≡ Menu

The Lithium Clue to Exoplanets

The exoplanet watch among our readers is clearly in full operation, to judge from the number of backchannel messages I received about the latest work from HARPS (High Accuracy Radial Velocity Planet Searcher). The remarkable ESO spectrograph attached to the La Silla 3.6-meter telescope now offers evidence that Sun-like stars that host planets will show a sparser lithium signature than stars without planets. Says Garik Israelian, lead author of the paper now appearing in Nature:

“For almost ten years we have tried to find out what distinguishes stars with planetary systems from their barren cousins. We have now found that the amount of lithium in Sun-like stars depends on whether or not they have planets.”


All of this helps us understand our own star, whose low levels of lithium (140 times less than it should have had when formed) have long been apparent. The HARPS work draws on an analysis of some 500 stars, including seventy known to host planets. Ace planet hunter Michel Mayor calls the sample the best available to date “to understand what makes planet-bearing stars unique.” The majority of the stars hosting planets showed less than one percent of the lithium possessed by other stars. Most stars should possess roughly the same amount of lithium unless the element, produced just after the Big Bang, had been destroyed inside the star.

Image: Artist’s impression of a baby star still surrounded by a protoplanetary disc in which planets are forming. Using ESO’s very successful HARPS spectrograph, a team of astronomers has found that Sun-like stars which host planets have destroyed their lithium much more efficiently than planet-free stars. This finding does not only shed light on the low levels of this chemical element in the Sun, solving a long-standing mystery, but also provides astronomers with a very efficient way to pick out the stars most likely to host planets. Credit: ESO.

So what exactly is going on here? How does a planet disturb the matter in a host star to rearrange the chemical elements found within its outer atmosphere, where lithium can exist? The Israelian’s team points to loss of angular momentum caused by the presence of planets, causing host stars to spin less rapidly and allowing their atmospheres to mix more freely. Lithium in the stellar atmospheres would thus be drawn to the stellar surface and destroyed. But stellar ages could also account for the difference, as this article in Nature News points out, noting comments to that effect by Jorge Melendez (University of Porto, Portugal).

The Israelian team disputes that idea, noting that all the stars in its sample are more than a billion years old, and arguing that its findings cannot be coincidence. If they’re right, we may have an observational tool that could speed up the exoplanet search. The paper is Israelian et al., “Enhanced lithium depletion in Sun-like stars with orbiting planets,” Nature 462 (12 November 2009), pp. 189-191 (abstract).


Comments on this entry are closed.

  • Mike November 12, 2009, 14:47

    It sure raises a lot of questions.
    First off how can the HARPS team be certain the lithium abundant stars really are without planets? Wouldn’t the best you could say is the lithium abundant stars in their sample don’t have giant planets?
    HARPS is very advanced but there is still an unavoidable observational bias toward Jovian mass or super-earth planets,preferably in close orbit.
    For all anyone knows at this time there maybe a large percentage of exoplanet systems that lack large planets because we would not have been able to detect only Earth to Mars mass planets with the radial velocity method.
    Funny enough that might still allow for their assertion about faster rotating stars retaining more lithium.I think the host stars angular momentum would not be transferred as much to a family of exclusivly low mass planets as it would to a system containing large mass planets.
    The HARPS team may have discovered a tool alright,perhaps just not the one they think they’ve found.
    Hopefully I’ve not gotten too far over my head with speculations about exoplanet systems with out gas giants,ice giants or super size terrestrial
    planets. But until we get SIM or some version of TPF observing who can say for sure they don’t exist.
    After all,did anyone predict the existence of hot Jupiters?

  • andy November 12, 2009, 16:07

    One of the most interesting things that seems to be coming out of the study of extrasolar planets is that the hot-Jupiters (i.e. the easiest type of planet to detect) seem to have fairly specific preferences for the types of environments they occur, which are not necessarily shared with the planet population as a whole. For example, so far no hot-Jupiters have been found around M-dwarfs: for the most part, the Jupiter-class planets around such stars are out beyond the snowline. There is a strong bias towards high-metallicity stars: the planetary systems known around low-metallicity stars contains giant planets on long period orbits. The hot-Jupiters are also apparently rare in star clusters.

    The point is: with this kind of study, it is important to be able to distinguish between hot-Jupiters and the planet population as a whole, for which the statistics are rather less robust. Is an apparently “bad for planets” environment really bad for planets, or just for planet migration? As the detection methods begin to probe the parameter space of Jupiter- and Saturn-analogues, I suspect quite a bit of the ideas about where planets occur developed based on where hot-Jupiters are found will have to be modified.

  • Matt Metcalf November 12, 2009, 16:09

    I agree with Mike… we should not jump to conclusions about these results. They show an apparent correlation, but correlation != causation. Much more information is needed to determine the importance of this new knowledge.

  • FrankH November 12, 2009, 16:10

    Lithium abundance is a good indicator of a star’s age. A young star with a high lithium content and a hot Jupiter would poke a hole in their theory.

    I think overall metallicity is a far better indication of planet-friendly star system, and there are papers tying higher metallicity with hot Jupiters.

    Other than one highly suspect “planet” found in a globular cluster star, I don’t know of any low metallicity stars with known planets.


  • philw1776 November 12, 2009, 17:02

    Right Mike, given lack of sensitivity to terrestrial mass planets, absense of evidence is not evidence of absense. I do think they’re correct though.

  • AlastairMayer November 12, 2009, 17:26

    I’d be interested to see any lithium-abundance comparisons with binary (or higher multiple) stars, especially where the companions are close enough to exert a similar gravitation and angular momentum influence that a planetary system would.

    In particular, it’s interesting to note that – at least according to the first paper I found online – Alpha Centauri B has significantly reduced lithium (comparable to Sol’s) but yet Alpha Centauri A is relatively abundant in lithium (at least with respect to a Cen B and Sol). Since B orbits A at Saturn/Neptune distances, I would have expected a similar lithium-reducing effect even if A didn’t have planets. Perhaps its a late capture? Or perhaps the “lithium test” isn’t the clear indication of planets (or other companions) that we’d like.

    (Since I’m in the middle of writing a novel where both Alpha Cen A & B have terrestrial planets, I half-hope for the latter, although from a more practical standpoint I’d love an easy test for “worth checking further”. I wonder what the rate of false positives or negatives is.)

  • Ronald November 12, 2009, 19:05

    Surely, sufficiently high metallicity is now widely accepted as a prerequisite for planets, but lithium would be the *indicator* of the actual presence of them.
    I would really like to know whether lithium content could also be an indicator of the *type* of planets and planetary systems, or at least of total planetary mass, or just of any planetary presence.

    It is a fact that lithium content diminishes with stellar age, so this is probably a very crude method anyway.

  • tacitus November 13, 2009, 3:07

    The results from Kepler should provide a much clearer picture of the relationship between lithium abundance and planetary systems.

  • Ronald November 13, 2009, 6:33

    AlastairMayer November 12, 2009 at 17:26:

    Very relevant points about Alph Cen A and B !

    First of all, a binary (or multiple) star system would probably exert a similar effect, hence the lithium ‘trick’ probably wouldn’t work (reliably) for those.

    Secondly, why this difference in lithium between A and B? Does it indicate planets for B but not for B? Or is it an effect between A and B, in which case I’d rather expect A to have a lower lithium content.

    No, B is most likely not a ‘late capture’. ‘Exogeneous’ binaries are exceedingly rare, virtually all binaries (and multiple) are ‘indigenous’. For A and B this is also confirmed by estimated age and chemical composition (spectroscopically). I can imagine some very wide binaries, such as Zeta 1 and 2 Reticuli to be secondary gravitational captures, especially if this is confirmed by significant spectroscopic differences.

    What I would really like to know (is this in the paper? So far I have only seen an abstract) is whether there is a kind of abrupt ‘cut-off’ lithium content, a limit below which lithium is indicative of planets. This would then show up as a sudden steep gradient (slope) in a graph, which could not only be explained by age depletion (which would probably be more gradual).

    With regard to Mike: yes, I agree that so far this correllation can mainly be between lithium and close-orbit (sub)giants, for obvious observational/technical limitation reasons.

    Note the following quote from the abstract (** added by me for emphasis): “We find that the planet-bearing stars have less than one per cent of the primordial Li abundance, while about *50 per cent of the solar analogues without detected planets* have on average ten times more Li. ”

    I.e. the strong correllation works one-sided only: identified planet-bearing stars are consistently low lithium, but stars not yet observed as planet-bearing can be either high or low lithium. It is extremely unlikely that within the researched group of stars all planet-bearing ones would have been identified already. This could indeed mean that within this group the low lithium stars will later also appear to have planets, not yet discovered. We just don’t know yet.

  • Ronald November 13, 2009, 8:00

    I checked various sources for the characteristics of the (most likely) present top-10 solar twins in the galactic neighborhood (identified as such by varuous astronomers, also using Hipparcos data and others), particularly their metallicity, Li content and estimated age (the last one is tricky, I took most recent/relaible, or average estimate).

    What is remarkable is the fact that, though most are roughly solar age, some have higher lithium content than the sun, notably Hip 100963.

    If indeed low Li is an indicator of planets, then one would expect solar type stars with (very) low metallicity (and hence small chance of having planets) to have relatively high Li and vice versa.
    Would be very interesting research.

    I hope the table shows well ordered;

    Star Metal*sun Li*sun Age(gy)
    Be Cn Venaticorum 0.62 1.71 3.8
    18 Sco 1.05-1.12 3.4 4.2
    37 Geminorum ? ? 5.5
    16 Cygni B 1.05-1.23 <1 7.5
    Hip 30104 (HD44594,HR2290)1.35-1.55 1 7
    Hip 100963 (HD195034) 0.72? 5.6-6.3 5.1
    Hip 55459 (HD98618) 1.05-1.12 2.82-2.95 4.3
    Hip 78399 (HD143436) 1? ? ?
    Hip 73815 (HD133600) 1.05 0.71 6.1
    Hip 56948 (HD101364) 1.02 0.95 5.6

  • andy November 13, 2009, 15:27

    High-metallicity is definitely accepted as a predictor for short-period giant planets, the jury seems to be still out for longer period planets. Steinn Sigurðsson of Dynamics of Cats fame mentions this in his post about the recent HARPS batch of planets. The lowest metallicity planet host listed in Extrasolar Planets Encyclopaedia has [Fe/H] = -1 (the double-superplanet system BD+20 2457), but the lowest metallicity for a hot-Jupiter host is [Fe/H]=-0.3 (the transiting planet host star CoRoT-1). The planet population of these low-metallicity systems appears to consist of Neptunes, super-Earths and giant planets on wide orbits, not hot-Jupiters.

    As for the globular cluster planet in the B1620-26 system being “highly suspect”, there’s definitely a lot of evidence it is there – last I recall the torque on the inner binary system had been observed. Whether it is a true planet or a sub-brown dwarf is still debatable of course. There’s a presentation by Sigurðsson available online here which discusses the parameters of this system and its evolution.

  • andy November 17, 2009, 16:22

    Interestingly enough, the abundances of the well-studied 16 Cygni system, which consists of two solar analogues (16 Cygni A and B) and a red dwarf (C) show significant differences in lithium: the planet host star 16 Cygni B is depleted in lithium by a factor of roughly 5 relative to 16 Cygni A, for which radial velocity searches have consistently failed to find planets… not sure of any other planet-host binaries that have these kind of measurements available for more than one component.

  • Ronald November 19, 2009, 6:29

    @andy: interesting indeed, your info ref. 16 Cygni.

    Remakably, both 16 Cygni A and Alpha Centauri A, both with relatively high lithium compared with the B component, are also the biggest and brightest components.

    For 16 Cygni, however, minimum separation between A and B is 843 AU, i.e. a (very) wide binary. With ref. to AlastairMayer (November 12, 2009 at 17:26), it seems unlikely in this case, that component B would perform the lithium depleting role of a (giant) planet for component A, such as suggested by him for Alpha Centauri. So in case of 16 Cygni, if lithium is indeed an indicator for planets (or at least giant planets in close orbit), the high lithium of A would rather indicate absence.

    It would be fascinating to learn more about planet formation and the presence of planets around *both* components of a binary star system. As far as I know, there are no cases known yet of both stars of a binary system having planets, but I would love to be corrected here.