Recent measurements have shown that the inorganic clathrate Sr8Ga16Ge30 has
good thermoelectric properties. This discovery has caused intense experime
ntal activity to synthesize and test other compounds in this class. It has
been conjectured that clathrates may be good thermoelectrics if they satisf
y several conditions. The Sr atoms, trapped inside the clathrate cages, sca
tter phonons efficiently, leading to low thermal conductivity. Electric con
ductivity takes place mostly through the clathrate frame and the conduction
electrons are not scattered by Sr vibrations. The compounds, being made of
atoms that are semiconductors in the solid state, may have a high Seebeck
coefficient. There has been no direct evidence, experimental or theoretical
, for this scenario. By performing density functional calculations we show
that these ideas are correct. The Sr atoms are weakly bound to the cage and
do undergo large-amplitude motion. An analysis of conductivity shows that
the largest contribution comes from a band in which the electrons are locat
ed on the clathrate frame. Bands originating from the Sr atoms contribute l
ittle to conductivity. There is very little charge transfer between the Sr
atoms and the frame, and as a result, Sr vibrations are weakly coupled to t
he conduction electrons. The calculated Seebeck coefficient is in reasonabl
e agreement with the measured one. We find that it is strongly affected by
the positions of the Ga atoms in the frame and by doping. (C) 1999 American
Institute of Physics. [S0021-9606(99)01431-2].