heating or infrared excitation condition, the response of the material mainly came from the Bi₂Te₃ microparticles, which is n-type. Latest report shows that carbon nanostructures are p-type because of its cyclic and conjugated structure [29,30]. Since the band gap of Bi₂Te₃, E_g, is very small, it is not surprising that the particle is only responsive to the infrared irradiation, while not sensitive to the visible light. Figure 6 also displays the decaying model, which is related to the charge recombination effect as observed in most of the photovoltaic materials.

To further validate the p-type behavior of the composite material under visible light excitation, we still used the test facility as shown in Fig. 5(a) and measured the responses of the specimen with the open-window mode followed by the close-window mode with a slower pacing. The results are shown in Fig. 7. Figure 7(a) illustrates the results when the window opened for a 10 s period followed by closed for 10 s in each complete cycle. It was clearly shown that the visible light reaching the specimen caused the voltage spiking followed by the recovery due to the recombination of the electron–hole charged pairs. To test the consistency in the photovoltaic response and the stability of the composite material under ambient storage conditions, another specimen was prepared by cutting the fiber into the same dimension of 15 mm × 50 mm from the same batch of the electrohydronamically manufactured composite mat. Figure 7(b) shows the open circuit voltage test results from the newly made specimen. Obviously, Figs. 7(a) and 7(b) show the same trend. Thus, it is concluded that the composite material is effective in converting the photon energy corresponding to the visible light spectrum into electrical energy. The photon valve behavior as shown by Fig. 7(a) or 7(b) also builds the foundation of the composite material for photon sensing application.

To observe the charge generation and extinction behavior, the short- and long-term exposures of the composite material to the same visible light source were experimented. The representative data were plotted in Fig. 7(c). At time t = 9.5 s, the window in the light-blocking plate shown in Fig. 5(a) was OPEN. This allows the photon energy response to be detected instantly, and it was found that the peak reached around 0.2 s (at time t = 9.7 s). When the peak was reached, the recombination was found more than 30 s (time period t = 9.7–40 s). This information reveals that the photosensitive of the composite material is high and fast enough for most of important applications, for example, passing traffic monitoring under the totally dark or twilight conditions.

Using the test facility setup as shown in Fig. 5(b), the photosensitive response of the composite material can be used to detect the deceleration of the fidget spinner. The test results were shown in Fig. 8. Figure 8(a) is the open circuit voltage measured for the composite specimen. At t = 0 s, the fidget spinner started the rotational motion. At t = 10 s, the spinner slow down obviously and the spinner stopped rotating at t = 17 s. After the first cycle, series of tests were performed. In the second cycle, the spinner started rotation at t = 18 s and stopped at t = 40 s. In the third cycle, the spinner started rotation at t = 41 s and stopped at t = 65 s. Each cycle took about 20–25 s. When the spinner was slowing down, the photon-induced voltage became higher and higher in the absolute value.

At t₁ = 61 s, the angular speed ω₁=2π × (1/3)/0.4 = 5.233 rad/s. At t₂ = 65 s, the angular speed ω₂ = 2π × (1/3)/0.9 = 2.326 rad/s. Therefore, using the above Eq. (2c), the acceleration for the spinner, α, can be determined as −0.7268 rad/s².

The hyperthermia responses of both the precursor and the composite were tested. The results were shown in Table 2. In the second column of the table, the notation: Bi₂Te₃/PAN nanofiber mat refers to the material just processed by the electrohydrodynamic casting without heat treatment. Sometimes we called it as the precursor for the Bi₂Te₃/CNF composite, the material after the oxidation stabilization and high temperature heat treatment. The data in the first row refer to the time intervals. In the second row, the hyperthermia behavior of the Bi₂Te₃/PAN nanocomposite fiber was revealed. After the PAN fiber was partially carbonized and converted into a polymeric carbon, the hyperthermia behavior does not change much as can be seen from the comparison of the data shown in rows 2 and 3. The reason for this behavior is that Bi₂Te₃ has stronger hyperthermia than either the PAN fiber or the polymeric carbon fiber. As the two composites were exposed to the microwave, the Bi₂Te₃ powders responded to the external electromagnetic field. This caused the temperature going up much faster than that either by the substrate (the tissue paper) or by the electrohydrodynamic cast PAN fiber.

Comparing the hyperthermia property of the composite fiber mat made in this work with that of other nanofibers was also made. As reported early in Ref. [53], magnetic responsive particles such as iron oxide and cobalt oxide were effective in generating hyperthermia due to the magnetic hysteresis associated with the permanent magnetic domains in such oxides. In Ref. [54], a Sb–Te alloy particle-loaded polymeric carbon nanofiber composite was prepared using the similar electrohydrodynamic casting method as described in this work. The conversion from electromagnetic wave energy to the heat energy was studied as well. In fact, it is found that either the iron or the cobalt-based oxide shows stronger hyperthermia behavior than that of the currently made PAN or CNF loaded with Bi₂Te₃ or Sb₂Te₃ powders. That is why in current biomedical applications, iron oxide, nickel oxide, and cobalt oxide are commercialized (for magnetic resonance imaging and invasive hyperthermia treatment). However, the polymeric carbon-based material has the unique property of the tunable electrical conductivity [55]. The other advantage is that carbon is considered biocompatible with human body tissues [56]. Such advantages may allow the composite material as made in this work to have some potential applications in biomedical field.

The electrohydrodynamic casting has been successfully used for manufacturing the Bi–Te/PAN nanofiber composite material. The subsequent processes, the intermediate oxidation stabilization, and high temperature heat treatment, allow the PAN nanofiber to be converted into the partially carbonized nanofiber (CNF). The hydrogen atmosphere can effectively prevent the Bi–Te alloy particles from oxidation.

The scanning electron microscopic analysis reveals the uniform distribution of the Bi–Te microparticles in the fiber mat. The CNF forms networks connecting the Bi–Te particles. The fibrous morphology of the composite mat shows the random orientation of the PAN fiber and CNF.

Thermoelectric response tests indicate that the Bi–Te alloy particle/CNF composite material is n-type. The absolute value of the Seebeck coefficient of the Bi–Te/CNF, 4 mV/K, is much higher than that of the currently available carbon-based thermoelectric materials or Bi–Te alloys.

Photovoltaic response measurement of the heat-treated composite mat specimen shows the p-type semiconducting behavior of the Bi–Te/CNF because only the CNF responds to the visible light. The composite material is sensitive to visible light due to the polymeric carbon network with the conjugated structure. It has the potential to be used as photoelectric sensors in the wide spectrum covering the frequency range from infrared to visible light.

The addition of Bi–Te into PAN or CNF can enhance the hyperthermia behavior of either PAN or CNF significantly, which is helpful for the electromagnetic energy to heat energy conversion of the composite fiber mat. Photoelectrical response of the composite fiber mat is demonstrated by measuring the open circuit voltage of the specimens under infrared or visible light. The maximum absolute value of the voltage generated is about 0.5 V. The composite fiber mat have multiple functions for potential applications such as photoelectric energy conversion, light sensing, mechanical motion monitoring, and electromagnetic wave signal detecting.

The SEM images were made possible through the NSF MRI Grant. We acknowledge the support by the 2016–2017 and 2017–2018 Provost's Teacher-Scholar program, the 2016–2017 SPICE grant program, and the 2016–2017 RSCA grant program. Mr. Anan Hamdan is appreciated for his assistance in the SEM experiments.

• •

  Directorate for Engineering, National Science Foundation (Grant No. CMMI-1333044).

• •

  Division of Materials Research (Grant No. DMR-1429674).