Direct Photons at PHENIX
By Gabor David
Like a complicated movie plot, relativistic heavy ion collisions are a sequence and interplay of very different physics processes, which nevertheless strongly shape each others outcome - later ones often masking the effects of previous ones. We are reasonably certain that the main stages of this evolution are: modifications of the accelerated nucleus even before it collides, initial hard scattering of partons, formation and equilibration of a dense medium, its expansion, transformation and cooling, and formation of the final state (observable) particles. Ideally one would like to take a "snapshot" at each time-slice, and view them separately, making sure that none of these snapshots are blurred by later developments in the system. Luckily, direct photons give exactly such snapshots, and better still, they have a unique property: no matter where and when are they produced, once born they leave the collision region unchanged, without further interaction. They are in principle the ideal "historians" - but unfortunately we can see their snapshots only simultaneously. Imagine watching all frames of a movie projected on top of each other and trying to understand the plot: measuring and interpreting direct photons is a comparable challenge.
Maybe even bigger, because not all photons we observe are "direct". In fact, the majority of them are not: they come from final state hadrons - mostly neutral pions and etas, which decay into photons long after the collision is over. While those decay photons enable us to study identified hadrons up to 20-25GeV/c transverse momentum, and led us to one of the first and biggest discoveries at RHIC (jet quenching), they are a background spoiling the direct photon measurements, a background that can give more than 90% of the all observed photons and hard to eliminate.
You certainly noticed (and maybe got annoyed) that so far we didn't even give a definition of "direct photons". That's because they are only defined indirectly, by exclusion. Whatever their source, the physics mechanism that produces them, we call those photons direct that come from the collision but are not hadron decay products. This cautious wording is warranted because with our increased understanding of the evolution of the collision theorists find (and calculate) new sources of direct photons, earlier not thought of or found negligible. On Fig. 1 (made in 2005) contributions from five different sources are shown, along with their their sum (top, solid curve) and the measured low transverse momentum direct photons (PHENIX data, black squares with error bars). Two decades ago only three sources were considered: the two steepest exponentials, labeled "Th-Th" and "HG", which are thermal radiation from the Quark-Gluon Plasma (QGP) and the hadron gas, respectively, and the blue curve labeled "prompt", or initial hard scattering, photons coming mainly from the so-called gluon Compton scattering process and becoming the dominant contributor above 6GeV/c transverse momentum. Now we ponder at least five sources - in fact, other photon production mechanisms not indicated here were also proposed. Often the curves are not uniquely predicted by the theory. For instance, we know that the "Th-Th" and "HG" curves should be exponential, but their slopes - related to the temperature of the system at the time the respective photons were emitted - is a free parameter in the theory, to be determined from the measured data.
(Ab)using our analogy the individual curves are the "frames" of the movie and their sequence tells the "plot", what happened in subsequent stages of the collision. Unfortunately all photons are alike, no matter what subprocess produced them, so what the experiment observes (after eliminating the huge hadron decay background, a quite involved analysis) is the sum of all curves - just as if all frames were projected on top of each other. While this sum cannot be unambiguously broken up into its components - current experimental errors and the free parameters in theories prevent us from doing so - something has already been learned. Note that on Fig. 1 at the lowest transverse momenta thermal contributions are dominant; the measured points now restrict the range of possible initial temperatures to [300,660]MeV and require a thermalization time in the 0.15-0.6fm/c range. While higher precision and thus tighter limits are desirable, this is already an important constraint; in fact, recently one of the most hotly debated questions is what mechanism can cause the system to thermalize that fast.
Our ultimate goal is not only to measure the overall direct photon yield (the sum curve) but to disentangle it into its components as uniquely as possible. Looking at Fig. 1 it is obvious that the best place to start is the high end of the spectrum where according to theory one component, hard scattering is overwhelming. Once this is measured with sufficient precision to resolve ambiguities in the theory, in principle one can calculate and subtract the hard scattering part from the measured (sum) spectrum, then turn to the next strongest component.
The high end of the spectrum is also somewhat easier to access experimentally than the lower region. In heavy ion collisions at high transverse momenta hadron production is strongly suppressed ("jet quenching"), as is then the background from their decay, but the direct photons come out of the collision region unscathed (they interact only electromagnetically, the strongly interacting medium is transparent to them), so the signal/background (direct/decay) photon ratio becomes very high (bigger than 3).
One very informative way to study what is specific to relativistic heavy ion collisions - like the effects of a new medium - is to measure the nuclear modification factors, which compare particle yields observed in nucleus-nucleus collision (like Au+Au) to properly scaled yields of nucleon-nucleon (p+p) collisions. Proper scaling means to take into account that nuclei have many nucleons, so whatever the probability of a particular process in an elementary p+p collision, it will be higher in Au+Au. The nuclear modification factor is then defined such that it is unity if (for a specific process) the Au+Au collision is nothing more than a straightforward superposition of p+p collisions. Deviations from unity, like enhancement or suppression indicate nuclear effects, the qualitative difference between p+p and Au+Au collisions.
On Fig. 2 we show the nuclear modification factors measured in central Au+Au collisions for three identified particles. Two are hadrons (the neutral pion and the eta) which are strongly suppressed: in Au+Au we see only a fifth of what we would expect from scaled p+p. Note that high transverse momentum hadrons are fragments of hard scattered partons, and the interpretation of the suppression is that a very dense, strongly interacting medium has been formed in which most of the hard scattered partons lose energy. The situation is quite different for direct photons, which are (mostly) produced by hard scattering. Albeit with large errors, the nuclear modification factor for photons is consistent with one almost everywhere except the last points. This is what one would expect if they are produced in hard scattering just as in p+p and once born they "sail through" the medium without further interaction. This may sound as a fairly dull observation, (no surprises, no new phenomena?) but it is actually very significant because it validates the concept of scaling p+p data to characterize medium effects.
But the plot is thicker than that. While the explanation given above is plausible, it is by no means the only possible one. Important as it is, the nuclear modification factor is a fairly "global" quantity, and it might well happen that in Au+Au collisions new processes are at play (absent in p+p), some of which enhance, others maybe deplete the direct photon yield, and the nuclear modification factor is unity only because they accidentally cancel each other's effect. In fact, we know from first principles (the different fractional quark charge content of protons and neutrons) that even in the absence of nuclear effects the scaling from p+p to Au+Au should not be perfect for photons - and the last two points on Fig. 2 may be a hint of this "trivial" effect. At medium transverse momenta it is almost certain that there are new processes at play, they just seem to balance each other's effect - we can investigate this by measuring azimuthal asymmetries of photon emission and the correlation of photons with other particles. (Read more about the current status and future of photon measurements here.) The challenge is enormous - this is one of the hardest analyses - but the prize is big, too. True to their name, photons can shed light on the most intricate, otherwise inaccessible inner details of what is happening in a relativistic heavy ion collision.