Introduction

Almost all organic semiconductors currently used as part of semiconductor materials are π-conjugated oligomers and polymers1,2,3,4. However, π-conjugated polymers are lacking in natural bio-compounds. Studies on naturally occurring bio-semiconductors are scarce5. We reported an n–type bio-semiconductor based on an amorphous kenaf cellulose nanofibre6 exhibiting rectification, n–type negative resistance, and DC/AC conversion. However, owing to their thixotropic properties that create aggregates containing entangled pores, producing dense films using long fibres with an aspect ratio of more than 100 is extremely challenging7. In this study, a prototype film composed of amorphous kenaf cellulose particles (AKCPs), which were defibrillated and milled to a diameter of ~ 11 nm, was produced, and the electronic properties of AKCPs were evaluated. Subsequently, electron spin resonance (ESR) measurements8,9, the only means of observing radicals in organic materials, were performed at 295 K to investigate the origin of electrons in the n–type semiconductors. Furthermore, electron mobility was quantified via Hall measurements. An analysis of the AC impedance elucidated the tissue-dependent conduction mechanism of AKCPs. Recently, compound semiconductor devices for the skin10, stretchable electronics11 and function of the retina12 with properties similar to those of bio-semiconductors, have been reported.

Results and discussion

Semiconducting characteristics with electric storage

The DC measurement method was used to determine the voltage-controlled I–V characteristics of AKCPs for a sample with a thickness of 14 µm at 298 K within the current range of 020 mA and voltage range of –100 to 40 V (Fig. 1a). The I–V curve exhibits a clear forward rectification effect. Figure 1b illustrates the I–V characteristics for a sample thickness of 25 µm. A clear N-type negative resistance appears between approximately 87 V and 71 V, as well as a small N-type negative resistance (inset of Fig. 1a). In excess of 0 V, an increase is observed in the current value in the forward direction. Meanwhile, the I–V characteristics of a sample with a thickness of 19 µm in a current-controlled measurement at 0.105 A are shown in Fig. 1c, where the current value increases sharply from approximately 88 V during an increase from 0 to 100 V. The enlarged curve shown in the inset of Fig. 1c shows an S-type negative resistance, as in the Ni–Nb–Zr–H amorphous alloy13. Based on the semiconducting theory, the negative resistance characteristics are classified into static negative resistance characteristics, such as those of tunnel diodes and thyristors14, and dynamic negative resistance characteristics considering the carrier transfer time and material band structure specificity, such as those of impact avalanche transit diodes and Gunn diodes15. This study considers the latter (see Supplementary Information (SI). S9). The semiconductor properties of AKCPs obtained from the aforementioned experiments are shown in the SI, Fig. S4, which shows a Schottky junction n–type semiconductor16. The observed S-type negative resistance (Fig. 1c), in addition to the N-type negative properties (Fig. 1b), may be attributed to the kenaf becoming granular instead of fibrous. Figure 1d presents the R–V characterisation on a logarithmic scale from 20 to 40 V for AKCPs with a thickness of 12 μm. The R–V curve exhibits a three-field change in magnitude between 0 and 10 V, indicating a switching effect. In contrast, a specimen with a thickness of 101 µm exhibits a storage effect (Fig. 1e) that is not observed in the case of amorphous kenaf cellulose nanofibres17. The storage capacity increases with voltage, and a storage capacity of 418.5 mJ/m2 is obtained at 450 V. The discharge curve at 450 V is shown in the inset. This value is less than half the values for amorphous kenaf cellulose nanofibres18 and amorphous alumina (AAO)19 (1416.7 and 1710.3 mJ/m2, respectively). These phenomena may be attributed to an increase in the capacitance of the AKCPs with increasing thickness. The analysis results of trace impurities in the AKCPs are presented in Table S1. However, the effect of trace impurities on semiconductor properties is currently unclear. The effects of bound water will be addressed in a subsequent study.

Figure 1
figure 1

Voltage-controlled I–V characteristics of AKCPs with a thickness of 14 (a), 25 (b), and 12 µm (d) at a sweep rate of 51.5 V/s. (c) I–V characteristics in a current-controlled measurement at 0.105 A for AKCPs with a thickness of 19 μm. Inset of (c) shows an enlarged S-type figure. (e) Discharging behaviour of the AKCP device with a thickness of 101 µm for a constant current of 1 μA after 2 mA10 V charging for 5 s. Inset of (e) shows discharging behaviour.

Origin of radical electrons derived from cellulose structure

The results of the ESR measurements are displayed in Fig. 2a. The curve obtained at 295 K is the isotropic peak of a singlet. As the unpaired electrons shown in the ESR spectrum are extremely sensitive to the molecular arrangement around the electrons, the g-value of the ESR signal presents a decisive guideline for identifying the organic radical species20. Based on the molecular structure of cellulose (C6H10O5)n, the g-values of 2.004 and 2.009 correspond to the alkoxy radical CO·21. In this study, we identify the position of the radical group for the cellulose molecule relative to the radical electron, which is the origin of electron conductivity. The alkoxy groups generated at the side chain Cs (C2, C3, C6) are more reactive than the alkoxy group generated at C1 on the main chain and are rapidly deactivated by secondary reactions22; therefore, the C2–O2, C3–O3, and C6–O6 groups in Fig. 2c are excluded as radical electron candidates. The radical electrons in the AKCPs appear in cellulose via the glycosidic bond between the two glucose units and O1. Attenuated total reflection-Fourier transform infrared (ATR-FTIR) measurements were performed to confirm the existence of glycosidic bonds. The results are shown in Fig. 2b. Peaks of COC stretching motion are observed at 1184 and 884 cm−1 (Ref.23). Considering the electronegativity values of 2.20, 2.55, and 3.44 for H, C, and O, respectively, the electron-induced effects are depicted in Fig. 2d when applied to cellulose molecules (see S7 in SI for details). The locations of the appearance of the radical electrons obtained from ESR are shown in Fig. 2c, as C1–O1·–C4. The electrons induced in conventional organic semiconductors are π-electrons from the C=C double bond24. The Hall coefficient measurements showed an electron mobility of 10.66 cm2/Vs, a carrier density of –9.89 × 1015 (1/cm3), and an electric resistivity of 9.80 × 102 Ωcm at 298 K. The mobility is two orders higher than 0.5–1.0 cm2/Vs25 for amorphous Si and 0.08–2.5 cm2/Vs26,27 for π-conjugated organic semiconductors. However, Farka et al.28 reported a relative high value of 40.8 cm2/Vs in p-doped polyethylene-(3,4-dioxythiophene) (PEDOT: sulfate).

Figure 2
figure 2

(a) Singlet symmetrical ESR spectra at 298 K. (b) ATR-FTIR spectroscopic analysis at 298 K for the AKCPs. (c) Cellulose comprising (C6H10O5)n with the green, pink, and light-blue dots representing carbon, oxygen, and hydrogen atoms, respectively. (d) Electron-induced effects for cellulose structure.

Complex evaluation of I–V characteristics

To non-destructively analyse the electronic contribution of the sample, its AC impedance was measured from 1 mHz to 1 MHz. The Nyquist diagram of the impedance data, corresponding to 14 μm in Fig. 1a, is illustrated in Fig. 3a and b. The impedance of the AKCPs with respect to frequency exhibits a linear slope with a change of π/4 rad, as shown in Fig. 3b, and a combined pattern of two semicircles. The π/4 rad region (Warburg region) can be attributed to the porous sample6,29,30. The two semicircles represent a tissue composed of two fibres, such as the bast and core in kenaf31. The peak frequency fmax of the semicircle is 1.48 mHz; therefore, a relaxation time of 107.6 s can be calculated using the relationship RCtotal = 1/(2πfmax). In the low-frequency region of the Bode diagram shown in Fig. 3c, the real impedance increases sharply to 8 MΩ, whereas the imaginary impedance peaks at 1.48 mHz and decreases at relatively low frequencies. The 1.48-mHz peak corresponds to dielectric dispersion owing to interfacial polarisation in the low-frequency range32. In the phase angle diagram of Fig. 3d, the capacitance behaviour for frequencies lower than 0.073 Hz (near zero phase angle) is clearly similar to that of a parallel RC circuit. The series capacitance Cs and parallel capacitance Cp increase as the frequency decreases; however, the increase in Cs is relatively rapid. Cs plays a vital role in determining the DC I-V characteristics of the sample. In the relationship between the time constant RC and frequency, as shown in Fig. 3e, in both logarithmic displays, the time constants RCs and RCp increase almost linearly with decreasing frequency; RCs and RCp at 1 mHz are 448 and 50 s, respectively. A larger duration (from 0.1 s to a few hours) is required for practical use.

Figure 3
figure 3

(a,b) Nyquist plots as a function of frequency for the AKCP device. (c) Frequency dependence of real and imaginary impedances, (d) phase angle, and series and parallel capacitances. (e) Frequency dependence of RCs and RCp.

Mechanism of electron conduction

Based on the Nyquist diagram (Fig. 3a) and Debye relaxation peak in the Bode diagram (Fig. 3c), the equivalent electrical circuit of the AKCP can be regarded as a series coupling of three equivalent parallel circuits, as shown in Fig. 4a. Because the sample used in this study consists of 18 nanofibrils and their boundaries, the total resistance of the sample including the electrodes (Fig. 4b) is the resistance Rf of the nanofibrils in the AKCP, the boundary resistance Rfb between the AKCPs (Fig. 4c), and the electrode interface resistance Rer (Fig. 4d). The capacitance Cf of the nanofibrils in the AKCPs, the boundary capacitance Cfb between the AKCPs, and the electrode interface resistance Cer shown in Fig. 4a also play an important role in determining the semiconductor properties in this study. However, because the bonds between nanofibrils and the structure of AKCPs are unknown, precise structural analysis using soft X-rays is required.

Figure 4
figure 4

(a) Equivalent circuit corresponding to the Nyquist diagram shown in Fig. 3a,b. (b) Image of a semiconductor composed of AKCPs and their boundaries measured via the direct current method. (c) Image figure of cellulose nanoparticles. (d) Equivalent circuit for semiconductor conduction comprising fibril resistance Rf, boundary resistance Rb and electrode interface resistance Rer.

Conclusion

If cellulose, the most abundant natural renewable compound, is found to exhibit semiconductor properties, it can be used to develop many applications. In the present study, we reported the rectification effect of n-type semiconductors with N- and S-type negative resistivity properties derived from the glycosidic bond, C1–O·–C4, between two glucose units and O1 of cellulose molecules in AKCPs. The electron mobility was found to be 10.66 cm2/Vs, which is two orders of magnitude higher than that of conventional polymer semiconductors. The N- and S-type negative resistance effects may open up new fields in place of conventional p–n junction devices. We are working on conifers and hardwoods with the aim of expanding into areas different from conventionally engineered semiconductors.

Methods

The AKCP specimen was fabricated on an Si substrate via spin coating, which was performed at a speed of 400 rpm for 5 s using a 2% (w/v) AKCP/water dispersion. The AKCP films were dried in a ventilated oven at 363 K. The specimens (12 mm wide, 14–101 μm thick, and 15 mm long) were mechanically sandwiched between an Al electrode and carbon electrode on the AKCPs (SI, Fig. S4). ESR measurements were performed at 295 K using a Q–band ESR spectrometer. Hall measurements were performed at 298 K using the conventional Van der Pauw technique. The currentvoltage (I–V) and resistivityvoltage (R–V) characteristics were measured within 30 min after sample preparation and under DC voltages ranging from 210 to 210 V in air at a sweep rate of 51.4 V/s, using a Precision Source/Measure Unit (B2911A, Agilent). The AC impedance and frequency were measured using a potentiostat/galvanostat (SP-150, BioLogic Science). I–V measurements were performed using a Precision Source/Measure Unit (B2911A, Agilent).