Creation of Flexible Heterogeneously-Doped Carbon Nanotube Paper PN Diodes to Enhance Thermoelectric and Photovoltaic Effects

Abstract

This study investigates the fabrication and characterization of flexible PN diode devices using phosphorus- and boron-doped carbon nanotube (CNT) paper, also known as Buckypaper (BP). The BP substrate is fabricated from multi-walled carbon nanotubes (MWCNTs) and doped with phosphorus and boron to form N-type and P-type semiconductors, respectively. Various experimental techniques, including Raman spectroscopy, Hall effect measurements, and scanning electron microscopy (SEM), are employed to analyze the properties of the doped BP. The results reveal that the current-voltage (I-V) and capacitance-voltage (C-V) characteristics preliminarily exhibit the basic electrical properties of a diode after doping with P-type and N-type carriers. Subsequently, optimized vertical stacking combined with parallel electrode configurations for the BP diode devices demonstrates that vertical series stacking gradually enhances the thermoelectric voltage, while horizontal parallel connections approximately scale up the thermoelectric and photovoltaic voltages proportionally. The findings underscore the critical role of creating heterogeneously doped CNT-paper PN junction electric fields in improving the performance of carbon-based semiconductor devices. Furthermore, we demonstrate that these directionally oriented energy devices, when stacked, can form modular systems with enhanced efficiency. This work highlights the potential of flexible carbon material-based devices for advanced thermoelectric and photovoltaic applications.

1. Introduction

Since the discovery of carbon nanotubes (CNTs) by Sumio Iijima in 1991 [1], their exceptional electrical, mechanical, and optical properties have attracted significant attention [2,3,4,5,6,7,8]. CNTs have been extensively studied for their potential applications in nanoelectronics, sensors, solar cells, and flexible devices [9,10,11]. However, the inherent variability among individual CNTs and their nanoscale dimensions poses challenges for fabricating functional devices from individual nanotubes. Therefore, the ability to assemble CNTs into macroscopic structures such as films, fibers, and paper while preserving their intrinsic properties is crucial for their integration into functional devices.

Buckypaper (BP), composed of randomly arranged CNTs, can be fabricated through vacuum filtration of CNT suspensions [12,13,14,15]. Among these structures, BP stands out due to its scalability, flexibility, and ease of fabrication, making it particularly suitable for electronic applications. For instance, in the modification of BP’s thermoelectric properties, immersing BP in concentrated sulfuric acid or nitric acid can enhance its thermoelectric Seebeck coefficient from 10 μV/K to 20 μV/K, while oxygen doping increases it to 40 μV/K [16,17].

In addition to thermoelectric enhancements, early studies have demonstrated CNTs’ potential in photovoltaic applications. For example, Liu et al. sputtered zinc oxide onto the surface of BP and subjected it to thermal annealing, which improved its photoelectric response [18]. Various strategies such as doping, functionalization, blending, and device engineering [19,20,21] have been employed to improve CNTs’ electrical properties. However, these approaches typically focus on uniform, global enhancements to the material, leading to improved performance but lacking a doping concentration gradient or the ability to induce an intrinsic electric field with directional characteristics within the material.

We hypothesize that creating anisotropic doping in CNT materials could enable the development of more powerful functional device modules by leveraging the principles of device property superposition. To this end, we investigated the fabrication of PN diodes using heterogeneously-doped CNT paper. By creating non-isotropic built-in electric fields at the PN junctions, we aim to enhance thermoelectric and photovoltaic performance.

In our experiments, optimized weight percentages of phosphorus and boron were doped onto opposite sides of the CNT paper to form a PN junction diode. Subsequently, thermoelectric performance was significantly enhanced by employing optimized vertical series stacking combined with parallel connections. While phosphorus and boron doping are commonly used in semiconductor applications, the use of PN junctions’ built-in electric fields, combined with superposition principles to enhance BP’s thermoelectric performance, is demonstrated here for the first time in soft electronic material.

The following sections detail the fabrication, characterization, experimental processes, results, and discussion of these advanced semiconductor devices.

2. Experimental Procedure

2.1. Fabrication Process

The fabrication process begins with preparing a standalone, flexible carbon-based semiconductor substrate known as Buckypaper (BP). Multi-walled carbon nanotubes (MWCNTs) purchased from CONJUTEK (CONJUTEK, Taipei City, Taiwan), with a purity greater than 95%, a diameter of 10–50 nm, and a length of 10–100 μm, were used. A total of 60 mg of MWCNT powder was placed into a beaker, and 500 mL of a dispersant solution consisting of deionized water and 1 wt.% Triton X-100 (EMPEROR chemical CO., LTD, Taipei City, Taiwan), surfactant was added. This mixture was sonicated for 30 min using an ultrasonic homogenizer with a power of 60 W to preliminarily disperse the CNT powder into the diluted surfactant solution. Subsequently, deionized water was added, and the solution was divided into five 500 mL portions, each containing CNT powder and trace dispersant. Each portion underwent an additional 20 min of ultrasonic dispersion. Proper dispersion, indicated by the absence of sedimentation, ensured the stability of the suspension.

After the dispersion process was complete, BP was collected on filter paper using vacuum filtration. The BP was then immersed in isopropanol to dissolve and remove the Triton X-100, followed by thorough rinsing with deionized water. Once dried, BP was peeled off the filter paper, forming a flexible BP with a diameter of 4.5 cm, thickness of 140 μm, resistivity of 2.29 × 10⁻² ohm-cm, P-type semiconductor carrier concentration of 2.29 × 10¹⁷ cm⁻³, and porosity of 78%. Its appearance and flexibility are shown in Figure 1a. Compared to Buckypaper fabricated using spray coating or powder pressing methods, which tend to lack flexibility and are prone to cracking, the vacuum filtration method employed in our approach offers distinct advantages. Specifically, this method ensures that individual carbon nanotubes are initially well-dispersed independently within the suspension, achieving a three-dimensional spatial distribution. Subsequently, the nanotubes are allowed to overlap and intertwine within this 3D space, resulting in exceptional flexibility. Notably, the resulting Buckypaper can be bent to angles exceeding 90 degrees without fracturing.

2.2. Device Fabrication

Next, we transformed the BP into functional semiconductor devices by doping with N-type and P-type charge carriers. First, phosphorus pentoxide (P₂O₅) powder was dissolved in deionized water to prepare a 7.5 wt.% phosphoric acid solution, which was uniformly sprayed onto the BP surface. The sample was then subjected to a thermal diffusion process at 450 °C for one hour under vacuum, allowing phosphorus atoms to diffuse into one side of the BP, forming an N-type semiconductor. Hall measurements indicated an N-type carrier concentration of 1.36 × 10²⁰ cm⁻³.

For P-type doping, boron trioxide (B₂O₃) was dissolved in deionized water to create a 5 wt.% boric acid solution. The other side of the BP was treated with this solution under the same thermal diffusion conditions (450 °C for one hour), increasing the concentration of P-type carriers. Hall measurements revealed a P-type carrier concentration of 6.21 × 10¹⁹ cm⁻³. The diode fabricated through this process is referred to as the Buckypaper diode (BP diode).

To further investigate the defect changes on the CNT surface and the crystallinity alterations caused by thermal doping, Raman spectroscopy (Andor BWII RAMaker SR-750) with a 532 nm laser was used to scan BP samples on glass slides. We utilized a Hall measurement system (Keithlink Tech., Taipei, Taiwan) with the van der Pauw method for four-point probe measurements to verify the carrier concentration changes induced by different doping levels.

The morphology and composition of doped BP were observed using field-emission scanning electron microscopy (FESEM, JEOL JSM-7000F, Tokyo, Japan), and the elemental composition of the doped carriers was analyzed by energy-dispersive X-ray spectroscopy (EDS, OXFORD X-ACT, Abingdon, UK). BP’s resistivity was measured using the four-point probe method and a Keithley 2400 (Cleveland, OH, USA) multifunctional meter. The thermoelectric voltage was measured using a thermoelectric module coupled with a Keithley 2000 multifunctional voltmeter, while photovoltaic characteristics were measured using a solar simulator (A.M. 1.5G, Forter Tech. LCS-100, Taichung, Taiwan) at an intensity of 1000 W/m². The photovoltage was recorded using a Keithley 2000 multimeter.

3. Results and Discussion

From Figure 1a, it is evident that the Buckypaper fabricated from carbon nanotubes exhibits excellent flexibility. Figure 1b shows the Raman spectra before and after doping with charge carriers. Typically, the D-band is associated with defects in the material, while the G-band serves as an indicator of the degree of graphitization in carbon materials. The 2D-band, a second-order scattering process of the G-band, provides information about interlayer interactions and multilayer structures.

The Raman spectra reveal that during the thermal doping process, high-temperature treatment leads to crystal reorganization, thereby enhancing crystallinity, as reflected by the increased intensity of the G-band. However, the introduction of doped carriers also results in an increase in defects within the carbon nanotubes, manifested by an increase in the D-band. As the power increases, the peaks of the D-, G-, and 2D-bands become broader, indicating the broadening effect due to changes in the material’s structure or chemical properties.

Further investigation suggests that additional defects or changes in the chemical structure of the carbon nanotube surface induced by thermal carrier doping may contribute to the observed peak broadening and the increased intensity of the defect-related peaks. The doping of carriers often introduces surface functionalization, adding functional groups or disrupting the existing structure of the carbon nanotubes. The broadening of peaks in the Raman spectrum may indicate an increase in structural disorder, which correlates with the introduction of defects. The increase in structural irregularity can lead to greater energy scattering, resulting in peak broadening. If the doping of carriers partially disrupts the graphite structure, changes in the shape of the G-band peak may be observed. Additionally, carrier doping could cause crosslinking between carbon nanotubes or the formation of new carbon-based structures, which may affect the 2D-band by altering electronic interactions and modifying the 2D-band characteristics.

These findings emphasize the critical role of carrier doping in modulating the properties of carbon-based materials.

From Figure 2a,b, it can be observed that during the dispersion process of the carbon nanotube powder, friction between the ultrasonic homogenizer probe and the material resulted in the release of trace metal elements, particularly titanium and aluminum. After doping with boron oxide, the analysis showed a boron content of 11.65 wt.%, which facilitated the formation of P-type semiconductors in the Buckypaper. On the other hand, doping with phosphorus oxide resulted in a phosphorus content of 2.90 wt.%, leading to the formation of N-type semiconductors in the Buckypaper. Consequently, this process formed a BP diode with a PN junction.

Generally, to ensure that the measured physical property trends of device samples are accurate, multiple samples are fabricated under the same process conditions, and each sample undergoes repeated measurements. In our study, we prepared eight identical samples and conducted multiple measurements on each. From these results, we categorized the data to identify trends that most accurately represent the physical characteristics of the devices in the experiment. Figure 3a shows the current–voltage relationship of the Buckypaper PN diode. The current and voltage equation for a non-ideal diode is as follows:

$$I=I_S·(e^{{\displaystyle \frac{q·v}{nKT}}}-1)$$

(1)

where v is the applied bias voltage across the diode, I is the current generated by the diode under the applied bias, I_S is the reverse saturation current, n is the ideality factor, q is the elementary charge (Coulomb’s charge), k is the Boltzmann constant, and T is the absolute temperature.

By substituting the constants and setting the temperature to room temperature (T = 300 K) into the equation, we can perform data fitting to extract the reverse saturation current (I_s) and the ideality factor (n). After fitting the data, the ideality factor was found to be 32.35. In comparison, the typical range for the ideality factor in silicon diodes is between 1 and 10. Additionally, the extracted reverse saturation current was 1.08 × 10⁻⁶ A, indicating a relatively weak junction between the P-type and N-type semiconductors, which leads to a slightly higher leakage current. However, the current–voltage trend clearly indicates the formation of a built-in electric field at the PN junction, resembling that of a typical diode.

Figure 3b shows that as the forward bias increases, the depletion region between the PN junction narrows, resulting in an increase in capacitance. Moreover, in the capacitance–voltage relationship, we performed a voltage sweep from negative to positive voltage. When initially applying a negative voltage, due to the large specific surface area of the Buckypaper, the capacitance tends to decrease first due to the charging effect of the bulk structure. Only after this does the capacitance reflect the expected trend of increasing as the PN junction capacitance rises with increasing voltage.

Figure 4b illustrates the thermoelectric measurement system. At the top, a cooled aluminum block connected to a chiller maintains a set temperature of 20 °C. This block can be vertically adjusted to make contact with the sample, ensuring a stable cold-side temperature on the upper surface of the sample. On the lower side, a heated metallic plate capable of reaching temperatures up to 200 °C provides the heat source. Both the cold and hot blocks are coated with a thin, thermally conductive yet electrically insulating layer. Additionally, silver foil electrodes are affixed to the surfaces of both blocks. When the sample is clamped between the two blocks, the system enables thermoelectric measurements to be performed under programmatic control.

Figure 4d depicts the AM1.5G-standard solar photovoltaic measurement system (A.M. 1.5G, Forter Tech. LCS-100, Taichung, Taiwan). A light source equipped with a shutter mechanism allows the sample to be exposed to light in periodic on-off cycles. The system is housed within a black, light-shielded metal enclosure designed to prevent electromagnetic interference. Photovoltaic measurements can also be conducted using programmatic control for precise operation.

From Figure 4a, it can be observed that after the formation of the PN junction in the device, the number of electron–hole pairs in the depletion region increases with the rising temperature gradient, driven by the thermal energy enhancement. Additionally, the built-in electric field at the PN junction further drives the carriers’ energy, thereby increasing the thermoelectric voltage across each unit. By vertically stacking single PN thermoelectric devices, from one layer to three layers as seen in trajectories (1) through (3), the thermoelectric voltage gradually increases with the number of PN junction layers. Although increasing the number of PN junctions helps improve the thermoelectric voltage, the temperature difference experienced by each layer is only one-third of the original, meaning the voltage does not increase proportionally to the number of layers. However, a key advantage is that stacking vertically does not increase the overall footprint of the device, and multi-layer PN thermoelectric structures can still achieve higher thermoelectric voltages.

On the other hand, when the PN thermoelectric devices are placed horizontally and connected in series using silver foil electrodes for thermoelectric measurements, each device experiences the full temperature gradient. After horizontally connecting the devices in series, the thermoelectric voltage increases roughly in proportion to the number of connected devices, as shown in trajectories (1), (4), and (7). This behavior follows the same principle as voltage stacking in battery series configurations. Finally, by vertically stacking three PN thermoelectric devices into a multilayer structure and then connecting each group horizontally, the maximum thermoelectric voltage can be achieved, as indicated by trajectory (9). From trajectory (1) to trajectory (9), the Seebeck coefficient increases approximately from 10 μV/K to 30 μV/K. It is believed that, by adhering to the principle of superposition and further stacking additional PN junction elements, the Seebeck coefficient could potentially exceed 100 μV/K.

In terms of photovoltaic measurements, vertical stacking of BP layers blocks light due to the opaque nature of the material, preventing the lower layers from enhancing the photoelectric effect. However, when PN photovoltaic diodes are connected horizontally in series, as shown in Figure 4c, the photovoltage increases approximately in proportion to the number of connected devices, following the same principle as voltage stacking in battery series configurations.

4. Conclusions

This study successfully demonstrates the fabrication of flexible devices using carbon-based materials, which are lighter in weight compared to solid rigid materials, offering significant advantages for portable devices. Their superior flexibility allows them to conform more effectively to various curved surfaces compared to rigid materials.

Additionally, PN junction devices were fabricated using phosphorus- and boron-doped Buckypaper to enhance thermoelectric and photovoltaic responses. By optimizing the series configuration of PN diode elements, we achieved significant improvements in both photovoltage and thermoelectric voltage. The findings highlight the critical role of PN doping in creating built-in electric fields and the importance of structural optimization in enhancing the performance of carbon-based semiconductor devices. Future work will focus on further optimizing these parameters, incorporating materials that enhance energy harvesting capabilities, and exploring broader applications in photovoltaic and energy collection systems.