Physics Department, University of York, Heslington, York YO10 5DD, UK
A paper with the above title was submitted to the Monthly Notices of the Royal Astronomical Society on 20th November 2000. It was seen by three referees, all of whom recommended against publication. The purpose of this presentation is to place into the public domain a contribution to cosmogony that otherwise would not be available. For completeness the objections that have been raised are also recorded here together with associated responses and comments by the authors. The presentation is in four parts:
One of the authors, Stephen Oxley, on whose D.Phil thesis much of this work is based, is no longer in the academic field and this presentation is made solely by M M Woolfson.
The paper above was a revised version produced in response to referee's reports on an earlier version. The first report received on the earlier version was:
MA409 is about the formation of planets and brown dwarfs through Woolfson's 'Capture Theory', in which massive interstellar clouds are tidally disrupted and induced to fragment by encountering a compact star in a young stellar cluster. The authors also present new SPH models of this process.
Every serious astronomer - except Woolfson and his collaborators - uses the planetesimal theory of planet formation. In this model solid matter in the circumstellar protoplanetary disc agglomerates into protoplanetary cores. Those that accrete substantial quantities of gas from the disc become gas giants, such as giant planets of the Solar system. Evidence for accretional processes in the early Solar system is well known and overwhelming. Recent (well known) observations of an extrasolar planetary transit also confirms standard planetary structure models. No serious astronomer believes there is any need to ditch the standard model in favour of Woolfson's 'Capture Theory', which has under gone very little detailed development since conception and has not been able to account for the details of any significant observation. I can only recommend that MA409 is not accepted for publication in MNRAS.
I should also note the following:
This report fell short of being an objective scientific critique of the paper itself and the editor approached another referee.
I have read through the paper and the previous referee's comments. First of all I cannot agree with the previous referee that we should not discuss alternative theories for planet formation as there is a standard model believed by most astronomers (the use of "serious" astronomers I find needlessly insulting and probably incorrect). It is clear that our ideas of planet formation based upon one object, the solar system, needs some revision in the light of the newly discovered extrasolar planets and their unexpected locations close to their parent stars.
Nevertheless I cannot recommend publication of the paper as I find that there are several minor flaws and one misleading major flaw that invalidates the results.
Firstly, the timescale derived for the interactions uses an encounter radius of 1500 au in order to achieve realistic numbers. In contrast, the encounters that are presented in the paper have encounter radii of order 100 au. Thus the encounter rate derived is wholly inappropriate and misleading to the reader. The relevant rate is at least 225 times less than the 5 per cent quoted. In fact, using actual number densities (such as the ONC) encounter rates at 100 au occur on timescales of > 1e08 years which explains why so many of the stars in the ONC (age 1e06 years) still have disks of this size. This would imply that BDs have a probability of encounters of ~1e-5, and thus in a cluster of 1000 stars is negligible.
Secondly the paper states that it has used an SPH code including radiative transfer yet not details of the calculations, methods or test cases to prove the code are presented. This would be necessary before the results can be published (or a reference to a refereed journal where the methods etc is published).
Minor flaws include the lack of any recent references for the parameters chosen. Indeed, densities of >10.4 stars/pc.3 are high for the majority of stellar clusters (except the core of the ONC). Furthermore, there is no convincing evidence for the low-mass stars having formed later than other stars in the cluster. Pre-main sequence ages are extremely difficult to determine and any reliance on PMS tracks is untrustworthy (see for example Tout et al 1999, MN, 310, 360). Along these lines the claim for free-floating planets in the ONC is unreliable as there is as yet no way to confirm the PMS evolutionary tracks used.
In general the idea seems interesting but the timescales involve argue strongly against a significant number of planets forming in this way. Details of the calculations would have to be made clear in order to determine if they were reliable. It does seem that more realistic approach is that of protostellar disks providing the mass for the planet formation (the work of Whitworth and collaborators published in MN). The advatages are clear in that the disks are sufficiently long-lived in order to have a reasonable chance to undergo a close interaction (~1e7 years), and that the mass is already orbiting the star.
This report was received on 26th July 2000. While some of the points made could be challenged the major point was accepted. The calculation of probabilities was made corresponding to a different linear scale to that used for the SPH calculations. The calculations were repeated at an appropriate scale and, at the same time, other pieces of analysis were finished off that were included in the revised version of the paper presented here. This was sent to MNRAS on 29th November 2000. On 4th March 2001 I wrote to the editor asking if there was any response from the referee. On 6th March the following report arrived by email.
This paper purports a new mechanism to explain the formation of planetary system via the disruption of proto-Brown Dwarfs in stellar clusters. The idea is interesting and there is certainly some validity in investigating alternative mechanisms for planet formation. Unfortunately I cannot recommend that this paper be published in MNRAS. I will list below my reservations about this paper in decreasing importance.
on pg 13 the authors state that a protostar is formed when a body of higher density than the surrounding medium is stable and able to being the slow process of contraction. This is misleading as it implies a non-dynamical nature of star formation.
In fact, a protostar is formed at the end of a gravitational collapse event (from unstable Ics), once the density has increased sufficiently such that infrared radiation can no longer escape freely from the object. This occurs on the object's free-fall time and is not slow, but in fact is generally faster than the interaction timescale for realistic Ics.
In conclusion I cannot recommend publication of this paper.
The policy of MNRAS is to base rejection on two referee's reports but, in view of the brief nature of report A the Editors gave us the opportunity of having the paper sent to a third referee if we chose. However they stated that they would send the report of Referee B to the third referee. We requested that if they did so then they should also send our response to the report. This follows.
The referee's comments are dealt with, by and large, in the order presented that is in his order of decreasing importance.
In his preamble to making specific points the referee refers to the paper as describing the disruption of brown dwarfs in a cluster. Actually the interactions considered are not restricted to those involving a brown dwarf and he should have noticed that the three computational examples most relevant to later parts of the paper involve protostars that have main-sequence masses. During the more-than three months the referee kept the paper we have concluded that in the embedded stage of cluster development proto-bodies at the lower end of the brown dwarf mass range may not play a role. This makes no significant difference to calculated probabilities - at most a factor of two. Probabilities depend much more, and linearly, on stellar number densities in the embedded state that can vary by three orders of magnitude.
In the top paragraph of page 10 (towards the bottom of p.6 in the present paper format), that must have escaped the notice of the referee in his scan of the paper, we refer to this testing quite specifically. I cannot believe that the referee wishes us to reproduce in detail what has occupied 19 pages of a previously-published piece of work.
We seem to differ in our definitions of a protostar. The referee's definition puts the start of the protostar stage somewhere close to the beginning of Kelvin-Helmholtz contraction whereas we take it to begin when there is a discrete identifiable mass of material that will evolve to form a star. Our definition seems more in accord with usual nomenclature - for example, Hayashi's original description of the evolutionary track of a protostar on an H-R diagram starts with a very diffuse body of radius about 4,500 AU that then begins a free-fall collapse. Anyway, we make it quite clear what we mean by a protostar; at the top of page 14 (middle of p.10 in present paper format) we say explicitly that we refer to a body at the beginning of its free-fall stage. The referee failed to notice this, assumed that we are referring to a protostar according to his definition and then points out the problems that this would introduce.
In our paper we did not raise the question of how, and in what state, protostars form but since this is an issue raised by the referee we must respond to it although I do not think that this material needs to be added to the paper. We refer to the paper, Woolfson (1979), given in the references. This describes star formation as due to the collisions of streams of matter within a turbulent collapsing cloud. Stars formed in this way have comparatively little angular momentum and, indeed, peripheral material in the colliding streams, which is not compressed, can be captured to give a surrounding disk within which almost all of the angular momentum resides. This is the kind of starting point from which planet formation can begin in the solar nebula approach. As the referee has stated, star formation is a dynamic process. Streams of matter with high Mach number can collide and form condensations anywhere. Clearly, in the vicinity of another star the initial shape would not be spherical but, in any case, our spherical starting model very quickly adopts the shape required by the prevailing equipotential surfaces.
The referee seems, by implication, to be disparaging the 1969 paper we quote on the analysis of young stellar clusters. This depended on locating stars on an H-R diagram and hence determining their ages and masses from theoretical evolution tracks on to the main sequence. It indicated one stream of development, starting at about class F5 or so and progressing with time to smaller mass stars and a separate stream in which progressively larger mass stars are produced. The 1979 paper interpreted the first stream as due to the primary formation process and, as indicated above, these stars spin comparatively slowly. The second stream was analysed in terms of mass increasing by accretion. The derived relationship of angular momentum to mass was in excellent agreement with that deduced from observations. Recent work on pre-main-sequence evolution uses up-to-date computer codes and considers factors such as mass loss or mass gain during the evolution. While, for stars neither gaining or losing mass, the new evolutionary tracks are better than the older ones I do not think that it would change the general character of the mass v time pattern as deduced in 1969. Even if it did it would not be greatly important to the model presented here; as long as compact stars are present while new stars are being formed in the embedded stage of the cluster, then the mechanism would apply whatever their respective masses.
As an example of how later work may not necessarily be better than older work I refer to a paper on star formation by Bhattal et al., 1998, MNRAS, 297, 435. This produces protostars by collisions of massive clouds. The protostars are in the form of disks and in a typical case, detailed in the paper, a stellar condensation has a mass of 40Msun and angular momentum 1011 times that of the Sun and more than 108 times as much angular momentum than that of a fast-spinning O or B star. How this is supposed to evolve into a main-sequence star is not explained.
More than three months passed before we received the following report on 19th June 2001. On the basis of this report the editor finally rejected the paper. Any opportunity to respond to the report was denied us by the editor (see §3) but the report is given here with responses added.
Omitting the brief re-statement of the manuscript's contents customarily given in referee reports (as this must now be well known to the editor) I will cut to the chase. I do not believe the crucial details of the proposed processes are described correctly.
The whole basis of the calculation in §6 is that the protostar is in a diffuse state and that the time available for a planet-producing interaction is at most the dynamical timescale (the free-fall time). This will be of the order of a few thousand years. Because of other rather conservative constraints introduced into the analysis regarding minimum distances etc. the actual time for a fruitful interaction is less. As far as we can judge from the comment the referee is suggesting that we used the initial radius of the protostar for several million years (we are not sure what he means by the lifetime of the cluster in this context). Our analysis shows that with an initial radius of ~1 000 AU, justified by the argument involving equations (5) to (7) (and not 'many-thousand AU'), together with a dynamic timescale gives the probabilities shown in Figure 10. If our analysis had been shown to be flawed then we could accept the referee's conclusion. However, we cannot accept a conclusion based on fantasy rather than the model we have presented. A generous interpretation is that the referee has not understood what we have presented. Less generous interpretations are possible.
There are two separate components in this comment. One is related to SPH as a general scheme, as used and developed since 1977. It is a free Lagrangian code with artificial viscosity introduced to handle shocks and it is widely used for astrophysical problems. No computational procedure can precisely simulate physical reality but SPH has been shown to reproduce the expected form of behaviour in collisions and tidal interactions without necessarily giving precise quantitative results. In fact collisions, particularly supersonic collisions are the most challenging applications and special procedures have been developed to handle these situations. Fortunately, collisions do not play a role in what we do and the normal SPH part of our application here follows standard procedures. Actually very similar results for the break-up of the protostar are found with 1 000 points or fewer but more points are needed to be able to follow the progress of the protoplanetary condensations. Again, wide variations of parameters (over several orders of magnitude in dimension) give similar behaviour patterns, suggesting that the simulation is a robust one and reproduces something that could occur in the physical world.
The other component relates to radiation transfer where the referee states that, for this, SPH is not a good scheme. We can concur with that. We have used the expression 'a novel and realistic model of radiation transfer, incorporated within our general SPH package.'. This is briefly described in the third paragraph of §5 and in some detail in the Oxley reference. It is not part of the SPH procedure as such but is interwoven into it. It should not be lightly dismissed without proper, or in this case any, examination. Its development and validation took nearly two years and it contains a great deal of physics. Just as an example, for opaque situations the algorithm leads to a diffusion form for radiation transfer with proper speed of propagation. It is also not a cheap procedure. Attaching it to the SPH package increases computation time by up to two orders of magnitude.
This comment is related to the following one (§3 in the referee's disjointed numbering system). The referee is clearly a supporter of the solar nebula concept and possibly someone working in that area. He/she therefore has a strong fixation on the conditions that occur when planets are produced by planetesimal accretion. Because planetesimals come together with a small hyperbolic excess the protoplanets are initially in more-or-less circular Keplerian orbits within a medium moving in similar fashion. In this rather quiet environment medium material needs to lose very little energy to be captured by the protoplanet. There may be various ways of losing this energy - for example, material striking the surface of the protoplanet may share its energy with surface material all of which is then moving at less than the escape speed. This naturally leads to accretion. Incidentally, Bate and Bonnell use SPH in their work!
A protoplanet produced by the capture model as described here is large diffuse object, with an escape speed from its surface of order 1 km s-1 or less, moving on a highly eccentric orbit. For most of its orbit, especially near periastron, its speed relative to the medium would be several times the escape speed from its surface (even more so for the surrounding disk) and the net result would be abrasion rather than accretion (see Woolfson, The Origin and Evolution of the Solar System, p.88). Without detailed analysis it would be difficult to predict what the net outcome of the exchange of matter with the medium would be - but intuitively it seems much more likely to be a loss by the protoplanet.
First the strongly sub-Keplerian rotation of the disk which I understand is an important part of the theory, is not demonstrated convincingly. The stellar wind able to remove the nebula (reminiscent old and no longer supported theories from 1970 for disk removal) is assumed. Then it is said to act on dust grains in the nebula (well, so is it a nebula or a wind or a radial wind permeating a nebula in circular motion or what?)
Then the miscalculated sizes of the secondary bodies ("planets") together with an inapplicable formula (from Dodd and McCrea, 1952) containing a huge gravitational focusing factor and not at all desribing the situation in a rotating disk, give rise to incredible and unreal reduction of the initial semi-major axis a~1000 AU to a fraction of AU (relevant to exoplanets). The only feasible mechanism for migration that goes so far is the so-called type II migration in disks (cf. Articles in Protostars and Planets III and IV volumes, Arizona Press 1993 and 2000 respectively). Unfortunately, at large distances like those cited as initial in the paper the type II migration may be outward instead of inward (if most of the disk mass is inside), so that even this mechanism is very uncertain.
Finally, eccentricity evolution is misunderstood as well, since the large "protoplanet" masses of 10 m_j migh cause eccentricity pumping of the orbits, not damping (cf. PP IV volume). The manuscript seems to neglect a substantial body of work started by Goldreich and Tremaine in the late 1970s, on disk resonances, torques and consequent secondary body evolution. They work in sometimes non-intuitive, if well physically grounded, ways. Modern theories show that gas drag is ineffective compared with gravitational torques at object masses larger than planetesimals.
The discovery of extrasolar planets close to their primaries and the longstanding problem of producing Uranus and Neptune on a sufficiently short timescale has stimulated solar nebula theorists to look at various mechanisms for planetary migration. A currently well supported idea involves the way that a planet in a circular Keplerian orbit interacts with a Keplerian disk. Spiral waves are generated, travelling both inwards and outwards from the planet's orbit. Those moving outwards carry angular momentum that is deposited when the energy in the spiral wave is dissipated. The reaction on the planet producing the wave is to cause it to lose angular momentum and so spiral inwards. Conversely those travelling inwards give a reaction that causes the planet to move outwards, a 'reverse migration'. This is the point of the referee's comment that if most material is interior to the planet then it will move outward. Another aspect is that if a wave generated by, say, Jupiter, moves outwards and interacts with another planet then angular momentum can be added to that planet so causing it to move outwards. A brief account of wave transport, with some mathematics will be found in Woolfson's The Origin and Evolution of the Solar System, pp.168-169. Spiral waves and other possible migration mechanisms are also briefly mentioned in the sixth paragraph of §9 of the present paper.
A defining characteristic of type II migration, mentioned by the referee as the only possible mechanism, is that a massive planet in a medium with Keplerian motion opens up a gap within which it moves in a circular orbit. The spiral waves that transmit angular momentum inwards and outwards depend for their formation on this particular arrangement of planet and medium. It is incredible that the referee could believe that a planet in a highly eccentric orbit could establish a gap (what would it look like?) or that the ever-changing relative motions of medium and planet could generate spiral waves. In short, a resistance model based on spiral waves cannot be applied to a body moving in a highly elliptical orbit through the disk. However, the factors giving rise to the Dodd and McCrea resistance will apply; the effects of the Keplerian pattern of motion of the medium will be completely swamped by the resistance due to the high relative speed of the planet relative to adjacent material. If there are additional forces at play then this will simply reduce the time for evolution of the orbit.
The referee has a strong fixation on what the solar nebula model involves and cannot seem to think outside it. The migration mechanisms considered by the solar nebula community are complex and, until recently, were barely understood. That such mechanisms must be considered in relation to solar nebula ideas does not mean that simple mechanisms, that are intuitively obvious, are invalid.
The referee may be right in saying that strong solar winds may not be a necessary feature of young stars. We simply do not know enough about the behaviour of YSOs. However, the idea of a strong solar wind to remove the nebula is still mentioned in recent papers, despite the referee's protestation. Actually the sub-Keplerian rotation of the disk is not critically important to the round-off process. Down to when the eccentricity is quite small the relative speed of body and medium is comparatively little affected by departures of the medium from Keplerian motion. Figure 11 shows that by the time the eccentricity has reduced to 0.1-0.2 the semi-major axis is down to a few AU or less. That is where the solar nebula theorists begin. We could go along with Goldreich and Tremaine at this stage!
Although not mentioned in the paper there is also a small but significant effect from stellar luminosity. For present solar luminosity at a distance of 5 AU the effect on the medium is equivalent to a 0.1% reduction in solar mass. If young stars are several times as luminous as the sun (which is suggested by theoretical PMS evolutionary tracks) then the effect on orbital evolution may not be negligible, but this has not been tested.
As far as eccentricity pumping is concerned, unless there has been some very recent new work that has changed the situation in the last couple of years, I understand that it only operates for bodies of brown-dwarf mass or above - probably well above according to the theory. Is this why the referee has attempted to promote our captured bodies to brown-dwarf status by stating that they are ~ 10 MJ in mass (7 out of 11 in our modelling are well below this) and will subsequently grow?
Another form of eccentricity pumping involves two planets in commensurate orbits (e.g. Melita and Woolfson, 1996, MNRAS 280, 854-862) but this is not applicable here.
Here I believe is the real problem - clearly raw nerves have been touched. The paper shows that various observations, especially those of recent origin (e.g. close exoplanets, free-floating planets) are explained by the present model and the model developed in Cardiff but not (yet) by the solar nebula model. The reference to "the history of conflict" suggests that a vigorous debate has been taking place over some period of time. On the contrary, the capture-theory model has been largely ignored, is comparatively unknown and is never included or even mentioned in conferences on planet formation. In addition it receives very harsh treatment at the hands of referees, which explains why papers on this subject have to be written in sufficient detail to try to anticipate unjustified criticism. The present referee has repeatedly suggested that we are ignorant about various matters or do not understand various theoretical approaches. Had we anticipated this and included material to show that we were not, then this too would have been slated as unnecessary and making the paper too long!
The table of orbital elements was to illustrate the types of orbit that existed. It was stated as "a sample" and that was all that was required for the purpose. Why was this made an issue?
It is the solar nebula theorists that make extraordinary claims and the evidence that they offer is hardly extraordinary since it has not solved major, indeed critical, problems after thirty years of sustained work by a large community. All that is claimed for the present work is that it offers a plausible alternative model - no more, no less. The referee rehearses here the objections he/she has previously raised and we can only refer to our refutation of those objections - in particular the first that is completely unrelated to what we have written.
The following is an extract from the final rejection letter:
"The decision as to whether a paper is published is not made by the referee but by the editor based on the advice of the referees as you suggest. I am sorry to send you disappointing news on this occasion but we are unable to give your submission any further consideration."
The Editors of journals published by learned societies perform an important service to the scientific community and they should be fully supported by the communities they serve. I acted for many years as co-editor of an international journal (not in astronomy), and handled well over 600 papers. The great majority of papers were straightforward - sometimes accepted, sometimes requiring either minor or major revision and occasionally rejected. The reason for rejection was, as for MNRAS, that two referees had recommended rejection. However, no paper was rejected absolutely before the author(s) had an opportunity to see the referees' reports and comment on them. The final letter of rejection was in the form that the paper was rejected unless it could be shown that the referees' reports contained demonstrable factual errors. This was rarely so, authors would accept the rejection and there was no further correspondence. I cannot remember the exact number of comebacks there were to this approach but it could probably be counted on one hand. This was, of course, extra work for me as Editor, but it happened so rarely that the overall increase in effort was negligible and I had the satisfaction of knowing that I had carried out the editorial task in a fair and even-handed way. I can recall at least one case where I used my editorial judgement in favour of the authors since they had made a good scientific case for what they were doing and the referees' reports were either biased or mistaken - I forget which at this distance in time.
The form of the rejection letter from the editor in the present case gave no opportunity for response. The unusual report from referee A might have rung an alarm bell that judgements made in this subject area might not be entirely objective, but clearly it did not do so. Again, when referees delay for more than three months before responding (perhaps only responding when reminded), as did both referees B and C, then this might betoken some problem with the refereeing process. The report from referee C, that led to the final rejection, contained glaring factual errors in what was probably his major criticism - his point 1. The referee boldly stated a conclusion based on the very opposite of what is in the paper and this was accepted without question! In virtually every other comment he was criticising the work based on conclusions drawn from one model, the one he/she favours, carried over to another model based on an entirely different scenario. There was no opportunity to point this out to the editor who merely accepted what he/she was given at face value and therefore presumably subscribed to it.
I was also very disappointed in the general approach of the referees. As a referee myself I am occasionally asked to look at papers with a solar nebula theory background. I never take that into account but merely judge the paper on the basis of the science as presented. When there are points about which I am doubtful I say clearly what they are and, where possible make suggestions about how they can be put right. My general approach is to be supportive of the author(s) although I do recommend the rejection of papers if they do not meet an acceptable scientific standard. There is no sense in which the reports of referees A, B and C could be described as constructive or helpful.
In the cosmogony field there is no way of knowing what is right - only what is wrong. One characteristic of a good theory is that it must be vulnerable, that is to say that it is detailed and specific enough to be subjected to tests that may prove it to be wrong. Those presenting any theory in this field must be prepared to face criticism. The way to meet criticism is either to show that it is wrong or to develop new theoretical approaches that counter it. However, it is not legitimate, and contrary to all proper scientific values, to protect a theory by suppressing alternative ideas.
In view of the present experience with the referees of this paper and the editorial process of MNRAS it seems unlikely that this work would ever be published in a refereed journal in its present form. Presenting it on the Web, seems to be the only, if a rather inefficient, way of making known what I think is an interesting contribution in this area.
Stephen Oxley's D.Phil thesis can be found here
Please email any comments to firstname.lastname@example.org
Last updated 24 August 2001
Prof M M Woolfson