Indeed, stars don’t age (and die) in the same way. While some of them (with masses larger than 10 solar masses) end their lives with a huge explosion, the vast majority (approximately 98%) of the stars in the Universe die out gradually, getting progressively cooler and less luminous because no source of energy is active in their interior. In this late stage of their lives, the stars are called “White Dwarfs” (hereafter WDs). This will be also the final destiny of our Sun.
In all astrophysical textbooks and in all academic courses, a WD is described as the former “naked” core of a star that, at the end of its life, has lost its envelope. Any thermonuclear activity is permanently concluded in these objects, and they are destined to just become cooler and cooler with time. In fact, they have no possibilities of producing energy, neither through thermonuclear reactions, nor by gravitational contraction (their density is so high that they are incompressible and can just cool down at constant radius). Thus, the WD evolution is generally described as a pure cooling process, during which these stars become cooler and fainter as function of time. As a matter of fact, all the WD are expected to cool down and dim with essentially the same rate, and the relation between their age and their luminosity/temperature is so stringent that astronomers derive the WD age from their (dimming) luminosity. Indeed, the WD cooling rate has been adopted as cosmic chronometer to constrain the age of several populations of our Galaxy, including the disk, globular and open clusters.
The discovery published today in Nature Astronomy substantially changes this well-established scenario. In fact, by using very deep high-resolution images obtained with the Hubble Space Telescope (HST) in two “twin” stellar systems (they are famous for being very similar in terms of total mass, metal content, structure, age, etc.), we have discovered that not all White Dwarfs age at the same rate: some of them age slower than others, because they still have a very thin residual layer of Hydrogen providing them with an extra source of energy that delays their dimming and cooling.
The story of this discovery starts in September 2019, when Jianxing Chen, the first author of the paper, arrived in Italy from China to start his PhD course in Astrophysics at Bologna University. Jianxing was selected and granted by the China Scholarship Council (CSC) to attend its PhD in Italy. He was interested in studying the physics of White Dwarfs, and I was happy to propose him a thesis on the properties of these objects in old star clusters. As starting point, I proposed to focus on a simple but intriguing case: the comparison of the WD observational properties in two well-known Galactic globular clusters (named M3 and M13) that are so similar in many aspects to be considered as a sort of “twins”. However, as in the human case, twins are similar but not identical, and can show a few specific differences. The most relevant and known difference in the case of M3 and M13 is the morphology of their Horizontal Branch (HB). This is the evolutionary phase during which stars produce energy by the thermonuclear fusion of Helium in their core. The distribution of stars on the Horizontal Branch depends on the mass of the envelope surrounding the stellar core (or, equivalently, on the total stellar mass, since the core has the same mass in all the stars). In turn, this sets the stellar surface temperature, because thin envelopes imply high surface temperatures (hence, blue star), while large envelopes mean relatively low surface temperatures (thus, stars with redder colours). The colour-magnitude diagrams (CMDs) of M3 and M13 clearly show that these twins have different Horizontal Branch morphologies: the HB of M13 hosts many blue stars, with higher temperature and lower mass than M3, while in M3 the distribution is peaked at much lower temperatures (redder colours), which indicate larger stellar masses. Because White Dwarfs are the latest stage before death, their properties depend on what happened earlier in their evolution, and M3 and M13 thus represent a perfect natural laboratory to test how fast White Dwarfs generated by different stellar populations cool down in time.
To perform the study, we analysed deep HST observations in the near-ultraviolet (UV). The UV approach to the study of stellar populations in old clusters offers the advantage that deep exposures do not incur in heavy saturation from the bright cool giants, which dominate the optical emission in these systems. We applied the so-called “UV- route”, a well-tested procedure optimized to search for hot faint stellar sources in crowded fields, that we are adopting since many years to the study of Blue Straggler Stars in dense stellar clusters. Briefly, the photometric analysis is first carried out by searching for the stellar sources in the near-UV images, then each source is forced to be fit also in the images at longer wavelengths. The application of this technique to M3 and M13 yielded an exquisite result: the Luminosity-Temperature Diagrams of these clusters are spectacular (see Figure 1). The superb quality of these diagrams provides a full view of the stellar populations in the two systems, allowing a clear definition of all the evolutionary sequences, including the brightest portion (the upper 5 magnitudes) of the WD cooling sequence. From the comparison of these sequences in the two clusters comes the big surprise: the WD cooling sequence of M13 appears to be more populated than that observed in M3.
To make the comparison between the WD cooling sequences of the two clusters as straightforward as possible, we shifted the CMD of M3 to match that of M13. We found that only a shift in magnitude of −0.55 is required, providing an impressive match of all the evolutionary sequences, and further confirming that the two systems have approximately (within a few 0.1 Gyr) the same age. At this point the two CMDs were fully homogeneous, and we could focus on the properties of the WD cooling sequences. To be conservative we also excluded from the WD sequences possible interlopers and stars affected by large photometric errors. Thus, two clean sequences have been finally selected, and the resulting CMDs (after alignment), zoomed on the WD region, are shown in Figure 2.
The difference was still there.
At this point, before reaching any conclusion, we explored a possible source of bias: the level of completeness of the two samples. Indeed, this is a very insidious problem when comparing star counts in crowded regions. The evaluation of the completeness of a photometric sample requires extensive simulations using artificial stars. Briefly, artificial stars of known magnitude and colours are added to the image and then the photometric analysis is completely repeated. The recovery rate of these stars provides the completeness level of the photometric catalogue. We created a list of artificial stars with magnitude and colours following the WD cooling sequence. The artificial stars thus generated were then added to the real images, and the entire photometric analysis was repeated following exactly the same steps. To avoid artificial crowding, the added stars were placed into the frames in a regular grid of 23×23 pixels, each cell containing only one artificial star during each run. The procedure was iterated several times and more than 150,000 artificial stars have been overall simulated in the entire field of view sampled by each cluster. Jianxing spent two months in performing detailed artificial star experiments. But at the end we were able to accurately determine the level of completeness of the observed samples at different levels of magnitudes and different distances from the cluster centre. The result was really intriguing: the impact of incompleteness turned out to be very marginal: the global correction to the adopted samples is smaller than of 15%, with a final completeness-corrected populations counting 467 in M13, and 326 in M3.
The difference in the number of WDs in the two clusters therefore was real! The brightest portion of the WD cooling sequence of M13 turned out to be much more populated than that of M3 in the same magnitude range.
The next question obviously has been: Why?
The WDs in M13 are cooling down more slowly than those in M3? Could be this difference linked to the difference detected in the HB morphology of the two clusters?
To answer this question, the discussion with two major theoretical experts of WDs and stellar evolution was fundamental. Indeed, some models for slowly cooling WDs had been computed, suggesting that the reason for the slowdown was the presence of a very thin residual hydrogen-rich shell around the dense core of the WD: although very thin (of the order of 0.0001 solar masses), this hydrogen layer allows quite thermonuclear burning, an active energy source that slows down the cooling process, producing a class of “slowly cooling WDs”. They look similar to their normal sisters but have significantly longer cooling times.
At this point, the physical mechanism was likely identified. But how can it be possible that the mechanism is active in M13 and not in M3? Which is the event responsible for the presence of a residual hydrogen envelope around the extinguished core of WDs, or the lack thereof?
This event is the so-called “third dredge-up”, a mixing process occurring during the Asymptotic Giant Branch (AGB), the evolutionary phase immediately following the HB and preceding the WD stage. This process efficiently mixes the material present in the envelope of AGB stars, bringing most of the hydrogen deep into the interior where it is then burned. As a consequence, the star arrives to the WD stage essentially exhausted, with no residual hydrogen to burn. By lacking any possibility of extra-energy source, the WD just cools down as usually thought. However, as described above, some star arrives on the (blue side of the) HB with lighter mass. And the mass can be so small that the third dredge-up cannot occur. By skipping this critical event, the star reaches the WD stage with a residual hydrogen envelope thick enough to allow stable thermonuclear burning, and this provides the WD with an extra-energy production that delays its ageing.
At this point the scenario was clear: we found the first observational manifestation of stable hydrogen burning in cooling WDs!
The observations also provided the first empirical evidence of a causal connection among different phenomena occurring in the final evolutionary stages of low mass stars: the mass of the star during the HB stage affects the occurrence or not of mixing processes during the subsequent (AGB) phase; in turn, this determines the quantity of residual hydrogen envelope that a WD can have at its disposal during the following evolution, allowing or not stable thermonuclear burning that slows down the cooling. In practice, if most of the stars have a very small mass during the HB (as in the case of M13, where the HB morphology shows a long extension to the blue), they will experience no mixing processes during the AGB; hence, they will have a massive enough residual hydrogen envelope to allow stable thermonuclear burning during the WD stage, thus appearing as slowly cooling WDs. The same cannot happen in M3 where, according to the HB morphology, stars have larger masses (thus, all experience the third dredge-up).
This discovery changes the concept of White Dwarf itself, imposing to update its definition in astrophysical textbooks: “A White Dwarf is the naked core of a star that, at the end of its life, has lost most of its envelope and is fated to a progressive cooling at constant radius. However, depending on the size of the residual hydrogen-envelope, as set by its previous evolution, some stable thermonuclear burning can still occur on its surface, with a consequent extension of the cooling time.”