(I hope that this is an appropriate forum for this message; if not, please move it to a better place)
One of my students is doing an observational project on the contact eclipsing binary V527 And. It has magnitude V = 10.8 or so, with a period of 16.8 hours. Based on its (B-V) = 0.78 value, the nearly equal components are probably late G stars. If so, their orbital velocities should be around 40 km/s.
We are measuring a multicolor light curve and looking for (O-C) variations, but we could make better models if we knew the orbital speeds of the stars. If anyone is willing to try acquiring some spectra for us, we’d appreciate it very much, providing full credit in my student’s undergrad thesis. There’s a small chance we might try to publish it in IBVS or JAAVSO; if so, we’d happily offer co-authorship to anyone who can provide spectroscopy.
Well, in order to measure the velocity of around 40 km/sec well, one would need a velocity resolution of around, say, 5 km/sec. If one is observing at a wavelength of 5000 Angstroms, that implies a wavelength resolution of 0.08 Angstroms.
That is the resolution per pixel you would need, but if you are getting RVs from cross correlations with templates you can get that RV error with much lower resolution, depending on the S/N, wavelength coverage, and features in the spectrum.
You are asking for R~60,000 spectra of an 11th magnitude object that varies over ~17 hrs. For reference, we do this for RR Lyrae at this resolution down to that magnitude and we use a 2.7m telescope to get integration times of about 0.5 hrs with a usable S/N.
Additionally, if this is 2 G type stars with an orbital period of ~17 hrs, the orbital speed would be much higher - Earth’s orbital speed is about 30 km/s at a distance of 1 AU. Assuming this is a contact binary, their separation can’t be more than a few times the stars’ radii (a few hundredths of an AU). And if you get spectra that are double-lined, which I assume you would if the 2 components are so similar, then the distance between the lines would be twice as big.
The only complication for this might be the inclination angle of the orbit, but if they are eclipsing we are looking almost directly along the orbital plane, meaning the inclination would be small (negligible?) and you’d be seeing essentially the full orbital speed.
I think the time resolution is more important than the spectral resolution. You would want integration times for sure less than 10% of the orbital period, and probably less than 5%, which means less than 1 hr integration.
Thank you for your very informative message. You’re right – I made an error in my estimate of the orbital speed. It’s closer to 240 km/s, so measuring velocities to a precision of around 30-50 km/s would probably be sufficient. That means R ~ 6,000, quite a bit easier.
SEGUE was able to get RV errors of ~4 km/s with R~2000 spectra of relatively bright stars (see this paper). This was because of the high S/N and wide wavelength coverage. If you have a lot of features to match, each individual pixel doesn’t need to be at the resolution you need, you combine information from lots of pixels to get a better answer.
Do you know if this is a spectroscopic binary? If there are 2 lines that need to be resolved you’d need a higher resolution to do that. But at maximum RV difference (largest line separation) they’d be twice the orbital speed apart making that easier. So if you don’t know if it’s a spectroscopic binary, maybe providing an ephemeris so people can take a spectrum at maximum line separation would be beneficial to see if it’s double-lined and if it can be resolved with their setup (which would depend on resolution and integration time)?
I’ve looked for spectroscopic measurements, but haven’t found any so far, so I’m not sure if it’s a single- or double-lined spectroscopic binary; I’d bet on the latter, based on the lack of color change over the course of the light curve.
Ephemeris data: Based on observations made UT 2025 Oct 04, there system has minimum light at
T(min) = 2460953.83
The period is P = 0.70065. Adding one quarter of that period to the time of minimum above yields
T(quad1) = 2460954.01 + N * (0.70065)
and adding another half period for the other time of quadrature
T(quad2) = 2460954.36 + N * (0.70065)
The problem is, as I understand it, that this is likely to be a double lined binary, not a single lined so it would need sufficient resolution to split the lines and I suspect this is too faint to do this with sufficient resolution and SNR with a typical amateur setup of up to say 0.5m. (At least this is what my exploratory measurements with 0.28m and R~4000 suggest). Single lined are indeed much easier as you can measure shifts typically say 1/20th of the resolution without difficulty
Correction. the resolving power was R =6500 or 45km/s so perhaps just enough to split the lines. The SNR in 40min was just 13 though. The spectra cam be seen here
Cross correlation is very tolerant of low SNR. I was able to get ~2km/s 1 sigma uncertainty on mag 9-10 pulsating delta Scuti variables at SNR ~30 using cross correlation eg
so if single lined some sort of measurement may well be possible. Cross correlating the two consecutive spectra I posted gave a feasible shift of 5km/s ±~5km/s even at this low SNR. The weather prospects here though are not good for a long run to determine if any orbital component can be picked up
Cheers
Robin
I guess “low” and “high” are relative terms in a lot of ways
SDSS was able to get decent stellar parameters down to a S/N~10. The “high” S/N they mention in the SEGUE paper was S/N>30.
I think even if it is double lined you wouldn’t have to necessarily completely resolve the 2 lines to find a velocity difference that size, you could fit lines with 2 profiles, but it would be much more difficult and require a lot of lines probably to get reliable results.
Yes you can get good results with cross correlation even at very low SNR. (cross correlating my two consecutive SNR 13 spectra gave a credible 5km/s shift)
there are programs to disentangle two line spectra eg TODCOR
but they do need higher resolution and SNR than conventional single lined cross correlation so they have something to work on.
As it is a contact binary does that mean the stars will be tidally locked and the rotational line broadening will be comparable to the orbital velocity ? (essentially the system is not two discrete objects but essentially a rotating elongated blob so what we see is just a broadening and narrowing of the lines with phase)
