Dramatic Changes in Thermoelectric Power of Germanium under Pressure: Printing n–p Junctions by Applied Stress

Controlled tuning the electrical, optical, magnetic, mechanical and other characteristics of the leading semiconducting materials is one of the primary technological challenges. Here, we demonstrate that the electronic transport properties of conventional single-crystalline wafers of germanium may be dramatically tuned by application of moderate pressures. We investigated the thermoelectric power (Seebeck coefficient) of p– and n–type germanium under high pressure to 20 GPa. We established that an applied pressure of several GPa drastically shifts the electrical conduction to p–type. The p–type conduction is conserved across the semiconductor-metal phase transition at near 10 GPa. Upon pressure releasing, germanium transformed to a metastable st12 phase (Ge-III) with n–type semiconducting conductivity. We proposed that the unusual electronic properties of germanium in the original cubic-diamond-structured phase could result from a splitting of the “heavy” and “light” holes bands, and a related charge transfer between them. We suggested new innovative applications of germanium, e.g., in technologies of printing of n–p and n–p–n junctions by applied stress. Thus, our work has uncovered a new face of germanium as a ‘smart’ material.


Results
A representative set of structural and optical data collected from the single-crystalline ingots of germanium is given in Fig. 1. All of these ingots crystallized in the diamond-type lattice (space group #227 -Fd m 3 ) (Fig. 1a,b) and showed characteristic Raman peak at ~300 cm −1 (Fig. 1c). The near-infrared absorption spectra demonstrated a very abrupt absorption edge at near 0.8 eV (Fig. 1d). Using a standard expression for absorption edges in semiconductors with nearly parabolic energy bands, as follows 25 (where α is the absorption coefficient, α 0 is a constant, E is the energy, C is an instrumental shift, and n ∼ 1 / 2 and ∼ 2 for direct and indirect band gaps, respectively), we established the direct band gap as of 0.8 eV (Fig. 1d). This value perfectly agrees with the literature data that address this gap to a direct electronic transition between the top of the valence band and the bottom conduction band at Γ point of the Brillouin zone [50][51][52][53][54] . However, germanium is known to have an indirect fundamental band gap of E g = 0.67 eV [50][51][52][53][54] . Indeed, our absorption spectra suggested the existence of the indirect band gap somewhere between 0.6 and 0.7 eV (Fig. 1d), but this absorption edge was masked by the very strong resonance effects in our double-side-polished samples.
Thermopower of germanium in the semiconductor and metal phases. At ambient pressure the two ingots of germanium, labelled as D and K, exhibited a compensated electrical conduction with comparable hole and electron contributions. Meanwhile, the D ingot showed a slight preference to p-type, likewise, the K ingot -to n-type (Fig. 2a). The third ingot, labelled as G, was characterized by more pronounced n-type conduction at ambient pressure (Fig. 2b). Notice here, that germanium with one dominant type of charge carries typically has the larger Seebeck coefficients of about several hundreds of μ V/K 67,68 . With pressure increase, the Seebeck coefficients of samples #K1 and #G1 demonstrated the n-p sign inversion at near 1 and 3 GPa, respectively (Fig. 2). Whereas, the sample #D1 displayed even a double p-n-p sign inversion at the beginning of compression to 2 GPa (Fig. 2a). In general, under applied pressure all three samples demonstrated the similar maxima of their Seebeck coefficients at near 3-4 GPa, followed by a progressive drop of the thermopower value (Fig. 2). In the thermopower curve of sample #D1 we detected a distinct kink at near 10 GPa, which may be attributed to the transition to the metallic β-Sn phase (upper inset in Fig. 2a). On the semiconductor-metal phase transition in silicon at a similar pressure value of 10 GPa, the thermopower exhibited the same feature 26 . In this metal β-Sn phase the Seebeck coefficient of germanium was weakly varied about S ≈ + 12 μ V/K (Fig. 2a). Upon the decompression cycle, the thermoelectric power in sample #D1 inverted its sign at near 1 GPa and tended to high negative values, suggesting a transition to a semiconducting phase (lower inset in Fig. 2a). On the contrary, the Seebeck coefficient of sample #G1, which was decompressed from 6 GPa, that is below the phase transition point of 10 GPa 30-49 , kept positive values and turned to S ∼ + 150 μ V/K after the pressure was released (Fig. 2b). On the re-pressurization cycle the sample #G1 behaved already as a p-type material (Fig. 2b).
Scientific RepoRts | 7:44220 | DOI: 10.1038/srep44220 Metastable phases of germanium. To determine the crystal structure of the recovered from high pressure samples we examined them by Raman spectroscopy and X-ray diffraction. In Fig. 3 we display these data on example of recovered sample #D1 which turned to a slightly textured polycrystal. The Raman spectra collected from different points at its surface exhibited peaks at 88,99,149,185,191,212,228,244,273, and 300 cm −1 (Fig. 3a). The intensities of these Raman peaks were strongly varied from point to point (Fig. 3a), thereby indicating that the spectra are highly sensitive to orientation of the crystal grains. These Raman spectra well agreed with those observed in previous works for a metastable polymorph of germanium, prepared either in diamond anvil cells 60 , or by a surface nanoindentation 61,62 . In the literature, these spectra were attributed to a simple tetragonal lattice with 12 atoms per unit cell (st12, space group #96 -P4 3 2 1 2) (also known as Ge-III) [60][61][62] . Whereas, other papers reported different Raman spectra, e.g., a strong peak at near 200 cm −1 , and addressed them to another metastable polymorph with a body-centred cubic lattice with 8 atoms per unit cell (bc8, Ge-IV) [63][64][65][66] . Earlier investigations noticed that the formation of the metastable polymorphs in germanium is controlled by both a decompression rate 58 and stress conditions 57 . The Rietveld refinement of the X-ray diffraction pattern collected from the recovered from high pressure sample #D1 confirmed the tetragonal P4 3 2 1 2 structure of Ge-III (Fig. 3d,e). We found its unit cell parameters and atomic coordinates as follows: a = 5.927(2)Å, c = 6.969 (6) (Fig. 3b). These parameters were similar to those reported earlier for this phase 39 . Remarkably, that besides this tetragonal Ge-III phase, the recovered sample #D1 exhibited no traces of any other phases. The other two samples, #G1 and #K1, after their recovery from high pressures were mixtures of both the original and the tetragonal st12 phases. These facts show that for preparation of a pure st12 phase one should apply a high pressure well above the phase transition point of 10 GPa [30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49] .
The thermopower curve for sample #D1 demonstrated that upon pressure releasing at below 1 GPa it turned to a semiconductor with the dominant n-type conduction (lower inset in Fig. 2a). Earlier band structure calculations predicted that the st12 phase of germanium should be a direct-band-gap semiconductor with an energy gap of E g = 0.7 eV 59 . The major part of recovered sample #D1 crystallizing in this st12 structure (Ge-III) presented a thin disc-shaped film of ~15 μ m in the thickness and of ~150 μ m in the diameter, and, hence, we could proceed with examination of its electronic band structure by means of near-infrared absorption spectroscopy (Fig. 3c). However, these spectra did not reveal any absorption edges. But it should be noted here that numerous defects and high concentrations of free carriers associated with these defects often impede direct observation of absorption edges in polycrystalline semiconductors. To obtain the metastable polymorph of germanium in the larger amount for more detailed investigations, we tried to prepare that by means of a high-pressure high-temperature synthesis at pressure near 20 GPa in large-volume presses. The bulk samples recovered from these syntheses were apparently multi-phased and seemed to be more appropriate for investigations of a "zoo" of germanium Tuning the thermopower in germanium. To examine the discovered pressure-driven thermopower inversion in detail, we measured pressure dependencies of the thermopower for several more samples of germanium (Fig. 4). We can summarize our findings and conclusions, as follows: 1. Tuning the thermopower by applied pressure up to 1 GPa seems to be reversible (Fig. 4a). It was seen, for instance, in sample #K2, in which we observed an n-p inversion at near 0.55 GPa under compression to 1.1 GPa and then upon pressure releasing (Fig. 4a). Sample #D2 decompressed from about 2.1 GPa showed a certain positive shift in the thermopower value after the pressure was released (Fig. 4b). In another sample recovered from a higher pressure of 3 GPa, this positive shift became more sizable and the Seebeck coefficient turned to a value of S ∼ + 100 μ V/K (not shown). 2. Treatment by the higher pressures, ranging between 4 and 9 GPa, that is, above the flat extremum in the thermopower curves, always results in the irreversible turn to the p-type conduction (Figs 2b and 4c,d). After the pressure is released, the Seebeck coefficient comes to values of S ∼ + (150-200) μ V/K. 3. High-pressure treatment somewhat above the semiconductor-metal phase transition point (10 GPa) 30-49 , leads to the formation of the metastable st12 polymorph (Ge-III), which is characterized by the dominant n-type conduction and by the Seebeck coefficient as of S ∼ − (150-250) μ V/K (Fig. 4e, lower inset in Fig. 2a).
Although, we found a very good consistency in the thermopower data collected for different germanium samples, the extrema around 2-4 GPa were noticeably varied from sample to sample (Figs 2,4). The thermopower curve for sample #K3 above 4 GPa showed a sizable deviation from the curve for sample #K4 (Fig. 4c), thereby indicating that the appearance of pressure gradients leads to smearing of the thermopower extremum. The appearance of these pressure gradients is related to a strong shifting of the sample from the central area of the limestone container (Fig. 5b) to one of its edges. For comparison we measured the Seebeck coefficients of two more samples, #K5 and #K5 under pressure up to 6-7 GPa in the high-pressure cell with concave anvils, which provided more uniform quasi-hydrostatic pressure conditions (Fig. 5a,b). The pressure dependencies of the thermopower measured in this cell also displayed the n-p sign inversion followed by the pronounced extremum (Fig. 5c). But compared to the data gathered in the cell with flat anvils from the samples cut from the same ingot K (#K1, #K3, #K4) (Figs 2a and 4c), these features in the thermopower behaviour were apparently shifted to the lower pressures (Fig. 5c). Thus, the Seebeck coefficient of germanium, measured in the cell with concave anvils demonstrated an anomalously high pressure derivative as of about 1000 μ V/GPa from ambient pressure to 0.5 GPa (inset in Fig. 5c). After the pressure was released the originally n-type germanium also turned to p-type and demonstrated the Seebeck coefficient as of S ~ + 200 μ V/K (Fig. 5c). Comparing the data collected in the two different high-pressure cells, we can conclude that the presence of minor non-hydrostatic stresses can partly suppress and smear the extremum in the pressure dependence of the thermopower (Figs 2 and 4a,b). Thermopower of the metastable polymorph of germanium. As we have verified in this work, the compression of germanium in our cell to pressure values somewhat above the semiconductor-metal phase transition point (10 GPa) 30-49 followed by gradual decompression, led to its transmutation into the metastable st12 polymorph (Ge-III) (Fig. 3). After the decompression cycles for the second pressure runs we could in-situ prepare the metastable st12 phase in samples #D3, #D4 and #K3 (Fig. 4e), and then, we performed the re-pressurization runs for this Ge-III polymorph (Fig. 6a). These dependencies for Ge-III polymorph demonstrated rather spectacular features, and at about 10 GPa they suggested a transition to the metal phase (Fig. 6a). This scenario looked resembling to the semiconductor-metal phase transition in the original cubic-diamond-type phase (upper insets in Figs 2a and 4e). It was interesting to note that in some pressure dependencies of the temperature difference (Δ T) along the sample thickness, one could see the pronounced bends at near 10 GPa (Fig. 6c). Since, the Δ T value depends on sample thickness (h) and its thermal conductivity (λ) as: Δ T ~ h/λ, this bend in the Δ T curves might be linked to enhancement of the thermal conductivity in the metal β-Sn-type phase.
The thermopower curves for the metastable st12 polymorph (Ge-III) apparently indicated the existence of an intermediate electronic (or structural) phase, which was observed between 4 and 8 GPa on pressurization, and between 7 and 2 GPa on decompression cycles (Fig. 6a). It is interesting to recall here, that in case of silicon, a return transition from a metal β-Sn phase to a semimetal rhombohedral r8 phase (Si-XII) 60,69-71 , was well detectable in pressure dependencies of the thermopower, by a gradual lowering in its value at near 7-9 GPa upon decompression 26 . A feature we discovered in germanium at near 7 GPa upon pressure releasing (insets in Figs 6a and 4e) looked very similar to that in silicon 26 , although, the rhombohedral r8 phase has not yet been observed in germanium. At about 2 GPa on pressure releasing, silicon transforms to a semimetal bc8 phase (Si-III) with a p-type electrical conduction. 72 Previous work showed that on this transition the positive Seebeck coefficient of silicon abruptly raises to magnitudes of S ∼ + (15-20) μ V/K 26 . On the contrary, the thermopower of germanium, below ~2 GPa drastically changed its sign and turned to high negative values (Fig. 6a). We could verify by Raman spectroscopy that these samples #D3 and #D4 after the pressure was released, crystallized in the st12 phase (Ge-III). It was reported in the literature that the β-Sn → st12 phase transition in germanium under decompression begins already at about 7-9 GPa 57,60 . These pressure values have correspondence with the above-discussed minor lowering in the thermopower value we observed in germanium at below 7 GPa, but not with the thermopower jump at below 2 GPa (inset in Fig. 6a). Thus, the crystal structure of the intermediate phase we found in germanium at pressures between 4 and 8 GPa on pressurization and between 7 and 2 GPa on decompression cycles (Fig. 6a), Scientific RepoRts | 7:44220 | DOI: 10.1038/srep44220 cannot be figured out at the moment. Potentially, it could be the same tetragonal st12 phase but with the dramatically modified electronic band structure and reduced or even closed energy band gap.

Discussion
The Seebeck coefficient (S) of an intrinsic non-magnetic semiconductor linearly depends both on its band gap value (E g ) and on the ratio of hole (σ p ) and electron contributions (σ n ) to electrical conduction, and in a simple two-band case, it comes as follows 73 : where, k is the Boltzmann's constant, e is the electron charge (k/|e| ≈ 86.4 μ V/K), T is the temperature, r n (r p ) and ⁎ m n ( ⁎ m p ) are the scattering parameters and the effective masses of density of states of electrons (holes), respectively. The samples of germanium we investigated are intrinsic semiconductors, and, hence, their behaviour can be analysed in the framework of this model. Both indirect and direct band gaps in the cubic-diamond-structured phase of germanium, were reported to widen with pressure, with the coefficients of about 4 meV/GPa for the indirect gap, 50,51 and of 120 meV/GPa for the direct one 52,54 . As seen from Eq. 1 these moderate changes in the band gaps cannot explain the anomalous pressure dependencies of the thermopower (Figs 2,4, and 5). Hence, these thermopower inversions may be attributed only to variations in the σ p /σ n and ⁎ ⁎ m m / p n ratios (Eq. 1). For germanium samples with pure n-type conduction (i.e., σ p = 0), and with the typical values of the scattering parameter as of r n − ½ and E g = 0.67 eV, [50][51][52][53][54] the Seebeck coefficient should be larger than − 1 mV/K. In sample #G1 (Fig. 2b) the thermopower value at ambient conditions was only S∼ − 270 μ V/K, thereby suggesting the σ p /σ n ratio as of 0.63 (Eq. 1). The electrical conduction of the other two bulk samples, #D and #K of which Seebeck coefficients were of about ± 50 μ V/K at ambient pressure (Fig. 2a), was apparently compensated (σ p ≈ σ n ).
Thus, the pressure-driven shift to the p-type conduction in germanium (Figs 2,4,5) should be related to enhancement of the hole partial contribution (Eq. 1). This hole contribution to the conduction is determined by σ p = μ p p, where p is the "effective" concentration of hole carriers and μ p is their "effective" mobility value 73 . As the fundamental band gap of germanium only slightly widens with pressure 50,51 , the concentration of charge carriers, which in intrinsic semiconductors, are linked to native point defects in crystal lattice, unavoidable impurities, and those carriers which are thermally-activated over a band gap, should not increase with pressure. Therefore, the high positive values of the Seebeck coefficient near 1-4 GPa (Figs. 2, 4, and 5) indicating that the p-type conduction becomes dominating, may be related to increase in hole mobility values.
As stated in the literature, the top of the valence band of germanium at Γ point of the Brillouin zone consists of two overlapping hole bands of so-called "light" and "heavy" holes with typical effective masses of about 0.043m 0 and 0.33m 0 , respectively. Several previous works claimed experimental observations of distinct crossovers in the electronic band structure of germanium under applied pressure of 2-3 GPa 74-77 . For instance, it was found that the electrical conduction of n-Ge is moderately diminished with pressure to 2 GPa 74 or 3 GPa 75 , in agreement with the minor widening in its band gap value 50,51 , but above this pressure point the electrical conduction begins to increase with pressure 74,75 . Another work discovered kinks at 1.8 GPa in pressure dependencies of phonon energies of germanium and addressed them to band structure reconstruction 76 . The last paper speculated that with pressure application the bottom of the Δ valley of the conduction band of germanium shifts below the bottoms of the Γ and L valleys, and hence, its fundamental indirect band gap becomes related to the transition between the bottom of this Δ valley and the top of the valence band at the Γ point of the Brillouin zone 76 . Dramatic changes in electronic transport properties of germanium found near 3 GPa in one more work, were also addressed to the intervalley transition 77 . Meanwhile, it should be also noted that some other studies of the electronic transport properties of germanium did not find any remarkable features across the above pressures 78 . One more paper, considering the anomalous behaviour of germanium in the cubic diamond phase, proposed a possibility of pressure-stimulated transfer of the hole carriers from the "heavy" holes band to the "light" one 79,80 .
The mobility values of carriers of the "light" holes band should be essentially higher than those of carriers of the "heavy" holes band, and hence, upon this transfer the hole partial contribution to the electrical conduction should be significantly enhanced. In a line with this conjecture, two recent studies on "compressively strained" by Sn-doping germanium 81 and strained films of pure germanium 82 clearly documented the above proposed splitting of the "heavy" and "light" holes bands.
The abrupt pressure-driven n-p inversion and the high positive values of the Seebeck coefficient we observed at pressures of 1-5 GPa (Figs 2, 4, and 5) indicated a dramatic enhancement of the hole partial conduction (Eq. 1). This feature may be well explained by the above-discussed splitting of the two holes bands under applied pressure and a following transfer of the carriers from the "heavy" holes band to the "light" one. This model can also explain the anomalously high pressure derivative of the Seebeck coefficient we documented (Inset in Fig. 5c) as well as the crucial influence of minor non-hydrostatic stresses, which was seen from the comparison of the data obtained in the two different cells (Figs 2a, 4c,d, and 5a). One can surmise that this fine reconstruction of the band structure in germanium should be limited by available free hole carriers. Therefore, the pronounced pressure-driven n-p inversions we revealed in this work (Figs 2, 4, and 5), may be well visible in samples with intrinsic semiconductor conductivity. Whereas, in strongly doped samples of n-type, such a pressure-driven n-p sign inversion is unlikely to be observable, although, some anomalies in the properties resulting from the splitting of the holes bands still may occur.
The irreversible shift to the p-type conduction observed in the samples recovered from high pressures below the Ge-I → Ge-II phase transition point, i.e., in the cubic-diamond-type phase (Figs 2, 4c,d, and 5a), is most likely related to the conservation of residual strains which can keep a splitting of the "light" and "heavy" holes bands after the pressure is released. But it should be also noted, that applied high pressures could produce a number of "damages" in the crystal lattice, and hence, a native defect structure of the crystals might be moderately modified under pressure. Earlier studies of an impact of fast-neutron bombardment of germanium revealed that point damages in its crystal lattice lead to p-type conduction 83,84 . These results are in line with our findings (Figs 2,  4, and 5). Theoretical investigations of potential point defects which may be formed under external mechanical impacts on the crystal lattice of germanium, found two energetically favourable self-interstitial defects, such as: (i) a "split-interstitial" configuration which is electrically neutral, and (ii) an "open cage" configuration which has a donor-type 85,86 . Thus, we cannot infer which sorts of defects could potentially contribute to enhancing the p-type conduction, but their contribution could not be significant. The n-type conduction we established in the st12 metastable polymorph (Ge-III) (Figs 2a, 4e, 6a) indicates that the native defects in its crystal structure are mainly of an electron type.
Scientific RepoRts | 7:44220 | DOI: 10.1038/srep44220 The dramatic changes in the thermopower of conventional germanium we revealed in this work, suggest novel possible innovative applications of this material. Among those, we can anticipate different micro-and nanoscale junctions with stress-controlled properties, embedded in various integrated circuits. A simple example of such junctions is a stress-controlled n-p switch. Using some designed printer-type device with a set of hard tips, one can "print out" circuits and zones of different conduction types on surface of germanium. The simplest examples of such 'printing' can be (i) a 'writing' of a thin p-type layer on a surface of n-type germanium (Fig. 7a), or (ii) a fabrication of a thin n-type layer of the metastable Ge-III polymorph on a surface of conventional germanium with the cubic-diamond-type structure (Fig. 7b). In the latter case, a stress distribution in the material should lead to the fabrication of an intermediate p-type layer of the cubic-diamond-type germanium between this n-type Ge-III layer and the substrate, as shown in Fig. 7c. Varying the geometrical parameters of the printing tips and conditions of load/unload, one can modify the profile depths of such multi-layered structures. Earlier investigations have already discovered that applied stress can remarkably tune the electronic properties of germanium. For instance, it was predicted that controlled tensions along < 111 > directions can turn germanium to a direct band gap semiconductor 87 ; experimentally, this strategy was realized in thin films 88 .
Above 10 GPa in the metal β-Sn-type phase of germanium, the Seebeck coefficient in different samples varied between S ~ + 5 and + 17 μ V/K (insets in Figs 2a, 4e and 6a). The lowest values were measured in the samples undergoing the transition to the metal Ge-II phase from the metastable Ge-III one, with a concurrent n-p sign inversion in the Seebeck effect (Fig. 6a). But upon the phase transition from the original cubic-diamond-type Ge-I phase with the p-type conduction, the thermopower values of the metal Ge-II phase were essentially higher, of S ~ + (11-17) μ V/K (inset in Fig. 4e). This difference can be explained by the fact that upon the reconstructive transition to the β-Sn-type metal phase, the samples passed via a region of the phase coexistence, and hence, above 10 GPa the contributions of either the p-type Ge-I or the n-type Ge-III phases were still considerable. Meanwhile, all the samples demonstrated the similar uptrends in their pressure curves of the thermopower in the metal β-Sn phase (insets in Figs 4e and 6a). This behaviour should be related to a gradual band structure reconstruction. For "simple" metals with weakly changeable band structures, the volume contraction is expected to lead to a moderate decrease in the absolute value of the thermopower because of a proportional increase in the "effective" free carrier concentration 89 . However, even elemental metals deviate from this trend 90 , thereby indicating that band structure modifications make a major contribution to pressure dependencies of their Seebeck coefficients. Re-visiting the thermopower data for the metal β-Sn-type polymorph of germanium, we can conclude the following: (i) the Ge-I → Ge-II phase transition was best seen in sample #D1 at 10 GPa (Fig. 2a), (ii) the thermopower value of the pure β-Sn-type phase is about S ≈ + 12 μ V/K, and (iii) after the phase transition is completed, the pressure behaviour of the thermopower of the β-Sn-type phase corresponds to the above predictions for "simple" metals (inset in Fig. 2a) 89 . The high values of the Seebeck coefficient of the metastable Ge-III polymorph we found in the present work (Fig. 6a,b) suggest that this phase can be a narrow-band-gap semiconductor with a certain potential for the thermoelectricity. It should be noted here that Ge-rich materials, and in particular, Si 1−x Ge x alloys, are known to be excellent thermoelectrics [91][92][93][94][95][96][97][98][99] . We have verified that this metastable st12 polymorph (Ge-III) in our thin samples recovered from high pressure persisted at least for several years. Probably, the local strains in the recovered samples help to retain this metastable high-pressure structure at ambient pressure.

Conclusions
We have measured the Seebeck effect of single-crystalline samples of germanium with intrinsic electrical conduction of both p-and n-types under high pressure to 20 GPa at room temperature. We have established that applied pressure strongly shifts the conduction to p-type in the original semiconductor phase, and then, the p-type conduction is further conserved in the metal β-Sn-type phase. Upon pressure releasing, the β-Sn-type phase transformed to the st12 metastable polymorph (Ge-III) with the n-type semiconducting conductivity. We have addressed the shift to the p-type conduction in the cubic-diamond phase to a pressure-driven splitting of the overlapped "heavy" and "light" holes bands, stimulating a charge transfer to the "light" band with more mobile hole carriers. In addition, we have verified that this n-p sign inversion is reversible if applied pressure is less than 2 GPa, and under higher applied pressures it becomes irreversible. Thus, our work has clearly demonstrated that the electronic transport properties of germanium may be dramatically tuned by a moderate applied stress. This finding can stimulate novel innovative applications of germanium as a 'smart' material. We have suggested that germanium may be utilized, for instance, in stress-controlled n-p switches and in technologies of 'printing' of n-p and n-p-n junctions by applied stress.

Experimental section
For investigations we used several conventional bulk single-crystalline ingots of germanium from different suppliers. For convenience, we labelled these bulk samples by D, G, and K letters, and consequently numbered small microscopic samples cut from these ingots (e.g., samples cut from ingot D as #D1 -#D4). The carrier concentrations in these bulk ingots were about 10 14 cm −3 . We also synthesized two bulk samples of germanium from conventional powder at 20 GPa and 600 °C using a 1200-tonne multi-anvil press at Bayerisches Geoinstitut. Both the original crystals and samples recovered after the high-pressure experiments were characterized by standard structural and optical techniques (Figs 1, 3). The crystal structure of the samples was verified in X-Ray diffraction studies performed on a high-brilliance Rigaku diffractometer (Mo Kα radiation) equipped with Osmic focusing X-ray optics and Bruker Apex CCD detector. In addition, we examined the crystal structure of the samples by Raman spectroscopy using two setups. In one of them the Raman spectra were excited with the 514.5 nm line of an Ar laser and analyzed by a Renishaw Ramascope; in another one the spectra were excited with the 632.8 nm line of a He-Ne laser and analysed by a Labam spectrometer. The electron structure of the samples was examined by near-infrared absorption spectroscopy using a Bruker IFS 120 Fourier transform spectrometer For the absorption studies the original samples were double-polished to the thickness of about 15-20 μ m; the samples recovered after the high-pressure experiments had similar thicknesses.
The measurements of the thermopower (Seebeck coefficient) under high pressures (Figs 2, 4-6) were carried out on a fully automated high-pressure setup 100 . This setup presented a mini-press that smoothly generated an applied force to a high-pressure cell with a sample. Several nanovoltmeters and other sensor devices were connected to the cell for recording of all relevant parameters of a sample and environment 100 . This type of measurements enabled to follow the properties evolution under nearly continuous variation in pressure. A force applied to the high-pressure cell was automatically measured in-situ by a digital dynamometer directly on the cell. Then, a pressure value on a sample was automatically estimated from a calibration load curve based on the well-known and distinctly observable phase transitions 101 .
We utilized two different anvil-type high-pressure cells of the modified Bridgman-type 102 . In these cells a sample container made of the limestone (soft CaCO 3 -based material) served both as a pressure-transmitting medium and as a gasket to keep a sample in the space between the anvils 103 . A high and uniform pressure was generated in the central area of the sample container. In visual examinations of sample containers recovered after the high-pressure experiments we verified the sample position (Fig. 5b). The majority of the experiments were performed in a cell with flat anvils made of sintered diamonds with typical working diameters (culets) of ~600 μ m 104 . We loaded in this cell a thin disc-shaped sample with typical sizes of about 200 × 200 × 30 μ m 3 . In another cell the both anvils had a semispherical cavity in their central parts that enabled to provide a more uniform pressure in a larger volume (Fig. 5a) 103,105 . In the latter cell we loaded bulk samples with typical sizes of about 150 × 150 × 150 μ m 3 . To produce a temperature difference (Δ T) of a few Kelvin along the sample thickness, the upper anvils in both cells, were heated up by electrical current heaters. This temperature difference was measured by means of thermocouples attached to the fixed points at the anvils. A relative uncertainty in determination of the Seebeck coefficient values by this method was related to a potential inaccuracy in estimation of the above mentioned Δ T value, and it was found to be less than 5% 106 . We monitored that the outcome thermoelectric signal was caused exclusively by the produced temperature difference, Δ T (Fig. 6b). Therefore, parasitic thermal and electrical signals did not make any noticeable contributions to the thermoelectric signal, and hence, the absolute error in determination of the thermopower should be less than 0.5 μ V/K 106 . Other details of the high-pressure thermopower technique were similar to those described in recent previous works 106,107 .