Reuven Ramaty, Benzion Kozlovsky and Richard Lingenfelter

The origin of cosmic rays has been a major problem in astrophysics for nearly a century¹. Any lingering doubt about whether the bulk of the cosmic rays (those with energies below about 10¹⁵ eV) are Galac tic or extragalactic has been settled in favor of a Galactic origin by gamma ray observations with the EGRET instrument on the COMPTON Gamma Ray Observatory (CGRO). These observations² showed that the cosmic ray energy density in a nearby galax y (Small Magellanic Cloud) is much lower than that found locally in our own Galaxy and is thus inconsistent with a uniform extragalactic density. This, of course, does not preclude an extragalactic origin for the very highest energy cosmic rays observed a bove about 10¹⁹ eV (See Physics Today, January 1998, page 31). The power of about 10⁴¹ erg s^-1 required to maintain the cosmic rays throughout the Galaxy is most likely supplied by supernovae. With a Galactic supernova rat e of roughly 3 per century, the required energy per supernova is ~10⁵⁰ erg, which is about 10% of the kinetic energy of the expanding supernova ejecta. Shock acceleration in the supernova blast wave driven by the ejecta could impart such a fra ction of the available kinetic energy to cosmic rays¹.

Although the source of the energy in the cosmic rays thus appears to be well understood, the source of the particles that become cosmic rays is still a matter of debate. It is not clear whether the bulk of the cosmic rays are accelerate d from interstellar gas and dust, from pre-accelerated particles originating from stellar coronae, or from fresh supernova ejecta before they mix into the interstellar medium.

In addition to providing a direct sample of cosmic matter with implications on processes of nucleosynthesis, the cosmic rays also play a major role in the Galactic nucleosynthesis of the light elements, Li, Be and B, which are largely b ypassed in the major processes of nucleosynthesis in stars. Hubert Reeves, Willy Fowler and Fred Hoyle showed in 1970 that Galactic cosmic rays interacting with the ambient interstellar medium on a time scale comparable to the age of the Galaxy could have produced most of these light elements. It is their extremely low abundances (<10^-5) relative to their C and O progenitors that make this spallogenic origin viable. In fact, of all stable elements and isotopes, only for Li, Be and B does cos mic ray spallation play a major role. Even though not all the ⁷Li and ¹¹B are cosmic ray produced, ⁶Li, ⁹Be and ¹⁰B are almost certainly the sole products of cosmic ray interactions.

The studies of the origin of both the light elements and the cosmic rays have recently acquired a whole new dimension, as a result of the extensive observations of light element abundances in old halo stars formed early in the history o f our Galaxy. These observations, at optical and UV wavelengths, now greatly expand the time scale of cosmic ray studies, from the ~10 million year mean age of the contemporary cosmic rays to the ~10 billion year age of the Galaxy. The fact that the earl y Galaxy was almost totally devoid of C and heavier elements (it contained mostly H and He produced in the Big Bang) has very important implications for the origin of early, as well as contemporary cosmic rays.

Li and Be abundances relative to H for stars of various ages, as a function of their Fe abundance are shown in Figure 2. B abundances relative to Be are shown in Figure 3. The metallicity [Fe/H], defined as the logarithm of the Fe- to-H abundance ratio, increases with time from the accumulated production of supernova nucleosynthesis, and thus provides a convenient (albeit nonlinear) representation of elapsed time since the formation of the Galaxy. Studies of Galactic chemical evolut ion⁴ have provided information on the age-metallicity relation. The halo phase of our Galaxy, for which [Fe/H] is less than about –1 (i.e. the period when the average Fe abundance of the Galaxy was less than 10% of the solar value), correspond s to a period of about a billion years preceding the formation of the Galactic disk. Light element abundances have been measured for stars whose Fe abundance is as low as one thousandth of that of the Sun. Such observations³ are very challengi ng because the spectral lines of these rare elements are very weak and they are usually blended with interfering lines from other more abundant elements. The observations therefore require large telescopes and very efficient, high resolution detectors. Wh ile the Li and Be lines can be observed from the ground, the B lines require observations from space.

The flat portion of the Li evolution, usually referred to as the Spite plateau (after the original work of François and Monique Spite in 1982), is generally believed⁵ to represent the Li abundance resulting from nucleo synthesis in the Big Bang. The subsequent increase in Li/H is due to nucleosynthesis in a variety of Galactic objects, including Type II supernovae, novae, and giant stars, as well as production by cosmic rays.

Unlike the Li abundance, the Be abundance (Figure 2) has no obvious flat portion. The approximately linear evolution of Be is equivalent to an essentially constant Be-to-Fe abundance ratio up to [Fe/H] of about -1 (Figure 3). For this e ntire period of early Galactic evolution, the bulk of the Fe is thought⁴ to be produced in Type II and Type Ib supernovae. These core collapse supernovae result from massive (more than about 10 solar masses) stellar progenitors which explode to leave compact remnants (neutron stars or black holes); they produce, on average, about 0.1 solar masses of Fe per supernova, independent of the epoch of Galactic evolution (i.e. the metallicity) in which their stellar progenitors were born⁶. T he decrease in Be/Fe seen in Figure 3 at later times (i.e. for [Fe/H]>-1) most likely results from additional Fe production in Type Ia supernovae⁴ which account for about half of the Fe made in the later phases of Galactic evolution, but wh ich do not eject much C and O, the progenitors of Be. Thus, the observed Be-to-Fe abundance ratio over the entire history of Galactic evolution, coupled with the theoretically derived constancy of the average Fe yield per core collapse supernova, strongly suggests that Be production is due to such supernovae, with an essentially constant Be yield per supernova.

As already mentioned, a variety of processes produce Li and neutrino-induced spallation reactions on C in Type II supernovae can contribute to ¹¹B production⁶. But the only viable process⁵ for producing the Galactic Be is spallation of interstellar C, N and O by cosmic ray protons and a particles and the spallation of cosmic ray C, N and O in collisions with interstellar H and He. This strong relationship to cosmic rays, coupled with the essentially constant Be production per core collapse supernova required by the data for the entire h istory of Galactic evolution, has major implications on the origin of the cosmic rays. It rules out the acceleration of cosmic rays, or at least the Be producing cosmic rays, out of the interstellar medium, because in that case the composition of the cosm ic rays would evolve proportionally to that of the interstellar medium, and the Be yield per supernova would increase as the interstellar abundances of C and O increase, contrary to the observations.

Independent evidence against cosmic ray acceleration purely out of the interstellar medium is provided by energetics. The constant Be-to-Fe abundance ratio of 1.4x10^-6 (Figure 3), combined with the average Fe yield of 0.1 M solar masses per core collapse supernova, requires that on average each such supernova produce about 2x10^-8 solar masses of Be. Calculations⁷ of Be production by cosmic rays of varying compositions and energy spectra yield (Figure 4) the energy in cosmic rays that must be supplied by an average core collapse supernova to produce the required Be. As can be seen, if all of the cosmic rays were accelerated from the interstellar medium and interacted in the interstellar medium (solid and dashed ISM curves), then an unacceptably large amount of energy, namely about 10⁵³ ergs or two orders of magnitude more than the total energy in supernova ejecta, would be needed per supernova in order to produce the observed Be-to-Fe abundanc e ratio at the lowest metallicities. Thus, a scenario in which all the cosmic rays are accelerated out of the interstellar medium can clearly be ruled out.

The simplest solution is that the shock from each supernova accelerates cosmic rays from its own ejecta (see the Box). In this case, it is reasonable to assume that the cosmic ray sources at all metallicities have, on average, the same composition and the same energy spectrum as those of the current epoch cosmic rays. Be production in the early Galaxy is then mostly due to fast C and O interacting with ambient H and He, the yield of the inverse reactions due to fast protons and a particles being very low because of the low C and O abundances of the ambient interstellar medium³. The CRS (cosmic ray source) curves in Figure 4 give the required energy in cosmic rays per supernova for this scenario. This energy of about 10 ⁵⁰ erg is now practically independent of metallicity and is essentially the same as that derived for the current epoch cosmic rays.

Another, more complex scenario invokes a separate low energy cosmic ray component, accelerated from fresh nucleosynthetic matter, to produce the bulk of the Galactic Be. The existence of such a component would allow the acceleration of the standard cosmic rays out of the interstellar medium at all epochs of Galactic evolution, including the current one. Before considering this option, we briefly discuss the B data which have important implications for neutrino nucleosynthesis in superno vae.

Observations with the Hubble Space Telescope of the B abundance show³ ( Figure 3) that the B-to-Be abundance ratio also remains essentially constant, implying a common origin for these two elements. It has often been mentione d⁵ that there is a problem with a pure cosmic ray origin for B in that its isotopic ratio, ¹¹B/¹⁰B=4.05± 0.2 measured in meteorites and ¹¹B/¹⁰B =3.4(+1.3,-0.6) in the interstellar medium⁸ , exceeds the calculated ratio (2-2.5) for production by the Galactic cosmic rays⁷. But the required additional ¹¹B production could be due to ¹²C spallation by neutrinos in core collapse supernovae⁶. As both the neutrino and cosmic ray induced spallation processes are related to such supernovae, the constancy of the B-to-Be ratio is assured. The neutrinos mostly make ¹¹B and not ¹⁰B because their temperature is not high enough for interactions above the higher threshold energy for ¹⁰B production. The required ¹¹B production per Type II supernova, about (2-7)x10^-7 solar masses⁷, is consistent with the supernova calculations⁶.

The possible existence of a distinct low energy component of cosmic rays which could not be observed in the inner solar system because of solar modulation, is a topic of major interest for cosmic ray research. The detection^{9 of MeV gamma rays with the COMPTEL instrument on CGRO from the Orion molecular cloud complex provided evidence for the existence of such cosmic rays. The observed spectrum in the 3-7 MeV region exhibits structure (Figure 5) that appears to be due to n uclear deexcitation lines of 12C and 16O. Such line emission can only be produced by accelerated particle interactions. The high energy gamma ray emission observed from Orion with the EGRET instrument on CGRO is consistent with pion production due to the irradiation of the molecular clouds by standard Galactic cosmic rays2. As such cosmic rays underproduce the observed line emission by at least three orders of magnitude, the gamma ray line production must be due to very la rge fluxes of low energy cosmic rays (up to 100 MeV/nucleon) whose origin is still a mystery. They deposit a minimum of a few times 1038 erg s-1 into the ambient medium in conjunction with the gamma ray line production10. The phenomenon is probably relatively short lived, because this huge deposited power leads in only about 105 years to an energy equal to the total ejecta kinetic energy of a supernova.}

C and O deexcitation lines have been observed from many solar flares, but there are very significant differences between the Orion and solar spectra (Figure 5). The lines from Orion appear much broader, suggesting that they come primari ly from accelerated C and O interacting with ambient H and He, rather than accelerated protons and a particles interacting with the ambient C and O. This requires a strong enhancement of the accelerated C and O abundances. The Orion upper limit on the 1-2 MeV emission⁹, where solar flare spectra show many lines from Ne, Mg, Si and Fe, also re quires the enrichment of the accelerated C and O relative to these heavier elements. The suppression of both the Ne-Fe and proton and a particle abundances relative to C and O could be understood if the seed particles that are accelerated come from the winds of Wolf-Rayet stars¹¹. These stars¹ have optical spectra dominated by emission lines of He, C, N and O. The w eakness of their H lines is caused by mass loss of the progenitor star. While the transfer of the H envelope to a binary companion is a possible mechanism for Wolf-Rayet star formation, the dominant process appears to be mass loss due to radiation driven winds in massive stars. The strong winds, by removing the stellar envelopes, reveal the products of nucleosynthesis and thus become enriched in C and O relative to H and He, and also relative to heavier elements, which are synthesized deeper in the star a s well as during the subsequent supernova explosion. These C and O enriched winds may then be accelerated to higher energies, by the supernova shock as well as by shocks and turbulence in the bubble created by the winds of the Wolf-Rayet and other massive stars¹².

Soon after the COMPTEL discovery, it was suggested¹³ that Orion-like low energy cosmic rays could make a significant contribution to the total Galactic light element inventory. An attractive feature of these low energy cosmic rays was their ability to yield a higher ¹¹B-to-¹⁰B ratio than that produced by the standard Galactic cosmic rays¹⁰. This could account for the meteoritic observations discussed above, but, as we already noted, the excess ¹¹B could also result from neutrino interactions in supernovae⁶. In addition, since the low energy cosmic rays in Orion are strongly enriched in C and O, as a consequence of the acceleration of the nucleosynthetic products in the wi nds of Wolf-Rayet stars before their mixing into the interstellar medium, one might expect that they could provide an explanation for the observed constancy of the Be-to-Fe abundance ratio as a function of metallicity. However, even though Wolf-Rayet star s could be the dominant injection source in Orion, they were not a major source of Be producing C and O in the early Galaxy because the generation of the Wolf-Rayet winds depends on metallicity. Specifically, both evolutionary calculations and observation s of the Large and Small Magellanic Clouds showed that the lower the metallicity of the environment the fewer Wolf-Rayet stars are produced¹⁴. As the Fe production in core collapse supernovae is independent of the metallicity of the progenitor stars, this would imply an increasing Be-to-Fe production ratio, contrary to the observations.

Thus, while the early Galactic Be data strongly suggests production by cosmic rays originating from supernovae accelerating their own ejecta, the most popular models for cosmic ray origin posit acceleration out of the ambient inters tellar medium¹⁵. These models are based on the inferred composition of the cosmic ray source material. This composition is significantly modified by nuclear spallation during cosmic ray propagation in the Galaxy^1,15, but the average source composition of the most abundant elements can be well determined. These cosmic ray source abundances relative to solar system abundances are shown in Figure 6 (from¹⁶) as a function of their atomic mass number. We see that there is a st rong enhancement of the abundances of the highly refractory elements relative to those of the highly volatiles. Alternatively, the enrichments may reflect a correlation with first ionization potentials (FIP), since the highly refractories also have low FI P, while the highly volatiles have high FIP. C and O are enriched as well, but not as much as the highly refractories.

The correlation with first ionization potential led to the suggestion that the cosmic ray source material originates from the atmospheres of stars. Based on the fact that the abundances of elements with low FIPs are enhanced in the sola r corona and in solar energetic particles, it was suggested that similar shock acceleration on low mass, cool stars could provide a particle injection source for acceleration by supernova shocks in the interstellar medium¹⁷. The correlation wit h volatility led to a model¹⁶ in which the refractory enrichments result from the preferential acceleration of the erosion products of refractory interstellar grains which have been pre-accelerated to energies of about 100 keV/nucleon by super nova shocks. Because of their suprathermal energies, the sputtered ions from these grains are preferentially accelerated relative to the volatile elements which are accelerated from thermal energies in the ambient interstellar gas. For C and O, it has bee n suggested¹⁶ that the bulk of these elements is accelerated from the winds of Wolf-Rayet stars, a scenario which is motivated by the strong cosmic ray enrichment¹⁵ in ²²Ne which is also overabundant in these stars. On th e other hand, C and O enrichments relative to the volatiles are expected, since a significant part of the C and O is locked in grains. The fraction of the O in highly refractory oxides (primarily Al₂O₃, MgSiO₃, CaO and Fe< SUB>3O₄) is sufficient to account for the observed O enrichment. C can also form refractory graphite grains, as observed¹⁸ in the Type II supernova SN1987A.

As the cosmic rays in both of the FIP- and volatility-correlated models are accelerated interstellar matter, they would produce insignificant amounts of Be and B in the early Galaxy and thus require a separate cosmic ray component accel erated out of fresh nucleosynthetic matter at all epochs of Galactic evolution. As already mentioned, the low energy cosmic rays discovered in Orion with COMPTEL indicate the presence of a separate component, but the Wolf-Rayet winds, which are likely so urces of these cosmic rays, cannot lead to cosmic rays capable of accounting for the constant Be-to-Fe abundance ratio in the early Galaxy because the formation of Wolf-Rayet stars is metallicity dependent¹⁴. For the same reason, Galactic cosm ic ray C and O originating from Wolf-Rayet stars, with the rest of the cosmic rays being accelerated out of the interstellar medium, also does not solve the problem.

What then are the arguments against supernovae accelerating their own ejecta and how strong are they? We summarize these in the Box. We see that the possibility of cosmic rays being accelerated from supernova ejecta is still wide open.< /P>

Our goal in this review has been to show how the recent atomic spectroscopy observations of light element abundances in old halo stars has brought exciting new insights to the question of the origin of the cosmic rays, a problem whi ch traditionally has been investigated by in situ cosmic ray observations. We have shown that the light element abundances, especially that of Be, require a critical reexamination of prevailing theories of cosmic ray origin. The cosmic rays in the early G alaxy, or at least their C and O, must have been accelerated from freshly nucleosynthesized matter rather than from the then extremely metal poor interstellar medium. It is still not clear, however, whether the contemporary cosmic rays are also accelerate d from such fresh matter or from the interstellar medium. If accelerated from the interstellar medium, a separate component accelerated from fresh nucleosynthetic matter at all epochs of Galactic evolution would be needed, but there is no evidence for suc h cosmic rays from in-situ (in the solar system) measurements. Nuclear spectroscopic observations of the Orion molecular cloud complex have revealed the existence of large fluxes of low energy cosmic rays, probably accelerated from enriched nucleosyntheti c matter in the winds of Wolf-Rayet stars. Many challenging questions remain concerning the origin of these low energy cosmic rays and their role in light element production. However, the simplest scenario for the origin of the Be and the bulk of the B is production by cosmic rays from sources which at all epochs of Galactic evolution have approximately the same composition and energy spectrum as those of the Galactic cosmic rays observed locally. This suggests that the bulk of the C and heavier elements in the Galactic cosmic rays are accelerated from supernova ejecta.

We wish to acknowledge Hubert Reeves who suggested to us the possibility of light element production in the early Galaxy by standard Galactic cosmic rays. We also acknowledge Martin Lemoine and Keith Olive for the compiled data that we used, Eric Gotthelf for the ASCA data (Figure 1), and Michel Cassé, James Higdon and Elisabeth Vangioni-Flam for important discussions.

The acceleration of the cosmic rays from supernova ejecta has long been an appealing scenario for cosmic ray origin. But several arguments have been raised in that appear to argue against this possibility. The most important of thes e are:

(i) The enrichment of the cosmic ray source abundances relative to solar abundances in refractory (or low FIP) elements relative to volatile (high FIP) elements. If these enrichments were due to the FIP selection, they would argue again st the supernova ejecta because the conditions for such selection (temperatures ~10⁴K) exist in the atmospheres of stars but not in supernova environments. If the enrichments are due to selection according to volatility, pre-acceleration of int erstellar grains could occur in the supernova environment. The grains could be interstellar, i.e. produced prior to the supernova explosion¹⁶. Alternatively, they could be produced as high velocity condensates in the expanding supernova ejecta< SUP>18. The latter scenario guarantees a fresh nucleosynthetic source for the refractory cosmic rays.

(ii) The similarity of the cosmic ray source and solar abundance for the refractory elements. Even though this suggests that the source is interstellar matter, it is not an argument against acceleration of supernova ejecta because super novae are the primary sources of these elements in the Galaxy⁴. However, in the supernova origin scenario the various types of supernovae should contribute to the cosmic rays in the same proportions as they do for the Galactic abundances.

(iii) The ultraheavy cosmic ray nuclei. The enrichment of r-process (nucleosynthesis via rapid neutron capture) nuclei in the cosmic rays (platinum) supports the supernova origin (the r-process is thought to occur in supernovae just abo ve the newly formed neutron star). The fact that s-process (slow neutron capture nucleosynthesis) elements (strontium and barium) are present in the cosmic rays but not produced in supernova explosions, is invoked against the supernova origin. However, s- process elements are made in the cores of stars, including the stellar progenitors of supernovae, and are ejected in supernova explosions. Overabundances of Sr and Ba relative to Fe have been observed¹⁸ in SN1987A.

(iv) The time between nucleosynthesis and acceleration. K-capture isotope decays are prevented if they are accelerated on time scales shorter than their lifetime (an accelerated atom is stripped and K-capture is prevented). Abundances o f such isotopes and their decay products can provide information on the time between nucleosynthesis and acceleration. The cosmic ray ⁵⁹Co/Ni is used¹⁵ (⁵⁹Ni ® ⁵⁹Co, t =1.1x10⁵ yrs). If ⁵⁹Ni decayed, acceleration must have occurred several tens of thousands of years after nucleosynthesis, ruling out acceleration of the ejecta before mixing into the interstellar medium. However, the contribution of ⁵⁹Ni decays amounts to only ~1/3 of the total ⁵⁹Co, the rest coming from nucleosynthesis in the explosion⁶ and spallation during propagation. Uncertainties in nucleosynthesis theory and spallation cross sections, and err ors (~30%) in the cosmic ray data, do not yet allow the determination of the fate of ⁵⁹Ni.

Figure 1. The supernova remnant SN1006 seen in X-rays. The bright regions show synchrotron emission produced by very high energy electrons accelerated by the supernova shock. The data were taken with the CCD cameras aboard the Advan ced Satellite for Cosmology and Astrophysics (ASCA). More details can be found at

Figure 6

 

To view postscript (.ps & .eps) files
obtain GhostScript & GSview here.