It is unclear how very close binary stars form , given that during the pre-main-sequence phase the component stars would have been inside each other .
One hypothesis is that they formed further apart , but were brought in closer after formation by gravitational interaction with a third member of the system .
If so , all close binaries should be members of triple ( or higher-order ) systems .
As a test of this prediction , we present a search for the signature of third components in archival spectra of close binaries .
In our sample of 75 objects , 23 show evidence for the presence of a third component , down to a detection limit of tertiary flux contributions of about 0.8 % at 5200 Å ( considering only contact and semi-detached binaries , we find 20 out of 66 ) .
In a homogeneous subset of 58 contact binaries , we are fairly confident that the 15 tertiaries we have detected are all tertiaries present with mass ratios 0.28 \lesssim M _ { 3 } / M _ { 12 } \lesssim 0.75 and implied outer periods P \lesssim 10 ^ { 6 } { d } .
We find that if the frequency of tertiaries was the same as that of binary companions to solar-type stars , we would have expected to detect about 12 tertiaries .
In contrast , if all contact binaries were in triple systems , one would expect about 20 .
Thus , our results are not conclusive , but sufficiently suggestive to warrant further studies .