For example, the closest (non-sun) star to Earth, Proxima Centauri, orbits Alpha Centauri A and B with a semi-major axis of ~8700AU, or ~0.14 light years. It's orbital period is more than half a million years, it's orbital speed is ~500m/s, and the gravitational acceleration that keeps it in orbit is just 0.17nm/s².
That 0.17nm/s² is tiny, but it really adds up over half a million years.
True. But we don’t have half a million years to make an observation. 0.17nm/s² over a year only adds up to a displacement of 85 km, and now we’re looking for a relatively small discrepancy in this expected movement? The radius of the sun is about 696000 km, for comparison.
I guess it's quite easy to measure periapsis and apoapsis even for stars or planets that have a nearly circular orbit.
I couldn’t find the software, so I couldn’t check a bit deeper.
That link came from the paper. Zenodo.org is a open science site curated by NASA.
The analysis is interesting. He is using Gaia space observatory data to identify "nearby" binary systems, filtering out those with a high likelihood of being compromised by additional bodies in the system, and those for which the bodies have poor mass estimates, then measuring the orbits of the remaining 26k systems that can be confidently projected to three dimensions and aren't too eccentric, etc. His results are that below some small level of gravitational acceleration (the stars are far apart with long orbital periods,) the orbits predictably deviate from Newtonian gravity, and that due to the small size of binary systems Dark Matter cannot be a factor.
It's a Nobel prize winner if correct. Good old fashioned falsifiable Science: future data sets and peer review of the work will tell us.
Too bad Gaia runs out of fuel in two years. The longer the data set the more confidence can be achieved in this sort of work. It's interesting that only the most recent of the so far three data sets from Gaia provides the data necessary (tangential and radial velocity,) to do this.
Er, it's actually maintained by CERN and funded by the European Union. NASA does somehow participate in it as part of its "Transform to Open Science" initiative.
No offense, but Zenodo is a CERN project. [0]
"Transform to OPen Science (TOPS) is a $40 million, 5-year mission, led by NASA's Science Mission Directorate's Open-Source Science initiative. Within the TOPS mission, NASA is designating 2023 as the Year Of Open Science, a community initiative to spark change and inspire open science... Curated by: nasatransformtoopen"
I don't know. Even if there are equivalent capabilities, it's a different instrument, so the data will not simply slot in as the next tranche of observations.
Aside from the question of whether the code is semantically/notionally correct, what about numerical precision issues? Python floats are 64-bit IEEE doubles. I've read through enough "floating point representations are fraught with peril" articles in the past to be default-skeptical of code written by people who haven't already run into--and addressed--computational issues specifically with them.
I've never used numpy or done any hard mathematical analysis either in python or anything else. Does numpy somehow change the float type in python to use some other representation than doubles? How much precision is in a double? Is there "enough" for these astronomically large(/small) calculations? I guess a 52-bit mantissa can represent 2^52 = 4,503,599,627,370,496 ~= 10^15.65356. Is that ~"15 and a half digits" of precision? How does error propagate through all the calculations?
edit: I don't want to just poke holes in this person's research. I'm crossing my fingers and hoping it's all correct both for their sake and for mankind's.
What is the famous physics dude's quote. "If you need statistics redesign you experiment"? Or was it "math"?