Hi, is there any proper meaning to the sentence, what are you asking?

virat1ew77 wrote:

Hi, is there any proper meaning to the sentence, what are you asking?

Who are you talking to?

Hi all,

I wonder how is implemented star classification in SpaceEngine. I just found today a weird case today: HD115678 is a M-type star (which is right, in accordance with Simbad). But also, it is a high-luminosity star (absolute magnitude of ~0, even less according to Gaia DR2). It is therefore a giant. However, SpaceEngine categorizes it as a Dwarf, and shows a main-sequence star (e.g., small granules). I suppose this kind of weird classification happens for other stars.

Including a HR-diagram type classification (which is kinda easy as we know the Teff and the luminosity of the star) would prevent this kind of weird thing to happen, and increase the realism of the software. Is it planned in future versions ?

I am sorry if this question has already been asked, I did not manage to find the right keywords to find it.

issun-sensei, I often met things like that. Stars that meant to be giants appeared as dwarfs and vice versa, very bright ones were dim and so on. Obviously the collected star data is corrupted somehow or the engine doesn't work with it properly.

issun-sensei, HD 116578 aka HIP 64935 is loaded by SpaceEngine from the main catalog, an HIPPARCOS csv file wich contains no temperature data (see data/catalogs/catalogs.pak/stars/HIPPARCOS.csv), SE star solver then is estimating data not directly provided by catalog scripts with algorithms, it's a fairly complex task so the sky in SpaceEngine is certainly filled with inconsistencies.
The only way to fix this is compiling fresh data with new star catalogs via scripts (.sc files) containing more info, anybody can do this using the manual and existing scripts as example, in fact many in the community have provided over time fixes or specific catalogs for their area of interest.

Thanks for your answers. After checking the HIPPARCOS.csv file, I can see that we basically have the distance, apparent magnitude and spectral class of the star. This allows to retrieve the luminosity and temperature of the star (with some assumptions of course, but that are not shocking at all for such software). With that data, we can get the HR diagram position of the star... Which would allow to instantly categorize HD115678 as a giant! There's no need for specific other input, and the algorithm to do so seems quite simple. Of course, I don't know the software in depth and I can possibly miss something that would make such algorithm impossible or difficult.

In my opinion, respecting what the HR diagram tells us on the evolution of the star is quite imporant. It would greatly improve the consistency of the SE universe.

issun-sensei wrote:

Source of the post In my opinion, respecting what the HR diagram tells us on the evolution of the star is quite imporant. It would greatly improve the consistency of the SE universe.

It would indeed, but there are several reasons why this is currently an outstanding scientific endevour. It would result in publishable papers and not only on SE improving its quality. Let me explain why this is so complicated with the example of ESA's Gaia mission.


Gaia has catalogued almost 2 billion stars by now. The result of this mission include distance estimates for more than 1.5 billion of them, and thus are potentially the future of the non-procedural part of SE in the Galaxy. We only need to know their classification and we will have enough information to have them inside SE with more or less consistent and realistic parameters. But for that the best thing is to gather their spectra. But we can't. Why? Taking spectra requires time, you need lots of photons so you can sort them out in terms of their colour, and have enough photons in each slot of the spectrum for a consistent signal. Gaia has the largest catalog of stars but it has it because it spends very little time observing each one (in fact it observes around 460 stars each single second). A modern telescope could gather spectral data for a star within minutes but if you do that you have to sacrifice the numbers since it would take thousands of years to gather a similar number of detections as Gaia. Currently there are almost a million times less spectra than stars are known, and this can hardly improve significantly for the moment.

But hey, maybe we don't need the entire spectra to classify a star, maybe we can make it only by using photometry (which Gaia has). We could use an HR diagram to discriminate between different populations, as you said. That seems fairly good in theory, but in practice it is not. Take for example OB stars; you have stars that look red and still are blue massive stars, but they look red because they generally are far away and interstellar dust filters their blue colors (it is difficult to tell the difference between a nearby small hot star like sub-dwarfs and a massive Ob star and a red supergiant not as far). In theory distance estimates should make this easier to disentangle (but distance estimates are not accurate enough for large distances and interstellar dust can present overdensity fluctuations across the sky that make the effort really unworthy). So in general we are unable to distinguish OB stars from some red supergiants. To accomplish that we need very accurate extinction laws (rules that tell us how the different frequencies of light are attenuated by the Galactic dust), but we currently lack any extinction laws that might be good enough for this goal. We still have huge unknowns in terms of what properties the dust has (grain size, different chemical mixtures and proportions, and density at every distance and direction in the sky). Worst of all, our knowledge on the galactic extinction laws is built by measuring the spectra of stars and reconstructing how the dust in the line of sight has changed that spectral footprint, by knowing before hand about stars with similar spectral features that are nearby, so we end up needing lots of spectra all over again, to solve the main problem.

If we knew the exact form of the Galactic extinction laws (or at least with enough accuracy), we could reconstruct the real position of the vast majority of stars in the Gaia HR diagram as they would appear if no extinction was involved. Then we could apply some classification criteria. But for now this is impossible for the vast majority of stars. Also notice that there are more stars farther than closer (following the square-law), because "closer to you" is a much smaller volume of space than "farther from you", this means that the majority of stars are behind some non-negligible level of interstellar dust extinction and thus photometry really needs to be accompanied by accurate Galactic extinction laws, to allow for classification.

And if that weren't enough, there are other issues. For example, stars with discs around them are reddened (because of the visible tail of the infrared emission), that is the case for Be stars for example, which means you would classify them wrongly. We are also not considering the problems related to binary systems. Some stars that Hipparcos includes are not single but binaries that weren't separated at the time. If you have a blue star and a red star in a binary system that you can't separate with Gaia or any telescope and rely only on photometry, you would think you are dealing with a single star with a classification that doesn't match either of the two components whatsoever. The HR diagram can't help you on this, and a lot of work has to be done first to discover what are the binaries. After you discover what are the unresolved binaries of Gaia (which is a huge scientific task on its own), you can do two things; either you ignore all those cases (but they are probably something like 30% to 40% of the sample considering how common binaries are) or you take spectra of the system and separate the two spectra by doppler shifts between the lines of one component and the other (something that relies on spectra again and on the convenience of the orbits of the system being not perpendicular to the line of sight).

In the end is a very complicated issue. It would be solved if we had spectra for all the stars, but we currently don't have telescopes with enough aperture to gather so much photons as quickly as needed to generate high-resolution spectra of each and every star of the large catalogs in a human lifetime. We also don't have the angular resolution needed to separate the huge amounts of binary systems that might escape us right now. But one day we might have that, and when we reach that goal, we won't only have the stars correctly classified but also chemical footprints particular of each one of these stars, their ages, and much much more.

What you currently see for Hipparcos en SE is the best we can do with Hipparcos. In general they are more or less accurate but you will find lots that are not, and if there are not specific scientific articles taking spectra for specific stars you won't have any basis to change the way it is. We can only improve one by one right now. Since that is impossible with the 2 billion Gaia stars we really need some automated way of dealing with it. Until then you have to know that HR diagrams are not the solution.

I'd like to see SpaceEngine create more procedurally generated interacting galaxies, colliding galaxies and small irregular galaxies as satellites of larger galaxies (like the Magellanic Clouds). For that matter, since most irregular galaxies are just gravitationally disrupted spiral galaxies, SE shouldn't be generating them as uniform clouds: they should have a distinctive ellipsoidal or tadpole-shaped structure. It might also be possible to have procedurally generated ring galaxies like the Cartwheel, where the ellipticity of the ring and the core, the degree to which the core is off-center, and the angle between the ring and the core can be randomized. A smaller galaxy that caused the collision would also need to be generated nearby.

FastFourierTransform wrote:

issun-sensei wrote:

Source of the post In my opinion, respecting what the HR diagram tells us on the evolution of the star is quite imporant. It would greatly improve the consistency of the SE universe.

It would indeed, but there are several reasons why this is currently an outstanding scientific endevour. It would result in publishable papers and not only on SE improving its quality. Let me explain why this is so complicated with the example of ESA's Gaia mission.


Gaia has catalogued almost 2 billion stars by now. The result of this mission include distance estimates for more than 1.5 billion of them, and thus are potentially the future of the non-procedural part of SE in the Galaxy. We only need to know their classification and we will have enough information to have them inside SE with more or less consistent and realistic parameters. But for that the best thing is to gather their spectra. But we can't. Why? Taking spectra requires time, you need lots of photons so you can sort them out in terms of their colour, and have enough photons in each slot of the spectrum for a consistent signal. Gaia has the largest catalog of stars but it has it because it spends very little time observing each one (in fact it observes around 460 stars each single second). A modern telescope could gather spectral data for a star within minutes but if you do that you have to sacrifice the numbers since it would take thousands of years to gather a similar number of detections as Gaia. Currently there are almost a million times less spectra than stars are known, and this can hardly improve significantly for the moment.

But hey, maybe we don't need the entire spectra to classify a star, maybe we can make it only by using photometry (which Gaia has). We could use an HR diagram to discriminate between different populations, as you said. That seems fairly good in theory, but in practice it is not. Take for example OB stars; you have stars that look red and still are blue massive stars, but they look red because they generally are far away and interstellar dust filters their blue colors (it is difficult to tell the difference between a nearby small hot star like sub-dwarfs and a massive Ob star and a red supergiant not as far). In theory distance estimates should make this easier to disentangle (but distance estimates are not accurate enough for large distances and interstellar dust can present overdensity fluctuations across the sky that make the effort really unworthy). So in general we are unable to distinguish OB stars from some red supergiants. To accomplish that we need very accurate extinction laws (rules that tell us how the different frequencies of light are attenuated by the Galactic dust), but we currently lack any extinction laws that might be good enough for this goal. We still have huge unknowns in terms of what properties the dust has (grain size, different chemical mixtures and proportions, and density at every distance and direction in the sky). Worst of all, our knowledge on the galactic extinction laws is built by measuring the spectra of stars and reconstructing how the dust in the line of sight has changed that spectral footprint, by knowing before hand about stars with similar spectral features that are nearby, so we end up needing lots of spectra all over again, to solve the main problem.

If we knew the exact form of the Galactic extinction laws (or at least with enough accuracy), we could reconstruct the real position of the vast majority of stars in the Gaia HR diagram as they would appear if no extinction was involved. Then we could apply some classification criteria. But for now this is impossible for the vast majority of stars. Also notice that there are more stars farther than closer (following the square-law), because "closer to you" is a much smaller volume of space than "farther from you", this means that the majority of stars are behind some non-negligible level of interstellar dust extinction and thus photometry really needs to be accompanied by accurate Galactic extinction laws, to allow for classification.

And if that weren't enough, there are other issues. For example, stars with discs around them are reddened (because of the visible tail of the infrared emission), that is the case for Be stars for example, which means you would classify them wrongly. We are also not considering the problems related to binary systems. Some stars that Hipparcos includes are not single but binaries that weren't separated at the time. If you have a blue star and a red star in a binary system that you can't separate with Gaia or any telescope and rely only on photometry, you would think you are dealing with a single star with a classification that doesn't match either of the two components whatsoever. The HR diagram can't help you on this, and a lot of work has to be done first to discover what are the binaries. After you discover what are the unresolved binaries of Gaia (which is a huge scientific task on its own), you can do two things; either you ignore all those cases (but they are probably something like 30% to 40% of the sample considering how common binaries are) or you take spectra of the system and separate the two spectra by doppler shifts between the lines of one component and the other (something that relies on spectra again and on the convenience of the orbits of the system being not perpendicular to the line of sight).

In the end is a very complicated issue. It would be solved if we had spectra for all the stars, but we currently don't have telescopes with enough aperture to gather so much photons as quickly as needed to generate high-resolution spectra of each and every star of the large catalogs in a human lifetime. We also don't have the angular resolution needed to separate the huge amounts of binary systems that might escape us right now. But one day we might have that, and when we reach that goal, we won't only have the stars correctly classified but also chemical footprints particular of each one of these stars, their ages, and much much more.

What you currently see for Hipparcos en SE is the best we can do with Hipparcos. In general they are more or less accurate but you will find lots that are not, and if there are not specific scientific articles taking spectra for specific stars you won't have any basis to change the way it is. We can only improve one by one right now. Since that is impossible with the 2 billion Gaia stars we really need some automated way of dealing with it. Until then you have to know that HR diagrams are not the solution.

I remember talking to Vlad about the H-R diagram and he liked the idea of including a small H-R diagram for stars in the field of view and when you mouse over individual stars it would highlight its position on the diagram and you could also mouse over a speck on the diagram and it would highlight the star.  This would also be a quick way of finding sunlike stars (or extreme stars) in your field of view.

Also to address your point about not having enough aperture on telescopes, cant we substitute long exposure cameras for that?  The cameras can collect many photons over long periods of time to compensate for not having enough aperture.

In addition to your hopeful outlook of attaining spectra of more stars when we have better telescopes, perhaps by then we'll also be able to find the chemical and atmospheric composition of planets that orbit them.

A-L-E-X wrote:

Source of the post I remember talking to Vlad about the H-R diagram and he liked the idea of including a small H-R diagram for stars in the field of view and when you mouse over individual stars it would highlight its position on the diagram and you could also mouse over a speck on the diagram and it would highlight the star.  This would also be a quick way of finding sunlike stars (or extreme stars) in your field of view.

Yeah, that would be a nice educational feature.

A-L-E-X wrote:

Source of the post Also to address your point about not having enough aperture on telescopes, cant we substitute long exposure cameras for that?  The cameras can collect many photons over long periods of time to compensate for not having enough aperture.

You can have longer exposures to acquire spectra (even if the signal to noise ratio wouldn't be as good), but then you must point your telescope to a region of the sky for longer time-spans and thus you have to give up on having spectra of lots of stars in a reasonable amount of time. My point was exactly that! there's currently a compromise between the amount of stars you can classify and the time taken to classify them using spectroscopy. Gaia has observed almost 2 billion stars in a few years, that would be an impossible number if you wish to acquire spectra precisely because it would take time and thus you would have two options: wait for a several millennia until all the stars are observed or have a modest tiny fraction of the stars classified. In either case we can't include all Gaia stars in SE without gross inaccuracies for now, because even if we have their brightness and distance more or less constrained we lack the tool to correctly classify them with Gaia alone. But there's hope since in the next Gaia releases (in a few years) we will have millions of low-resolution spectra (for the brightest stars in the catalog), which is not bad at all.

A-L-E-X wrote:

Source of the post In addition to your hopeful outlook of attaining spectra of more stars when we have better telescopes, perhaps by then we'll also be able to find the chemical and atmospheric composition of planets that orbit them.

I've always wondered what would could be done if we had 1000 times more research funding for a century. Imagine having 6 Gaia like telescopes lunched in 3 orthogonal directions (each pair in both ways) to the distance of Eris. We could get instantaneous parallaxes (not the ones where we have to wait for half a year and then decouple the true proper motion of the stars from the parallax motion in the sky, but just pure and instantaneous parallax) and thus we could observe the true proper motion with extreme accuracy. Imagine also that these are not regular Gaia missions, but UVI (Ultraviolet-Visible-Infrared) cameras. The Gaia catalog would not only be a thousand times more precise (allowing us to map the entirety of the Galaxy), but also 100 times more entires could be added. The Milky Way dust wouln't be a problem at all and we could probe the other side of the Galaxy. Even more, we could measure distances to individual stars in hundredths of nearby galaxies using parallax (we are currently able to measure the distance to Andromeda and the Milky Way satellites by averaging the parallaxes of many stars), That in turn would allow to better understand Cepheids and thus better measure distances to far away galaxies, making cosmology an entirely different things from what it is now. We could detect gravitational waves of nearly any frequency and on a daily basis just by examining the slight deviations of light rays of background stars. Imagine even further; imagine these Gaia's to have 50 cameras instead of 1, pointing to different patches of sky to cover a whole-sky field of view with the same resolution. They could observe continuously each and every object of their 100 billion catalog. That would allow to detect all kind of variable stars (many never seen before), detect basically all transiting planets in the sky (billions basically, including hundredth thousand terrestrial planets around Sun-like stars in just a few years of observations), flaring events (that would allow to understand precisely what are the habitability prospects in other systems), millions of objects in the far reaches of the Solar System (perhaps the first ever direct evidence of an Oort Cloud) just by watching transits, millions of rogue planets (that thanks to the deployment of these 6 Gaia could allow for distance measurements and follow-up). These Gaia's could "turn on spectroscopic mode" (the prism could move on the optical path) for a few minutes every hour and thus we could learn not only to classify stars but to understand the particular chemical footprint of each one of them, their changing behavior, their multiplicity and the orbits of the multiple systems, the presence of many millions of additional plats that can only be revealed by radial velocity measurements, the composition of their atmospheres and tracking seasonal changes on them. We could understand the kinematics of the Galaxy in such a way that would allow us to reconstruct the entire history of it, its rotation curve through time, the interaction between star-forming regions, the collisions and mergers with smaller galaxies, the distribution of dark matter in 3D, the motion of the Solar System and thus past encounters and possibly even the siblings of the Sun, the chemical reservoir where our own planet came from, now sparse and mixed throughout the galaxy, it would in turn probably explain the origin of life.
I've wondered a lot about such a powerful system. It would require pharaonical amounts of efforts from the human species and probably the entire GDP of a developed country, but after that century we would have become semi-gods in terms of omniscence, compared to what we are now.
Sorry, this is getting off-topic, so next time let's try to discuss this in another thread.

Oh I love this discussion!  What I find so vexing about the dust regions is there are large parts of our own galaxy that we simply dont know very well, including the center!  I wonder if we could discover more massive stars than our theories even thought possible.  And more luminous stars than we thought could exist!  I wonder if we could also discover habitable planets inside globular clusters (always a fantasy for me since the skies visible on these planets would be absolutely spectacular!)  With the cameras arranged as you propose we could also locate spectroscopic binaries which cannot be separated optically.  And have a much more accurate 3D model of our own galaxy.  When you mentioned mergers I thought of actually being able to witness black hole mergers, what a phenomenal site that must be!  And detecting the first signs of gamma ray bursts just before they are about to happen!  And the original stellar uterus from where the sun originated may point us in a direction where there is other life similar to that on our planet!  I envision a Star Trek like journey to the center of the galaxy (except in this case it would probably be closer to the outer edge where life is more likely to find hospitable conditions.)  Right now what we have is only accurate out to about 100 light years correct?  

I guess for the short term few years from now we'll have to settle for the low resolution models (still a few million stars) which is far better than what we have right now.  And that might be for the best, because what you suggest would require extremely rigorous computers to be able to map out on our end, I dont think any current high end systems can even handle the vast amount of data and graphics we've been talking about.  Hopefully our computer technology advances will be in sync with our astronomical discoveries so we will actually be able to utilize the incoming mountain of data when it does start coming in.

issun-sensei wrote:

Hi all,

I wonder how is implemented star classification in SpaceEngine. I just found today a weird case today: HD115678 is a M-type star (which is right, in accordance with Simbad). But also, it is a high-luminosity star (absolute magnitude of ~0, even less according to Gaia DR2). It is therefore a giant. However, SpaceEngine categorizes it as a Dwarf, and shows a main-sequence star (e.g., small granules). I suppose this kind of weird classification happens for other stars.

Including a HR-diagram type classification (which is kinda easy as we know the Teff and the luminosity of the star) would prevent this kind of weird thing to happen, and increase the realism of the software. Is it planned in future versions ?

I am sorry if this question has already been asked, I did not manage to find the right keywords to find it.

If you know how to change it in space engine's catalogs, you can export the properties of the star and change the type of the star from "V" to "M0 Ia" 

There is a small detail I wish I could modify in SpaceEngine, and so far I was not able to. I have been using it just for a few days, so maybe I was clumsy about it. It is the intensity of lines in the sky, specially those for constellations. I wish I could make them really faint, as it is when I switch the constellation lines on, they are so bright that stars get eclipsed by them. I wish I could tweak those lines to make them barely visible. I found no menuy in the settings for this, my hope is that somewhere there is a config file where those intensities could be modified.

Yes I want a setting like this too.  In Starry Night I change the color of the lines to make them dark grey which makes them barely visible (and very thin)