Effect of rapid solidification on the synthesis and thermoelectric properties of Yb-filled Co4Sb12 skutterudite

Article abstract of DOI:10.1016/j.jallcom.2019.04.337

In this work the effect of different processing routes on the kinetics of formation of Co₄Sb₁₂ phase and Yb filling in the skutterudite cage, as well as thermoelectric behaviour was evaluated. As prepared slowly cooled ingot and rapidly solidified ribbons show the presence of multiple phases, as a consequence of the complex solidification path characterized by two peritectic reactions. As solidification rate increases, grain refinement is promoted. Annealing of rapidly solidified ribbons induces a faster solubilisation of Yb due to increased grain boundary diffusivity. For this reason, rapid solidification can be considered a promising intermediate step for lowering the total processing time of skutterudites. Sintering of powders obtained from the slowly cooled ingot and rapidly solidified ribbons leads to the formation of massive samples with different densities, depending on the particle size distributions of the starting powders. The effect of Yb content on the thermoelectric properties was critically analysed, considering both data from the literature and this work.

Full text of DOI:10.1016/j.jallcom.2019.04.337

Accepted Manuscript
Effect of rapid solidification on the synthesis and thermoelectric properties of Yb-filled
Co Sb skutterudite
12
4
F. Aversano, S. Branz, E. Bassani, C. Fanciulli, A. Ferrario, S. Boldrini, M. Baricco, A.
Castellero
PII:
S0925-8388(19)31647-0
DOI:
https://doi.org/10.1016/j.jallcom.2019.04.337
JALCOM 50521
Reference:
To appear in: Journal of Alloys and Compounds
Received Date: 30 October 2018
Revised Date: 20 March 2019
Accepted Date: 30 April 2019
Please cite this article as: F. Aversano, S. Branz, E. Bassani, C. Fanciulli, A. Ferrario, S. Boldrini, M.
Baricco, A. Castellero, Effect of rapid solidification on the synthesis and thermoelectric properties of
Yb-filled Co Sb skutterudite, Journal of Alloys and Compounds (2019), doi: https://doi.org/10.1016/
4
12
j.jallcom.2019.04.337.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Effect of rapid solidification on the synthesis and thermoelectric properties of Yb-filled
Co4Sb12 skutterudite
F. Aversano1, S. Branz1, E. Bassani2, C. Fanciulli2, A. Ferrario3, S. Boldrini3, M. Baricco1,*, A.
Castellero1
1University of Turin, Department of Chemistry & NIS Centre, Turin, Italy
2CNR – ICMATE, Lecco Unit, Lecco, Italy
3CNR – ICMATE, Padova Unit, Padova, Italy
*corresponding author (M. Baricco). E-mail address: marcello.baricco@unito.it; tel.: +39
0116707569; fax: +39 0116707855
Abstract
In this work the effect of different processing routes on the kinetics of formation of Co4Sb12 phase
and Yb filling in the skutterudite cage, as well as thermoelectric behaviour was evaluated.
As prepared slowly cooled ingot and rapidly solidified ribbons show the presence of multiple
phases, as a consequence of the complex solidification path characterized by two peritectic
reactions. As solidification rate increases, grain refinement is promoted. Annealing of rapidly
solidified ribbons induces a faster solubilisation of Yb due to increased grain boundary diffusivity.
For this reason, rapid solidification can be considered a promising intermediate step for lowering
the total processing time of skutterudites.
Sintering of powders obtained from the slowly cooled ingot and rapidly solidified ribbons leads to
the formation of massive samples with different densities, depending on the particle size
distributions of the starting powders. The effect of Yb content on the thermoelectric properties was
critically analysed, considering both data from the literature and this work.
Keywords: thermoelectricity; skutterudite; phonon scattering; rapid solidification; diffusion.
1

ACCEPTED MANUSCRIPT
Introduction
The interest in thermoelectric materials raises from their ability to convert wasted heat into
electricity in reliable and long standing solid state devices, improving energy efficiency. The
thermoelectric performance of materials is characterized by the dimensionless figure of merit
ꢁ
ZT = ꢀ ꢂ T, where S is the Seebeck coefficient, σ is the electrical conductivity, T is the absolute
ꢃ
temperature and κ is the total thermal conductivity, containing electronic, lattice and bipolar
contributions (κ = κꢄꢅ + κꢅꢆꢇ + κꢈꢉ). The ideal thermoelectric material should have a high power
factor, PF = Sꢊσ, and low thermal conductivity in a wide range of temperature. However, the
interdependence of S, σ and κel leads to an intrinsic maximum limit of ZT in correspondence of a
certain charge carrier concentration [1]. ZT can be further improved by reducing the lattice thermal
conductivity since it can be decoupled from the electronic contribution without varying the charge
carrier concentration [1].
Filled skutterudites can have high electrical conductivity like a crystal and low thermal conductivity
like a glass, which make them excellent thermoelectric material according to the phonon glass
electron crystal (PGEC) concept [2]. In fact, the presence of two large interstices inside the
elementary cell allows to insert interstitial atoms such as lanthanides [3,4] or alkaline earth metals
[5,6], acting as phonon scattering centres without affecting electrical conductivity. The filler content
is ruled by electronic issues [7-11]. Skutterudites based on the elements of the group 9 (Co, Rh, Ir)
and pnictogen elements (P, As, Sb) are electronically stable and can be only partially filled
according to the valence state of the filler ion. In the case of Co4Sb12, Yb is one of the ideal filler
species, due to its heavy mass and small radius compared to others. Conversely to trivalent
lanthanide ions, Yb allows to optimize electronic properties as a consequence of the mixed valence
state (Yb2+ and Yb3+) [3, 9]. It was shown that the contribution of Yb2+ ion increases with the Yb
occupancy factor in the cell [11], allowing the extension of the solubility range of the filler. The
effect of the filler amount in YbxCo4Sb12 on the thermoelectric properties was studied by several
authors [12-24]. The solubility of Yb in the skutterudite cell depends on the synthesis process [13],
2

ACCEPTED MANUSCRIPT
the temperature and the Co/Sb ratio [19].The actual amount of Yb in the cell is usually lower than
the nominal one [9, 16], due to its volatility and tendency to form oxides, affecting thermoelectric
performance. The desired amount of Yb in the compound can be attained by adding an excess of
Yb (30-50 mol.%) [9, 16].
A step forward for the commercial exploitation of skutterudites in thermoelectric applications is to
establish a reliable, reproducible and affordable processing route. Skutterudites processed directly
from the melt are affected by solidification segregation, originating from the peritectic reactions in
the binary phase diagram (see for instance Ref. [25] for Co-Sb), that induces porosity, chemical
inhomogeneity, secondary phase formation, incomplete filling of the skutterudite cell, and,
consequently, poor thermoelectric performances [13, 22]. Possible solutions to this problem are 1)
performing a solid/liquid reaction between CoSb2 and Sb [22], 2) stirring the liquid before casting
[13], 3) cooling following an oscillating temperature gradient [26], 4) direct formation of the
skutterudite phase by deep undercooling of the liquid through rapid solidification techniques [27].
Rapid solidification allows to obtain refined microstructures [28] and, in some cases, even
amorphous phases [29]. The introduction of a high density of interfaces in grain refined materials
leads to a reduction of the lattice thermal conductivity with respect to coarse microstructures
obtained by conventional processing routes [30].
Scope of this work is to employ rapid solidification as an intermediate step in the processing of Yb-
filled Co4Sb12 and to evaluate its effect on the kinetics of single phase formation and Yb filling. The
effect of the Yb filling on thermoelectric properties is critically discussed in the frame of a literature
survey.
Experimental
Elemental Co (rod, 99.995 %), Sb (shots, 99.999 %) and Yb (powder, 99.9 %) were weighed
according to the nominal stoichiometry (Yb0.25Co4Sb12) and placed into graphite coated quartz
ampoules. The ampoules were sealed under vacuum (10-5 mbar) and placed in a muffle furnace for
the following heat treatment: 1) slow heating (6 hours) from room temperature to 1473 K, 2)
isotherm at 1473 K for 5 h (the ampoule was shacked every hour to guarantee homogenization of
3

ACCEPTED MANUSCRIPT
the melt), 3) free cooling in the furnace to room temperature. In this way, ingots of the master alloy
were obtained.
Rapid solidification of the alloy was performed using a planar flow casting apparatus (Edmund
Bühler GmbH), where the master alloy ingot was induction melted in a BN crucible under Ar
atmosphere and the melt was injected under a pressure of 0.2 bar onto a copper wheel rotating at
20 m/s, forming ribbons 20-30 µm thick. The cooling rate during rapid solidification is estimated to
be 105 K/s [28].
Phase evolution in master alloy ingots (bulk) and rapidly solidified samples (RS) was studied by
isothermal annealing at 898 K for 1, 2, 4 days.
Powders for sintering were obtained by hand grinding in an agate mortar the following precursors:
-
-
-
Master alloy ingots annealed for 4 days at 898 K (powder 1);
rapidly solidified ribbons annealed for 4 days at 898 K (powder 2);
as quenched rapidly solidified ribbons without annealing (powder 3).
Finally, powders 1, 2 and 3 were sintered by Open Die Pressing (ODP) [31] obtaining samples
ODP1, ODP2 and ODP3, respectively. The powders were loaded in Fe sheath, preheated at 823 K
for 10 min before sintering between two heated plates at 748 K for 10 min applying a load of 200
MPa. At the end of the sintering procedure, load was released and heating of the plates was
switched off. In the case of samples ODP1 and ODP2, the samples were slowly cooled between
the plates passing from 748 K to 373 K into 2 hours, while in the case of ODP3 the sample was
subjected to free cooling, covering the same temperature gap in 30 min.
Structural characterization was carried out by X-ray diffraction (XRD) using a Panalytical X’Pert
instrument with Cu Kα radiation (λ = 1.5406 Å) working at 40 kV and 30 mA. The values of lattice
parameter and Yb occupancy factor at 2a site in the skutterudite cage were refined by Rietveld
method using the software MAUD [32]. The microstructure and chemical composition were
monitored by Scanning Electron Microscopy (SEM) using a Zeiss EVO 50 XVP – LaB6 equipped
with an Oxford Instruments INCA Energy 250 Energy Dispersive X-ray analyzer (EDX).
Thermoelectric properties of the sintered samples were measured as a function of temperature.
ODP 2 was measured up to 873 K, while ODP3 was measured up to 688 K (below the sintering
4

ACCEPTED MANUSCRIPT
temperature) in order to avoid the annealing of the sample and preserve the microstructure
obtained after sintering. Seebeck coefficient, S, and electrical conductivity, σ, were measured by
means of a custom built apparatus [33] applying a heating ramp of 2 K/min and working under
helium flow (200 sccm/min) . The thermal conductivity (κ) was calculated according to the formula
κ = αρCꢉ, where α is the thermal diffusivity, ρ is the mass density and Cp is the specific heat
capacity. The thermal diffusivity (α) of 10 mm x 10 mm samples was measured by the laser flash
method (LFA 457 MicroFlash, Netzsch, Selb, Germany) under Ar flow. The mass density (ρ) was
evaluated by Archimedes' principle with a pycnometer and specific heat capacity (Cp) was
determined by differential scanning calorimetry, DSC (Perkin Elmer DSC 7), according to the
following procedure. Heating ramps at 10 K/min were performed in a temperature range of 30 K.
Each ramp was preceded and followed by a long isotherm in order to stabilize the signal. The
measurement was repeated with the empty alumina pans, a standard (zaffire) and the sample. The
s
specific heat of the sample, Cp , at different temperatures was obtained by comparing the signals
measured for the sample and the standard using the “height method” and the “area method”
through the relationships [34]
ꢍ
ꢒ
ꢒ
∆ꢎ ꢐ∆ꢎ
ꢔ
ꢔ
ꢏ
ꢑ
ꢑ
ꢓ
Cꢉꢋ = Cꢉꢌ ꢍ
(1)
∆ꢎ ꢐ∆ꢎ
ꢓ
ꢏ
z
where Cp is the specific heat of the standard, ∆Ys and ∆Yz are the calorimetric signals of the
sample and the standard, respectively, ∆Y0 is the calorimetric signal of the empty pans, mz and ms
are the mass of the standard and the sample, respectively, and
ꢍ
ꢒ
ꢒ
∆ꢕ ꢐ∆ꢕ
ꢔ
ꢔ
ꢏ
ꢑ
ꢓ
Cꢉꢋ = Cꢉꢌ ꢍ
(2)
∆ꢕ ꢐ∆ꢕ
ꢓ
ꢑ
ꢏ
z
where Cp is the specific heat of the standard, ∆Hs and ∆Hz are the integrated areas under the
calorimetric signals of the sample and the standard, respectively, ∆H0 is the integrated area under
the calorimetric signal of the empty pans, mz and ms are the mass of the standard and the sample,
respectively. Measurements were performed on ODP2 and ODP3 samples and the values of
specific heat obtained for the two samples with the two methods were averaged at each
temperature.
5

ACCEPTED MANUSCRIPT
Results and discussion
EDX analysis of all the samples revealed a ratio Sb/Co slightly larger than the stoichiometry (3:1),
as reported in table 1 and table 2, indicating that equilibrium is attained in the two phases region
CoSb3 + Sb [19], where the solubility of Yb in the skutterudite depends on the Sb/Co ratio. Even if
close to the detection limit of the technique, the total amount of Yb measured by EDX (1.0-1.2
at.%) is lower than the expected equilibrium value (1.5 at.%), indicating that a fraction of the Yb
powder was not successfully alloyed and confirming the difficulty to maintain the Yb stoichiometry
in the alloy, as already highlighted in Ref. [9] and Ref. [16].
The XRD pattern of the as prepared master alloy ingot (bulk), Figure 1(a), shows only the
crystallographic reflections of the CoSb3 phase. However, backscattered electrons SEM
micrographs of the same sample (not shown) reveal the presence of additional phases (CoSb2,
CoSb, YbSb2) dispersed in the CoSb3 matrix, indicating that the alloy did not reach the equilibrium
because of the slow kinetics of the peritectic reactions
CoSb + liquid → CoSb2
CoSb2 + liquid → CoSb3
(3)
(4).
After annealing at 898 K for 4 days, the XRD pattern of the bulk shows only the crystallographic
reflections of the CoSb3 phase, Figure 1(b). Again, backscattered electrons SEM micrographs of
the annealed bulk, Figure 2(a), show the presence of the same additional phases (CoSb2, CoSb,
YbSb2) observed in the corresponding as prepared sample, indicating that the alloy did not reach
the equilibrium, even after annealing at high temperature.
In the case of the as quenched ribbon, the XRD pattern, Figure 1(c), reveals the presence of five
phases (CoSb3, CoSb2, CoSb, YbSb2 and Sb) indicating that the system is out of the equilibrium,
as a consequence of the high cooling rate experienced by the alloy during rapid solidification.
Figure 2(c) shows the SEM micrograph of the longitudinal cross section of the ribbon, obtained
with the backscattered electrons detector. The lower edge of the ribbon is smooth while the upper
edge is rough, indicating that the former solidified in contact with the copper substrate (wheel side)
and the latter solidified in contact with the argon atmosphere (free side). The micrograph is
characterized by different levels of grey, indicating the presence of several phases, in agreement
6

ACCEPTED MANUSCRIPT
with the XRD pattern, Figure 1(c). The microstructure becomes progressively coarser from the
wheel side to the free side, because of the progressive reduction of the cooling rate from the wheel
to the free side.
After annealing the ribbon at 898 K for 4 days, the formation of a single phase is promoted, as
shown by the XRD pattern, Figure 1(d), and by the backscattered electrons SEM micrograph,
Figure 2(b), that shows uniform grey contrast.
Table 1 shows the lattice parameter, the Yb occupancy factor in the 2a site and the ratio Yb2a/Ybtot
in bulk and ribbon samples as a function of annealing time at 898 K. The bulk shows a progressive
increase of the two parameters during annealing without levelling off at a constant value,
suggesting that equilibrium is not reached even after 4 days of annealing, in agreement with
literature data reporting that complete solubilisation of Yb is accomplished after at least 7 days of
annealing at 1073 K [14, 16]. In the case of the ribbon, the values reported in table 1 drop after one
day of annealing, subsequently increase reaching a plateau after 2-4 days of annealing, indicating
attainment of the equilibrium. The larger amount of solubilized Yb in the as prepared ribbon is due
to the quenching of the sample from a higher temperature where the Yb solubility is larger with
respect to the annealing temperature [19]. The maximum amount of Yb solubilized in the
skutterudite (around 0.19 for the ribbon annealed for 4 days) is lower than the stoichiometry (i.e.
0.25), indicating that about 25 % of Yb was lost during processing, in agreement with the results of
EDX analysis.
The filling of the skutterudite cage occurs faster in the ribbon than in the bulk, as a consequence of
the presence of a finer microstructure. The apparent diffusion coefficient, Dapp, is given by [35]
ꢖ
Dꢆꢉꢉ = Dꢅ + Dꢈ ꢗ
(5)
where Dl is the lattice diffusion coefficient, Db is the grain boundary diffusion coefficient, δ is the
grain boundary width and d is the grain size. In the case of the rapidly solidified ribbon, the grain
size is around 0.1 µm, while for the bulk, grain size is of the order of several tenths of microns.
Thus, considering a constant value of the grain boundary width, the value of the δ/d ratio for the
rapidly solidified ribbons is about two/three orders of magnitude larger than in the bulk, increasing
the contribution of the grain boundary diffusion and the apparent diffusion coefficient in the ribbons.
7

ACCEPTED MANUSCRIPT
As a consequence, the filling of the skutterudite cell with Yb atoms during annealing occurs faster
in the ribbon than in the bulk, confirming that rapid solidification can be used as an intermediate
step to reduce the total processing time.
The XRD patterns of the sintered samples show exclusively the crystallographic reflections
belonging to CoSb3, as shown in Figure 1(e) for ODP2 that was taken as a representative case.
However, SEM backscattered electron micrographs, Figures 2(d-f), reveal the presence of residual
unreacted phases that are progressively less evident in ODP3 (CoSb, CoSb2, Sb, YbSb2), Figure
2(f), ODP1 (CoSb2, YbSb2), Figure 2 (d), and ODP2 (CoSb2, Sb, YbSb2), Figure 2 (e). In the case
of ODP3, sintering of the starting multiphase powder (powder 3) promoted the homogenization of
the sample forming an almost single phase dense massive material where some unreacted phases
are still present, as shown in Figures 2(f). Table 2 shows the values of the lattice constant and the
Yb occupancy factor in 2a site for the sintered samples. In the case of ODP1 and ODP2, the
values obtained after sintering are slightly lower than those of the starting powders (powder 1 and
powder 2), Table 1, because the data refer to different batches. However, the ratio between
Yb2a(ribbon)/Yb2a(bulk) (about 1.30) is maintained in the corresponding ODP samples. The amount
of Yb dissolved in the skutterudite is not expected to be affected by the sintering process because
this was performed at a temperature lower than the annealing of the starting materials. The value
of Yb occupancy factor in ODP2 is very close to the maximum solubility at 900 K (0.19) reported in
Ref. [19], indicating that the system is almost at the equilibrium. In fact, only small traces of
unreacted phase (CoSb2, Sb, YbSb2) are still present after sintering, as shown in Figure 2(e). In
the case of ODP1, Figure 2(d), the amount of secondary phases is larger than in ODP2, as a
consequence of the slower diffusivity in the bulk. The presence of secondary phases different from
those predicted by the phase diagram [19], indicates that phase equilibrium as well as equilibrium
Yb solubility were not attained in ODP1.
A similar consideration can be made for ODP3 where non equilibrium secondary phases are also
present, Figure 2(f). In addition, the lower values of lattice constant and Yb occupancy factor with
respect to ODP1 and ODP2 can be further explained by the lower solubility of Yb in the
skutterudite at the temperature at which the sintering was performed [19].
8

ACCEPTED MANUSCRIPT
The sintered samples show different values of relative density, as reported in Table 2. The
significantly lower density of sample ODP1 (75 %) with respect to ODP2 (95 %) and ODP3 (90 %)
can be explained by the different particles size distributions of the corresponding starting powders,
estimated by SEM image analysis, that are reported in Figure 3. In fact, a multimodal distribution of
the particles size, here represented by the broader peak (ODP2 and ODP3), allows a better
packing of the powders before the sintering process. The slightly larger value of relative density of
ODP2 (95 %) with respect to ODP3 (90 %) is probably due to the slower cooling rate of the
sample, that reduced the effect of internal stresses induced by the different thermal contractions of
the metallic sheath and the thermoelectric sintered powders. Conversely ODP3, experiencing a
faster cooling, suffered a compressing action at low temperature inducing a loss in compaction of
the grown grains, inducing a residual micro-porosity in the bulk.
Thermoelectric properties were measured only on ODP2 and ODP3 because ODP1 shows a value
of relative density too low (75 %). ODP2 and ODP3 show similar values of the Seebeck coefficient
as a function of temperature, Figure 4(a), indicating an n-type behaviour and suggesting that the
presence of secondary phases in ODP3 and the different levels of filling with Yb between the two
samples is too small to significantly affect the thermopower. As recently reported in Ref. [36],
secondary phases affect thermoelectric properties of Yb-filled Co4Sb12 skutterudites when they
exceed a critical value (i.e. around 8 wt.%). Both ODP2 and ODP3 show decreasing values of the
electrical conductivity, σ, as a function of temperature, revealing a behaviour typical of heavily
doped semiconductors, Figure 4(b). This can be explained by the increased carrier concentration
due to the filling with Yb [16, 21, 24]. However, ODP2 shows larger values of σ than ODP3. The
effective electrical conductivity depends on the relative density of the sample and was calculated
applying the Bruggeman symmetric model, where porosity is described as spherical inclusions [37,
38]. For relative densities of 90 % (ODP3) and 95 % (ODP2), the calculated effective electrical
conductivities are 0.85 and 0.925 times the electrical conductivity of the fully dense material,
respectively. The ratio between the calculated effective electrical conductivities of ODP3 and
ODP2 is 0.92, while the ratio between the corresponding measured electrical conductivities is 0.89
at room temperature and 0.92 at 688 K. The agreement between the calculated and measured
9

ACCEPTED MANUSCRIPT
values suggests that the different relative densities of ODP2 and ODP3 are responsible for the
different measured electrical conductivities.
Figure 4(c) shows thermal diffusivity, α, as a function of temperature. Both samples show similar
trends with a minimum around 600 K, however ODP2 shows lower values of α with respect to
ODP3. Figure 4(d) shows the values of the specific heat at different temperature. Each point was
obtained averaging four values obtained by means of the “height” and “area” methods on ODP2
and ODP3 samples. The error bars represent the standard deviation and the relative error is lower
than 5 %, in agreement with the typical values for this technique. We can see that at room
temperature the specific heat approaches the limit given by the Dulong-Petit approximation (Cp =
3R). To extend the temperature dependence of the specific heat across a wider range, the
experimental data were fitted according the following typical expression [39]
Cꢉ = 3R + aT + bTꢊ
(6)
where R is the gas constant (8,314472 Jmol-1K-1), a and b are fitting parameters, which turn out
equal to -3.13 10-3 K-1 and 9.15 10-6 K-2, respectively.
From the data of thermal diffusivity, density and specific heat, the thermal conductivity was
determined at different temperatures according to the relationship κ = αρCꢉ, as shown in Figure
4(e). It can be seen that ODP2 and ODP3 show similar values between 300 K and 500 K, then
above 500 K thermal conductivity of ODP2 becomes lower than the one of ODP3. The total
thermal conductivity, κ, is given by κ = κꢄꢅ + κꢅꢆꢇ + κꢈꢉ , where κel, κlat and κbp represent the
electronic, lattice and bipolar contributions, respectively. The electronic contribution to the thermal
conductivity, κel, was calculated through the Wiedemann-Franz law κꢄꢅ = LσT, where L is the
Lorentz constant (2·10-8 V2K-2), σ is the measured electrical conductivity and T is the absolute
temperature. The bipolar contribution, κbp, is due to the thermal excitation of charge carriers (i.e.
holes for n-type semiconductors) opposite to the dominant ones (i.e. electrons for n-type
semiconductors) above a critical temperature [40].
In order to separate the contributions of κlat and κbp, we plotted the difference between κ and κel as
a function of 1/T, as shown in the inset of Figure 2(e). The contribution of the bipolar thermal
10

ACCEPTED MANUSCRIPT
ꢍ
ꢒ
conductivity [40], that is favoured at high temperature, becomes relevant when κ − κꢄꢅ vs. 1/T
deviates from linearity (dotted lines are a guide to the eye), explaining the reason of the minimum
ꢍ
ꢒ
in thermal conductivity, κ, observed around 600 K. ODP2 shows lower values of κ − κꢄꢅ than
ODP3, because of the larger amount of Yb hosted in the skutterudite structure, that leads to an
increase of phonon scattering and, consequently, to a reduction of lattice thermal conductivity In
ODP2. The phonon scattering due to the larger Yb filling fraction compensates the increased
electronic thermal conductivity caused by the larger density, leading to an overall lower thermal
conductivity with respect to ODP3. Power factor was calculated from the measured values of the
electrical conductivity and Seebeck coefficient as a function of temperature, as reported on the left
hand side of Figure 4(f) (open and filled diamonds). The higher values shown by ODP2 with
respect to ODP3 are due to the larger values of the electrical conductivity, related to higher density
of the sample, since the two samples show the same values of the Seebeck coefficient. Finally, the
figure of merit, ZT, as a function of temperature is reported on right hand side of Figure 4(f) (open
and filled circles). ODP2 shows larger values of ZT with respect to ODP3 as a consequence of the
larger power factor and the lower thermal conductivity. In the case of ODP2, ZT reaches a
maximum value of 0.9 at 773 K, in agreement with the values reported in Refs. [16, 21] for similar
values of Yb.
Figure 5 shows the thermoelectric properties (electrical conductivity, σ, Seebeck coefficient, S,
lattice thermal conductivity, κtot - κel, and figure of merit, ZT) measured at 673 K as a function of the
Yb content in Co4Sb12. The squares, diamond and triangles represent data from the literature (the
reference is indicated by the corresponding number) while the stars are the data from this work.
Data from the literature were grouped according to different symbols, depending on the technique
used for determining the Yb content in the skutterudite phase. The data show some scattering
likely due to presence of small amounts of secondary phases, different density of the samples,
difference between nominal and actual Yb content. As a general trend, when the Yb content
increases, electrical conductivity, σ, increases, while the absolute value of the Seebeck coefficient,
|S|, decreases, as shown in Figure 5(a) and Figure 5(b), respectively. This trend is in agreement
with the increased density of charge carrier due to the insertion of Yb atoms in the skutterudite
11

ACCEPTED MANUSCRIPT
cage [16, 21, 24]. The actual Yb content in Co4Sb12, measured by XRD and EDX/EPMA (squares
and diamonds, respectively), never exceeds 0.35, setting this value as the solubility limit [18, 19]. It
was shown for Fe2Ni2Sb12 (isoelectronic with Co4Sb12) that, as the Yb occupancy factor at 2a site in
the skutterudite cage increases, the valence state of the Yb ion progressively decreases from
trivalent to divalent, leading to the observed maximum solubility of the filler [9, 11]. This solubility
limit was recently exceeded by using non equilibrium processing such as ball milling [36]. Since the
samples obtained in Ref. [36] contained relevant amounts of secondary phases, that considerably
affects thermoelectric properties, they were not taken in account in the present discussion.
Considering the triangles, representing the thermoelectric property as a function of the nominal Yb
content, we observe that, when x < 0.35, they follow the same trend of the squares and diamonds,
while, when x > 0.35, they are shifted to the right with respect to the squares and diamonds,
representing the thermoelectric property as a function of the actual Yb content. When the solubility
limit of Yb is exceeded, the YbSb2 secondary phase forms [18, 19] and its amount increases as the
total Yb content increases. Thus, for x > 0.35, the triangles represent multiphase materials rather
than single phase materials, and show an increasing metallic behaviour (i.e. increasing electrical
conductivity, σ, and decreasing Seebeck coefficient, |S|) as the amount of secondary phase
increases. The opposite effect of Yb filling on electrical conductivity and Seebeck coefficient leads
to an optimized power factor, S2σ, for 0.2<x<0.3 (not shown). The effect of the Yb solubility limit is
even more evident when the lattice thermal conductivity, taken as the difference between κtot and
κel, is considered, Figure 5(c). In this case, the presence of a minimum of κtot - κel around x 0.35
indicates that phonon scattering is maximized in correspondence of the Yb solubility limit and that
further addition of Yb is counterproductive. Finally, the combined effect of power factor and thermal
conductivity leads to an optimized value of the figure of merit, ZT, for 0.20<x<0.25, as shown in
Figure 5(d).
Conclusions
In this work, we explored the effect of different processing routes on the structure, microstructure
and thermoelectric properties of Yb-doped CoSb3 compound. Uncontrolled cooling of the melt
12

ACCEPTED MANUSCRIPT
leads to the formation of a porous bulk with coarse microstructure. Rapid solidification by planar
flow casting allows the formation of fragmented ribbons with a refined microstructure and no
porosity. In both cases, as prepared samples show multiple phases, thus annealing was needed to
promote homogenization. The solubilisation of Yb in the skutterudite structure is significantly faster
for the rapidly solidified ribbons as a consequence of the increased grain boundary diffusivity with
respect to the bulk sample. Thus, rapid solidification reveals to be a convenient intermediate step
to decrease the total processing time for the synthesis of filled skutterudites.
Powders obtained by grinding the bulk show a narrower particle size distribution with respect to the
powders obtained from the ribbons, leading to a lower density in the sintered samples. Sintering of
the powders obtained from as quenched and annealed ribbons leads to almost single phase dense
samples, however Yb filling is higher when annealed ribbons are used.
Thermoelectric properties of the sintered samples depend on porosity and Yb filling. As expected,
larger values of electrical conductivity were found in denser samples. Concerning thermal
conductivity, the interplay between density and Yb filling is more complex. On the one hand, less
dense samples are expected to show lower values of thermal conductivity. On the other hand,
denser samples show a larger Yb filling that leads to a lower lattice thermal diffusivity and, overall,
to a lower thermal conductivity with respect to less dense samples. A survey of literature data
shows a direct correlation between thermoelectric properties and Yb content when the latter is
below the solubility limit. When the Yb content exceeds the solubility limit, thermoelectric properties
show a deviation from the observed trend due to the presence of secondary phases affecting
thermoelectric behaviour.
Acknowledgements
A. Castellero, S. Branz and M. Baricco thank Fondazione CRT for financial support (grant n.
2015.1530)
References
[1] G.J. Snyder, E.S. Toberer, Complex thermoelectric materials, Nature Mater. 7 (2008) 105-114.
13

ACCEPTED MANUSCRIPT
[2] B. Sales, D. Mandrus, B. Chakoumakos, V. Keppens, J. Thompson, Filled skutterudite
antimonides: Electron crystals and phonon glasses, Phys. Rev. B 56 (1997) 15081-15089.
[3] B.C. Sales, D. Mandrus, R.K. Williams, Filled Skutterudite Antimonides: A New Class of
Thermoelectric Materials, Science 272 (1996) 1325-1328.
[4] G.S. Nolas, M. Kaeser, R.T. Littleton IV, T.M. Tritt, High figure of merit in partially filled ytterbium
skutterudite materials, Appl. Phys. Lett. 77 (2000) 1855-1857.
[5] L.D. Chen, T. Kawahara, X.F. Tang, T. Goto, T. Hirai, J.S. Dyck, W. Chen, C. Uher, Anomalous
barium filling fraction and n-type thermoelectric performance of BayCo4Sb12, J. Appl. Phys. 90
(2001) 1864-1868.
[6] M. Puyet, B. Lenoir, A. Dauscher, M. Dehmas, C. Stiewe, E. Müller, High temperature
transport properties of partially filled CaxCo4Sb12 skutterudites, J. Appl. Phys. 95 (2004) 4852-4855.
[7] C. Uher, Skutterudite-based thermoelectric, in: D.M. Rowe (Ed), Thermoelectrics Handbook -
Macro to Nano, Taylor and Francis, Boca Raton, 2006, chapter 34.
[8] B. Bourgoin, D. Bérardan, E. Alleno, C. Godart, O. Rouleau and E. Leroy, Preparation and
thermopower of new mischmetal-based partially filles skutterudites MmyFe4-x(Co/Ni)xSb12, J. Alloys
Compd. 399 (2005) 47-51.
[9] D. Bérardan, E. Alleno, C. Godart, O. Rouleau and J. Rodriguez-Carvajal, Preparation and
chemical properties of the skutterudites (Ce-Yb)yFe4-x(Co/Ni)xSb12, Mater. Res. Bull. 40 (2005) 537-
551.
[10] C. Artini, G. Zanicchi, G.A. Costa, M.M. Carnasciali, C. Fanciulli and Riccardo Carlini,
Correlations between Structural and Electronic Properties in the Filled Skutterudite
Smy(FexNi1−x)4Sb12, Inorg. Chem. 55 (2016) 2574- 2583.
[11] A. Kaltzoglou, P. Vaqueiro, K.S. Knight and A.V. Powell, Synthesis, characterization and
physical properties of the skutterudites YbxFe2Ni2Sb12, J. Solid State Chem. 193 (2012) 36-41.
[12] X. Y. Zhao, X. Shi, L. D. Chen, W. Q. Zhang, S. Q. Bai, Y. Z. Pei, X. Y. Li, T. Goto, Synthesis
of YbyCo4Sb12/Yb2O3 composites and their thermoelectric properties, Appl. Phys. Lett. 89 (2006)
092121.
14

ACCEPTED MANUSCRIPT
[13] Z. Yao, X.Y. Li, Y.S. Tang, L.D. Chen, Genomic Effects of the Quenching Process on the
Microstructure and Thermoelectric Properties of Yb0.3Co4Sb12, J. Electron. Mater. 44 (2015) 1890-
1895.
[14] Z. Xiong, X. Chen, X. Huang, S. Bai, L. Chen, High thermoelectric performance of
Yb0.26Co4Sb12/yGaSb nanocomposites originating from scattering electrons of low energy, Acta
Mater. 58 (2010) 3995-4002.
[15] H. Li, X. Tang, X. Su, Q. Zhang, C. Uher, Nanostructured bulk YbxCo4Sb12 with high
thermoelectric performance prepared by the rapid solidification method, J. Phys. D: Appl. Phys. 42
(2009) 145409.
[16] J.R. Salvador , J. Yang , X. Shi , H. Wang , A.A. Wereszczak , H. Kong, C. Uher, Transport
and mechanical properties of Yb-filled Skutterudites, Phil. Mag. 89 (2009) 1517–1534.
[17] X. Shi, H. Kong, C.-P. Li, C. Uher, J. Yang, J. R. Salvador, H. Wang, L. Chen, W. Zhang, Low
thermal conductivity and high thermoelectric figure of merit in n-type BaxYbyCo4Sb12 double-filled
skutterudites, Appl. Phys. Lett. 92 (2008) 182101.
[18] S. Wang, J.R. Salvador, J. Yang, P. Wei, B. Duan, J. Yang, High-performance n-type
YbxCo4Sb12: from partially filled skutterudites towards composite thermoelectrics, NPG Asia Mater.
8 (2016) e285.
[19] Y. Tang, S.-W. Chen, G.J. Snyder, Temperature dependent solubility of Yb in Yb-CoSb3
skutterudite and its effect on preparation, optimization and lifetime of thermoelectrics, J.
Materiomics 1 (2015) 75-84.
[20] H. Liu, X. Zhao, T. Zhu, Y. Gu, Thermoelectric properties of YbxCo4Sb12 system, J. Rare
Earths 30 (2012) 456-459.
[21] K.-H. Park, W.-S. Seo, D.-K. Shin, I.-H. Kim, Thermoelectric properties of Yb-filled CoSb3
skutterudites, J. Korean Phys. Soc. 65 (2014) 491-495.
[22] H.Y. Geng, S. Ochi, J.Q. Guo, Solidification contraction-free synthesis for the Yb0.15Co4Sb12
bulk material, Appl. Phys. Lett. 91 (2007) 022106.
[23] T. Dahal, Q. Jie, G. Joshi, S. Chen, C. Guo, Y. Lan, Z. Ren, Thermoelectric property
enhancement in Yb-doped n-type skutterudites YbxCo4Sb12, Acta Materialia 75 (2014) 316-321.
15

ACCEPTED MANUSCRIPT
[24] J. Yang Q. Hao, H. Wang, Y.C. Lan, Q.Y. He, A. Minnich, D.Z. Wang, J.A. Harriman, V.M.
Varki, M.S. Dresselhaus, G. Chen, Z.F. Ren, Solubility study of Yb in n-type skutterudites
YbxCo4Sb12 and their enhanced thermoelectric properties, Phys. Rev. B 80 (2009) 115329.
[25] H. Okamoto, Phase diagrams for binary alloys, ASM International, Materials Park (OH), 2000,
p. 258.
[26] B. Alinejad, A. Castellero, M. Baricco, Full dense CoSb3 single phase with high thermoelectric
performance prepared by oscillated cooling method, Scripta Materialia 113 (2016) 110-113.
[27] C. Artini, A. Castellero, M. Baricco, M.T. Buscaglia, R. Carlini, Structure, microstructure and
microhardness of rapidly solidified Smy(FexNi1-x)4Sb12 (x = 0.45, 0.50, 0.70, 1) thermoelectric
compounds, Solid State Sciences 79 (2018) 71-78.
[28] M. Baricco, E. Bosco, E. Olivetti, M. Palumbo, P. Rizzi, A. Stantero, L. Battezzati, Rapid
solidification of alloys, Int. J. Mater. Prod. Technol. 20 (2004) 358-376.
[29] M. Baricco, A. Castellero, M. Di Chio, Zs. Kovacs, P. Rizzi, M. Satta and A. Ziggiotti, “Thermal
stability and hardness of Mg–Cu–Au–Y amorphous alloys”, J. Alloys Compd.,434–435 (2007) 183-
186.
[30] S.K. Bux, J.-P. Fleurial and R.B. Kaner, Nanostructured materials for thermoelectric
applications, Chem. Commun. 46 (2010) 8311–8324.
[31] S. Ceresara, C. Fanciulli, F. Passaretti, D. Vasilevskiy, Texturing of (Bi0.2Sb0.8)2Te3
Nanopowders by Open Die Pressing, J. Electron. Mater. 42 (2013) 1529-1534.
[32] http://maud.radiographema.com
[33] S. Boldrini, A. Famengo, F. Montagner, S. Battiston, S.Fiameni, M. Fabrizio and S. Barison,
Test rig for high-temperature thermopower and electrical conductivity measurements, J. Electron.
Mater. 42(2013) 1319-1323.
[34] R.F. Speyer, Thermal analysis of materials, Marcel Dekker Inc., New York, 1994, pp. 79-80.
[35] D.A. Porter, K.E. Easterling, Phase Transformations in Metals and Alloys, 2nd edition,
Chapmann & Hall, London, 1992, p. 101.
16

ACCEPTED MANUSCRIPT
[36] B. Ryll, A. Schmitz, J. de Boor, A. Franz, P.S. Whitfield, M. Reehuis, A. Hoser, E. Müller, K.
Habicht and K. Fritsch, Structure, Phase Composition, and Thermoelectric Properties of
YbxCo4Sb12 and Their Dependence on Synthesis Method, Appl. Energy Mater. 1 (2018) 113-122.
[37] D.A.G. Bruggeman, The calculation of various physical constants of heterogeneous
substances: I. The dielectric constants and conductivities of mixtures composed of isotropic
substances. Ann. Phys., 24 (1935) 636-664.
[38] J.M.V.A. Koelman, A. de Kuijper, An effective medium model for the electric conductivity of an
N-component anisotropic and percolating mixture, Physica A,247 (1997) 10-22.
[39] A.T. Dinsdale, SGTE data for pure elements, CALPHAD 15 (1991) 317-425.
[40] S. Wang, J. Yang, T. Toll, J. Yang, W. Zhang, X. Tang, Conductivity-limiting bipolar thermal
conductivity in semiconductors, Sci. Rep. 5 (2015) 10136.
17

ACCEPTED MANUSCRIPT
Tables captions
Table 1. Lattice parameter, a, Yb occupancy factor in the 2a site, Yb2a, ratio (Yb2a/Ybtot) between
Yb in the skutterudite cage, Yb2a, and total amount of Yb, Ybtot, Co/Sb ratio as a function of the
annealing time at 898 K (Ybskutterudite). Lattice parameter and Yb2a occupancy factor were
determined by Rietveld refinement of the XRD patterns of Figure 1, the Co/Sb ratio was
determined by EDX analysis.
Table 2. Lattice parameter, Yb2a occupancy factor, relative density and Co/Sb ratio of the as
sintered ODP1, ODP2 and ODP3 samples.
18

ACCEPTED MANUSCRIPT
Table 1
Annealing
time
Bulk
Ribbon
(days)
a (Å)
Yb2a
Yb2a/Ybtot Co/Sb
a (Å)
Yb2a
Yb2a/Ybtot
0.98
Co/Sb
0
1
2
4
9.0459 0.12(2)
9.0458 0.12(1)
9.0471 0.12(9)
9.0486 0.14(4)
0.76
0.75
0.80
0.90
3.40
9.0530 0.19(2)
9.0482 0.14(2)
9.0522 0.18(4)
9.0527 0.18(8)
3.30
0.72
0.93
0.98
19

ACCEPTED MANUSCRIPT
Table 2
Sample
Lattice parameter
(Å)
Yb2a occupancy
factor
Relative density
Sb/Co
(%)
75
95
90
ODP1
ODP2
ODP3
9.0478
0.13(6)
3.40
3.24
3.35
9.0512
0.17(6)
9.0462
0.12(6)
20

ACCEPTED MANUSCRIPT
Figures captions
Figure 1. X-ray diffraction (XRD) patterns. (a) as prepared bulk ingots. (b) annealed (4 days at 898
K) bulk ingots. (c) as prepared rapidly solidified (RS) ribbons. (d) annealed (4 days at 898 K)
rapidly solidified (RS) ribbons. (e) ODP2 sintered sample. Black points, red curves and green
curves represent experimental data, refined patterns and differences between experimental and
refined patterns, respectively.
Figure 2. Backscattered electron SEM images. (a) annealed (4 days at 898 K) bulk ingot. (b)
annealed (4 days at 898 K) rapidly solidified (RS) ribbon. (c) as prepared rapidly solidified (RS)
ribbon. (d), (e) and (f) sintered ODP1, ODP2 and ODP3 samples, respectively.
Figure 3. Particle size distribution for powder 1, obtained by grinding the annealed bulk ingot,
powder 2 and powder 3, obtained by grinding annealed and as quenched rapidly solidified ribbons,
respectively.
Figure 4. Thermoelectric and thermophysical properties of sintered ODP2 (open symbols) and
ODP3 (filled symbols) samples as a function of temperature. (a) Seebeck coefficient, S. (b)
Electrical conductivity, σ. (c) Thermal diffusivity, α. (d) Averaged experimental values of the
specific heat, Cp, measured on ODP2 and ODP3 samples (black squares) with the corresponding
fitting curve (continuous line); on the basis of the fitting result, specific heat is extrapolated in a
larger temperature range (dashed line). (e) Thermal conductivity, κ; κtot and κel represent the total
and electronic thermal conductivities, respectively; the inset shows the difference between κtot and
κel as a function of the inverse of temperature: the deviation from linearity at high temperature
indicates the onset for bipolar conduction. (f) Left hand side (diamonds): power factor, S2σ. Right
hand side (circles): figure of merit, ZT.
21

ACCEPTED MANUSCRIPT
Figure 5. Thermoelectric properties at 673 K as a function of the Yb content in the Co4Sb12 cell. (a)
electrical conductivity, σ. (b) Seebeck coefficient, S. (c) lattice thermal conductivity, assessed as
κtot-κel. (d) figure of merit, ZT. The squares, diamonds and triangles represent data literature (the
numbers indicate the corresponding references), while the stars represent data from this work. The
legend for the different symbols indicates the method how the actual Yb content was determined
(XRD, EDS/EPMA, nominal composition).
22

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
•
•
•
Faster Yb filling in as quenched of Co4Sb12 due to larger grain boundary diffusion
Multimodal particle size distribution allows better sintering of the powders
Phonon scattering due to Yb filling compensates enhanced electrical conductivity