I. Introduction
The continued demand for reliable sources of energy can be addressed by the discovery of new sustainable energy sources as well as increasing the efficiency of power generation technology. Thermoelectric materials are one approach towards alternative energy.
Thermoelectric materials comprise a wide range of solid compounds distinguished by their ability to convert thermal and electrical energy. This property gives rise to two distinct technological applications: the development of temperature gradients for heating and cooling devices and the generation of electrical energy from heat.
The conversion efficiency of a thermoelectric material is related to the dimensionless “figure of merit”, ZT :
ZT = S2σT/κ (1)
where S is Seebeck coefficient, σ and κ are electrical and thermal conductivity, respectively, and T is absolute temperature.
The Seebeck coefficient, or thermopower, is a measurement of the amount of voltage generated per unit of temperature difference and is typically given in μV/K. Accordingly, obtaining a high figure of merit thermoelectric requires maintaining high electrical conductivity and large thermopower while simultaneously limiting thermal conductivity.
As shown in Fig.1, a typical thermoelectric module contains both n-type and p-type thermoelectric materials connected in series. n-type materials possess electron charge carriers and have negative Seebeck coefficients; conversely, p-type materials possess positive Seebeck coefficients and have hole charge carriers. Applying a temperature gradient across the module causes the carriers to diffuse towards the cold side, generating a thermoelectric voltage.
The value of ZT is dependent on the module operating temperature and has remained approximately one for the past several decades for “archetype materials” at all temperature ranges. These materials include antimony and bismuth tellurides for room temperature applications (300 K), lead telluride at moderate temperatures (~650 K), and silicon-germanium alloys at high temperatures (~1000 K). In Fig. 2 ZT-values for a number of thermoelectric materials are presented as a function of temperature.
Fig. 1: Schematic diagram of a typical thermoelectric module. Small legs of n-type (red) and p-type (blue) materials are connected (usually by Copper) in series and then sandwiched between ceramic substrates. In the case of electrical generation, heat is applied to one side of the module, causing the charge carriers to diffuse across the module and generating an electrical current.
Fig. 2: Figure of merits (ZT) for various thermoelectric materials at different temperatures.
In the present investigation we performed a reverse engineering study of the “soldered” interconnects between oxygen-free Copper (OFC) and n- and p-type components of the thermoelectric converter. The ultimate goal of this Project is to develop a cost-effective joining technology for fabricating ”(p-n) thermocouple” Cu / connectors in the “Fokker thermoelectric convertor”.
II. Experimental procedure
“As-received” samples were embedded in conductive resin and ground until the total sample “thickness” was reduced to approximately half of the original size. After subsequent polishing with diamond suspensions (down to 1 μm), interconnect cross-sections suitable for metallographic examination (Optical and Scanning Electron Microscopy (SEM) and Electron Probe Microanalysis (EPMA)) can be produced. It is important to keep in mind that somewhat excessive (“rough”) grinding may result in undesirable artifacts in the final microstructure.
III. Results and Discussion
- Thermoelectric materials used
Two sample cross-sections were investigated:
1) Sample MT 12-187 (Mat-Tech designation) is the n-type semiconducting material based on PbTe connected to Copper.
2) Sample MT 12-186 (Mat-Tech designation) is based on Copper and “basis” semiconducting material (p-type) containing according to EPMA, Ge, Te, Ag and Sb (Fig. 3). It follows from the information summarized in Fig. 2 that the thermoelectric used can be AgSbTe2-GeTe (TAGS).
Fig. 3: a) Back-scattered Electron Image (BEI) of the “basis” thermoelectric material in the sample MT 12-186; b) X-ray spectrum taken from the matrix (exhibiting “grey” contrast in the BEI) and c) corresponding results of the (standardless) quantification. (Note that intensities in the X-ray spectrum are presented in a “logarithmic scale” and there is a severe “overlapping” of the L-characteristic lines of Sb and Te).
It is also apparent from the microstructure presented in Fig. 3a that the “basis” thermoelectric material is not homogeneous. Some inclusions exhibiting “dark” contrast in BEI are clearly visible (Fig. 3a). These particles are rich in Germanium (Fig. 4a). It is interesting that these precipitates are surrounded by the “rim of the light grey phase” which contains mainly Te and Ag (Fig. 4b).
Fig. 4: X-ray spectra taken from a) inclusions exhibiting “dark” contrast in Fig. 3a and b) from the “light grey” rim surrounded “black” inclusions (indicated in Fig. 3a by arrow). Note that all intensities in the X-ray spectrum are presented in a “logarithmic scale”. The presence of the characteristic lines of Cu in the spectrum can be an artifact originated from the sample preparation (“smearing” from the Cu-containing plastic during polishing procedures).
Another telling example of inhomogeneity of the “basis” semiconducting material in the sample MT 12-186 sample is afforded by the micrographs taken in the vicinity of the “sample corner” (Fig. 5). It remarkable that although “black” inclusions in Fig. 5a are rich in Germanium
(Fig. 6a), which is similar to the case described before, the “dark grey” areas surrounded precipitates contain mainly Copper and Tellurium (Fig. 6b).
Fig. 5: a) BEI and b) optical (bright-field) image taken in the vicinity of the “sample corner” showing inhomogeneity of the “basis” semiconducting material in the sample MT 12-186.
Fig. 6: X-ray spectra taken from a) inclusions exhibiting “dark” contrast in Fig. 5a and b) from the “dark grey” area surrounded “black” inclusions. Note that all intensities in the X-ray spectrum are presented in a “logarithmic scale”.
As to the “homogeneity of the “basis” semiconducting material in the second, MT 12-187 sample, the electron micrograph (BEI) shown in Fig. 7 gives clear evidence that the PbTe-matrix contains some inclusions of virtually pure Lead. It seems that these precipitates are situated along boundaries of crystallites in the “sintered PbTe”. It is possible that during sintering procedure some of the starting Lead Telluride can be decomposed.
Fig. 7: BEI image taken from the “basis” semiconducting material in the sample MT 12-187 revealing Pb-inclusions in the PbTe-matrix.
If this is the case, then Tellurium being rather volatile will evaporate “leaving” liquid Lead behind. In systems exhibiting isotropy of solid-liquid interfacial energy, any region of liquid phase will aggregate at grain boundary junctions (or in the case of sintering, at the contact surfaces of particles within the powder compact) and take up a uniform shape controlled by a unique dihedral angle.
One sees from the microstructure shown in Fig. 7 that in our case of Lead Telluride sintering, dihedral angle has a value between 0° and 60°. This implies that most, if not all the grain (particles) edges and corner were replaced by continuous channels of liquid Lead.
To conclude this part of the demonstration two things should be mentioned.
1. Both “basis” thermoelectric materials used in the samples to be examined are not completely homogeneous. Perhaps, this is an outcome of the not well-controlled sintering procedure.
2. The combination of “basis” thermoelectric materials found in the samples indicates that operating temperature of the thermoelectric module is somewhere around 600 K (see Fig. 2).
- Microstructure of the transition zone in the thermoelectric /Cu interconnects
Let’s start with the interfacial region of the sample MT 12-187. In fact, nothing special (more than might have been expected from the available information provided by the “customer”) was found. This can be appreciated just by looking through the micrographs shown in Fig. 8. The BE images were taken somewhere in the central part of the sample (joint).
Fig. 8 BEI’s of the transition zone between Cu and thermoelectric material in the sample MT 12-187. The “irregular” Mo-layer is clear visible between Cu-connector and Lead, likewise Ni,Sb-based interlayer is present at the PbTe/Pb interface. (Note that some SiC-particles (“grinding debris”) are embedded in the soft Pb-layer)
Despite severe overlapping between characteristic L-lines of Molybdenum and M-lines of Lead in X-ray spectra, it is still possible to identify Mo-interlayer (however, much thinner than 3 μm !) at Cu/Pb contact surface (Fig. 8). Sometimes this interlayer is rather irregular, although at this moment, it is not clear whether this is merely an artifact resulted from the sample preparation or a genuine microstructural feature.
As to the Ni,Sb-based interlayer detected at the PbTe/Pb-interface, it is for me unclear (at least, at this moment) why this type of layer is introduced between the PbTe-semiconducting material and Pb (“solder”)! What for do we need Ni and Sb as constituents of the layer? What is the point? Is this merely a certain sort diffusion (reaction) barrier between PbTe and Pb? If this layer must be there, then how was it produced?
On the contrary, a transition zone in the sample MT 12-186 is much more complicated. A typical morphology of the central part of this sample is presented in Fig. 9. Again, remnants of the very thin (< 3 μm) layer of Mo is visible between Cu and “filler metal”. In some areas of the transition zone where Mo-interlayer was disintegrated, the formation of Cu-Sn intermetallic compounds (IMC) was detected (Figs. 9a and 10e).
Fig. 9: BEI’s of the central part of the transition zone between Cu and thermoelectric material in the sample MT 12-186. The often intermittent Mo-layer is clear visible at Cu-connector contact surface (Note that some SiC-particles (“grinding debris”) are embedded in the soft Pb-based layer)
Fig. 10: X-ray spectra taken from various locations of the transition zone between Cu and thermoelectric material in the sample MT 12-186 together the results of the (standardless) quantification. a), b), c) and d) correspond to the microprobe measurements in “Points” 1, 2, 3 and 4 in Fig. 9 b. Results presented in e) correspond to the measurements in the area in Fig. 9a indicated by “white” arrow. (Note that intensities in all X-ray spectra are presented in a “logarithmic scale”).
The question now is: “Where is tin coming from?” One sees from Fig. 9b and 10a that next to the basis thermoelectric material there is a SnTe-containing layer of about 3-5 μm thick. It can be noticed that grains of SnTe within this layer exhibit columnar morphology. A similar question arises: “How was this layer produced?”
One can also see that significant amount of PbTe-based material (“light grey islands” in the BEI given in Fig. 9b) are presented inside the Pb-based “connecting seam” (see Figs. 10b, c, d). In same domains of this interconnect these islands are in contact with the columnar grains of the SnTe-based layer. Close inspection of the transition zone revealed a two-phase morphology of this layer. It looks that in some areas of the interfacial region grains of both constituents (SnTe-based as well as PbTe-based phase) are routed to the Ge(Ag,Sb)Te-based (“basis”) thermoelectric material.
One more interesting feature is to be noted here. The original contact surface between “basis” thermoelectric material and SnTe-containing layer is revealed by the row of pores as indicated by “black” arrow in Fig. 9a.
Clearly, the observed morphology is a result of chemical reactions occurring within the layered structure of the initial connector assembly during joining (presumable soldering at temperature somewhat higher than melting pint of Pb, viz. 327 °C). It is important to realize that thermodynamic stability of PbTe is somewhat higher than that of SnTe. For example, at 600 K (near the melting point of Pb, let say, “soldering temperature”) the values of Gibbs free energy of formation (ΔfG0) for these monotellurides are -98 and -93 kJ/mole for PbTe and SnTe, respectively. This implies that from thermodynamic stand point it is possible that solid-state displacement reaction of type Pb(l,s) + SnTe(s) = Sn(l,s) + PbTe(s) can occur upon joining, and perhaps, during subsequent service. It is now becoming clear why in some areas of the transition zone, where Mo-interlayer was has discontinuity, the formation of Cu-Sn intermetallics was detected.
It seems not profitable in this MEMO to go further with the thermodynamic analysis of the possible interactions as well as with the any attempt to rationalize a topological arrangement of the product phases inside the reaction zone. Suffice to say that there is every reason to believe that “customer” does not aware of such complicated structure of the transition zone. The observed reaction morphology can be summarized as follow:
Cu|Mo or Cu-Sn IMC|Pb(Sn)+PbTe|PbTe+Pb|PbTe+SnTe|Ge(Ag,Sb)Te-based material
Perhaps, the most unexpected findings here are the occurrence of the chemical reaction between Pb and SnTe, rather poor quality of the PVD Mo-interlayer and, as a consequence, the formation of Cu-Sn intermetallics in the areas of the Cu-connector exposed to the filler metal. The latter may affect structural and electrical integrity of the joints and hence the overall performance of the thermoelectric device.
IV. Final Thoughts
Looking at the overall picture a number of things are coming in mind:
I have a strong feeling that we need to understand more about “technological steps” involved in the preparation of the “thermoelectric parts” of the device. The questions like “Why do we need Ni,Sb layer on the PbTe?” and “Why SnTe should be present at the top of the Ge(Ag,Sb)Te-based material?” should be clarified.
Perhaps, one academic question: “What would be (in general) the effect of inhomogeneities on performance of thermoelectric materials?”
It seems to me that in our particular case Lead can be “a material of choice” for soldering. However, quality of the Mo diffusion (reaction) barriers is very critical. Therefore, it is vital to understand why we need layer of SnTe on the top of the Ge(Ag,Sb)Te-based material. The presence of Sn might cause a problem in a sense of the reaction with Cu.
