Introduction

Recent advancements in thermoelectric materials research have created new opportunities for thermoelectric power generation and cooling. These advances may enable higher efficiency terrestrial waste heat recovery [1–3] and space thermoelectric power generation technologies [4, 5]. However, with the exception of radioisotope thermoelectric generators (RTG), use of terrestrial thermoelectric power generation technology has been limited. Radioisotope thermoelectric generators use a radioisotope heat source to generate >500 °C thermal gradients that are sustained for multiple decades [1–3]. Additionally, RTGs operate in vacuum, thus oxidation of the high temperature components does not occur. In terrestrial waste heat recovery, particularly when coupled to internal combustion engines, exposure to elevated temperatures (up to 600 °C) in the presence of air will likely cause oxidation [6–9]. Similarly, the degree of oxidation will depend on the TE materials’ compositions.

Skutterudites (SKDs) are promising thermoelectric materials, because they have a relatively high ZT of >1.0 [1–6] and are amenable to machining and metallization [7]. For example, Shi et al. [6] have shown that multiple filled n-type SKD has a ZT of 1.7 at 580 °C. Similarly, Tang et al. [10] have demonstrated that InSb-doped n-type SKD has a ZT of 1.4 near the same temperature. Additionally, Rogl et al. [11] recently reported a ZT > 1.0 for Misch-metal filled p-type SKD at temperatures above 500 °C. However, while SKD-based materials show promise, they are susceptible to oxidation or sublimation, leading to performance degradation during operation. A schematic diagram (Fig. 1) shows how degradation can occur in an n/p power generation couple. A typical couple consists of n- and p-type legs connected on the hot side with an interconnect (Fig. 1a). Heating a couple, consisting of air sensitive n- and p-type legs, in air results in oxidation (Fig. 1b). It has been shown that the rapid degradation of thermoelectric modules consisting of n-type YbxCo4Sb12 and p-type CexCoFe3Sb12 was due to oxidation from testing in air [12]. While heating a couple in an inert atmosphere can prevent oxidation, sublimation of species such as Sb from SKD [3, 7] also causes couple performance degradation (Fig. 1c). A protective coating could prevent both oxidation and sublimation (Fig. 1d). Promising coating materials include enamel [13], composite glass [14] and aerogel [8, 15]. It has been reported that an enamel coating on n-type CoSb3 is effective in suppressing oxidation up to 600 °C [13]. Similarly, a composite glass coating on n-type Yb0.3Co4Sb12 at 550 and 650 °C has also been demonstrated [14]. However, there have been no reports demonstrating oxidation suppression of p-type SKD. We believe that oxidation of p-type is considerably more challenging compared to n-type SKD owing to the higher weight fraction of oxygen-sensitive-elements such as Ce, Fe, and La. It follows that the purpose of this work was to demonstrate the efficacy of enamel-based oxidation suppression coatings with a specific focus on characterizing oxidation suppression in p-type SKD. The enamel was selected based on its glass transition temperature and flow-ability to make an airtight seal around SKD elements. Additionally, the enamel has a relatively low thermal conductivity to minimize parasitic heat loss during operation.

Fig. 1
figure 1

A schematic diagram showing how degradation can occur in a typical n- and p-power generation couple

Full size image

We investigated the oxidation suppression behavior and reactivity between the enamel and SKD for both n- and p-type at 500 and 600 °C and subsequently thermally cycled between 600 °C and room temperature. Scanning electron microscopy (SEM), energy dispersive spectroscopy (EDS) and optical microscopy (OM) were used to characterize changes of microstructures or composition before and after heating in air and to compare uncoated and enamel-coated SKD samples. These data along with the enamel processing conditions will be discussed.

Materials and methods

The n-type Ba0.05Yb0.025CoSb3 and p-type Ce0.9Co0.5Fe3.5Sb12 were prepared by a conventional melt, quench, and anneal process [8]. All pure elements were purchased from Alfa Aesar: Sb ingot (99.9 %), Co powder (99.9 %), Fe powder (99.999 %), Ba pieces (99.9 %), Ce pieces (99.9 %), Yb pieces (99.9 %). 100–300 g ingots were pulverized with an automated mortar and pestle, followed by planetary ball milling to further refine the particle size. The powder was loaded into 50 mm diameter graphite dies and hot pressed at 650 °C and 40 MPa pressure under flowing argon. After hot pressing, the pellets were diced into 3.5 × 7.0 × 6.0 mm parallelepipeds using a diamond saw.

Using a tube furnace (Lindberg Blue M), n-type (Ba0.05Yb0.025CoSb3) and p-type (Ce0.9Co0.5Fe3.5Sb12) SKD samples were heated to 100, 200, 300, 400, and 500 °C in air at a rate of 100 °C/hour. The samples were maintained at each temperature for 1 h, and cooled to room temperature at a rate of 100 °C/hour. An alumina tube with 1.9 cm inner diameter and 60 cm long (CoorsTek Inc.) was utilized to obtain a uniform temperature distribution of the sample in the tube furnace. A silica-based enamel was used in powder form primarily consisting of silicon and sodium oxide (66at.% oxygen, 11at.% silicon, 11at.% sodium, 6at.% fluorine, 2at.% aluminum, 2at.% nitrogen, 1at.% barium, and 1at.% calcium). The enamel powder size ranged from 0.1 to 1.0 μm. To characterize the melting and flowing behavior, enamel powder was melted on copper samples. Cu samples were placed in a crucible, covered with enamel powder and heated to 650 and 700 °C for 1 h in argon gas.

To coat SKD elements with the enamel, the SKD samples were placed in a graphite crucible and then covered in enamel powder. The sample, enamel powder, and crucible were placed in a furnace (MTI KSL-1100X-S), which was housed in an argon filled glovebox, and then heated to 700 °C for 1 h. The oxygen content in the glovebox was maintained at less than 0.3 ppm. To reduce stress and its effects due to orthogonal edges, SKD samples were chamfered manually using 240 grit sanding paper. The approximate chamfer radius was 0.5 mm. After enamel coating, the samples were heated in air at 500 °C for 1 h. Extended (8 days) isothermal air-stability tests were conducted at 500 and 600 °C, respectively. Thermal cycling tests were also conducted and consisting of 20 cycles, where one cycle entailed heating from room temperature to 600 °C at a rate of 100 °C/hour, holding at 600 °C for 1 h, and then cooling to room temperature at a rate of 100 °C/hour (Fig. 2).

Fig. 2
figure 2

Thermal cycling time–temperature profile

Full size image

Microstructural analysis was conducted using OM (LEICA EZ4D) then SEM (ZEISS EVO LS25). The atomic percent composition was measured using EDS analysis. For EDS analysis, standardless analysis using the library standards provided by the EDS software vendor was used. The software was TEAM (Texture and Elemental Analytical Microscopy) version 3.2. To identify the antimony oxide phase, X-ray diffraction (XRD) measurements were conducted using a Bruker D8 DaVinci diffractometer equipped with Cu-Kα X-ray radiation operating at 40 kV and 40 mA. Laser flash was used to measure the enamel thermal conductivity according to the relationship:

$$ \kappa = D \cdot \rho \cdot C_{p} , $$
(1)

where κ is thermal conductivity, D is the thermal diffusivity, ρ is the density, and C p is the isobaric heat capacity. Thermal diffusivity was measured using a Netzsch (LFA447, laser flash system) up to 250 °C. To obtain the room temperature density, the geometric volume and mass of the test specimen were measured. The heat capacity was measured using a DSC (TA Instruments, Q2000) with a temperature interval of 50 °C. The linear coefficient of thermal expansion (CTE) was measured using a thermal mechanical analyzer (TA Instruments, Q400).

Results and discussion

n-SKD stability versus temperature analysis

The n-type SKD surface topography and composition were analyzed as a function of heating temperature (Figs. 3 and 4). The n-type SKD surface appeared relatively featureless for the control sample, which was not heated (henceforth referred to as RT), and samples heated at 100, 200, and 300 °C (Fig. 3a–d). Above 300 °C, distinct topographic feature changes were apparent (Fig. 3e–f). At 400 °C, 1–2 micron features were observed, likely consisting of antimony oxide as described in more detail below (Fig. 3e). At 500 °C, the topographic features further increased in size to 3–5 microns, compared to the sample heated to 400 °C (Fig. 3f). Significant compositional changes of oxygen and antimony were also observed above 300 °C (Fig. 4). The oxygen and antimony concentration of the RT, 100, 200, and 300 °C were similar in that relatively little change in the surface composition was observed. Above 300 °C, EDS analysis indicated the atomic percentage of oxygen increased and that of antimony and cobalt abruptly decreased, which is consistent with the formation of antimony oxide [14]. In cross-sectional SEM analysis (Fig. 3g–h), 0.1 and 1.8 μm thick oxide layers were observed in the 400 and 500 °C samples.

Fig. 3
figure 3

SEM analysis of n-type Ba0.05Yb0.025CoSb3 after 1 h in air. a RT, top-view, b 100 °C, top-view, c 200 °C, top-view, d 300 °C, top-view, e 400 °C, top-view, f 500 °C, top-view, g 400 °C, vertical-view, and h 500 °C, vertical-view

Full size image
Fig. 4
figure 4

The element composition change of n-type Ba0.05Yb0.025CoSb3 after 1 h in air heated at up to 500 °C

Full size image

p-SKD stability versus temperature analysis

The p-type SKD surface topography and composition were also analyzed as a function of heating temperature (Figs. 5 and 6). Compared to the n-type, the p-type SKD microstructure was not as homogenous before heating as indicated by the presence of higher levels of porosity (Fig. 5a). Upon heating, the p-type SKD exhibited a greater degree of oxidation compared to the n-type SKD (Fig. 5b–h). The EDS analysis indicates an abrupt increase in oxygen above 200 °C (Fig. 6); similar to what was observed in the n-type SKD, but at a lower temperature. The simultaneous decrease in antimony and iron above 200 °C also indicates that the oxide layer may consist of antimony oxide, which is consistent with the observations made by Qiu et al. [16]. Additionally, severe discoloration and cracking occurred at and above 400 °C (Fig. 5e–f). Optical microscopy was used to characterize the surface topography for the samples heated to 400 °C that are shown in Fig. 5e and f, because the features were too large for analysis using SEM. Relatively large (100 μm) cracks occurred in the 400 and 500 °C p-type SKD samples. Compared with the n-type, the p-type samples were more air sensitive, as surmised by the severely cracked and deformed surfaces above 400 and 500 °C. In cross-sectional analysis, 0.1, 0.4, and 0.8 μm thick surface oxide coatings were observed after heating to 100, 200, and 300 °C (Table 1). Thus, it is apparent that the oxide layer grew faster in the p-type compared to the n-type SKD for the same given temperature. The oxidation of SKD is governed by an oxygen-diffusion-limited mechanism [17–19]. Additionally, our results are in good agreement with related work by Chen et al. [20, 21], where the oxidation of p-SKD (CeFe4Sb12) was observed to readily oxidize above 200 °C and the n-SKD (YbyCo4Sb12) oxidized above 300 °C.

Fig. 5
figure 5

SEM analysis of p-type Ce0.9Co0.5Fe3.5Sb12 after 1 h in air: a RT, top-view, b 100 °C, top-view, c 200 °C, top-view, d 300 °C, top-view, e 400 °C, top-view, f 500 °C, top-view, g 100 °C, vertical-view, and h 300 °C, vertical-view

Full size image
Fig. 6
figure 6

The element composition changes of p-type Ce0.9Co0.5Fe3.5Sb12 after 1 h in air heated at up to 500 °C

Full size image
Table 1 Oxidation behavior of n- and p-SKD after heating for 1 h in air
Full size table

The Pilling–Bedworth relationship (R PB) determines the ratio between the volumes of an oxide coating grown on a metallic substrate from which the oxide is grown. The R PB can be used to characterize the oxidation behavior of the n- and p-type SKD in this work [22]. R PB is expressed as follow

$$ R_{\text{PB}} = \frac{{V_{antimony oxide} }}{{V_{SKD} }} = \frac{{M_{antimony oxide} \times \rho_{SKD} }}{{M_{SKD} \times \rho_{antimony oxide} \times n}}, $$
(2)

where V antimony oxide is the molar volume of antimony oxide, V SKD is the molar volume of SKD, M antimony oxide is the molecular mass of antimony oxide, M SKD is the molecular mass of SKD, ρ antimony oxide is the density of antimony oxide, ρ SKD is the density of SKD, and n is the number of SKD atoms per formula unit of antimony oxide, respectively.

To obtain the R PB, the antimony oxide species was identified by XRD analysis. Figure 7 shows that the n-SKD and p-SKD heated in air formed antimony trioxide (Sb2O3) and antimony pentoxide (Sb2O5), respectively. These observations are consistent with those made by Zhang et al, and Ozawa et al. [24]. MSb2O3 and ρ Sb2O3 are 291.52 g/mole and 5.58 g/cm3, respectively [25]. MSb2O5 and ρ Sb2O5 are 323.52 g/mole and 3.80 g/cm3, respectively [26]. Mn-SKD and Mp-SKD were determined to be 435.41 and 1812.15 g/mole, respectively. ρ n-SKD and ρ p-SKD were determined to be 7.63 and 7.60 g/cm3 at room temperature, respectively. For the R PB-n-SKD calculation, we assumed the number of antimony atoms in the SKD was equivalent to the antimony atom in SKD formula. Thus, the n values are 2/3 in n-SKD and 2/12 in p-SKD, respectively. In the case of n-SKD, the R PB-n-SKD was 1.37, which is in the protective R PB range (1 < R PB < 2). Thus, the antimony oxide provided somewhat of a protective barrier against further surface oxidation for the n-SKD compared to the p-type. However, in the case of p-SKD, the R PB-p-SKD was 2.14, which is in the non-protective range (R PB < 1 or R PB > 2). The relatively larger R PB-p-SKD indicates the oxide volume is larger in the p-type compared to the n-type SKD likely causing cracking and spallation.

Fig. 7
figure 7

XRD spectra for n-type Ba0.05Yb0.025CoSb3 and of p-type Ce0.9Co0.5Fe3.5Sb12 heated in air at 500 °C for 1 h. The circles highlight peaks consistent with Sb2O3 ([23] and JCPDS 11-689) and the squares highlight peaks consistent with Sb2O5 ([24] and JCPDS 33-111)

Full size image

Enamel physical property characterization

The physical properties of the enamel were measured and are summarized in Table 2. The thermal conductivity is 0.93 W/m K at 50 °C, which is approximately 10 times lower than the n-type CoSb3 and six times lower than the p-type CeFe3CoSb12 [4, 27–29].

Table 2 The physical properties of the enamel at various temperatures. D, ρ, α linear, C p, and κ are thermal diffusivity, density, linear CTE, specific heat, and thermal conductivity, respectively
Full size table

Assuming a 0.1 mm thick enamel coating on 3.5 mm × 3.5 mm square cross section SKD leg that is 6 mm long (dimensions used in previous SKD generator prototype generators [8] ), the parasitic heat loss through can be estimated using the following relationships.

$$ U = \kappa \frac{A}{L} $$
(3)
$$ \frac{{U_{enamel} }}{{U_{SKD} }} = \frac{{\kappa_{enamel} \times A_{enamel} }}{{\kappa_{SKD} \times A_{SKD} }} = \frac{{\kappa_{enamel} }}{{\kappa_{SKD} }} \times \frac{{(3.7^{2} - 3.5^{2} )}}{{3.5^{2} }} $$
(4)
$$ \frac{{U_{enamel} }}{{U_{n - SKD} }} = \frac{1}{10} \times \frac{{(3.7^{2} - 3.5^{2} )}}{{3.5^{2} }} = 0.01, $$
(5)
$$ \frac{{U_{enamel} }}{{U_{p - SKD} }} = \frac{1}{6} \times \frac{{(3.7^{2} - 3.5^{2} )}}{{3.5^{2} }} = 0.02, $$
(6)

where U is the thermal conductance, A is area, and L is length. Based on these calculations the parasitic heat loss is 1 and 2 percent of the total thermal conductance for the n- and p-type SKD elements. These calculations confirm that the impact on device efficiency will likely be low. Further work is required to determine how thin the enamel coatings can be made and still provide adequate oxidation protection to better estimate the parasitic heat loss.

The CTE was measured using dilatometery and was determined to be 6.7 ppm/K at 50 °C, which is approximately 6.8 and 6.0 ppm/K lower (or about 100 % lower) than the n- and p-type SKD, respectively [27]. This mismatch in the CTE could be advantageous in that the coating will be under compressive stress during operation. This approach is analogous to tempering glass to suppress surface crack propagation [30].

DSC analysis was conducted to determine the enamel melting or flowing temperature. It was determined that the two endothermic reactions at 526 and 648 °C could be attributed to melting (Fig. 8a). It is possible that an inhomogeneous distribution of constituents could explain the observation of two melting temperatures.

Fig. 8
figure 8

a Enamel DSC analysis, b enamel flow test on a Cu coupon at 650 °C in argon, and c enamel flow test on a Cu coupon at 700 °C in argon

Full size image

Melting and flowing behavior was characterized by melting enamel on copper samples. This particular enamel composition is known to be compatible with Cu, thus it served as a control to demonstrate that the enamel could adhere to a metallic component with dimension comparable to a thermoelectric element. At 650 °C, the enamel did not sufficiently flow as indicated by the relatively rough or granular surface (Fig. 8b). However, at 700 °C, the enamel flowed sufficiently to provide a uniform, transparent coating around the copper sample (Fig. 8c). Thus, 700 °C was used to flow enamel around SKD samples.

Enamel coating

In the initial tests with SKD, the enamel coatings cracked upon heating. We, therefore, chamfered the SKD coupon edges to reduce stresses and, therefore, cracking. Characterizing the efficacy of the chamfering was studied by comparing n-SKD samples with chamfered and square edges (Fig. 9). After the initial enameling process, both samples had crack-free coatings. However, after heating both samples at 500 °C for 1 h in air, the enamel coating on the chamfered sample was not cracked, while the enamel coating on the sample with square edges was (Fig. 9c–d). It is apparent that cracks are mainly generated at sharp corners where stresses are concentrated. Several chamfered and non-chamfered SKD samples were coated with enamel. In general, the enamel coatings on the non-chamfered samples cracked, while no cracking was observed on the chamfered samples.

Fig. 9
figure 9

Effect of chamfering of n-type Ba0.05Yb0.025CoSb3: a chamfered orthogonal edge, low magnification, b chamfered orthogonal edge (magnified view), c acute angle edge, low magnification, and d acute angle edge (magnified view)

Full size image

After the enamel coating process (heat to 700 °C in argon), the n-type Ba0.05Yb0.025CoSb3 and the p-type Ce0.9Co0.5Fe3.5Sb12 were successfully coated as shown in Fig. 10. The ratio of grams of enamel to grams of SKD was approximately 1. The appearance of the enamel-coated SKD samples was similar to the previous experiments using copper samples in that a uniform enamel coating formed.

Fig. 10
figure 10

Enamel coating on n-type Ba0.05Yb0.025CoSb3 or p-type Ce0.9Co0.5Fe3.5Sb12. a Chamfered n-SKD, before coating, b chamfered n-SKD, after coating, c chamfered p-SKD, before coating, and d chamfered p-SKD, after coating

Full size image

Enamel coating efficacy testing (short term −1 h at 500 °C)

To characterize the high temperature stability of enamel-coated SKD samples, they were heated in air at 500 °C for 1 h. There were no appreciable changes after the oxidation test. After testing, the samples were cross-sectioned and characterized using OM and SEM (Fig. 11). No significant signs of n- or p-type SKD oxidation were observed. We believe this is the first report demonstrating an enamel coating that can suppress oxidation in air at 500 °C for both n- and p-type SKD.

Fig. 11
figure 11

OM and SEM analysis after 500 °C 1 h air test. Enamel-coated and chamfered n-type Ba0.05Yb0.025CoSb3: a OM, b SEM. Enamel-coated and chamfered p-type Ce0.9Co0.5Fe3.5Sb12: c OM, d SEM

Full size image

Enamel coating efficacy testing (longer-term thermal annealing at 500 and 600 °C for 8 days, and 20 cycle tests between 600 °C and room temperature)

Table 3 shows the isothermal and thermal cycle test results of the enamel protected n- and p-type SKD samples. Figure 12 shows that no surface oxide was observed in the extended isothermal and thermal cycle tests. Some lines and dots in the enamel near the SKD surface (Figs. 11 and 12) are due to charging effects. EDS analyses indicated that oxygen and antimony concentrations at the surfaces were similar with those in the center of the samples as shown in Fig. 13. In addition, the oxygen concentrations closely matched with un-heated SKD samples (Figs. 4 and 6). The results indicate that the enamel coating is effective in preventing oxidation of SKD at 500 and 600 °C in air. Although the p-SKD-enamel interface appears to have reacted more than the n-SKD-enamel interface, the coating remained intact over the 8 days of continuous heating and 20 cycles between room temperature and 600 °C. Evaluating the longer-term stability over hundreds of thermal cycles and weeks of isothermal testing is still on-going. Additionally, although the heating and cooling rates are likely lower than what would be expected in a Federal Standard Test (FTP) cycle, these preliminary data suggest that despite the difference in CTE, the enamel coatings remained intact throughout the 20 cycle test.

Table 3 Efficacy testing of enamel-coated n- and p-SKD under various conditions
Full size table
Fig. 12
figure 12figure 12

OM and SEM analysis of enamel-coated SKD. n-SKD heated at 500 °C for 8 days: a OM, b SEM. p-SKD heated at 500 °C for 8 days: c OM, d SEM. n-SKD heated at 600 °C for 8 days: e OM, f SEM. p-SKD heated at 600 °C for 8 days: g OM, h SEM. n-SKD heated between 600 °C and room temperature for 20 cycles: i OM, j SEM. p-SKD heated between 600 °C and room temperature for 20 cycles: k OM, l SEM

Full size image
Fig. 13
figure 13

The element composition as a function of position across the enamel-SKD interface after the thermal stability air test. a n-SKD heated at 500 °C for 8 days, b p-SKD heated at 500 °C for 8 days, c n-SKD heated at 600 °C for 8 days, d p-SKD heated at 600 °C for 8 days, e n-SKD heated between 600 °C and room temperature for 20 cycles, and f p-SKD heated between 600 °C and room temperature for 20 cycles

Full size image

Summary

This work demonstrated the efficacy of enamel coatings to suppress oxidation of n-type Ba0.05Yb0.025CoSb3 and p-type Ce0.9Co0.5Fe3.5Sb12 SKD. This is the first report demonstrating the efficacy of enamel coatings to suppress oxidation in p-SKD containing oxygen-sensitive elements such as Ce, La, Fe, and Sb. Enamel coatings prevented oxidation of n- and p-type SKD when tested up to 600 °C in air continuously for 8 days or thermally cycled for 20 cycles. It was also determined that chamfering SKD elements was necessary to lower stressed at the edges and prevent cracking of the enamel coating. The enamel physical and processing properties were also characterized and were determined to be compatible with SKD-based technology fabrication and operation conditions.