# For binary stars, what is the average semi-major axis?

Note I'm exclusively considering binary stars, not 3+ star systems.

Binary stars come in all extremes. Wikipedia says that their orbital periods can be a few hours, a few days, or hundreds of thousands of years.

Is there any data, however, on the average semi-major axis of their orbits? I'm trying to get an idea of how far apart the stars are from each other, and would like an average across all measured binaries in the Milky Way.

• What sort of average:- mean, median? The trouble with mean would be that the distribution of semi major axes would be highly skewed with a long tail, so the mean might not be the best or most stable. Oct 8, 2023 at 13:42
• There's also going to be lots of difficulty with sample bias. Binary stars are easier to observe when they are widely spaced, many close binaries are only known to be binaries by spectroscopy. orbital determination of these requires extensive follow-up observations, and hence won't be done for many such stars. This results in observational bias But see browse.arxiv.org/pdf/1401.6827.pdf (and other works of that author) for more about the statistics. But let me throw a number out: 16AU: from "lognormal distribution of period with median 100yr" and accounting for lower mean mass. Oct 8, 2023 at 16:20
• @JamesK I think your comment with the reference to this paper is worth an answer - even without much or any expansion. It has the direct answer to the question asked as quote from the paper's §5 summary - and it shows, that the question is not well-posed due to the nature of the distribution and the answer will mis-lead, and shows how one can do better (by following what the author of the paper does). Regardless: thanks a lot for making me aware of that very interesting paper! Oct 8, 2023 at 21:06
• The answer is mass-dependent. Oct 8, 2023 at 23:08
• different but related: What is a thermal distribution of eccentricities?
– uhoh
Oct 9, 2023 at 22:43

• Yeah there's always issues with distribution and biases. Like imagine if the sample size is 100, and the first half have a value of 3, and the last half have a value of 80. Then even taking the median is not gonna give a good idea of the "average" as you prolly imagine it. However there is one bias I'm concerned about more than others: Observation bias. It is far easier to observe wide binary pairs than close ones. Can you at least give me a concrete number for the difference between "wide" and "narrow"? Where is that threshold? 5 AU? 10 AU? Something else? Oct 9, 2023 at 3:33