Introduction

Germanium has regained attention as a material for next-generation electronics due to its excellent electrical and optical properties. Ge-on-insulator structures are crucial for advanced Ge-based transistors1,2,3, solar cells4,5, optical communications6,7,8, and thermoelectric devices9,10. While mechanical transfer methods yield high-quality Ge-on-insulator structures11,12, the complexity of the process limits its application. Consequently, Ge thin films have been synthesized directly on insulators using various techniques. Single-crystalline Ge layers are formed through oxidation-induced condensation13,14, rapid melting15,16,17,18, and epitaxial growth on Si-on-insulator19. Low-temperature synthesis of polycrystalline Ge layers has been explored using solid-phase crystallization20,21,22,23, laser annealing24,25,26,27,28,29, chemical vapor deposition30,31,32, lamp annealing33,34,35, plasma irradiation36, seed layer technique37, and metal-induced crystallization38,39,40,41,42,43. Notably, recent progress in low-temperature solid-phase crystallization has seen remarkable success. Modulating the deposition temperature (Td) of the precursor amorphous Ge layer and adding impurities significantly increased grain size in resulting polycrystalline Ge layers44,45,46. Carrier mobility reached peak values (690 and 450 cm2 V−1 s−1 for holes and electrons, respectively) for polycrystalline Ge films, even on a glass and flexible plastic substrates47,48. These carrier mobilities are higher than those of Ge thin films epitaxially grown on single-crystal Si substrates19.

The primary challenge of employing Ge layers in semiconductor devices is the presence of acceptor defects49,50. The high hole concentration (p) in the range of 1017–1019 cm−3, attributed to acceptor defects in Ge thin films, complicates Fermi level control, critical in fabricating most semiconductor devices. For instance, in thin-film transistors utilizing Ge as a channel, high leakage currents due to elevated p have been problematic51,52. Studies have sought to reduce p by adding group IV elements, such as Sn53,54,55,56 and Si57,58,59, to Ge. We achieved a reduction in p to the order of 1016 cm−3 using solid-phase crystallization of GeSn, the lowest reported value for a Ge-on-insulator structure formed by crystal growth and even lower than a single-crystal Ge layer56. While SiGe can reduce p by increasing Si composition, hole mobility (μ) has remained below 50 cm2 V−1 s−1, due to increased effective mass and grain boundary barrier height58. Achieving both low p and high μ in polycrystalline group-IV semiconductor thin films has proven challenging.

Hydrogen atom addition is a common method for compensating defects in semiconductors. In polycrystalline Si thin films, hydrogen passivates defects, leading to significantly improved electrical properties60,61. Hydrogen is also effective in reducing p in silicide semiconductors, where defects act as acceptors62. There have been some experimental reports of hydrogen effects on Ge, including surface modification21,63,64. However, hydrogen has not produced a positive effect on the electrical properties of Ge, which has been supported as reasonable by theoretical studies65. This study investigates the effect of hydrogen addition on solid-phase crystallized Ge layers, updating the highest carrier mobility. Hydrogen considerably deactivated acceptor defects in high-quality Ge layers, while no effect was observed in conventional Ge layers. Controlled hydrogen addition and heat treatment conditions reduced p to 1014 cm−3, two orders of magnitude less than the lowest conventional value, while maintaining high μ in the polycrystalline Ge layers.

Results

Figure 1a is a schematic of the sample preparation process. Amorphous Ge layers were deposited on SiO2 glass substrates using a Knudsen cell in a molecular beam deposition system. Subsequently, the samples were annealed to induce solid-phase crystallization. This process resulted in the formation of polycrystalline Ge layers on the glass substrates. Hydrogen was then introduced to the polycrystalline Ge layers using a plasma generator with a mixed gas of Ar and H2. Figure 1b shows that the Ge layers are uniformly formed on glass substrates. For both Td = 50 and 125 °C samples, the appearance changes from dark before crystallization to reddish after crystallization. This is because the amorphous state absorbs light more effectively, whereas the crystalline state does not absorb long-wavelength light, which is instead reflected. No color differences are observed due to hydrogen irradiation or Td, suggesting that these parameters do not affect the fundamental properties such as the band structure.

Fig. 1: Experimental procedure.
figure 1

a Schematic of sample preparation. b Photographs of the samples with Td = 50 and 125 °C before and after solid-phase crystallization (SPC) and hydrogen plasma treatment.

Full size image

Figure 2a and b illustrate that the grain size of the Ge layer after annealing strongly depended on Td. The Td = 125 °C sample exhibited a grain size more than one order of magnitude larger than that of the Td = 50 °C sample. The Td dependence can be explained in terms of the atomic density of the amorphous precursors (Fig. S1)45. For both Td = 50 and 125 °C, the Ge layer is in an amorphous state at the time of deposition. A low-density amorphous Ge layer is formed at Td = 50 °C (Fig. S1(a)), whereas a high-density amorphous Ge layer is formed at Td = 125 °C (Fig. S1(b)). For the low-density amorphous Ge, because the atomic spacing is large, another nucleation occurs before lateral growth progresses, resulting in numerous grain boundaries and small grains. Conversely, the high-density precursor undergoes rapid lateral growth before another nucleation event, producing larger grains with fewer boundaries. Thus, Td = 125 °C produces large-grained poly-Ge with high crystallinity, leading to superior electrical performance. According to the EBSD measurement, hydrogen plasma irradiation did not affect the grain size of the Ge layer (Fig. S2). Figure 2c and d demonstrates that the effect of plasma treatment on the electrical properties varies dramatically with Td. For the Td = 50 °C sample, both p and μ decreased slightly with increasing hydrogen plasma irradiation time (tH). The decrease in p suggests the passivation of acceptor defects by hydrogen. As in general polycrystalline semiconductor thin films, the µ of the current Ge layers is mostly limited by the grain boundary scattering of carriers45,54. Therefore, the decrease in µ with decreasing p is reasonable since the grain boundary barrier is higher at lower p66. The effect of hydrogen plasma irradiation was more pronounced for Td = 125 °C: p decreased dramatically with increasing tH, reaching the order of 1015 cm−3. The degree of the decrease increased with increasing plasma power (WP). These observations suggest that hydrogen plasma treatment effectively passivated acceptor defects in the high-quality polycrystalline Ge layer. The μ value of the Td = 125 °C sample initially decreased and then increased with increasing tH. The fact that µ improves while p decreases suggests that plasma treatment has reduced grain boundary carrier traps. Thus, hydrogen plasma treatment had little effect on conventional low-quality polycrystalline Ge layers, whereas it exhibited a dramatic acceptor defect compensation effect on high-quality polycrystalline Ge layers.

Fig. 2: Hydrogen plasma treatment effects on polycrystalline Ge layers.
figure 2

a, b Crystal orientation maps for the samples with Td = (a) 50 and (b) 125 °C, where the colors indicate the crystal orientation according to the inserted color key. c p and (d) µ for the samples with Td = 50 and 125 °C as a function of tH with WP of 150, 300, and 450 W.

Full size image

We investigated the hydrogen state in polycrystalline Ge layers. As shown in Fig. 3a, secondary ion mass spectrometry (SIMS) measurement reveals that hydrogen is present with a concentration gradient in the depth direction of the Ge layer. The Td = 50 °C sample showed a higher hydrogen concentration than that of the Td = 125 °C sample. This result is reasonable given that the Td = 50 °C samples have a high grain boundary density (Fig. 2a) because impurities generally diffuse faster at the grain boundaries than in the crystal. However, Fig. 3b shows that there was no change in the Raman spectra before and after plasma treatment for the Td = 50 °C sample, while the Raman peaks due to the bonding of Ge and H (Ge-H at 1987 cm−1, Ge-H2 at 2020 cm−1, and Ge-H3 at 2060 cm−1) were observed for the Td = 125 °C sample67. Figure 3c demonstrates that the hydrogen concentration in Ge increases with increasing WP. At a WP of 450 W, the hydrogen concentration peaked at approximately 10 nm from the surface of the Ge layer. This behavior suggests the formation of platelet defects due to the excessive penetration of hydrogen atoms68, which may contribute to the μ drop in Fig. 2d at a WP of 450 W. Figure 3d shows Raman peaks due to Ge-H bonding were observed in all samples with tH ≥ 5 min. The Raman peaks depend on tH: only the Ge-H peak for tH = 5 min, a peak attributed to Ge-H and Ge-H2 for tH = 10 min, and a broad peak that appears to be a combination of Ge-H, Ge-H2, and Ge-H3 for tH = 20 min. This behavior suggests that the number of hydrogen bonds per Ge atom increases with increasing hydrogen concentration in Ge. When considered in conjunction with the p behavior shown in Fig. 2c, the more pronounced the bonding between Ge and H, the more p is reduced, strongly suggesting that hydrogen atoms passivated the acceptor defects in Ge.

Fig. 3: Hydrogen properties in the Ge layers.
figure 3

a SIMS depth profiles of hydrogen concentration and (b) Raman spectra for the Ge layers with Td = 50 °C and 125 °C, where WP = 300 W and tH = 20 min. c WP dependence of the SIMS depth profiles of hydrogen concentration for the Ge layers with Td = 125 °C, where tH = 20 min. d tH dependence of Raman spectra for the Ge layers with Td = 125 °C, where WP = 300 W.

Full size image

The Raman measurement revealed that distinct Ge–H bonds were observed in high-quality (large-grain) Ge films, whereas no clear Ge–H signals were detected in low-quality (small-grain) films. This contrast can be attributed to both structural and electronic factors. In high-crystallinity Ge films, the well-ordered lattice structure provides a favorable environment for the formation of stable Ge–H bonds. In contrast, low-crystallinity films contain a high density of grain boundaries and structural defects, where hydrogen atoms are likely to be trapped. This trapping may inhibit the formation of stable Ge–H bonds. Moreover, differences in grain boundary density may also affect hydrogen diffusion behavior; in high-quality films, hydrogen is more likely to remain within the crystalline lattice, thereby promoting Ge–H bond formation. Furthermore, it has been reported that hydrogen in Ge exists predominantly in a negatively charged state (H−)65. The stability of this H− state strongly depends on the local electronic environment. In high-crystallinity films, the periodic potential and electronic uniformity provide favorable conditions for the stabilization of H− species, enabling the formation of stable Ge–H bonds. However, in low-quality films, potential fluctuations and local electronic disorder at grain boundaries and defects may destabilize the H− state. As a result, Ge–H bond formation may be suppressed, or such bonds, even if formed, may not exist in a sufficiently stable configuration to be detected. These findings suggest that the presence or absence of Ge–H bonds in Ge thin films is not only governed by the structural quality of the films, but also closely related to the charge state of hydrogen and its electronic stability within the host lattice.

We subjected Td = 125 °C samples, treated with hydrogen plasma, to low-temperature post-annealing (PA). Figure 4a shows the dependence of the electrical properties on the PA temperature (TPA). For TPA = 100 °C, no change in the electrical properties was observed compared to those before PA. In contrast, for TPA = 200 °C, p decreased by one order of magnitude to 4 × 1014 cm−3 while maintaining μ. The p value increased significantly for TPA = 300 °C, while both p and μ reverted entirely to their pre-plasma treatment values for TPA = 400 °C. These outcomes suggest that PA at suitable temperatures enhances the passivation of acceptor defects and grain boundary carrier traps, whereas PA at higher temperatures results in hydrogen desorption from Ge. For the Td = 50 °C sample, the low-temperature PA also reduced p and enhanced μ, but the effect was less pronounced compared to the Td = 125 °C sample (Fig. S3). To elucidate the origin of the improved electrical properties for TPA = 200 °C, we evaluated the detailed crystal structure using transmission electron microscopy (TEM). Figure 4b–d indicate that hydrogen plasma treatment introduced defects into the surface region. As WP increased, the thickness of the surface defect layer also increased, which is consistent with the SIMS results in Fig. 3. Figure 4(e) shows that the surface defects decreased after PA. Therefore, we infer that the recovery of the crystal structure led to the improved electrical properties.

Fig. 4: Effects of post-annealing (PA) on the polycrystalline Ge layers (Td = 125 °C) after hydrogen plasma treatment.
figure 4

a p and μ as a function of TPA (4 h), where dotted lines show p and μ of Ge before hydrogen plasma treatment. b–e Dark-field TEM images for the samples with WP = (b) 150, (c) 300, and (d) 450 W before PA and with WP = (e) 300 W after PA at 200 °C.

Full size image

Figure 5 presents the electrical properties of undoped polycrystalline Ge, GeSn, and SiGe thin films (<150-nm thickness) grown on insulating substrates using various techniques. Typically, polycrystalline Ge layers exhibit p above 1017 cm−3, irrespective of the technique used, due to acceptor defects. The addition of small amounts of Sn in Ge slightly reduces p and improves μ. Increasing the Si composition in SiGe (>50%) significantly reduces p but leads to a substantial reduction in μ, attributable to an increase in carrier effective mass and grain boundary barriers. The polycrystalline Ge layers obtained in this study exhibit by far the lowest p in polycrystalline Ge-based thin films and maintain relatively high μ.

Fig. 5: Comparison of p and µ.
figure 5

For undoped polycrystalline group-IV semiconductor thin films (<150-nm thickness) grown on insulating substrates using various techniques. The circular, square, and triangular symbols indicate Ge, GeSn, and SiGe, respectively. The numbers near each symbol show the reference number.

Full size image

Discussion

We explored the impact of hydrogen plasma treatment on polycrystalline Ge layers by appropriately controlling the plasma treatment conditions. Hydrogen atoms were introduced into the polycrystalline Ge layer, as revealed by Raman measurements. Bonding between Ge and H occurred in the large-grain Ge layers, a phenomenon not observed in the small-grain Ge layers. Consistent with the bonding of Ge and H, the p of large-grain Ge layers dropped from 1017 cm−3 to the order of 1015 cm−3, marking the lowest reported value for p in a polycrystalline Ge layer. Following the nature of polycrystalline semiconductor thin films, μ decreased as p reduced, but low-temperature PA improved μ while further reducing p. These observations suggest that hydrogen atoms in Ge passivate not only acceptor defects but also grain boundary carrier traps. The resulting Ge layer simultaneously achieved high μ (170 cm2 V−1 s−1) and low p (4 × 1014 cm−3), an accomplishment unattainable with conventional polycrystalline group-IV semiconductor thin films. The findings of this study address the longstanding issue of acceptor defects hindering the application of polycrystalline Ge-based thin films in device applications, enabling a wide variety of advanced Ge-based devices.

Methods

Sample preparation

Ge precursors were deposited on SiO2 glass substrates using the Knudsen cell of a molecular beam deposition system (base pressure: 5 × 10−7 Pa). The Ge thickness was 100 nm and Ge deposition rate was fixed at 1.7 nm min−1. During deposition, the substrate was heated at Td of 50 or 125 °C45. Subsequently, the samples were loaded into a conventional tube furnace in a N2 (99.9%) atmosphere and annealed to induce solid-phase crystallization at 450 °C for 5 h. Hydrogen was introduced to the polycrystalline Ge layers using a plasma generator, Yamato PDC 210 (Fig. S2). To ensure safety, a mixed gas of Ar and H2 (Ar:H2 = 95:5) was used for the introduction of hydrogen69. The gas flow rate and pressure were fixed at 5 sccm and 10 Pa, respectively, while the WP ranged from 150 to 450 W. Following the plasma treatment, PA was conducted at a TPA of 100 – 400 °C for 4 h in a vacuum furnace (base pressure: 1 × 10−3 Pa).

Material characterization

Electron backscatter diffraction analyses were carried out using a Hitachi-high-tech SU7000 (voltage: 15 kV) equipped with an Oxford AZtec analysis attachment. Hall effect measurements were recorded using the Van der Pauw method with a Lake Shore M91-EV system, where both μ and p were averaged over five measurements for each sample. SIMS measurements were performed using a PHI ADEPT1010 instrument. Raman spectroscopy was conducted using a JASCO NRS-5100 (laser wavelength: 532 nm; spot size: 5 μm), and the laser power (0.5 mW) was kept sufficiently low to avoid influencing the crystal phase and peak shift. An FEI Tecnai Osiris instrument operating at 200 kV was used for TEM analysis. Cross-sections of the samples were prepared using a focused ion beam (Helios Nanolab 600i).