Post-Processing Thermal Activation of Thermoelectric Materials Based on Germanium

Abstract

After the deposition process, the lattice structure of doped germanium remains low. Post-processing annealing reorders the structure and increases the output parameters. Thin films of germanium doped with gold (Ge:Au) and vanadium (Ge:V) were magnetron-sputtered on glass substrates. The course of the activation process was monitored in situ. Two different methods of post-processing thermal activation of the films were studied. The first method was to place the structure at an elevated temperature for a specified period of time. The second method involved placing the structure on a heating table and cycling the heating and cooling several times from room temperature to about 823 K. Both methods fulfill their function well. The differences come down to research aspects. The best thermoelectric parameters were achieved for germanium doped with 0.95 at.% vanadium. The Seebeck coefficient of 212 μV/K and the power factor of 1.24 mW·m⁻¹·K⁻² were obtained at 500 K.

1. Introduction

1.1. Thermoelectric Parameters

Thermoelectric devices are able to direct convert the heat to electricity. Thermoelectric generators (TEGs) are increasingly being used to recover energy from lost heat and use it to power small electronic systems such as sensors, transmitters, microcontrollers or RFID tags [1,2,3,4]. Such microcircuits often require only tens or hundreds of microwatts to operate. Thermoelectric generators can provide the required amount of energy and make the circuits fully autonomous. TEGs have no moving parts, so they are not susceptible to damage.

The most important parameters describing the performance of thermoelectric materials are the Seebeck coefficient, $\propto \left[\mathrm{V}/\mathrm{K}\right]$, thermal conductivity, $\lambda \left[\mathrm{W}/\left(\mathrm{m}·\mathrm{K}\right)\right]$ and electrical conductivity, $\sigma \left[1/\left(\mathsf{\Omega }·\mathrm{m}\right)\right]$ or electrical resistivity ρ$\left[\mathsf{\Omega }·\mathrm{m}\right]$. The electrical conductivity/resistivity determines the final resistance of the structure that will be made from a given material. Seebeck coefficient and thermal conductivity determine the output voltage of the thermocouple, which depends on the temperature difference between its ends. In order to facilitate the comparison of materials among themselves, these parameters have been related using a quantity called Figure of Merit (Z), expressed by Equation (1) [5,6].

$$Z={\displaystyle \frac{{\propto }^2·\sigma }{\lambda }}\left[{\mathrm{K}}^{-1}\right]$$

(1)

Since all component parameters depend on temperature, it is more common to use the Dimensionless Figure of Merit, ZT, calculated by (2).

$$ZT={\displaystyle \frac{{\propto }^2·\sigma }{\lambda }}·T$$

(2)

Z and ZT are widely used when comparing bulk materials. In the case of film materials deposited on a substrate, this becomes more debatable. The substrate conducts heat, so not only λ of the material determines the temperature difference between the ends of the structure. This is one of the reasons (in addition to the difficulty in measuring λ for films on the substrate) why, in the literature, the Power Factor parameter, PF, is also widely used (Equation (3)).

$$PF={\propto }^2·\sigma \left[{\displaystyle \frac{\mathrm{W}}{\mathrm{m}·{\mathrm{K}}^2}}\right]$$

(3)

By definition, the Seebeck coefficient is determined as the quotient of the measured electromotive force, E_T, and the temperature difference between the hot and cold ends of the investigated material:

$$\propto ={\displaystyle \frac{E_T}{\mathsf{\Delta }T}}$$

(4)

1.2. Thermoelectric Materials Based on Germanium

Semiconductors are materials of particular interest for thermoelectric applications, especially due to their high Seebeck coefficient, high ZT and doping capabilities. It is possible to modify their type of electrical conductivity (the same thermoelectric material can be of either p- or n-type), for example, carrier concentration or thermal conductivity, allowing for material ZT engineering [5,7,8]. High ZT materials (ZT > 1) have been known and used for several decades [9,10,11,12,13]. Bi₂Te₃ and Sb₂Te₃ are the commercial low-temperature thermoelectric materials most commonly used [14,15,16]. But there is also a lot of research on materials based on germanium and tellurium, such as TAGS (Te:Ag:Ge:Sb) and GeTe, as well as on silicon.

The thermoelectric properties of TAGS materials made as solid structures by powder sintering were studied in [17,18,19]. Chang et al. [19] obtained PF = 2.5 mW/(m·K²) and ZT = 0.8 at 473 K for the structure (GeTe)₈₅(AgSbTe₂)₁₅ (TAGS-85). Schröder et al. obtained a 50% higher Seebeck coefficient for TAGS-80 but also more than twice lower electrical conductivity at 668 K. It translated to PF = 2.1 mW/(m·K²) and ZT = 1.75 [18]. Six years earlier, similar results were presented by Yang et al. [17]. He examined TAGS materials with different compositions—(GeTe)_x(AgSbTe₂)_100−x (x = 75, 80, 85 and 90; TAGS-x). TAGS-85 had the best parameters: PF = 3.15 mW/(m·K²), ZT = 1.5 at 720 K.

On the other hand, GeTe-based materials doped by other elements are often reported in the literature. They are one of the most promising thermoelectric materials for medium temperatures, mainly thanks to a combination of their unique structural, thermal and electronic properties. They are fabricated using a variety of methods—hot pressing of the powders [20,21,22], fusing and quenching the melts in air or in water [23,24], and magnetron sputtering [25]. Ivanova et al. reported hot-pressed (GeTe)_0.97(BiTe)_0.03 samples doped with 2 at.% Cu [20]. The Power Factor was at the level of 3.63 mW/(m·K²) and ZT = 1.5 at 600 K. Luo et al. doped GeTe with 1% Cu₂Se [21]. He achieved the Power Factor equal to 4.4 mW/(m·K²) at 600 K, but the high thermal conductivity decreased ZT to the level of 0.73 at 600 K. Suwardi et al. investigated the enhancement of thermoelectric and mechanical properties of germanium telluride via rhenium doping. GeTe and GeSbTe were doped with 2 at.% of Re [22]. The dopant noticeably improved the mechanical properties; however, the influence on the thermoelectric parameters was negligible. The final parameters at 760 K were PF = 4.27 mW/(m·K²) for GeTe and PF = 4.37 mW/(m·K²) for GeSbTe and the ZT 1.0 and 1.8, respectively. Morgan et al. fabricated germanium telluride thin films for wearable energy harvesting. GeTe was directly sputtered by radio frequency (RF) magnetron co-sputtering [25]. The achieved Power Factor was 0.76 mW/(m·K²).

Unfortunately, many of the notable thermoelectric materials have some drawbacks, such as high production costs, complicated manufacturing processes, poor compatibility with CMOS technology, and use of rare or toxic constituents. In an era of high interest in green electronics and green materials, the latter feature becomes of added importance. Unfortunately, both antimony and tellurium are classified as toxic elements. Therefore, there is a systematic increase in interest in Group IV materials, which are non-toxic and relatively easy to obtain. We are mainly talking about silicon and, again, germanium, but without dangerous additives. For example, SiGe alloy is a well-known, reliable, and nontoxic material. It is appreciated by NASA and has been used in space missions for decades [13,14]. But not only NASA is interested in developing SiGe. Xu et al. manufactured the Si₈₀Ge₂₀ alloy doped with P and Ga using hot isostatic pressed powders. The samples were fabricated with different Ga/P ratios from 1:1 to 1:7, with a total P concentration varying between 2 mol% and 3 mol% and a total Ga content concentration ranging from 0.5 mol% to 1.8 mol% [26]. They achieved the PF = 4 mW/(m·K²) at room temperature. Miyata et al. prepared an amorphous thin film by the alternate deposition of Si and Ge doped heavily with Au in an ultrahigh vacuum chamber [27]. After post-processing annealing, they achieved PF = 4 mW/(m·K²) at 600 K. Perez et al. investigated Si₈₀Ge₂₀ thin film fabricated using DC sputtering on highly ordered porous alumina matrices [28]. The resulting structure has PF = 0.44 mW/(m·K²) and ZT = 0.09 at 300 K. Taniguchi et al. prepared both SiGe and Ge epitaxial films on Si substrates using molecular beam epitaxy (MBE) [29], where the highest PF = 4.7 mW/(m·K²) exhibits the epitaxial Ge film.

The fact that Ge-containing films, with only minor doping of other materials, are promising thermoelectric materials is confirmed in the literature [30,31,32,33,34]. The environmentally friendly, pure Ge thin films are characterized by particularly high Seebeck coefficients at room temperature, but as they are undoped, their electrical conductivity is low. Consequently, the Power Factor is quite low. However, even slight doping strongly increases these parameters. Miyata et al. investigate the thermoelectric properties of amorphous Ge: Au thin films prepared by alternating evaporation of Ge and Au in the ultra-high vacuum chamber [30]. The samples differed in Au concentration. The best parameters were found for Ge:Au (10 wt.% of Au), PF = 1.2 mW/(m·K²) at 1000 K. Nowak et al. used magnetron sputtering from circular alloy targets onto glass substrates to prepare a thin film Ge:Au (5 wt.% of Au). The PF reached 0.15 mW/(m·K²) at 500 K [31]. Kusano et al. reached PF = 0.24 mW/(m·K²) on glass substrate and PF = 0.16 mW/(m·K²) on flexible polyimide substrate, both at room temperature, for pure Ge thin film [32]. 100 nm thick polycrystalline Ge_1-xSn_x films grown by magnetron sputtering versus Sn composition (0.09 ≤ x ≤ 0.15) were studied by Portavoce et al. [33]. The Power Factor of the best structure, Ge₀.₈₅Sn₀.₁₅, was 0.056 mW/(m·K²) at room temperature. Ishiyama et al. [34] described the fabrication process and thermoelectric performance in polycrystalline Yb₃Ge₅ thin films deposited onto SiO₂ glass substrates at room temperature using RF magnetron sputtering. He received PF = 1.5 mW/(m·K²) at room temperature.

Furthermore, the authors of this publication investigated magnetron-sputtered germanium thin films doped by various materials [35,36,37,38,39,40]. The Ge:Sb:W, Ge:Sb:Ta, Ge:Au:V, Ge:Au:Hf, Ge:Au and Ge:V films were fabricated and characterized. The content of dopants did not exceed 10 at.%. The PF of the films was in the range of 0.1 ÷ 0.6 mW/(m·K²), and the best parameters show a Ge:Au film (6.5 at.% of Au).

The electrical properties of Ge films are strongly dependent on their crystalline structure. The application of amorphous films is limited because of their very high resistance. With an appropriate choice of deposition conditions and the admixture introduced to the Ge film, it is possible to obtain films with embedded crystalline grains [37]. The grains generally have much higher electrical conductivity than the amorphous phase. The grain dimensions and overall contribution of the crystalline phase in Ge films can be changed by the annealing process at elevated temperatures. It eliminates metastable states (relaxation of metastable states), especially concerning the vacancy distribution.

The post-process annealing process is widely used by scientists and the electronics industry to decrease the electrical resistivity of films [23,25,31,32,34,36,38].

The main objective of this work was to study the effect of post-processing thermal activation on film performance. The study was conducted on germanium films doped with two different metals (gold or vanadium) at different concentrations to determine how activation affects different materials. Two different post-annealing processes leading to the thermal activation of Ge-based thin films are described in the paper. In Section 2, the fabrication process of Ge:V and Ge:Au structures using magnetron sputtering is described. Five structures were prepared. Three were activated using the method described in Section 3.1. The other two were activated using the second method (Section 3.2).

2. Materials and Methods

Films based on doped germanium, deposited by the standard method of magnetron sputtering in an argon atmosphere, are characterized by disordered internal structures with numerous defects. It is also possible to use non-standard deposition—on substrates heated to high temperatures (above 850 K) [36]. This significantly improves the performance of the films, but at the same time is technologically more difficult. The results presented in the paper are related to the possibility of improving the thermoelectric parameters of films obtained by standard deposition through post-processing thermal treatment.

Ge:Au and Ge:V test structures (doped film contents in the range of 0.6 ÷ 4.17 at.%) were fabricated by reactive magnetron sputtering in a low-pressure argon plasma. The compositions were applied to Corning 7059 glass substrates through metal masks. The process was performed in a Pfeiffer PLS 570 HV-type vacuum deposition system (Pfeiffer Vacuum+Fab Solutions, Aßlar, Germany). A WMK-100 magnetron, a Dora Power System (Wrocław, Poland) power supply and a germanium target of 99.99% purity doped with Au, as well as a germanium target of the same purity on which V was placed, were used. A circular magnetron was supplied with unipolar pulses of 1100 ÷ 1200 V at the peak and power of 400 ÷ 500 W under Ar pressure ranging from 4 to 6.5 × 10⁻² Pa. In the unipolar pulse of 160 kHz mode, the power was controlled with the length of the current pulse train and intervals between them (frequency of 1.7 kHz). The films were deposited at the rate of about 0.33 to 0.42 nm/s. The distance of the substrates from the target surface did not exceed 100 mm. The composition of the films was determined by EDS analysis. The thicknesses of the deposited films were measured by an optical profilometer. They ranged from 0.8 μm to 1.3 μm. NiCrSi/Ag contact fields with thicknesses ranging from 1.2 to 2.2 µm were deposited on the films. Figure 1 shows exemplary test structures fabricated on Corning 7059 substrates.

Two methods of post-processing activation of thermoelectric structures were studied. The first method was to place the structure at an elevated temperature for a specified period of time. After this, the basic output parameters of the structure were measured. Temperatures of 723 K or 823 K and times of several minutes were used.

The second method involved placing the structure on a heating table and cycling the heating and cooling several times from room temperature to about 823 K. The resistance of the structures was monitored in situ, allowing moments of film activation to be observed.

Five different structures were investigated. They differ in the kind of dopant and its quantity. The parameters of all structures are summarized in

Table 1. The parameters of all fabricated and investigated structures.

Symbol	S1	S2	S4	S3	S5
Material	Ge:V (0.95 at.%)	Ge:V (1.47 at.%)	Ge:Au (3.13 at.%)	Ge:V (0.6 at.%)	Ge:Au (4.17 at.%)
Activation method	1	1	1	2	2
Max temp. of activation [K]	723	823	823	823	823
Film thickness [μm]	1.1	0.9	1.0	0.8	1.3
Related figure (characteristics)	Figure 2	Figure 3	Figure 4	Figure 5	Figure 6
ρ  [µΩ·m] at 300 K, after activation	64	32	375	140	85
ρ  decrease at 300 K, after activation	600×	30×	40×	1000×	2.5×
α  [μV/K] at 375 K/500 K	197/212	180/190	118/144	199/210	132/147
α  increase at 375 K, after activation	–	200×	20%	–	38%
PF  = α2/ρ [mW·m−1·K−2] at 375 K/500 K	0.75/1.24	0.14/0.20	0.04/0.07	0.34/0.53	0.21/0.28

.

Before and after the post-processing activation, S1–S5 samples were investigated using a scanning electron microscope (SEM) with an energy-dispersive (EDS) detector. The results are presented in Appendix A.

3. Results

3.1. The First Method of Post-Processing Activation

Two different structures based on germanium doped with 0.95 at.% or 1.47 at.% of vanadium and one doped with gold at 3.13 at.% were selected for testing. The structures were marked S1, S2 and S4, respectively. Their parameters are given in Table 1.

3.1.1. Structure S1

The internal resistance, R, of the S1 structure (germanium doped by 0.95 at.% of vanadium, film thickness of 1.1 μm) was initially measured at room temperature. Next, it was placed on a heating table and preheated to 723 K. After time intervals (1, 2, 5, 10, 20, 50 min), the structure was removed, and its electrical resistance was measured at room temperature. Then, it was placed on the table again for the next time interval. The results of the measurements are shown in Figure 2a. The electrical resistance (and therefore resistivity) increases for the first 2 min, then continuously decreases up to the 20th minute. The last measurement, after 50 min, no longer showed the change in resistivity. The shape of the characteristics suggests that initially, the lattice structure is highly disordered, and electrical conduction occurs on its defects. In the first minutes of annealing (1–2 min), the number of defects decreases, but the crystal lattice is not yet ordered enough for the resistivity to decrease. In the results the increase of structure resistance was observed. Between 10 and 20 min, the ordering of the network was completed, and the resistance dropped to a final level of 320 Ω. The electrical resistivity of the film before the activation process was 38,200 μΩ·m (at 300 K). After activation, it dropped to 64 μΩ·m (at 300 K). It is almost 3 orders of magnitude. After the activation process, the characteristics of R(T) were measured and the characteristics of ρ(T) were determined for the tested film in the temperature range of 298 ÷ 587 K (Figure 2b).

After the activation process, the electromotive force of the structure was measured. The temperature of the cold end was stabilized at 303 ÷ 313 K; the hot end was heated up to 587 K. The generated electromotive force versus ΔT was characterized. Based on it, the Seebeck coefficient was determined. The results are shown in Figure 2c,d. The Seebeck coefficient was 197 μV/K for T_HOT = 375 K and 212 μV/K for T_HOT = 500 K. The characteristics indicate the linear nature of the relationship α(T_HOT). This repeated also with the results of the subsequent S2–S5 structures. The main source of inaccuracy was the location of the temperature measurement—not directly on the tested film but on the heating table/radiator under the structure. Due to higher measurement inaccuracies at the low-temperature difference between the ends of the tested structure, the results were read for 375 K, where the characteristics stabilized.

The PF coefficient was 0.75 mW/(m·K²) for T_HOT = 375 K and 1.24 mW/(m·K²) for T_HOT = 500 K (Figure 2e).

3.1.2. Structure S2

The scope of the test was expanded for the S2 structure (germanium doped with 1.47 at.% of vanadium, film thickness 0.9 μm). Not only was the electrical resistance measured, but the characteristics of E_T(ΔT) and R(T) were measured before the activation process. The structure was then placed on a heating table and preheated to 823 K, a temperature 100 K higher than in the S1 case. After time intervals (2, 5, 10 min), it was removed, and the characteristics E_T(ΔT) and R(T) were again measured in the range of 298 ÷ 500 K. The structure was then placed on the table again for the next time interval. The results of the measurements are shown in Figure 3.

The E_T(ΔT) characteristic for the inactivated film was measured first. It is shown in Figure 3a. For T_HOT temperatures around 390 K and 450 K, changes in its slope can be seen (T_COLD was about 310 K)—marked by red arrows in Figure 3a. This should be interpreted as points of irreversible transformation of the semiconductor’s internal structure. At these points, the first stages of film activation occurred. They translated into an increase in electromotive force and then a decrease in electrical resistivity. This means that the activation process already began during the initial measurements of the structure, even before the main annealing at 823 K. During characterization, the hot part of the film was gradually heated to 500 K, which already caused the first changes in the internal structure. It is likely that only in a part of the film since the cold part was stabilized to near-room temperature all the time. The characteristic in Figure 3a has a few more, less obvious, ripple points. These could be further transformation points, but could also be simply due to measurement inaccuracies.

In the second case, the characteristic R(T) was measured for the inactivated film. It is shown in Figure 3b. For T_HOT temperatures around 340 K, 380 K and 450 K, changes in its slope can be seen (T_COLD was about 310 K)—marked by red arrows in Figure 3b. This should be interpreted as further points of irreversible transformation of the semiconductor’s internal structure. The film passes through subsequent activation thresholds, which each time affect its parameters. It should be noted that when the R(T) was measured, the film was already preactivated as a result of the E_T(T) measurement, as described above. However, during E_T(ΔT) characterization, only a hot part of the film was heated. The entire structure was then heated up to 500 K. If the E_T(ΔT) characteristic were measured again in the next step, it would probably have a shape different from the original (E_T at a higher level).

The S2 structure was then placed on a heating table at 823 K. After 2 min, it was removed from the heater, and the characteristics of E_T(ΔT) followed by R(T) were measured again, as described above.

The characteristics of E_T(ΔT) are linear (Figure 3c). After a step change after the first activation cycle (t = 0 min), no further changes in the level of generated E_T were observed. The temperature of the cold end was stabilized at 304 ÷ 311 K; the hot end was heated to 500 K. Based on the measured thermoelectric force, the Seebeck coefficient of the S2 structure was determined at different stages of the activation process. The results are shown in Figure 3d. The Seebeck coefficient was 1.75 µV/K for T_HOT = 375 K and 5.6 µV/K for T_HOT = 500 K before the activation process. After activation, it increased to 180 µV/K for T_HOT = 375 K and 190 µV/K for T_HOT = 500 K. The α coefficient increased by two orders of magnitude at 375 K but only 35 times at 500 K. This is due to the progressive activation of the film during the initial measurements, as described above. Characteristics indicate the linear nature of the α(T) relationship, with a clear deviation from this shape at lower temperatures. This should be regarded as measurement inaccuracy, which has the greatest impact on the results at low temperatures (because of the low generated thermoelectric force).

The characteristics of R(T) before the activation process (t = 0 min) have a shape gently deviating from linear (also for the plots after t = 2 ÷ 10 min, which is faintly visible on the graph shown)—Figure 3e. The electrical resistivity of the film at 300 K before the activation process was 10,000 μΩ·m After activation; it dropped to 315 μΩ·m, it is more than 30 times.

Electrical resistance (and, therefore, resistivity) decreased significantly after the first activation cycle. The generated electromotive force increased by two orders of magnitude at the same time. This means that the activation process started faster than in the case of the S1 structure. Measurements after the second time interval (5 min) led to full activation of the structure. A decrease in electrical resistance was still observed (Figure 3e), but the changes in electromotive force were already negligible, comparable to the inaccuracy of the measuring instruments (Figure 3c). The last time interval (10 min) has resulted in only minor changes in the thermoelectric parameters of the S2 structure.

The PF coefficient of the activated film was 0.14 mW/(m·K²) for T_HOT = 375 K and 0.2 mW/(m·K²) for T_HOT = 500 K (Figure 3f). The characteristic has a near-linear shape, which is due to the small temperature dependence of the α coefficient.

3.1.3. Structure S4

The S4 structure (film thickness 1.0 µm) was made of a different material—germanium doped with 3.13 at.% of gold. The activation procedure was similar to that of S2. The initial measurements were made before the annealing process. The E_T(ΔT) characteristic for the inactivated layer was first measured, followed by the R(T). The structure was then placed on a heating table and preheated to 823 K. After time intervals (2, 5, 10 min), the structure was removed from the table, and the characteristics of R(T) and E_T(ΔT) were measured in the range of 298 ÷ 500 K. The structure was then placed back on the table for the next time interval. The results of the measurements are shown in Figure 4.

In the characteristic E_T(ΔT) before the annealing process (t = 0 min), only very slight transformation points were observed (Figure 4a). For T_HOT temperatures around 360 K, 400 K and 420 K, changes in its slope can be seen (T_COLD was about 310 K)—marked by red arrows in Figure 4a.

The R(T) characteristic for the inactivated film (t = 0 min) was then measured. It is shown in Figure 4b. For T_HOT temperatures around 340 K, 350 K, 390 K and 450 K, changes in its slope can be seen (T_COLD was about 310 K)—marked by red arrows in Figure 4b. This could mean that the temperature dependence of the resistivity is such, but also that such a shape could be due to the gradual activation of the film already during the initial measurement. This is suggested by the shape of the other characteristics (after t = 2, 5, 10 min), which are strongly close to a straight line (Figure 4c). The transformation points are consistent with the E_T(ΔT) characteristic (Figure 4a) and with the structure S2 (Figure 3b). Nearly complete activation occurs in the first activation cycle. As a result of the activation, the electrical resistivity at 300 K decreased from 15,000 μΩ·m to 375 μΩ·m; it is 40 times.

All other E_T(ΔT) characteristics are linear (Figure 4d). Unlike the case of electrical resistivity, the electromotive force does not increase to a maximum level after the first activation cycle. It increased by about 5%. After the second time interval (a total of 5 min at 823 K), it increased by another 23%. In subsequent cycles, the changes were small. Finally, after 10 min at 823 K, E_T increased by 23% compared to the pre-activation state at 375 K, and by 34% at 500 K. This proportionally translates into the Seebeck coefficient.

The Seebeck coefficient was 96 µV/K for T_HOT = 375 K and 107 µV/K for T_HOT = 500 K prior to the activation process (Figure 4e). After activation, it increased to 118 µV/K for T_HOT = 375 K and 144 µV/K for T_HOT = 500 K. The characterization shows the linear nature of the relationship α = f(T_HOT).

The PF coefficient of the activated film was 0.04 mW/(m·K²) for T_HOT = 375 K and 0.07 mW/(m·K²) for T_HOT = 500 K (Figure 4f).

The procedure of first way of post-processing activation of the thermoelectric films was to place the structure at an increased temperature (723 or 823 K). At every specified time interval, the structure was cooled to room temperature, and its thermoelectric parameters were measured. For the S1 structure, simple tests were performed at 723 K. They showed that activation begins in the first minute of annealing. Full activation takes 10 ÷ 20 min. During the first two minutes of activation, the resistivity of the film increased, which was related to the removal of electrical conductivity on defects in the semiconductor structure. It was not observed for the S2 structure. This could mean that the S2 structure was less defective after the film deposition process; consequently, conductivity on defects did not occur.

The study carried out for structure S2 was much more accurate and gave a better picture of the activation process. It can be seen that it started at a temperature of about 390 K when the characteristic E_T(ΔT) was measured (Figure 3a). Admittedly, only part of the structure was at this temperature (the cold part was close to room temperature), but in this part of the semiconductor, the first irreversible changes took place, altering the internal structure. The activation did not occur at one point in time but occurred in steps. With further heating of the hot part of the film, more transformation points appeared. Additionally, during the next measurement of the R(T) characteristic (still before the first annealing cycle), further transformation points were observed (Figure 3b). Further thermoelectric parameters measurements were made after the first annealing cycle at 823 K (2 min). They showed a step change in both the electrical resistivity (a decrease of 30 times) and the Seebeck coefficient (an increase of 200 times at 375 K). A further annealing cycle (total of 5 min) brought some improvement in electrical conductivity (generated electromotive force no longer increased). Further annealing did not have an effect on the structure. This means that the S2 structure was fully activated faster (2 ÷ 5 min) than the S1 structure (10 ÷ 20 min). This is mainly due to the higher amount of energy supplied to the film (annealing temperature 100 K higher), but it could also mean that the structure was less defective after the deposition process. The S1 and S2 structures differed in the percentage of vanadium dopant, which may have contributed to this difference. However, this issue was not thoroughly analyzed in this work.

The S4 structure activated in the first annealing cycle. Subsequent cycles only slightly affected the resistivity of the film but resulted in an increase in the generated electromotive force. The activation times of structures S2 and S4 appear to be very close to each other, despite the different dopants. The irreversible transformation points are for almost the same temperatures. They differed in the order in which they activated the parameters. For S2, the Seebeck coefficient reached the final level faster than the electrical resistivity. For S4, it was the other way around.

The Seebeck coefficient of the S4 structure before activation was high (almost 100 μV/K). During activation it increased by only 23% at 375 K. Otherwise for S2. α before activation was low (less than 1 μV/K). Upon activation, it increased 200 times. This means that Ge:V films absolutely require post-processing activation. Ge:Au films can be used immediately, even without post-processing annealing, in applications where the resistivity of the film is not important (e.g., certain types of sensors).

3.2. The Second Method of Post-Processing Activation

The second method consisted of placing the structure on a heating table and cycling heating and cooling several times between 333 K and 823 K. The resistance of the structures was monitored in situ. Two structures were selected for testing. One was based on germanium doped by 0.6 at.% of vanadium, and the other was based on germanium doped by 4.17 at.% of gold. Structures were marked S3 and S5, respectively. Their parameters are given in Table 1.

3.2.1. Structure S3

The S3 structure (germanium doped with 0.6 at.% of vanadium, film thickness 0.8 µm) was placed on a heating table with an initial temperature of 293 K and heating to 773 K began. During heating, the electrical resistance was measured in situ. The resulting characteristics R(T) and ρ(T) are shown in Figure 5a (heating_1). The characteristics are non-linear, which may indicate a strong temperature dependence of the resistivity. However, closer inspection indicates that the characteristics can be divided into 3 sections. Each of them is close to a straight line. This means that two points of irreversible transition in the semiconductor structure occurred during heating, around temperatures of 533 K and 623 K. During cooling_1 (773 K → 333 K), Figure 5b, there was a clear change in the characteristic curve. The resistance of the structure increased. This is an effect analogous to that shown in Figure 2a for structure S1. The initially disordered lattice structure is dominated by the electrical conductivity on the structural defects. A reduction in their density results in an increase in resistivity. There are no other obvious transformation points. During the second heating cycle (heating_2), up to 823 K, the characteristic curve almost perfectly coincides with cooling_1. However, at 725 K, intensive changes in film resistivity begin. Between the measurement points for 763 K and 803 K, this process accelerates rapidly. A step change is observed, a decrease in resistivity of one order of magnitude. Subsequent heating/cooling cycles (in the range of 333 ÷ 823 K) no longer cause any change. Finally, the electrical resistivity of the film reached 140 μΩ·m (at 300 K), an improvement of 3 orders of magnitude. The shape of the characteristics after the activation process indicates a much smaller temperature dependence of the electrical resistivity. The shape is similar to that of the activated S2 structure.

After the activation process the electromotive force was measured. The temperature of the cold end was stabilized at 302 ÷ 309 K, the hot end was heated to 544 K. The generated electromotive force versus ΔT was characterized. On the basis of it, the Seebeck coefficient was determined. The results are shown in Figure 5c,d. The Seebeck coefficient was 199 µV/K for T_HOT = 375 K and 210 µV/K for T_HOT = 500 K. Characteristics indicate the linear nature of the relationship α(T_HOT).

The PF coefficient was 0.34 mW/(m·K²) for T_HOT = 375 K and 0.53 mW/(m·K²) for T_HOT = 500 K (Figure 5e).

3.2.2. Structure S5

In the case of the S5 structure (germanium doped by 4.17 at.% of gold, film thickness 1.3 µm), the tests started by measuring the E_T(ΔT) characteristics of the inactivated film. The temperature of the cold end of the structure was stabilized at 297 ÷ 306 K; the hot end was heated to 500 K. The generated electromotive force versus ΔT was characterized. The measurement results are shown in Figure 6a (before the activation graph). The characteristic is close to a straight line, with no clear changes in slope, which could indicate activation of the structure. The Seebeck coefficient determined from the first few measurement points was approximately 75 µV/K. For the last few measurement points, it was about 110 µV/K.

After this initial measurement, the S5 structure was placed on a heating table with an initial temperature of 293 K and heating up to 833 K was started. During heating, the electrical resistance was measured in situ. The resulting characteristics R(T) and ρ(T) are shown in Figure 6b (heating_1). The figure shows the change in the slope of the characteristic at 353 K. This is a point of transformation in the internal structure of the film. There is a step decrease in the electrical resistivity of the material. A second, much stronger change occurs in the temperature range 583 ÷ 613 K. The film activates in the first annealing cycle, which is faster than in the case of the S3 structure. However, it should be noted that for structure S5, the E_T(ΔT) characteristic was investigated before annealing, which could already pre-activate the film. No analogous measurements were performed for S3. During the subsequent annealing cycles (cooling_1, heating_2, cooling_2), no more significant changes were observed in the electrical resistivity, which stabilized at the level of 85 μΩ·m (at 300 K). It is an improvement of 2.5 times compared to the first measurement. The shape of the ρ(T) characteristic is close to a straight line. The shape is similar to that of the activated S4 structure.

After the activation process, the electromotive force was measured (Figure 6a, after the activation graph). The temperature of the cold end was stabilized at 298 ÷ 306 K, and the hot end was heated to 500 K. The generated electromotive force versus ΔT was characterized. On the basis of it, the Seebeck coefficient was determined. The results are shown in Figure 6c. The Seebeck coefficient was 132 µV/K for T_HOT = 375 K and 147 µV/K for T_HOT = 500 K. For lower temperatures (below 375 K), the graph is non-linear, similar to the previous α(T_HOT) characteristics. This is probably due to measurement inaccuracies. At higher temperatures, the function is linear.

The PF coefficient was 0.21 mW/(m·K²) for T_HOT = 375 K and 0.28 mW/(m·K²) for T_HOT = 500 K (Figure 6d).

Tests have shown that the second method of post-processing activation also has the intended effect. The activation process for the S3 structure is analogous to that for the S1 structure. After the first portion of heat energy is supplied, the resistivity increases, and the film requires further activation (Figure 5b). In the case of the S1 structure, an increase in resistivity was observed during the first 2 min of annealing at 723 K. In the case of S3, the heating process of 293 → 773 K took approximately 5 min, but the sample remained at the maximum temperature of 773 K for several seconds only. The structure was then cooled. The heating_1 characteristic (Figure 5a,b) shows two obvious transformation points. This results in a change in the shape of the characteristics, as can be seen by comparing the heating_1 and cooling_1 waveforms (Figure 5b). Finally, during the next heating cycle (heating_2, Figure 5b), at a temperature of about 770 K, there is a step decrease in resistivity of one order of magnitude. Full activation of the S1 structure occurred after 20 min of continuous holding at 723 K. For the S2 structure, this was 5 min at 823 K. For S3, the process took about 20 min, but the sample was only kept at peak temperature only for a few tens of seconds (773 K, then 823 K). The heating_1 and heating_2 cycles lasted about 5 min each, and cooling_1 about 10 min. The results obtained are in line with predictions—holding at a higher temperature accelerates the activation process, as the thermal energy is delivered in a stronger flux.

The activation of the S5 structure occurred noticeably faster than that of S3. During the first heating_1 cycle, above 623 K, the characteristics take on an almost final shape. Subsequent heating/cooling steps cause only small changes in the resistivity of the film. However, it should be noted that for S5, the characteristic E_T(ΔT) was measured earlier (before the heating_1 cycle). Part of the structure was heated to 500 K, resulting in partial activation.

4. Discussion

The presented methods of post-process thermal activation of semiconductor films fulfill their function. For each of the structures tested, the thermoelectric parameters improved significantly. The electrical resistivity measured after the activation process was between 2.5 and even 1000 times lower (depending on the structure) than that measured immediately after the film deposition process. This is a large discrepancy in the degree of improvement. However, it is not due to the use of different activation methods. The type and concentration of dopant in germanium (Au or V) may be of some importance, but it is unlikely to be the key factor. Such large differences in the degree of decrease in resistivity are most likely due to large differences in the degree of defectification of the films immediately after the deposition process. Post-processing annealing of a heavily defect semiconductor causes ordering of its internal structure—defects are removed, and grain growth occurs. If the film was already of much better quality after the deposition process, the changes due to activation are smaller. The quality of the film immediately after deposition depends on a number of factors, both process-related (e.g., vacuum quality, substrate temperature, deposition rate) and also related to the type and concentration of the dopant. For structures S2, S4 and S5, the electrical resistivity decreased from 2.5 to 40 times, which is significantly less than for structures S1 and S3 (250 ÷ 1000 times, see Table 1). This is due to the order in which the material parameters were measured. For S2, S4 and S5, the characteristic E_T(ΔT) was measured before R(T). The films started to activate already during the electromotive force measurements. This means that the first measurement of the internal resistivity involved an already pre-activated semiconductor. This is also the reason for the smaller jump in the electrical resistivity value.

After the deposition process, a smooth, amorphous Ge layer is obtained. SEM images of exemplary Ge:Au and Ge:V surfaces are provided in Figure A1 in Appendix A. The crystal lattice is defected by residual gases. Electrical conductivity can be attributed to embedded impurities such as oxygen or nitrogen [40].

During the annealing process, the internal structure ordering takes place. The dopant (Au or V) builds into defect sites, filling in the gaps in the lattice structure. According to Tsao [40], post-processing annealing also causes the formation of isolated crystalline grains that reduce the volume of the continuous amorphous germanium lattice, which explains the initial increase in electrical resistivity. This agrees with our observations. This is well shown in the activation illustrations for the S1 and S3 structures (Figure 2a and Figure 5b). The grains conduct well in their volume, but the flow of electric charge carriers across their boundaries is hindered because they are insulated from each other. This results in high resistivity of the layer and a decrease in PF during the initial stage of annealing. Further annealing causes the grains to grow and be bridged by metallic doping. At some point, the percolation threshold is exceeded. Electrical conduction begins to occur directly between the grains. The annealing also caused the transformation from the amorphous phase to the crystalline (111), typical for Ge [30,35,36,41,42,43].

From the point of view of the study, the most significant finding is that thermal activation is an effective method to sort out the internal structure of the film and leads to a significant improvement in the output parameters. The electrical conductivity of the S5 structure (Ge:Au, 4.17 at.% of Au) is approximately 11,800 S/m and is very close to the results reported in the literature: 7000 ÷ 16,000 S/m [27,30,31,36]. At the same time, it is obtained in a technologically simpler way—it does not require maintaining a high substrate temperature (e.g., 850 K in [36]) during the deposition process.

The initial Seebeck coefficient of the investigated films (before the activation process) was determined only for structures S2, S4 and S5. Therefore, only in their case is it possible to determine the effect of activation on this parameter. As in the case of ρ, there was a clear improvement for each structure. For the S2 (Ge:V) structure, the Seebeck coefficient increased by about 2 orders of magnitude. For Ge:Au films, the increase was not as spectacular—23% and 38% for the S4 and S5 structures, respectively. The gold-doped films were already characterized by a relatively high α (close to 100 µV/K) after the deposition process. In the case of vanadium doping, these were a single µV/K. This leads to the conclusion that Ge:Au films without the activation process are characterized by α at a high level, which allows their practical application without post-processing annealing (e.g., in sensors, where the internal resistance is less important). Ge:V films absolutely require activation.

In the paper two methods of postprocessing thermal activation were presented. In practice, it does not matter which method is used to reorder the internal structure of the film. From a physical point of view, both works similarly—they provide thermal energy to the film’s crystal lattice. Differences come down to research aspects. The second way (heating cycles) allows us to observe the sample continuously—gradual heating smoothly transitions to gradual cooling after reaching the maximum set temperature. This makes it possible to accurately see all changes in the resistivity of the film. The disadvantage is that the electromotive force can be measured before and after the heating cycles only. Using the first way of activation, measurements of E_T(ΔT) and R(T) can be made alternately after each heating cycle. This allows the observation of changes in both parameters. Unfortunately, not continuous, which is a disadvantage of the method from a research point of view. The second disadvantage is the exposure of the film to thermal shocks, a rapid temperature drop from 723/823 K to room temperature. Here, it should be noted that, despite such shocks, the structures were not damaged, which positively demonstrates the reliability of the Ge:Au and Ge:V films.

In the case of the solution presented in the work [36], thermal energy was already provided during the film deposition process (the substrate was on a table heated to 700 ÷ 850 K). The internal structure was ordered on the fly, and after the deposition process, the film was ready for use. The disadvantage of this solution is the technological complications associated with the need to use a table heated to a high temperature inside the vacuum chamber. Post-process heating achieves similar results without complicating the vacuum system. The results of the presented studies show that such a procedure can be successfully applied to various germanium-based films doped with gold or vanadium in different concentrations.

5. Conclusions

The level of the thermoelectric parameters obtained from the tested films is less relevant in the context of this paper. However, they were collected in Table 1. The best parameters were achieved for the S1 structure, germanium doped by 0.95 at.% of vanadium. Its PF coefficient is 0.75 mW/(m·K²) at 375 K and 1.24 mW/(m·K²) at 500 K. The Seebeck coefficient of the film is about 197 µV/K at 375 K and 212 µV/K at 500 K. In comparison, for the bulk TAGS-85 alloy reported in the literature [19], the PF value reaches 2.5 mW/(m·K²) at 473 K, with a lower α value of 158 µV/K.

The films studied can be used as thermoelectric material for medium temperatures, in sensing and energy harvesting applications. It should be emphasized that the simplicity of their fabrication technology and the ability to improve parameters with a simple process, despite imperfect deposition conditions, is not without limitation. They show high stability and improvement of properties with long-term use, as indicated by the activation processes carried out, which can be regarded as basic aging (temperature) tests. For long-term operations at elevated temperatures, such behavior is an obvious advantage. Such conditions are the basic working environment of thermocouples.