1. Composition

As mentioned before, the composition of the deposited material can be controlled through the applied current density or deposition potential. Higher current densities or more negative potentials lead to a decrease in Te content in the deposit, a relation, which has been well established in various publications. Examples for this relationship can be found in virtually every publication mentioned, see e.g., Tittes et al. for galvanostatic conditions^{Reference Tittes, Bund, Plieth, Bentien, Paschen, Plötner, Gräfe and Fischer8} and Michel et al. for potentiostatic conditions.^{Reference Michel, Diliberto, Boulanger, Stein and Lecuire32} Achieving the stoichiometric material composition can be considered to be standard procedure, with the only requirement of having an appropriate electrolyte.

2. Morphology

The most frequently encountered morphology for stoichiometric Bi₂Te₃ is commonly described as needle-like and occurs for both current as well as potential-controlled deposition methods. Other morphologies can be achieved, e.g., granular or dendritic structures,^{Reference Naylor, Koukharenko, Nandhakumar and White34} but are typically related to either incorrect composition or unreasonably high-growth rates.^{Reference Michel, Diliberto, Boulanger, Stein and Lecuire32,Reference Wen, Corderman, Seker, Zhang, Denault and Blohm38} Exemplary images for well-formed Bi₂Te₃ films with the needle like structure, deposited by Ma et al.,^{Reference Ma, Ahlberg, Sun, Iversen and Palmqvist42} are shown in Fig. 2.

FIG. 2.

Needle-like surface morphology of electroplated Bi₂Te₃ films (images taken from Ref. Reference Ma, Ahlberg, Sun, Iversen and Palmqvist42 reproduced by permission of The Electrochemical Society).

3. Thickness

The deposited material thickness varies greatly, with reported values between 100 nm and 800 µm, which are likely due to the different foci of the individual research. For integration into TE generators, a minimum thickness of some tens of microns is required, with higher values being beneficial.^{Reference Lim, Snyder, Huang, Hennan, Ryan and Fleurial43,Reference Glatz, Schwyter, Durrer and Hierold44} Thick films have been achieved under both potentiostatic and galvanostatic conditions, as reported by Li et al.,^{Reference Li, Toprak, Soliman, Zhou, Muhammed, Platzek and Muller11} as well as by applying a pulsed deposition method, as reported by Glatz et al.^{Reference Glatz, Durrer, Schwyter and Hierold12} Therefore, while maybe not necessarily state-of-the-art, the deposition of thick films of Bi₂Te₃ has been proven to be feasible. An original approach was proposed by Maas et al. with the use of a soluble anode made of bismuth telluride.^{Reference Maas, Diliberto, de Vaulx, Azzouz and Boulanger18} Thus, dense and homogenous films up to 400 µm were obtained with low-internal stress.

4. Crystallinity

XRD experiments conducted on the deposited films universally confirm the existence of a Bi₂Te₃ phase, proving that the deposition process leads to at least partially crystallized materials. Most reports also indicate a preferential (110) orientation of the deposited material, as long as it is close to the correct stoichiometry. The crystallization process was studied in detail by Chaouni et al.,^{Reference Chaouni, Bessieres, Modaressi and Heizmann45} who investigated the orientation of the films as a function of substrate, material composition, and deposition thickness. The investigation revealed that various textures are possible for small growth thicknesses (<10 µm), which are dependent on both stoichiometry and substrate material. However, for films with a thickness of >10 µm, the (110) preferred orientation always appeared, regardless of the initial conditions. This effect is attributed to the higher in-plane electrical conductivity of the highly anisotropic Bi₂Te₃ crystallites. Similar results were obtained by Richoux et al.^{Reference Richoux, Diliberto, Boulanger and Lecuire37} in a study investigating the details of the nucleation mechanism for pulsed deposition methods. Additionally, they concluded that pulsed methods lead to smaller aggregates and, therefore, films with decreased roughness. Additionally, the effect of the pH, in the 1–3.5 range, on the structural properties of the material was investigated by Suresh et al.^{Reference Suresh, Chatterjee, Sharma, Ganguly, Kargupta and Banerjee46} Their results show that an increase in the pH of the solution results in a decrease in the grain size and dislocation density.

1. Characterization methods

The electrical characterization of electroplated films is complicated by the unavoidable existence of the electrically conductive seed layer underneath the deposited material. The three main approaches to circumvent this problem, which are proposed in the literature, are to remove the deposited layer from the substrate and seed layer, to extract the properties of the deposited layer from the combined measurement or finally to use a thin seed layer composed of the same material as the deposited one.

Removing the deposited material from the seed layer was already proposed by Takahashi et al. in 1994,^{Reference Takahashi, Katou, Nagata and Furuta47,Reference Takahashi, Kojima, Sato, Ohnisi, Nishiwaki, Wakita, Miyuki, Ikeda and Muramatsu48} who used an epoxy resin to transfer the film onto a glass plate. Similar processes were described in detail by Heo et al.^{Reference Heo, Hagiwara, Ichino and Okido10} and Ma et al.^{Reference Ma, Ahlberg, Sun, Iversen and Palmqvist14} The deposited material can then be characterized with standard metrology equipment. A potential disadvantage of this method is that the films can be damaged by the mechanical stress applied during the peeling process. Nonetheless, this method is probably the most commonly applied approach in the literature.

Calculating the film properties based on the measured properties of the seed layer and the measurement of the combined properties was proposed by e.g., Nguyen et al.^{Reference Nguyen, Su, Wang, Vullers, Vereecken and Fransaer49} and Deb et al.^{Reference Deb, Isoda, Caballero-Calero, Diaz-Chao, Martin-Gonzalez and Shinohara50} While this method is convenient, it requires an additional measurement of the seed layer properties and is reliant on several simplified assumptions. An experimental investigation of this approach was conducted by Rostek et al.^{Reference Rostek, Kottmeier, Kratschmer, Blackburn, Goldschmidtböing, Kröner and Woias51} for Bi₂Te₃ layers on Platinum seed layers with varying thickness for the individual layers. It was concluded that reasonable results with accuracy of better than ∼15% can be achieved for the measurement of the electrical resistance and the carrier concentration, as long as the deposited film thickness is significantly larger than the seed layer thickness (ratio of >50). The determination of the Seebeck coefficient was also possible, however, due to the influence of the thermal properties of the substrate material, is somewhat less accurate, and requires a higher thickness of the TE material (ratio of >125).

Using a seed layer of the same material as the deposited one was proposed by Yoo et al.,^{Reference Yoo, Huang, Lim, Herman, Ryan, Fleurial and Myung9} who used a seed layer of sputtered Bi₂Te₃ for the experiments. The measurements of the film can then also be conducted with standard metrology. However, the method relies on the assumption that both layers have similar properties, which might not necessarily be the case. Each of these three methods can be used to obtain the TE properties of electroplated materials. Removing the deposited material from the substrate, if carefully conducted, is, however, probably both the most used as well as the most reliable method.

2. Transport properties

An overview of the obtained TE properties of electrodeposited Bi₂Te₃ is given in Table I. It is obvious that not every publication contains a full TE characterization of the synthesized material. This is presumably partly due to the different foci of some of these works and partly due to the described difficulties in obtaining these measurements. The most frequently reported value is the Seebeck coefficient, as it represents the most important TE property. This is followed by the electrical resistivity, and finally the carrier concentration, which is only reported in a minority of publications.

3. Seebeck coefficient

The Seebeck coefficients obtained for electroplated materials are typically on the order −40 to −80 µV/K, with maximum reported values around −125 µV/K (e.g., Refs. Reference Heo, Hagiwara, Ichino and Okido10, Reference Diliberto, Richoux, Stein and Boulanger13, and Reference Wu, Nguyen, Vullers, Vereecken, Binnemans and Fransaer16). The majority of publications report that only n-type material is obtained through the electroplating process, regardless of the stoichiometry of the deposited material.^{Reference Tittes, Bund, Plieth, Bentien, Paschen, Plötner, Gräfe and Fischer8} This is in direct contradiction to the well-established results achieved on materials prepared from melts,^{Reference Scherrer and Scherrer21} indicating that Te-deficient material should exhibit p-type conductivity. This dominant n-type doping in electroplated materials strongly indicates a different, or possibly additional, doping mechanisms than the antistructure disorder typically associated with Bi₂Te₃. This effect is discussed e.g., by Diliberto et al. and Ma et al.,^{Reference Diliberto, Richoux, Stein and Boulanger13,Reference Ma, Ahlberg, Sun, Iversen and Palmqvist14} however, to date no full explanation is available in the literature. While some reports include the fabrication of p-type material,^{Reference Miyazaki and Kajitani7,Reference Glatz, Durrer, Schwyter and Hierold12,Reference Jung and Ko26,Reference Nguyen, Wu, Su, Vullers, Vereecken and Fransaer27} they remain the exception rather than the rule.

4. Carrier concentration

The comparatively low values for the Seebeck coefficient are related to the high-carrier concentration associated with electroplated materials, which are typically on the order of 10²⁰ to 10²¹ 1/cm³ (Table I). These values are about one order of magnitude higher than the ones obtained for melted material and are responsible for the comparatively overall low TE quality. In combination with the predominantly n-type Seebeck coefficient, the high carrier concentration indicates an additional doping mechanism, which might be related to the high density of grain boundaries in electrochemically deposited materials.^{Reference Ma, Ahlberg, Sun, Iversen and Palmqvist14}

5. Electrical resistivity

The reported values for the electrical resistivity of near-stoichiometric materials are typically around 8–15 µΩ∙m, with worst-case values around 100 µΩ∙m. While these values are comparable to optimized materials, they are mostly related to the high-carrier concentration. The mobility of electroplated films is correspondingly rather low, with typical values around 5–20 cm²/Vs.^{Reference Magri, Boulanger and Lecuire4,Reference Huang, Wang and Murphy5,Reference Ma, Ahlberg, Sun, Iversen and Palmqvist14,Reference Fleurial, Snyder, Herman, Giauque, Phillips, Ryan, Shakkottai, Kolawa and Nicolet52} These values are about one order of magnitude lower than the ones obtained for optimized material^{Reference Scherrer and Scherrer21} and possibly also caused by the high density of grain boundaries.

An overview of the TE properties of electroplated materials in comparison to state-of-the-art, optimized materials, fabricated by a thermal bulk synthesis method and by coevaporation is given in Table I.

1. Additives

Since the as-deposited films are rough or dendritic in some cases, several publications are focused on the effect of additives, and primarily surfactants intended to improve the morphology of the deposited films, i.e., to make them more homogeneous. Examples are given in the work of Lim et al.,^{Reference Lim, Kim and Oh66} who used ethylenediaminetetraacetic acid, Kuleshova et al.,^{Reference Kuleshova, Koukharenko, Li, Frety, Nandhakumar, Tudor, Beeby and White68} who used sodium ligninsulfonate, and Yoo et al.,^{Reference Yoo, Song, Lim, Myung, Lee, Oh, Lee, Kim, Kim, Choa, Lee, Lee and Lim78} who investigated the effect of cetyltrimethylammonium bromide. For all cases, the addition of the surfactant led to a significantly improved morphology of the deposited films.

1. Composition

The composition of the binary compound can essentially be controlled by adjusting the ratio of ions in the electrolyte (see e.g., Ref. Reference Huang, Kellock and Raoux80) or by adjusting the deposition conditions, i.e., either the deposition potential or the applied current density. This work is analogous to the deposition of binary Bi₂Te₃ and is mostly unproblematic.^{Reference Del Frari, Diliberto, Stein, Boulanger and Lecuire73} One challenge is the strong dependence of the composition on the applied potential or current, which is more pronounced than for the deposition of Bi₂Te₃. This can possibly be attributed to the significantly lower binding energy of Sb₂Te₃, as described by Schumacher et al.^{Reference Schumacher, Reinsberg, Akinsinde, Zastrow, Heiderich, Toellner, Rampelberg, Detavernier, Broekaert, Nielsch and Bachmann70} The deposition of binary materials with the correct composition, while challenging, can be considered to be a state-of-the-art process. The deposition of ternary compound is significantly more difficult, as the correct balance between three individual components needs to be achieved. This typically requires the adjustment of the electrolyte composition in addition to finding the correct deposition conditions. One possible procedure to fully control the film composition is outlined in the work of Del Frari et al.^{Reference Del Frari, Diliberto, Stein, Boulanger and Lecuire73} and is based on adjusting the Sb to Bi ratio in the electrolyte. Unlike the deposition of the binary compounds, the full compositional control in the ternary compounds cannot be considered to be state-of-the-art and remains a challenge.

2. Morphology

Even if the film composition of the Sb₂Te₃ and (Bi, Sb)₂Te₃ compounds can be finely tuned, their morphology is one of the main challenges. While several different variations of the growth mechanism and resulting film morphology have been established in the literature (for examples see Refs. Reference Golgovici, Cojocaru, Anicai and Visan28, Reference Jung and Myung79, and Reference Park, Xiao, Yoo, Rheem and Myung81), they are nearly universally inadequate for practical applications, as they lead to discontinuous and extremely rough films.

Several different approaches to improve the morphology have been reported in the literature, with the three most prominent being the addition of surfactants (e.g., Refs. Reference Kuleshova, Koukharenko, Li, Frety, Nandhakumar, Tudor, Beeby and White68 and Reference Yoo, Song, Lim, Myung, Lee, Oh, Lee, Kim, Kim, Choa, Lee, Lee and Lim78), the application of suitable pulsed electroplating protocols (e.g., Refs. Reference Del Frari, Diliberto, Stein, Boulanger and Lecuire64 and Reference Richoux, Diliberto and Boulanger69) and finally, the application of controlled stirring conditions (e.g., Refs. Reference Schumacher, Reinsberg, Akinsinde, Zastrow, Heiderich, Toellner, Rampelberg, Detavernier, Broekaert, Nielsch and Bachmann70 and Reference Jung and Myung79). All of these approaches have led to significant improvements in the deposited films; however, there is no obvious agreement in the literature about the best method. The identification of a simple, reliable, and repeatable approach for improving the morphology of the deposited materials is, therefore, a challenge that does not seem to be fully solved yet.

3. Thickness

The typical reported deposition thickness for the Sb₂Te₃ and (Bi, Sb)₂Te₃ materials is around a few micrometers and, therefore, significantly lower than for Bi₂Te₃ compounds. These low thickness values are related to bad morphology, as thin layers are typically still interconnected and can, therefore, be characterized.^{Reference Li, Soliman, Zhou, Toprak, Muhammed, Platzek, Ziolkowski and Muller65} The highest thickness values of 20 µm were reported by Del Frari et al.,^{Reference Del Frari, Diliberto, Stein, Boulanger and Lecuire64} however, with a very large associated roughness of ∼12 µm. Structured Sb₂Te₃ thermolegs with a thickness of 20 µm and good morphology were reported by Snyder et al.^{Reference Snyder, Lim, Huang and Fleurial82} and Kim and Oh.^{Reference Kim and Oh83}

4. Crystallinity

The crystallinity of the deposited materials is analyzed in the majority of the available publications, with varying and not consistent results. Most publications report the existence of a crystalline phase of the deposited material, usually with a random orientation. This is in strong contrast to the results obtained for the deposition of Bi₂Te₃, which typically shows a strong preferred (110) orientation. Several reports indicate partly or even fully amorphous materials (e.g., Refs. Reference Ma, Ahlberg, Sun, Iversen and Palmqvist42 and Reference Lim, Kim and Oh66). The crystallinity can be significantly improved by applying a pulsed deposition protocol, as described by Schumacher et al.^{Reference Schumacher, Reinsberg, Akinsinde, Zastrow, Heiderich, Toellner, Rampelberg, Detavernier, Broekaert, Nielsch and Bachmann70} These results indicate that the growth mechanism and the applied deposition conditions have a strong influence on the crystallinity of the material.

It seems likely that the crystallinity and morphology of the deposited materials are to a certain extent interrelated and also need to be optimized in conjunction. However, no ideal recipe is readily available in the literature, and the growth mechanisms do not seem to be fully understood yet.

1. Seebeck coefficient

Regardless of the exact composition, all reported Seebeck values for both the binary Sb₂Te₃ and the ternary (Bi, Sb)₂Te₃ compounds are positive, indicating the expected p-type conductivity for these materials. The range of reported values for the Seebeck coefficient is, however, very large with values between 40 and 330 µV/K. This large spread is probably related to the different crystallinities of the deposited materials, as discussed in the previous section. For fully amorphous films, or films with a higher content of amorphous material, the Seebeck coefficient would be expected to be higher, while fully crystallized materials should typically exhibit lower values (e.g., Refs. Reference Lim, Kim and Oh66 and Reference Yoo, Song, Lim, Myung, Lee, Oh, Lee, Kim, Kim, Choa, Lee, Lee and Lim78). Other effects, like the precise composition, the deposition conditions and the different morphologies will, however, also influence the measured values, and therefore, general trends are difficult to identify (e.g., Refs. Reference Li and Wang67 and Reference Schumacher, Reinsberg, Akinsinde, Zastrow, Heiderich, Toellner, Rampelberg, Detavernier, Broekaert, Nielsch and Bachmann70).

2. Carrier concentration

The determination of the carrier concentration, as well as the electrical resistivity, is complicated by the common lack of a homogeneous, interconnected layer. The few available articles, which report on the carrier concentration for the p-type materials, consistently place the values in the range of 10²¹ to 10²²/cm³ (e.g., Refs. Reference Lim, Kim and Oh66 and Reference Kuleshova, Koukharenko, Li, Frety, Nandhakumar, Tudor, Beeby and White68). These values are consistent with the theoretical expectations for highly disordered and partly amorphous materials. Similar to the Bi₂Te₃ films discussed previously, the high-carrier concentration in Sb₂Te₃ and Bi_0.5Sb_1.5Te₃ films is frequently coupled with low carrier mobility, indicated by the high values for the electrical resistivity.

3. Electrical resistivity

The values reported for the electrical resistivity range from 10 to 1000 µΩm, with one extreme value of 25,700 µΩm.^{Reference Ma, Ahlberg, Sun, Iversen and Palmqvist42} The most commonly encountered range of values lies between ∼50 to 250 µΩm. This large spread is consistent with the high variance of the other electrical properties and is again probably related to the difference in crystallinity and the different morphologies resulting from the various fabrication conditions. The carrier mobility is correspondingly low, with reported values on the order of 10–50 cm²/Vs for both Sb₂Te₃ and Bi_0.5Sb_1.5Te₃ films.^{Reference Kuleshova, Koukharenko, Li, Frety, Nandhakumar, Tudor, Beeby and White68,Reference Lim, Park, Lim, Yoo, Lee, Myung and Lee84} For many materials with high resistivity values, it can be expected that the mobility is even significantly lower. This low mobility and the correspondingly high electrical resistivity in these compounds is, in addition to the morphology, the second main challenge for their application in TE applications.

4. Annealing

Annealing of the deposited materials offers a possible solution for the low mobility of the materials, as it would be expected to increase the crystallinity and, therefore, the carrier mobility. As a consequence, the electrical resistivity should also improve, even though this could potentially be partly countered by a corresponding decrease in carrier concentration through the decreased defect density.

The effect of an annealing treatment on the crystallinity of Sb₂Te₃ was investigated by Huang et al.,^{Reference Huang, Kellock and Raoux80} and it was found that an annealing temperature of 120 °C is already sufficient for the crystallization of amorphous starting material. Park et al.^{Reference Park, Xiao, Yoo, Rheem and Myung81} also observed a transition from amorphous to crystalline material for (Bi, Sb)₂Te₃ films, although they used an annealing temperature of 250 °C. They additionally reported a corresponding increase in the mobility of their films. Li and Wang^{Reference Li and Wang74} observed an improvement in both the Seebeck coefficient as well as the electrical resistivity after a temperature treatment at 250 °C for 4 h in inert N₂ atmosphere.

An increase in the average grain size, identified by XRD measurements, and an improvement in the electrical resistivity were reported by Lim et al.,^{Reference Lim, Park, Lim, Myung, Lee, Jeong, Yoo and Lee85} and similar results were also reported by Ma et al.^{Reference Ma, Ahlberg, Sun, Iversen and Palmqvist42} and Schumacher et al.^{Reference Schumacher, Reinsberg, Akinsinde, Zastrow, Heiderich, Toellner, Rampelberg, Detavernier, Broekaert, Nielsch and Bachmann70}

An annealing treatment is, therefore, beneficent for the electrical properties of the electroplated Sb₂Te₃ and (Bi, Sb)₂Te₃ materials. Overall, the improvements seem most pronounced in the electrical resistivity and carrier mobility, while the change in the Seebeck coefficient of the materials is less distinct. However, similar to Bi₂Te₃, the evaporation of tellurium from the compound was reported to slightly change the composition of the compound,^{Reference Ma, Ahlberg, Sun, Iversen and Palmqvist42} indicating that a controlled annealing might be required.

This approach was experimentally investigated by Schumacher et al.^{Reference Schumacher, Reinsberg, Rostek, Akinsinde, Baessler, Zastrow, Rampelberg, Woias, Detavernier, Broekaert, Bachmann and Nielsch36} by annealing (Bi_xSb_1−x)₂Te₃ films in a saturated Tellurium atmosphere. The results demonstrated that this equilibrium annealing not only prevents the evaporation of Te but can also be used to control the composition of the material. The TE properties of the materials were significantly improved with final Seebeck coefficients on the order of ∼180 µV/K carrier concentrations around 8∣10¹⁹/cm³ and carrier mobility of ∼30 cm²/Vs.

1. Type C: Cross-plane devices (vertical/vertical)

In the category of vertical/vertical devices, several different fabrication processes for the fabrication of micro-TE generators, with different fundamental principles, have been established in the literature. There are three basic different approaches for creating the necessary structures.

One is the use of a single substrate and a positive type photoresist, which can be structured several times and can finally be removed. This allows for the successive deposition of the materials and an eventual restructuring of the lower contact structure. The second approach is the use of a permanent photoresist, which doubles as a substrate, and allows the device to be accessed and structured from both sides. This allows for a selective deposition by using structured seed layers, and a full restructuring of this seed layer into the lower contact structures. The third one is the inclusion of a flip-chip bonding step, which allows for a separate structuring of the n- and p-type materials and contact structures.

Each of these approaches has its advantages and disadvantages, and can potentially be used for different applications. Key differences can be found in the choice of substrate materials, process complexity, choice of applicable photoresists, and number of lithography steps. None of the demonstrated published processes offers the “ideal” combination of high-thickness thermocouples, simple fabrication, and a fully removable photoresist template.

The first study carried out to confirm the potential of TE compound electroplating for the entire fabrication of microcoolers was presented in 1998.^{Reference Fleurial, Borshchevsky, Ryan, Phillips, Snyder, Caillat, Kolawa, Herman, Mueller and Nicolet6} The process developed at the Jet Propulsion Laboratory (JPL) at the California Institute of Technology is based on the successive deposition of the n- and p-type materials into the same photoresist template on a single wafer.^{Reference Lim, Snyder, Huang, Hennan, Ryan and Fleurial43,Reference Snyder, Lim, Huang and Fleurial82,Reference Fleurial, Snyder, Herman, Smart, ShakkottaI, Giauque and Nicolet104,Reference Borshchevsky, Caillat, Fleurial, Kolawa, Olson, Philips and Ryan105} The fundamental concept is based on the use of a positive type photoresist, which allows for multiple consecutive structurization steps, as long as all processing is conducted under yellow light conditions. The process was intentionally developed with the goal of electrochemical deposition for the TE and contact materials, and therefore, takes full advantage of the possibility of template-based growth.

The schematic process flow, as published by Snyder et al.,^{Reference Snyder, Lim, Huang and Fleurial82} is shown in Fig. 5. First, the lower contact material, a 3-µm thick gold layer, is electroplated into a conventional photoresist template, using a thin chrome layer as the seed layer [Fig. 5(a)]. A thick positive type photoresist, with a thickness of up to 40 µm, is then spin-coated on the substrate, and the template for the deposition of the first set of TE legs is lithographically opened [Fig. 5(b)]. The first TE material is then deposited into the template [Fig. 5(c)]. Next, a thin layer of photoresist is used to cover the already deposited material, and then the holes for the second TE material are opened. The second set of thermolegs can then be deposited into the template, while the first set of legs is protected from the electrolyte by the thin second photoresist [Fig. 5(d)]. The second thin photoresist layer is then removed by a short flood exposure and development, and a thin layer of gold is sputtered onto the structure to serve as seed layer for the upper contacts. A new layer of photoresist is then deposited and structured, and serves as the template for the deposition of the upper contact, which consists of 3 µm of low stress nickel [Fig. 5(e)]. The unnecessary layers (upper photoresist, thin-sputtered gold layer, thick photoresist, and chrome seed layer) are then removed by a suitable combination of solvents and etching steps, creating the final freestanding device [Fig. 5(f)].

FIG. 5.

Fabrication process flow schematic for the generator fabrication process applied by Snyder et al. (figure taken from Ref. Reference Snyder, Lim, Huang and Fleurial82 reproduced by permission of Nature Publishing Group).

This process offers several key advantages, e.g., that the fabrication is fully conducted on a single substrate, uses mainly conventional microfabrication techniques, and finally leads to a fully freestanding TE device without any thermal shunts. However, there are also some challenges, mainly the necessity of positive type photoresists, which offer only limited chemical resistance. Additionally, the overall process is fairly complex, as it involves several, sometimes unconventional, photolithography steps. The fabricated demonstration device contained 126 TE elements (n-Bi₂Te₃ and p-Sb₂Te₃, with a diameter of 60 µm and a height of 20 µm, and overall dimensions of ∼2 × 2 mm. The combined Seebeck coefficient of a thermocouple was ∼160 µV/K.

Huang et al.^{Reference Huang, Li, Chen, Zeng and She98} have followed a similar process integration by using electroplating and MEMS technology with four positive masks and photoresists. The column-type micro-TE cooler is constituted by 50 pairs of Bi₂Te₃ and Sb₂Te₃ pillars, with a diameter of 30 µm and a thickness of only 5 µm. The principal differences in comparison with the former JPL process relate to the rotation of the cathode during the electrodeposition and the galvanostatic process with an accurate current controller, both to enhance the density and the uniformity of TE deposits. The infrared thermal image measuring system showed an average cooling of about 0.7 K for the integrated µ-TEC under 5 volts driven. A variation of this process was proposed by the same group in 2008.^{Reference Huang, Lin, She, Li, Chen and Kuo99} A μ-TEC is fabricated with clustered multipillar structures (20 µm wide—8 µm high) in a 6 × 6 mm chip area. Each leg involves 100 parallel Bi₂Te₃ micropillars with 100 parallel Sb₂Te₃ micropillars [Figs. 6(a) and 6(b)]. A temperature difference of 1.2 K was measured, a higher value than the temperature difference of the µ-TEC with legs made of single pillars.

FIG. 6.

(a) Schematic design of the column-type Bi₂Te₃/Sb₂Te₃ µ-TEC with clustered multipillar structure and (b) fabrication process flow schematic for the generator fabrication process applied by Huang et al. (Figures taken from Ref. Reference Huang, Lin, She, Li, Chen and Kuo99 reproduced by permission of Elsevier).

A group in Korea^{Reference Jung and Ko26} fabricated a thermoelectric generator (TEG) based on n- and p-type bismuth telluride legs according the JPL process. The synthesis by pulse plating of n-type is performed in an aqueous medium while the p-type compound is obtained using a solution with 30% v/v Ethylene Glycol. TEGs made of 5 thermocouples (150 µm in height and 300 µm in diameter) were structured in a SU-8 photoresist on a silicon substrate (covered by SiO₂ 300 nm/Ti 10 nm/Au 300 nm). The thermocouple has a Seebeck voltage of 141 µV/K, and the generator produced an output power of 24.36 nW/TC at temperature difference of 10 °C. These moderate performances are caused by a high internal electrical resistance (20 Ω per thermocouple).

More recently, Uda et al.^{Reference Uda, Seki, Saito, Sonobe, Hsieh, Takahashi, Terasaki and Homma106} proposed the fabrication of a microdevice (8 arrays of 110 TCs, 20 µm in height, 50 × 50 µm²) through a patterning process, schematically described without any experimental details about the patterning. Both legs of the TCs are made of BiTe compounds. If the composition of the n-legs is Bi₂Te₃ type, the composition of p-legs is far from a V₂VI₃ compound (Bi 62.4 and Te 37.6 at.%). The work focuses on the size of the upper electrodes connecting the TE legs. The authors claimed that a benefit for enhancing the performances of the device is observed with an umbrella-type configuration, the term meaning that the upper electrode size is larger than the cross section of legs. The first result evidenced that the open circuit voltage is twice as large and the generated power (0.96 µW) 4.5 times larger than that of the device with a smaller electrode. Nevertheless, the performance remains weak, certainly due to the high resistance (total resistance of 79 Ω) and to an incorrect stoichiometry for TE compounds. This preliminary work needs to be confirmed with optimized electroplated compounds.

The fabrication process developed at the ETH Zürich, later greenTEG as a spin-off of the ETH Zürich,^{Reference Glatz, Schwyter, Durrer and Hierold44,Reference Glatz, Muntwyler and Hierold107,Reference Schwyter, Glatz, Durrer, Hierold, Bright, Bourouina, Courtois, Desmulliez, Karam and Wang108} is based on utilizing a chemically very resistant and mechanically stable photoresist, SU-8, as both deposition template and “substrate” for the final device. This photoresist layer is processed on a handle wafer with a sacrificial adhesion layer but is removed from the wafer during the later processing steps. The final device is embedded in the SU-8 layer and even offers a certain degree of mechanical flexibility.

The process flow is schematically depicted in Fig. 7 and was initially validated by electrochemical deposition of Cu and Ni.^{Reference Glatz, Muntwyler and Hierold107} First, the plating electrodes are deposited and structured on top of the sacrificial layer [Fig. 7(a)]. These electrodes basically serve as the seed layer but are structured in a way, which allows the selective deposition of the n- and p-type TE materials. The next step is the application and structuring of the 190 µm thick SU-8 mold, which contains the holes for both types of TE legs [Fig. 7(b)]. The p- and n-type TE materials are then successively deposited into the respective cavities, and any overgrowth is removed by a polishing step [Figs. 7(c) and 7(d)]. The 20-µm thick electrical copper contacts between the thermolegs are then electroplated and structured via chemical etching [Fig. 7(e)]. To access the plating electrodes on the other side of the device, the sacrificial layer is then dissolved to release the SU-8 layer [Fig. 7(f)]. As the plating electrodes are not in the desired configuration, they are removed and the electrical interconnects necessary for a device are plated and structured instead, creating the final device [Fig. 7(g)]. In the case of Bi₂Te₃-based devices, gold is used instead of copper as the interconnect material, due to its higher chemical resistance and lower contact resistances.^{Reference Glatz, Schwyter, Durrer and Hierold44} The demonstrated Bi₂Te₃-based devices^{Reference Glatz, Schwyter, Durrer and Hierold44} contained 99 thermocouples with leg diameters of 210 µm and a height of 140 µm. The overall dimensions of the generator were ∼10 × 10 mm. The reported combined Seebeck coefficient of one thermocouple was 91 µV/K.

FIG. 7.

Fabrication process flow schematic for the generator fabrication process applied by Glatz et al. (a) Patterning of seed layer electrodes; (b) structuring of SU-8 mold; (c) selective ECD of Cu-thermolegs; (d) ECD of Ni-thermolegs; (e) the bottom interconnects between the different thermolegs are etched into a 20-_m thick electroplated Cu-layer; (f) release of SU-8 layer by removing the sacrificial layer. Flipping the SU-8 disk gives access to the plating electrodes; (g) patterning of top interconnects and bond pads. (figure taken from Ref. Reference Glatz, Muntwyler and Hierold107 reproduced by permission of Elsevier).

The greenTEG process is unique in several ways and offers some advantages over other fabrication methods. Due to the possibility of accessing and especially restructuring the lower layers during the process, it is possible to use negative type photoresists and separately structured deposition electrodes for the two materials. This avoids several photolithographic steps when compared to the previously discussed JPL process. Due to the use of SU-8, very high structures are possible, and the final device retains a certain amount of flexibility. However, there are also some inherent disadvantages, with the main one being the unavoidable thermal shunt through the resist template. The absence of an electrically insulating and mechanically stable substrate might also be problematic in some applications, although the device can obviously be bonded onto a substrate if desired.

In 2012, ETH Zurich and greenTEG enhanced the electroplated microscale TE generators and proposed an optimization concerning the output power performance^{Reference Wojtas, Schwyter, Glatz, Kuhne, Escher and Hierold109} by integrating a microfluidic heat transfer system. This integration reduces the interfacial thermal losses. The heat transfer system consists of copper microchannels and a polymer PMMA manifold distribution layer. The heat transfer system assembled onto the TE generator evidenced an output power of 1.47 mW/cm² at an applied temperature difference of 50 K. This was confirmed in 2013^{Reference Wojtas, Ruthemann, Glatz and Hierold110} with a device exhibiting a net output power of 126.3 mW/cm² with a module ZT of 0.1 and an applied temperature difference of 95 K, with the thickness of TE legs being reduced to 60–90 µm.

The process developed by Kim and Oh. from Hongik University—Korea^{Reference Kim and Oh83,Reference Kim and Oh111} is mostly based on a similar concept to the one applied by JPL, however, the upper contact structure is fabricated on a separate substrate, and then flip-chip bonded to the rest of the device. The fabrication process is shown in Fig. 8. The n- and p-type materials are deposited into a positive type photoresist, onto a nonstructured seed layer, and with successively opened deposition cavities for the thermolegs [Figs. 8(a)–8(e)]. To achieve a precisely controlled height, the legs are then polished, and a nickel/tin capping layer is then deposited for the bonding process [Figs. 8(f)–8(h)]. The seed layer, which also serves as the lower contact material, is then etched to form the lower contact structure by a third patterning of the thick photoresist layer [Fig. 8(i)]. The photoresist is then dissolved, forming the bottom substrate device structure without the upper contact [Fig. 8(e)]. The upper contact on the top substrate is structured in a similar manner and then bonded to the bottom substrate by use of an anisotropic conductive adhesive, forming the final device [Figs. 8(k)–8(m)].

FIG. 8.

Fabrication process flow schematic for the generator fabrication process applied by Kim and Oh (a) sputtering of the Ti/Cu/Ti metallization, (b) photoresist (PR) patterning, (c) electrodeposition of the BiTe legs, (d) PR patterning, (e) electrodeposition of the SbTe legs, (f) planarization, (g) PR patterning, (h) electrodeposition of the Ni/Sn capping layer, (i) PR patterning, (j) bottom electrode patterning, (k) PR patterning and electrodeposition of Sn on a Top substrate, (l) top electrode patterning, (m) flip chip bonding (figure taken from Ref. Reference Kim and Oh111 reproduced by permission of the Japan Institute of Metals and Materials).

This process offers some advantages over the ones already described above. One is the direct implementation of an upper substrate, which is potentially advantageous for some application scenarios. The second is the use of a fully thickened lower contact layer as the seed layer, which possibly makes the deposition process more reliable as the electrical resistance is lowered. However, the process is fairly complex, as it involves numerous photolithography steps in addition to planarization and flip-chip bonding. The final device contained 242 thermocouples, with a leg diameter of 100 µm and a height of 20 µm. The overall dimensions were ∼5 × 5 mm². The Seebeck coefficient for a thermocouple was, based on the material properties, estimated to be ∼545 µV/K (−59 µV/K for n-Bi–Te and 485 µV/K for p-Sb–Te).

According to the same concept of a flip chip bonding, a µTEG was developed by Roth et al. in 2013.^{Reference Roth, Rostek, Lenk, Kratschmer, Cobry and Woias112} The generator is obtained from electroplated material (n-type Bi₂Te₃ and Cu) on a silicon substrate using a process based on two negative photoresists (SU-8 and Ordyl). The upper contacts of 71 TCs (300 µm wide and 135 µm high on a total area of 1 cm²) were formed by a flip-chip process of silicon chip with a solder paste Bi₅₇Sn₄₂Ag₁. One year later, Copper is replaced in the generator by Sb₂Te₃.^{Reference Roth, Rostek, Cobry, Kohler, Groh and Woias113} In this work, a thickness of 300 µm for electroplated antimony telluride legs is reported, which is larger than the previous one in the range of several microns. This large thickness with a good morphology and density of antimony telluride deposits was obtained by pulse plating (on-pulse for 20 ms at −32.2 mA/cm² and an off-pulse for 175 ms at 0 mA/cm²) and with two stirring modes during the plating. A first period up to 5 µm legs was conducted without stirring to ensure a good growth on the starting layer. After 5 µm, electroplating is carried out with stirring the electrolyte but without changing the current densities and the pulse durations. After the evaporation of 100 nm Ti and 100 nm Au as barrier/bonding layer on TE legs, a thermal annealing was performed for 100 h at 250 °C in tellurium atmosphere to reduce the electrical resistance by inducing a phase change from amorphous to crystalline state. The measured power output was equal to 166 µW at a temperature difference of 46 K (i.e., 0.08 µW/K²), a value nevertheless inferior to a TEG made with annealed Bi₂Te₃ and Cu.

2. Type B cross-plane devices (vertical/lateral)

An alternative approach of a vertical assembly of planar-electroplated TC modules was investigated by Wang et al.^{Reference Wang, Ji, Xu, Li, Visan and Golgovici114} to increase the integration density of thermocouples in a given volume. In this configuration (Fig. 9), the thermal gradient is applied at the extremities of the films. This work can be anchored in the concepts of Stordeur and Stark.^{Reference Stordeur and Stark88,Reference Stordeur117} The first step is the electrodeposition of n-type BiTeSe and p-type BiTeSb strips (3.6 mm long × 0.4 mm wide × 50 µm thick) and copper (as connecting layer) on a metallic substrate structured by photolithography. After electrodeposition, the chains of thermocouples are peeled off from the seed layer and bonded together with epoxy resin. The next steps are the successive in-series connections of two adjacent foils (by electroless plating of copper) at the end of the connecting layers by silver wires. To create the final device, the stack is placed between two hard thermal plates, contacting to heat sink and source. The bonding of thermal plates is performed by thermal conductive layers (the lithography process, the nature of hard plates, and thermal conductive layers are unspecified). The proposed multilayer generator consists of a stack of 8 layers, each of them with a number of 20 p-types legs and 21 n-type legs on it. The final result is a small compact generator (4 mm high × 25 mm wide × 1 mm thick) with 160 thermocouples. The characteristics of the presented device are its internal resistance (1.1 kΩ), its open-circuit voltage (660 mV), maximum output power, and corresponding power density under the temperature difference 20 K (77 µW–770 µW/cm³) at 25 °C. The principle interest of the work lies in the synthesis of high-quality TE legs due to a gyration of thin strips after horizontal plating. Nevertheless, the peeling of TE elements, which induces high-internal resistance, is a certain disadvantage.

FIG. 9.

Schematic fabrication process applied by Wang et al. (a) Views of the film-like TE component; (b) schematic view of the TE aggregation; and (c) schematic view of the partly encapsulated TE aggregation (captions and figures taken from Ref. Reference Wang, Ji, Xu, Li, Visan and Golgovici114 reproduced by permission of Elsevier).

3. Type A cross-plane devices (lateral/lateral)

Very few examples of horizontal assembly of electroplated Peltier modules can be founded in the literature. The concept was tested in 2001 by Qu et al.^{Reference Qu, Plotner and Fischer115} for an in-plane Bi–Sb generator, and more recently, for an in-plane BiTe–SbTe thermopile sensor (Fig. 10) by Kim and Oh,^{Reference Kim and Oh116} the latest piece of work that we would like to present in this section. After a metallization (sputtered Ti) onto a glass substrate [Fig. 10 (a)], p-SbTe and n-BiTe strips (2 and 5 µm thick) were potentiostatically plated [Fig. 10(b) and 10(c)] and patterned with photolithography. The titanium seed layer was etched in hydrofluoric acidic medium [Fig. 10 (d)]. The strips were embedded in a cured photoresist as an insulation layer [Fig. 10 (e)], and the device was finalized by the insertion of a heater element composed of trilayers (Ti/Cu/Ti) connected to the strips at the center of the sensor [Fig. 10 (f)]. Two devices were designed with either 196 pairs of legs of 50 µm-width or 130 pairs of 100 µm-width. The first one showed a sensitivity of 24.8 mV/K, while the second one reached 15.4 mV/K, the decrease is linked to the smaller number of pairs on the same area. Thicker strips (5 µm) enhanced the sensitivity up to 36.5 mV/K.

FIG. 10.

Fabrication process flow schematic for the generator fabrication process applied by Kim and Oh (a) sputtering of the Ti seed layer, (b) electrodeposition and patterning of the p-type Sb-Te legs, (c) electrodeposition and patterning of the n-type Bi-Te legs, (d) removal of the Ti seed layer, (e) formation of a photoresist insulation layer after exposing the sensor electrode pads, and (f) formation of a Ti/Cu/Ti thin film heater. (figure taken from Ref. Reference Kim and Oh116 reproduced by permission of Japan Institute of Metals and Materials).

Although out of the scope of the review, some preliminary attempts of TE nanowire-based device fabrication processes have been done. The works are based on the successful fabrication of arrays of nanowires through the template synthesis approach. The reader can find related information in the review paper by Martín-González et al.^{Reference Martín-González, Caballero-Calero and Díaz-Chao118} As example, the synthesis feasibility of n-type and p-type arrays of NWs has been demonstrated in 2002^{Reference Sander, Gronsky, Sands and Stacy119} and 2003,^{Reference Martin-Gonzalez, Prieto, Gronsky, Sands and Stacy120} respectively. On opposite to the works related, the film based devices, less work, have been performed on the development of nanodevices. It can be mentioned, nevertheless a device made of electrodeposited n-type Bi₂Te₃ and p-type BiSbTe nanowires in alumina,^{Reference Lim, Whitacre, Fleurial, Huang, Ryan and Myung121} the design of nanomicrogenerator composed of n-type and p-type Bi₂Te₃ nanowire arrays in alumina,^{Reference Wang, Jia, Huang and Zhang122} a hybrid nanowire BiSb in porous alumina-bulk Bi_0.4Sb_1.6Te₃ TE device,^{Reference Keyani, Stacy and Sharp123} a thermoelectrically cooled giant magnetoresistance sensor p-type BiSbTe or n-type Bi_xTe_y nanowires grown in polycarbonate or alumina templates.^{Reference Davis, Bellamkonda and Mannam124}