Liquid silicon and its application in electronics

CPS can be synthesized by the three-step reaction shown in Fig. 1. This is called the Kipping method.^8–10) In our experiment, CPS has been synthesized by the same reaction as follows. Diphenyldichlorosilane was added to a stirred suspension of lithium in tetrahydrofuran, and the resulting mixture was stirred and poured into cold water. The resultant precipitate was collected by filtration, washed with water, dried in vacuum, and recrystallized to give a white powder (decaphenylcyclopentasilane). Its suspension and aluminum chloride in cyclohexane were mixed and stirred at room temperature with the bubbling of dry hydrogen chloride gas to give decachlorocyclopentasilane. The remained hydrogen chloride was removed by N₂ bubbling from the solution. The solution of decachlorocyclopentasilane in cyclohexane was added dropwise to a suspension of lithium aluminum hydride in diethyl ether at 0 °C, and the resulting mixture was stirred for 12 h and then filtered. The filtrate was concentrated, and the residue was distilled under reduced pressure to afford pure CPS as a colorless liquid. The properties of the CPS are listed in Table I. Note that CPS has a relatively high boiling point, indicating a stable molecule to handle.

Zoom In Zoom Out Reset image size

In order to determine the exact structure of the CPS molecule, we have conducted a computational study using ab initio and density-functional methods.¹¹⁾ The calculations were based on Hartree–Fock theory, Mollet–Plesset perturbation theory, and density-functional theory (DFT). The geometrical structures of CPS were optimized within the given symmetry point group. After different structures of Si₅H₁₀ with different symmetries were analyzed, three stable structures of CPS were determined by our calculations, namely, planar (D_5h), envelope (C_s), and twist (C₂), which are shown in Fig. 2. The results of our calculations at all theoretical levels employed show that the twist structure is the most stable structure for CPS, and that the two other structures of CPS, i.e., twist (C₂) and envelope (C_s), have similar energies at all the theoretical levels employed (the difference between their energies is on the order of 1.0 meV). The relative energies of each structure are summarized in Table II. These results are in good agreement with the results of previous theoretical studies.^12,13)

Zoom In Zoom Out Reset image size

Si ink is a mixture of poly(dihydrosilane) and an organic solvent. CPS can be converted to poly(dihydrosilane) by ring-opening polymerization, which occurs when UV light is irradiated to CPS. Figure 3 shows the optical absorption spectra of CPS just after distillation, CPS kept at 50 °C for 30 min after distillation, and poly(dihydrosilane). We usually use UV light of 365 nm wavelength; however, CPS just after distillation does not absorb light of 365 nm wavelength. Hence, it is interesting to investigate whether it is polymerized by the UV light that it cannot absorb. (This had been mystery for a long time.) We happened to observe that immediately after distillation, CPS is difficult to polymerize by UV light of 365 nm wavelength. In contrast, after some period at room temperature after synthesis, CPS can be polymerized at a much higher rate. We confirmed that a small amount of poly(dihydrosilane) can be formed in CPS by thermal polymerization. As shown in Fig. 3, poly(dihydrosilane) absorbs UV light of 365 nm wavelength. Figure 3 also shows that CPS kept at 50 °C for 30 min after distillation tends to absorbs longer-wavelength UV light than CPS immediately after distillation. These facts suggest that CPS polymerization is a two-step process: (1) poly(dihydrosilane) absorbs UV light, and its energy is transferred to CPS, and (2) the transferred energy is used for the ring-opening polymerization of CPS. This mechanism is just a presumption, and a more detailed study is needed.

Zoom In Zoom Out Reset image size

In order to determine the polymerization behavior, we characterized the molar mass distribution and structure of poly(dihydrosilane) by size-exclusion chromatography (SEC) combined with multi-angle laser light scattering (MALLS) analysis and viscometry (SEC-MALLS-viscometry).¹⁴⁾ The SEC-MALLS system, viscometer, and refractive index (RI) detector were connected in series by stainless-steel tubes, as shown schematically in Fig. 4. An injector was set in the glove box. Readings of the excess Rayleigh ratio R(θ) at a scattering angle θ, specific viscosity η_sp, and RI response were acquired every 0.5 s during elution. A Shodex KF-805 column was used with a guard column attached to it. The molar mass (M) and root-mean-square radius (radius of gyration, R_g) were estimated using a Guinier–Zimm plot. The intrinsic viscosity [η] was determined by calculating η_sp/c directly for each elution volume. Here, c is the concentration of the solute [poly(dihydrosilane)]. For the analysis of the polymer structure, we adopted the sphere model, which is based on the viscosity theory of colloidal particles, to analyze the molar mass dependence of [η].¹⁵⁾ In this model, [η] is described as

$$\begin{equation} [\eta] = \frac{5}{2}\frac{N_{\text{A}}}{M}\frac{4\pi}{3}R_{\text{h}}^{3}, \end{equation} \tag{ 1 }$$

where R_h is the viscosimetric radius and N_A is Avogadro's number. Starting from Eq. (1), we describe the scaling behavior of [η] vs M.

Zoom In Zoom Out Reset image size

Figure 5(a) shows the RI intensities for four samples synthesized with irradiation times of 0, 30, 60, and 240 min. The sample subjected to no irradiation (0 min) corresponds to CPS, and the inverted peak at a elution volume of 13 mL marks the end of sampling. The RI chromatogram for CPS exhibits a sharp peak at an elution volume of 12 mL. The spectrum for the sample with the 30 min irradiation time exhibits a broad peak at the lower elution volume in addition to the weak CPS peak. For the sample with the 60 min irradiation time, the broad peak at the lower elution volume grows significantly, whereas the CPS peak disappears. This suggests the completion of CPS photo-polymerization within 60 min of irradiation.

Zoom In Zoom Out Reset image size

The differential molecular weight fraction of poly(dihydrosilane) is plotted as a function of the molar mass for the four samples in Fig. 5(b). The sample with the irradiation time of 0 min has a sharp peak at a molar mass of 150 g/mol, confirming the existence of CPS. Hence, the decrease in the intensity of the peak at 150 g/mol indicates how the molecular weight distribution spreads out when CPS is converted into poly(dihydrosilane) during the first 60 min of irradiation. The molar mass for poly(dihydrosilane) ranges broadly from 10² to 10⁶ g/mol. The molar mass distribution changes significantly for the samples with irradiation times of up to 60 min, however, minor changes in the molar mass distribution are observed for the samples with irradiation times longer than 60 min.

Here, we explore the structure of poly(dihydrosilane) on the basis of the scaling behaviors of [η] and R_g.¹⁴⁾ Logarithmic plots of [η] vs M and R_g vs M for the sample with the irradiation time of 240 min are shown in Figs. 6(a) and 6(b), respectively. These plots exhibit a scaling feature, which was fitted by the linear functions described in Eqs. (2a) and (2b):

$$\begin{align} [\eta] & = 0.414M^{0.206}, \end{align} \tag{ 2a }$$

$$\begin{align} R_{\text{g}} & = 0.033M^{0.470}. \end{align} \tag{ 2b }$$

The scaling relation [η] = KM^α is known as the Mark–Houwink–Sakurada equation.¹⁶⁾ The data fitted by Eq. (2a) was in good agreement with the scaling relation. The solid line in Fig. 6(a) was fitted to [η] between 10³ and 3 × 10⁶ g/mol, whereas the solid line in Fig. 6(b) was obtained by fitting R_g between 1.099 × 10⁶ and 4.457 × 10⁶ g/mol. Here, we focus on the fitting of [η] vs M and interpret the scaling behavior on the basis of the sphere model.¹⁵⁾ Since [SiH₂] (mass M_m = 30.102 g/mol) is considered to be a monomer unit of poly(dihydrosilane), the degree of polymerization (number of monomer units: N) of a polymer with molar mass M is given by N = M/M_m. Thus, the radius of gyration may be written as

$$\begin{equation} R_{\text{g}} = aN^{\nu}, \end{equation} \tag{ 3 }$$

where a denotes the "effective length" of [SiH₂]. By substituting Eq. (3) into Eq. (1) and by using the parameter ρ = R_h/R_g, the Mark–Houwink–Sakurada equation of [η] = KM^α is obtained, where α = 3ν − 1 and the coefficient K is given by

$$\begin{equation} K = \frac{10\pi}{3}\left(\frac{\rho a}{M_{\text{m}}^{\nu}}\right)^{3}N_{\text{A}}. \end{equation} \tag{ 4 }$$

Zoom In Zoom Out Reset image size

According to the present sphere model, the radius of gyration exponent is estimated to be ν = 0.402 for α = 0.206. Since ν = 0.410 in Eq. (2b), the agreement between the two results suggests that the sphere model for the viscosity can be appropriately applied to poly(dihydrosilane) with the monomer unit [SiH₂]. If the poly(dihydrosilane) has a straight-chain structure, α should be in the range from 0.5 for Flory theta solvent to about 0.8 in a good solvent.¹⁶⁾ However, the experimental value of 0.206 was much smaller than the cutoff value for a straight-chain polymer. Thus, we can presume that the poly(dihydrosilane) exists as a branched-chain structure.

With respect to the polymer shape, ρ = R_h/R_g gives a measure of the compactness of the molecule. By using Eqs. (2a) and (2b), ρ = 1.09 and ρ = 1.08 were determined for M = 1.099 × 10⁶ and 4.457 × 10⁶ g/mol, respectively. This also suggests that the poly(dihydrosilane) has a particle-like compact shape. For ρ = 1.09, by using Eq. (4) and K = 0.414 mL/g in Eq. (2a), a was estimated to be 0.145 nm. This value is in good agreement with the bond lengths between Si and H in various molecules.^17,18)

It is concluded that poly(dihydrosilane) has a branched structure with a particle-like compact shape in cyclohexene at 25 °C. The existence of a branch point (SiH) was also predicted and confirmed with using ¹H-NMR and Fourier-transform infrared spectroscopy (FT-IR).¹⁴⁾

Once Si ink is synthesized, the next process involves coating the Si ink onto a desired substrate to make a poly(dihydrosilane) film. However, the formation of a film having homogeneous thickness is not straightforward. Immediately after Si ink is spin-coated, the solvent starts to evaporate, leaving behind the solute [(poly(dihydrosilane) in this case] on the substrate. During this stage, various types of dried solid patterns can be generated depending on the conditions of several aspects of the process. In some cases, a pattern such as a coffee stain forms, whereas in others, a multi-lined or dotted pattern appears. Uniform film formation is not always guaranteed.

Whether or not a good film is formed strictly depends on the behavior of the solute substance on the substrate during drying. To clarify the phenomenon, we investigated the stability of poly(dihydrosilane) films on solid substrates by a comparison between the observed micrographs at room temperature and the calculated values of the Hamaker constant (A_ALS) for the air/liquid/solid substrate. To efficiently compare the observations and A_ALS values, we selected five types of SiO₂-based substrates (quartz, glass, and optical glasses A, B, and C) and a TiO₂ substrate, because SiO₂ is a popular substrate for silicon devices. Since the evaluation of the Hamaker constant is essential for determining the stability of polymer films, we adopted a simple spectrum method (SSM),^19,20) which is adequate for a system composed of materials with different resonance frequencies, such as solids and liquids. For comparison, we also calculated the A_ALS values basis of the Tabor–Winterton approximation (TWA).²¹⁾ We measured the optical parameters of several substances by an Abbe refractometry, ellipsometry,²²⁾ and FT-IR to determine the optical parameters in the London-dispersion (LD) spectra of the SSM.

The formation of Si ink is as follows. The wavelength, intensity and irradiation time were 365 nm, 1 mW/cm², and 60 min, respectively. Under these conditions, CPS was completely converted to poly(dihydrosilane) with a broad molar mass distribution (M_W = 10²–10⁶ g/mol). Then it was dissolved in cyclooctane at 1.5 wt %. The film was coated on a solid substrate by spin-coating at 2000 rpm for 30 s. The resultant film thickness was 40 ± 5 nm. All processes were carried out in a glove box, where the oxygen concentration and dew point in the glove box were less than 0.5 ppm and −75 °C, respectively. First, the coated-liquid film was allowed to stabilize overnight at room temperature in the glove box; then, only the film surface was oxidized slowly to freeze the pattern in the film and deactivate it. The resulting film, which was regarded as a poly(dihydrosilane) film covered with a thin SiO₂ layer, was taken out into the atmosphere, and the films were observed using an optical microscopy and SPM.

Figure 7 shows the micrograph of the poly(dihydrosilane) films coated on six types of substrates.²³⁾ The A_ALS value calculated by the SSM is noted on each image. Two types of prominent features, i.e., dot arrays and continuous figures, are observed in Fig. 7. A large number of cyclic dots in the liquid films on the quartz [Fig. 7(a)] and glass substrates [Fig. 7(b)] indicates film rupture. The separations among the dots in the film on glass were slightly longer than those among the dots in the film on quartz. The film on optical glass A showed two types of patterns: a patch like configuration of dots in Fig. 7(c) and a continuous pattern in Fig. 7(c'). In the present case, the two types of patterns coexisted in the same film. The images of the continuous films on optical glasses B and C and TiO₂ [Figs. 7(d)–7(f)] were observed with high reproducibility. The profile of a single dot in the array was measured by SPM. Figure 8 shows a three-dimensional (3D) image obtained by atomic force microscopy (AFM) for a typical dot on the glass substrate shown in Fig. 7(b).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We calculated the Hamaker constants based on the basis of the model shown in Fig. 9 by the SSM and TWA methods. The stability of the liquid film is determined by the free energy per unit volume F(L) = γ_SL + γ_LA + P(L), where L is the film thickness, γ_SL and γ_LA are the interfacial energies of the solid/liquid and liquid/air boundaries, respectively, and P(L) is the intermolecular interaction. When L is less than 1 µm, the energy term arising from gravity is ignored. For low-polarity materials, P(L) is dominated by van der Waals energy: P(L) = W_ALS. The W_ALS in the nonretarded Lifshitz theory^24,25) is given by

$$\begin{equation} W_{\text{ALS}} = \frac{A_{\text{ALS}}}{12\pi L^{2}}. \end{equation} \tag{ 5 }$$

According to this formula, a liquid film with A_ALS > 0 is stable, and that with A_ALS < 0 is unstable.²⁶⁾ The A_ALS values determined by the SSM (closed circles) and TWA (open squares) are plotted in Fig. 10 as a function of the refractive index of each substrate measured at 589 nm. The results for both the SSM and TWA indicate that the liquid films on quartz and glass are unstable (A_ALS < 0), and the liquid films on optical glasses B and C and TiO₂ are stable (A_ALS > 0). The TWA and SSM gave opposite signs for A_ALS for the film on optical glass A. Our observed micrographs are consistent with the stability of the liquid films determined from A_ALS, except for the film on optical glass A. The stable films with A_ALS > 0 showed a continuous figure, while the unstable films with A_ALS < 0 showed an array of cyclic dots resulting from rupture. The observed image of the liquid film on optical glass A showed the so-called metastable state caused by the appearance of both a continuous figure and a patchlike configuration of dots. These results will provide basic and important information on how to prepare uniformed poly(dihyrosilane) films on solid substrates.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 11 shows a photographic image of the pyrolyzed films with a thickness of 50 nm, coated on a quartz substrate 2 × 2 cm² in size at pyrolysis temperatures (T_p) between 270 and 420 °C. The films prepared at T_p ≤ 270 °C are transparent, and a drastic change from transparent to yellow color appears when T_p is increased from 270 to 300 °C. This change in color can be attributed to a reduction in E_g from that of poly(dihydrosilane) of 6.5 eV to lower values. The transparency of films prepared at T_p ≤ 270 °C corresponds to an insulator material, whereas the yellow and brown color of films pyrolyzed at T_p = 300–420 °C indicates the presence of semiconductor materials. From this, a transformation from a polymer to a solid amorphous semiconductor can occur near T_p = 270 °C.

Zoom In Zoom Out Reset image size

To study the pyrolysis directly, TG and DTA measurements of poly(dihydrosilane) were conducted. Figure 12 shows the TG signal (bold solid line) and the DTA signal (thin solid line) as a function of heating temperature T_h. Also shown is the temperature/time derivative DTG (dashed line) of the TG signal. The TG data indicate a weight loss of about 58% with heating up to 360 °C, and consequently only 42% of the material remains in the form of hydrogenated amorphous silicon. In agreement with thermal desorption spectroscopy (TDS) measurements.¹⁾ One may assume that the weight reduction is caused by the desorption of H₂ and SiH_x. According to the DTG curves, major weight losses take place at T_h = 100, 200, and 300 °C. A broad exothermal peak is found in the DTA signal extending from 80 to 450 °C with a maximum near 320 °C. These results suggest that the transformation from a polymer to a cross-linked amorphous silicon network (i.e., Si–Si bond construction) is particularly active between 300 and 360 °C.

Zoom In Zoom Out Reset image size

Figures 13(a) and 13(b) show the first-order Raman spectra of the films prepared at T_p in the range of 270–330 and 330–420 °C, respectively. The Raman data were averaged over 10 measurements for each sample in a backscattering geometry using a He–Ne laser line (633 nm). In Fig. 13(a), typical phonon bands of a-Si:H^28,29) are visible at T_p = 330 °C, however, the films prepared at T_p = 270 and 300 °C show no clear peaks corresponding to the amorphous silicon phonon bands. In the literature, the a-Si:H Raman phonon bands at 476, 387, 288, and 154 cm⁻¹ have been assigned to the transverse optical (TO), longitudinal optical, longitudinal acoustic, and transverse acoustic bands, respectively. Figure 13(b) shows the Raman spectra of the films prepared at T_p = 330, 360, 390, and 420 °C, where the intensity was normalized to the same TO height. The full width at half maximum (FWHM) of the TO band vs T_p is plotted in Fig. 13(c).

Zoom In Zoom Out Reset image size

Figures 14(a) and 14(b) show the IR absorption coefficient α(ω) in the wavenumber ranges of 1950–2200 and 550–750 cm⁻¹, respectively, for the annealed films prepared at T_p = 270–420 °C. The absorption peaks near 2000 and 2070–2090 cm⁻¹ at T_p ≥ 360 °C are assigned to the stretching modes of Si–H bonds embedded in the bulk material and to the stretching modes of Si–H/Si–H₂ bonds at the void surface, respectively.³⁰⁾ The absorption band near 640 cm⁻¹ in Fig. 14(b) is attributed to the Si–H wagging mode.³²⁾ Only that of the sample prepared at 270 °C deviates significantly from this value. As seen in Fig. 14(a), the Si–H stretching modes of all the samples are centered near 2080 cm⁻¹, except for the samples prepared at the rather low temperatures of 270 and 300 °C, which shifted to a higher wavenumber. We evaluated the sample quality using the infrared microstructure factor R expressed as

$$\begin{equation} R = \frac{I_{2090}}{I_{2000} + I_{2090}}, \end{equation} \tag{ 6 }$$

where I₂₀₉₀ and I₂₀₀₀ are the integrated intensities of the oscillation modes near 2070–2090 and 2000 cm⁻¹, respectively. From the spectral data in Fig. 14(a), the estimated R ranges from 0.91 (T_p = 360 °C) to 0.80 (T_p = 420 °C), however, it is known that device-grade a-Si:H films show R < 0.2^30,31) because of the domination of the 2000 cm⁻¹ peak. The large values of R for our films imply a void-rich structure. Although the origin of the peak shift at T_p ≤ 330 °C is not yet clear, it may be related to the oxidization or the change in the local environment of the Si–H dipole.^33,34) This suggests that the films prepared at T_p ≤ 330 °C are of low quality.

Zoom In Zoom Out Reset image size

The degrees of hydrogenation and oxidation were determined by FT-IR and SIMS. Figure 15 shows the hydrogen and oxygen contents as a function of T_p. The open circles show N_H/N_Si determined by FT-IR, in which the density of silicon is assumed to be N_Si = 5 × 10²² cm⁻³. The density of hydrogen, N_H, is evaluated from α(ω) in Fig. 14(b) using Eq. (7) with the proportional constant A₆₄₀ = 2.1 × 10¹⁹ cm⁻²,³⁵⁾

$$\begin{equation} N_{\text{H}} = A_{640}\int\frac{\alpha(\omega)}{\omega}\,d\omega. \end{equation} \tag{ 7 }$$

Since the hydrogen content of poly(dihydrosilane) is 67 at. %, the hydrogen concentrations of 9.2–10.6 at. % (by SIMS) of films prepared at T_p ≥ 360 °C indicate a remarkable reduction in the hydrogen content during the pyrolysis. The void-rich structure concluded from the large R value of films prepared at T_p ≥ 360 °C is presumably due to the 3D shrinkage due to the weight loss. The high oxygen content of the films prepared at T_p ≤ 330 °C indicates that the construction of the amorphous silicon network is not dense enough, thus it oxidizes rather easily in air.

Zoom In Zoom Out Reset image size

Figure 16 shows the results of absorption measurements in the UV–vis range for the films prepared at T_p = 300 to 420 °C. The data are depicted in a Tauc plot³⁶⁾ as

$$\begin{equation} (\alpha E)^{1/2} = B^{1/2}(E - E_{\text{g}}), \end{equation} \tag{ 8 }$$

where α is the optical absorption coefficient, E is the photon energy in eV, B is a constant, and E_g is the optical gap (called the "Tauc gap"). B is found to be 5.7 × 10⁵ and 7.7 × 10⁵ eV⁻¹ cm⁻¹ for T_p = 300 and 330 °C, respectively, and 6.7 × 10⁵ eV⁻¹ cm⁻¹ for T_p ≥ 360 °C. The T_p dependence of E_g in Fig. 16(b) can be understood by the dehydrogenation during pyrolysis, as shown in Fig. 14. For T_p exceeding 360 °C, the absorption is no longer dependent on the hydrogen content, and E_g = 1.64 eV is obtained. This is consistent with E_g = 1.5–1.6 eV observed in the vacuum deposited counterparts.^37,38)

Zoom In Zoom Out Reset image size

Typical values of the dark conductivity (σ_d), photoconductivity (σ_p), and spin density (N_s) at room temperature are plotted in Fig. 17 as a function of T_p. With increasing T_p from 300 to 330 °C, σ_d increases slightly, followed by a marked decrease at T_p = 360 °C, and reaches 5.1 × 10⁻¹¹ S/cm at T_p = 420 °C, while σ_p is on the order of 10⁻⁷ S/cm, except for the minimum at T_p = 360 °C. N_s, starting from 2.5 × 10¹⁶ cm⁻³ at T_p = 300 °C, exhibits a maximum in the film prepared at T_p = 360 °C, and reaches 2.8 × 10¹⁷ cm⁻³ for the high-T_p films. Since the paramagnetic spin density corresponds to the number of dangling bonds, which act as trapping sites, the reduction in N_s may lead to an increase in σ_p.³⁹⁾ The increase in N_s in the range from T_p = 300 to 360 °C is attributed mainly to the increase in the number of dangling bonds due to the breaking of the Si–H and Si–Si bonds. In contrast, the decrease in N_s in the range from T_p = 360 to 420 °C suggests that the high T_p has an annealing effect to reduce the number of dangling bonds [see also the FWHM data in Fig. 13(c)]. The photosensitivity (i.e., ratio of σ_p to σ_d) of the films prepared at T_p ≥ 360 °C is on the order of 10³, whereas that of device-grade vacuum-processed a-Si:H films is 10⁵–10⁶.^37,40) This low photosensitivity indicates the rather poor quality of the solution-processed films, which is presumably related to the high N_s > 10¹⁷ cm⁻³ for T_p ≥ 360 °C. Note that device-grade vacuum-processed a-Si:H films generally have N_s ≤ 10¹⁶ cm⁻³.³⁷⁾

Zoom In Zoom Out Reset image size

For the fabrication of the p–i–n junction structure, we prepared doped and non-doped a-Si:H films. For the p-type a-Si:H (p-Si) film, p-type poly(dihydrosilane) was synthesised by the photoinduced ring-opening polymerization of CPS, in which decaborane was dissolved at 80 °C. The dopant concentration of the film was adjusted by changing the quantity of decaborane in the CPS. The wavelength, intensity, and irradiation time for the UV-light-induced reaction were 365 nm, 15 mW/cm², and 10–60 min, respectively. The resultant polymer was dissolved in distilled CPS as the solvent. This p-type silicon ink [poly(dihydrosilane) solution] was spin-coated on a substrate and pyrolyzed at T_p = 390 °C for 30 min in a sealed chamber to prevent the desorption of decaborane. For the n-type a-Si:H (n-Si) film, n-type silicon ink was prepared using white phosphorus as the dopant.⁴⁴⁾ The wavelength, intensity, and irradiation time were 405 nm, 300 mW/cm², and 5–120 min, respectively. Distilled cyclooctane was used as the solvent for the n-type silicon ink. The ink was spin-coated on the substrate and pyrolyzed at T_p = 390 °C for 30 min on a hot plate. For the intrinsic a-Si:H (i-Si) film, nondoped silicon ink was also prepared using distilled cyclooctane as a solvent. The wavelength, intensity, and irradiation time were 365 nm, 15 mW/cm², and 10 min, respectively. The i-Si film was fabricated similarly to the n-Si film.

These p- and n-Si doped films were confirmed to be in the amorphous state from the Raman spectrum of phonon bands. The σ_d at 25 °C for the p- and n-Si films with a thickness of 50 nm are plotted as a function of boron or phosphorus concentration in Fig. 19. The dopant concentration was measured by SIMS. The σ_d values of both the p- and n-Si films show a significant increase with dopant concentration. These results suggest that boron and phosphorus in the doped a-Si:H films act as the acceptor and donor, respectively. The conductivity of our solution-processed doped a-Si:H films is one to three orders of magnitude smaller than that of device-grade vacuum-processed a-Si:H films.⁴⁵⁾ The maximum concentrations of boron and phosphorus are limited by the solubility of decaborane and white phosphorus in CPS, which are 7 × 10²¹ and 2 × 10²¹ cm⁻³, respectively.

Zoom In Zoom Out Reset image size

The film quality was found to be improved by hydrogen-radical treatment via hot-wire chemical vapor deposition. The apparatus used was nearly identical to that described in the literature.⁴⁶⁾ A tungsten wire with a diameter of 0.5 mm and a length of 100 cm was used as the catalyst and was heated at 1850 °C. Hydrogen gas (99.9999% purity) was introduced into the chamber at a flow rate of 15 sccm for 15 min. The chamber pressure and substrate temperature were maintained at 0.2 Pa and 200 °C, respectively. The properties of the i-, p-, and n-Si films with and without the hydrogen-radical treatment are listed in Table III. Typical properties of the vacuum-processed i-Si films are also listed for reference. An increase in hydrogen content and a marked reduction in spin density by the hydrogen-radical treatment were observed in all of the silicon films. The depth profile obtained by SIMS measurement showed that the hydrogen contents increased not only on the surface but also entire film. This indicates that hydrogen radicals diffused in the film and terminated the defects. As a result, with this treatment, a large increase in σ_p and a decrease in σ_d in the i-Si film were achieved, and the σ_d values of the p- and n-Si films were increased. These properties of the resultant a-Si:H films were comparable to those of device-grade vacuum-processed ones.³⁷⁾ These improvements may be attributed to the reduction in spin density.

Hydrogen-radical treatment reduced the spin density of the silicon films, while the use of heat treatment at temperatures of 100–400 °C in a hydrogen atmosphere had little effect on defect reduction in the films. We clarified that hydrogenation using hydrogen radicals is effective for the improvement in film quality.

The solar cell shown in Fig. 18 was fabricated by the following four steps: (1) a flat ZnO layer with a sheet resistivity of 100 Ω/sq and a thickness of 200 nm was sputtered onto a glass substrate with an area of 5 × 5 cm², (2) a p-Si layer (thickness: 30 nm) was fabricated on the ZnO substrate followed by the stacking of i-Si (thickness: 120 or 400 nm) and n-Si layers (thickness: 30 nm), (3) hydrogen-radical treatment was performed after fabricating the p–i–n structure, and (4) a 200-nm-thick Al electrode with an area of 5 × 5 mm² was deposited on the p–i–n structure by vacuum evaporation.

Three types of solar cell (cells 1, 2, and 3) were fabricated by changing the thickness of the i-Si layer and the dopant concentration of the p-Si layer, as listed in Table IV. Figure 20 shows the depth profiles for boron, phosphorus, zinc, carbon, oxygen, and nitrogen for cell 1. It can be observed that boron and phosphorus diffused in the i-Si layer, and that the concentration of these elements in the i-Si layer fell below the detection limit within 20–40 nm from each interface. We also confirmed that boron diffuses extensively into the i-Si layer at T_p >390 °C (data not shown). Therefore, the upper limit of the fabrication temperature for a p–i–n structure with a satisfactory interface is 390 °C. Figure 20(b) shows that carbon, oxygen, and nitrogen are distributed in the p–i–n structure at a concentration of 10¹⁹–10²¹ cm⁻³, which increases at each interface.

Zoom In Zoom Out Reset image size

Figure 21 shows the current density–voltage (J–V) characteristics at 25 °C of the solution-processed solar cells with and without the hydrogen-radical treatment. The J–V curves denoted by c1, c2, and c3 and c'1, c'2, and c'3 correspond to cells with and without the hydrogen-radical treatment, respectively. Moreover, photovoltaic parameters such as the open-circuit voltage (V_oc), short-circuit current (J_sc), fill factor (FF), and energy conversion efficiency (η) of the cells are shown in the figure. It is found that the hydrogen-radical treatment significantly improved J_sc from <0.2 to >1.7 mA/cm². Accordingly, η was improved from <0.01 to >0.30% for all cells. This enhancement in J_sc by the hydrogen-radical treatment may be due to the large improvement in photoconductivity for the i-Si layer, as shown in Table III.

Zoom In Zoom Out Reset image size

Cell 3 shows the highest J_sc among all the cells and is characterized by a low V_oc, it also has the lowest FF among the cells. The highest J_sc of cell 3 originated from the thicker i-Si layer of this cell than those of cells 1 and 2. The lowest FF value might have been related to the film quality of the i-Si layer. Observation of micrographs revealed that cell 3 had a few cracks, which is associated with pyrolysis of the thick (400 nm) i-Si layer. These defects may become a leakage current path and result in the degradation of the insulation of the cell.

In conclusion, solar cells with an a-Si:H p–i–n structure were successfully fabricated using doped and nondoped poly(dihydrosilane) solutions. We have demonstrated the effectiveness of hydrogen-radical treatment for improving the conductivity of a-Si:H layers by the reduction in spin density. The solution-processed a-Si:H solar cells were shown to have η of 0.31–0.51% under AM-1.5G (100 mW/cm²) illumination.