Transcript Slide 1

Direct photons at RHIC (and other issues…)
or
a quest for the forest
G. David, BNL
Even good ideas can get too much ingrained in our thinking
1
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Real photons are EM probes, too 
…sharing many advantages with dileptons – like virtually infinite
mean free path after created
Less (read: no) handle on when during the system evolution they
were produced (as opposed to dileptons, the “mass dimension”)
But produced at much higher rates (high pT reach) ,
and the mechanisms are (somewhat) better understood
(qualifiers, qualifiers everywhere…!)
Irreplaceable in proving that in-medium energy loss studies make sense
(and measuring things like gluon PDFs)
Revealing some real puzzles on thermalization and collective expansion
Apologies in advance: I’ll spend about a third of this talk on an issue
which at first glance is completely unrelated – but it is in fact a burning
issue where we need all the help the community can give
(and not even completely unrelated to photons)
2
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
High pT photons in p+p  pQCD testbench
PRD 86, 072008 (2012)
Good agreement with pQCD
slight preference for pT/2 scale
Two methods
Isolated/all direct photons:
 small contribution from fragmentation
3
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
High pT photons in p+p  pQCD with flying colors
Simply put:
one of the most glorious and
beautiful plots
PRD 86, 072008 (2012)
Picture-perfect* agreement between
theory and data, over many orders
of magnitude in collision energy
(and it even includes PHENIX low pT)
Just one outlyer (E706) – maybe
understood (but that’s a different talk)
Universal n=4.5 exponent in coll.
energy dependence  LO dominates,
(would give n=4), PDF stable, …
(*) It isn’t, but with the current experimental
errors it would be arrogant to complain 
Here we certainly seem to see both the
trees and the forest
4
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
If photons in p+p are understood, try heavy ions
(“yesterday’s discovery is today’s calibration”)
Or as the old Viennese saying goes:
“Warum denn einfach, wenn es auch kompliziert geht?”
jet fragment photon
jet
annihilation
compton
scattering
v2 > 0
Bremsstrahlung
(energy loss)
v2 < 0
…and this is only the high pT, i.e. the “easy” part…
5
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Blessing or curse? -- my 1994 cartoon, updated
hard scatt
“Historians”,
but very-very
hard to read!
jet Brems.
parton-medium interaction
jet-thermal
sQGP
hadron gas
g*  e+evirtuality
0.5
1
Mass
1
10
107
log t
(fm/c)
By selecting masses, hadron decay backgrounds are
significantly reduced. (e.g., M>0.135GeV/c2)
(GeV/c2)
6
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Blessing or curse? -- what does measured “T” mean?
The temperature-history:
interpretation of experimentally
fitted T is not trivial (depends on
model). May be OK as lower
limit.
Dielectrons (mass-dependent T)
to the rescue…
7
arXiv:1304.7030
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
The Ianus-faced photons in heavy ion collisions
The most direct observables
from the medium itself
PRL 104, 132301 (2010)
The cleanest probes of pQCD, IS:
they couldn’t care less about the
medium
PRL 109, 152302 (2012)
arXiv:1212.3995
8
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
The low pT (“thermal”) region – from p+p to A+A
arXiv:1208.1234
No excess in p+p,
apparently no excess
in d+Au,
substantial excess in Au+Au
in the pT region where
thermal radiation would be
expected
Note: lack of “thermal”
radiation in d+Au 
isn’t this evidence against
collectivity (in the hydro
“flow” sense)?
9
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Jet quenching
Relies on “binary scaling” and experimental handle on collision geometry,
which in turn is “proven” by direct photons
The most quoted single
result RHIC paper
Direct photon RAA proves that
binary scaling makes sense!
10
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
The high pT (“pQCD”) region – from p+p to A+A
Scaling is not perfect – partially explained by isospin effect
arXiv:1208.1234
PRL 109, 152302 (2012)
11
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
A big relief: Ncoll scaling makes sense
(at high pT)
PRL 109, 152302 (2012)
The basic tenets behind all “Eloss”,
“jet quenching” and “tomography”
- hard probes are produced before
any medium, collectivity emerges
- for hard probes A+A is an incoherent
superposition of p+p collisions
- the proportionality (Ncoll) can be
derived from simple geometry and s
(analytic or MC Glauber)
Since photons (almost) don’t interact
with the medium, they should be
uneffected  as they apparently are
Small perturbations (like isospin effect)
possible, but the fundamental picture
seems to hold
12
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Additional evidence
If the basic tenets hold, “flow” of high pT photons should be about zero
(fragmentation, jet-medium photons may modulate the picture)
And indeed, they are:
PRL 109, 122302 (2012)
The sources are predominantly jets.
RP measured “close” to the jet: bias
RP measured “far” from the jet
13
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photons at low pT – rates only
Shown in a zillion different versions, same conclusion: direct photonspectra alone,
while important, not sufficient constraint
14
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photon flow at low pT – is it real?
Initially treated with a liberal dose of
scepticism, but finally got accepted
for publication (around the same
time when ALICE made the similar
observation in Pb+Pb)
PRL 109, 122302 (2012)
QM’12, arXiv:1212.3995
15
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photon flow at low pT – confirmation
Before QM’11, when we released the
photon flow paper, PHENIX was also
worried, since the analysis is tricky.
So we didn’t release anything until we had
- for internal consumption only back then –
a completely independent confirmation:
External conversions: low rates,
but excellent resolution, good
particle ID.
While challenging, all difficulties
are – so to speak – “orthogonal”
to the other method; really
independent confirmation
16
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photon flow – where does it come from?
PRC 79, 021901 (2009)
The easiest way to get high
rates is high (early)
temperatures  but no flow
there yet, just acceleration
The easiest way to get high flow
is late (long acceleration),
just before kinetic freeze-out
but lower (thermal) rates
Having both high rates and high
flow is something like
“having your cake and eating
it, too”, tantalizing theorists
for years now.
The mantra: you have to explain yield and flow simultaneously!
17
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photon flow – play with aT (fireball acceleration)
If true, “QGP window” is essentially gone (QGP is not
the dominant source at any pT), and the large
apparent temperature is mostly of hadronic
(+ blue shift) origin.
Van Hees, Gale, Rapp
PRC 84, 054906 (2011)
18
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photon flow – play with time
F.-M. Liu
Early hydro initial time, QGP forms considerably later
(0.6 f/c vs QGP formation times up to 2.1 f/c)
 early emission (no flow part) was overestimated
arXiv:1212.6587
Q: what is the emission between thydro and tQGP? Apparently unanswered
(looks a bit like a “fiat” type theory so far  where’s the forest?
19
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photon flow – play with magnetic field
PRL 110, 192301 (2013)
Only a few words, since the author is going to
speak tomorrow
Origin (at least in part) of the large photon
flow could be the strong magnetic field?
I loved this paper, because
it explicitely told what could
disprove the theory!
A simple exercise is here:
(it could use smaller
error-bars…)
Also, v4 is in the works,
coming soon
20
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Direct photon flow - PHSD
“Large direct photon v2 … attributed to intermediate hadronic
scattering channels and resonance decays not subtracted
from the data”
21
arXiv:1304.7030
Runs counter “conventional” wisdom (which gets most of the flow from the
hadronic phase). Interesting, if it holds up. Centrality dependence?
Also, claiming a good description of the rate may be a stretch.
(And yes, omega is subtracted.)
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Orwell’s geometry:
all collisions are NOT created equal
arXiv:1304.3410
Ratios of identified hadron spectra
in peripheral Au+Au and central d+Au
Both Npart and Ncoll virtually identical
(eccentricity of course is not)
The ratios are constant (up to the highest pT)
but not one! (0.65)
Isn’t the Glauber counting too simplistic?
Is a “collision” in Au+Au the same thing
as in d+Au (or p+Au)?
Of course it isn’t…
22
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Now some results to lose sleep over
PHENIX preliminary, QM’12
2008 (high) statistics d+Au data, nuclear modification factors vs centrality
Is it possible that p0, h production at high pT in peripherals is enhanced???
23
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Jets, p0, h – central to peripheral
For p0, h this is true pT,
for jets it is total jet energy.
There is no unambiguous
transformation, but 1./0.7
is a reasonable compromise,
and would put the points
on top of each other.
Important: RCP is independent
of any p+p reference!
The only “external” quantity here
is the Ncoll value attributed to
the individual centrality classes
Note that RCP drops sharply, indicating major shape change from peripheral
to central
24
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Centrality: thinking out loud
The theorist tends to think in terms of impact parameter (b), or Npart, Ncoll, TAB, e, …
none of which is directly accessible in the experiment
The experimenter is concerned whether a/ the event is taken at all (trigger bias/efficiency)
b/ there are some global observables that can be tied to the theorists’ quantities and
while they are correlated to those quantities, they are as uncorrelated as possible
to the specific features of the event (like presence of jets, flow, etc.)
Assuming such observable(s) exist, a model is agreed upon that makes the translation
between experimental observables and theoretical quantities
Since you want to avoid introducing biases as much as possible, the model is tuned
with a large number of (more or less) average events, in regions preferably “far”
from the regions with the “specific features” studied (like a large h gap)
The correlation between the global observable and the theoretical quantity is typically
wide: events on the average will be properly classified – but not necessarily individually.
25
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
The verifiable case: p+p
Triggering and event characterization:
looking for activity (e.g. charged particle production Nch,
transverse energy ET)
preferably close to the beam and far from the
region of interest (mid-rapidity)
Typical Nch dist.
close to the beam
for average p+p
Now study those distributions as a function of
the activity observed at h~0
“Activity” here is the highest pT for any particle
seen around h~0; could be jet energy, etc.
Can be done both in simulation and in data!
Mean and RMS of the Nch dist. vs max pT
in the center
Trigger efficiency vs max pT
in the center
Note the characteristic
rise initially (well-known:
higher activity when
hard scattering occurs)
However, at higher pT
they start to drop slowly.
They have to, at least
asymptotically, for simple
kinematic reasons.
Of course other mechanisms can deplete forward activity way before kinematics does!
26
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Glauber-model and centrality in p+A, d+A, …
Straight path, independent collisions with the
same probability (cross section)  Ncoll, Npart
Folding with the average response observed in
p+p can tie Ncoll, Npart to observed Nch statistically
Weather or not fluctuations are taken into account
is irrelevant here
For instance:
20-40%
40-60%
60-88%
Charge distribution in BBC
(South, gold going direction)
0-20%
Experimentally defined centrality classes
Ncoll distribution for each class
from the model
Based on average responses, does not take into account possible special features
of rare events (like high pT particle or jet in the central region)
27
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Will this always work without further corrections?
Not necessarily.
For instance, as we have seen for p+p, the trigger efficiency decreases with increasing
energy in the center. Since the trigger requires coincidence on both sides and in
pA, dA on one side there are at most two nucleons, a similar drop in efficiency is
expected. This is well known and usually taken into account.
Centrality is usually defined in the direction where the large ion goes. Assume the
projectile makes N collisions, one of them with very high pT. Then the expected multiplicity
forward is only (N-1) times the average plus one reduced response
 the multiplicity observed by the experimenter (forward) is smaller than it would be
for an event that is identical except that no high pT is present
If centrality is defined with fixed multiplicity thresholds based on the average events
but applied to the rare, special ones, those rare events may be (mistakenly) classified
as lower centrality (lower average Ncoll) than they really are.
At higher pT this effect typically shifts to lower multiplicity (i.e. lower centrality) classes
events that Ncoll-wise – i.e. from the point of view of how probable a rare,
hard collision is – would belong in a higher centrality class.
28
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Illustration: shift between multiplicity classes / 1
Here is your average,
higher centrality event
True b, Ncoll
But now a very hard scattering happened (one in a
million!), with reduced fwd. response, therefore…
Expected
fwd. mult
True b, Ncoll
Expected
fwd. mult
…this is how you classify
the event…
Observed
fwd. mult
Percieved b, Ncoll
…and when you calculate RAA,
the denumerator (Ncoll * spp)
will be smaller than it should be
 RAA increases
(There can be other, even more
serious effects, as we’ll theorize later)
29
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Illustration: shift between multiplicity classes / 2
Charge distribution in BBC
(South, gold going direction)
This is where
the event
should be
20-40%
60-88%
Lost!
Trig. ineff.
40-60%
This is where
it is actually found
0-20%
This is where
it is actually found
This is where
the event
should be
If (experimental) centrality is determined with fixed (forward) multiplicity thresholds,
irrespective of what happened at h~0, events may end up in the wrong centrality
class – and attributed an incorrect <Ncoll>
30
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
More exotic possibilities
Indavertent confusion from the dual use of Ncoll (???)
We use it both to estimate the average soft response by folding the p+p distribution
(which assumes that the likes of Ncoll average p+p collisions in fact do happen in
the event)
but then we also use Ncoll to estimate how much an extremely rare p+p
process (hard scattering) is enhanced in p/d+A,
where it is still very-very rare (<<1/event)
But in those very rare instances when hard scattering did in fact happen,
will the d/p nucleon for the rest of its path interact with the remaining A nucleons
aa the original, intact nucleon (i.e. with the same spp a la Glauber?)
If not, what will happen?
Will it keep interacting, but with reduced cross-section (like spp)?
Will it be completely out of the pool (no more soft production whatsoever?)
Something in between? If so, what? Wounded or amputated nucleon?
31
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Can this be tested?
Reduced/vanishing cross-section in a different context:
Papp, Levai, Barnafoldi, Zhang, Fai -- nucl-th/0203075
High pT biases
Renk, arXiv:1212.0646
Would comparison to LHC help?
If (with similar centrality determination) LHC would see no effect in our pT range,
but similar effect at higher pT, the “kinematic” effect (depletion of available energy
forward) could be the culprit (or dominant)
If LHC would see a similar effect already in our pT range, the “dynamic” effect
(reduced or vanishing cross section) could be the dominant contributor
32
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Just to avoid confusion / misinterpretation
The Glauber-model is adequate and working for what is was originally meant
(soft physics, average events and / or very large systems)
The fact that the presence of a high pT particle biases distributions far away in
rapidity, is not only a kinematic triviality, but also proven by data
In A+A getting one nucleon “out of the pool” barely changes the global event
(not even in peripheral)
However, in d+A (or even worse, in p+A) once a hard collision happened,
one nucleon (or the nucleon!) of the projectile may be “out of the pool”,
 the global event changes drastically. Applying the same centrality
classification as for the average event may be misleading!
This is a very serious problem since we know little, if anything about what
does a nucleon do after making a hard collision – while currently we treat
this case as if nothing happened, kept interacting like an intact one…
33
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL
Summary
Direct photons are a perfect tool to understand various phases of the collision –
if we only knew how to interpret them
At high pT the appear to “behave” – reasonably well described from p+p to A+A
 increased experimental precision may lead to disentangling finer effects
(like positive/negative flow for isolated/non-isolated photons). No imminent
panic here, just years of work ahead
However, the way we characterize event geometry, seems to fail in extreme cases,
like very asymmetric systems, large pT “one in a million” type events
Time to re-think how we use
the Glauber-model?
Unexpected photon v2,
long-range jet correlations in d+Au,
rapidly rising RAA, …
Nature punishes us if we
get complacent , nevertheless
Don’t cut it:
put in proper perspective!
Even good ideas can get too much ingrained in our thinking
34
EM probes… ECT*, Trento, May 20-24, 2013 -- G. David, BNL