List for the 2014 Hot QCD White Paper Writing Group

["Steffen A. Bass" <bass AT phy.duke.edu>]

Dear WP writing committee,

since I won’t be at the Town Hall meeting this afternoon, I’d just like to 
share my personal thoughts about some important items for our WP with you:

Generally I am very confident that we (as community) and you (as writers) 
have a well-developed plan for the exciting science we wish to do over the 
next decade; for me a useful starting point in formulating this vision would 
be our community WP from two years ago that can be found at the website for 
this meeting (and in case you are looking for the latex source, let me know 
and I’ll be happy to send it to you).

There are two items that were not touched upon in detail in that WP, but that 
I strongly feel should be strengthened in this forthcoming WP:

        * the significant progress we’ve made in developing tools for 
quantitative model-to-data 
          comparisons and how these new tools/techniques will enable us to 
rigorously quantify the
          properties of QCD matter that we are promising as deliverables for 
the next phase of 
          RHIC operations

        * the role that synergy between experiment and theory play in this 
endeavor. Essentially we
          now have three pillars: experiment, lattice/first principles QCD 
and dynamical modeling 
          with effective models of QCD. The synergy is important - missing 
any of the three pillars
          puts the success of our field at risk. However, each pillar comes 
with its resource needs
          and in particular the third pillar has often played 2nd (or rather 
3rd) fiddle in that respect
          (quite often when we talk about computational needs we only think 
Lattice, but that’s 
          absolutely not true anymore). I’m attaching a WP regarding the 
computational needs for 
          RHIC phenomenology to this email that outlines developments and 
needs in that domain
          in significant detail (the WP was drafted by U. Heinz, B. Schenke 
and myself and also has
          Raju’s blessing - it was originally written as input for the 
Computational Nuclear Physics
          Town Hall Meeting).

I’ve also tried to synthesize 3 paragraphs from the model-to-data comparison 
WP and the computational RHIC phenomenology WP that I’m attaching as well, as 
a suggestion in terms of the language/content that I would suggest to include 
in our community WP.

Thank you very much for doing this job - having been involved in the last 
such exercise two years ago, I can fully appreciate the amount of time and 
work that you are investing for the future of our field (and the little 
amount of thanks/recognition you’re going to get…). Please let me know if you 
need anything else I can do for you.

all the best,

Steffen


        
The physics goal for the next decade is to characterize the properties of the 
quark-gluon plasma liquid by quantitative extraction of important medium 
parameters from precision measurements of sensitive observables, including 
hadron spectra, angular distributions and correlations, jet observables, and 
electromagnetic probes. To achieve this goal, detailed comparisons of 
theoretical calculations with a variety of experimental observables are 
necessary. 

These theoretical calculations all require significant computational 
resources and are absolutely essential for the success of the overall RHIC 
program. To perform  meaningful comparisons, for many observables 
event-by-event calculations are required. For example, the azimuthal 
anisotropy in the produced particle distributions is highly sensitive to 
event-by-event fluctuations. Furthermore, the details of fluctuations need to 
be under control when studying observables sensitive to the details of the 
phase transformation and the presence of a critical point.

Experimental data sets for a single beam energy and projectile combination 
tend to surpass several petabytes. The modeling community must not only 
reproduce results extracted from these data sets for numerous classes of 
observables covering numerous collisions, but must also investigate a 
high-dimension (of order two dozen) parameter space. In the last few years 
modelers have taken on the effect of fluctuations of the initial state, which 
requires analyzing hundreds of initial state configurations for a single 
impact parameters. Additionally, as the field begins to analyze and interpret 
data from the beam energy scan, modelers must forego the two-dimensional 
descriptions applied at the highest energies and consider approaches that 
model the dynamical evolution in three spatial dimensions, without the 
simplification of boost-invariance along the beam direction. Combined with 
the increased data number and size of the experimental data sets resulting 
from measurements at many different beam energies and from an increased 
number of collision systems, the numerical demands facing the modeling 
community will grow by 3 to 4 orders of magnitude. 

During the last decade strategies have been developed expressly for comparing 
complex compute-intensive theoretical models to large heterogenous data sets. 
These strategies utilize a combination of state-of-the-art methodologies in 
the statistical sciences and high throughput computing technology that are 
just now becoming part of the scientific toolset in theoretical nuclear 
physics. They integrate Bayesian inference and model surrogate algorithms 
with the modeling environment to constrain the multi-parameter model space by 
comparison with the high-dimensional data from RHIC and LHC. High-throughput 
computing techniques, combined with Gaussian process surrogate models to 
rapidly explore a model’s parameter space, will allow for the computational 
model to data comparison within a feasible amount of time and provide the 
means for multiple iterations of this process. However, obtaining 
quantitative conclusions requires one to faithfully calculate what has been 
measured and to assign uncertainties for comparing experimental observables, 
thus requiring a sustained effort in the development of realistic 
computational models of heavy-ion collisions.

Attachment: RHIC_computing_2014.pdf
Description: Adobe PDF document