(High-energy Theory Event Analyzer for collider processes)

The modern way of doing physics


This site is the public face of a novel approach for distributing and analyzing the results of fixed order NNLO calculations for LHC processes. Here are the main topics:



The basic idea


The idea is very simple: the existing calculations use Monte Carlo methods to integrate over the phase space of the processes of interest. Such MC samplings, called events, are usually directly binned in histograms and then discarded.

This approach to doing high-precision NNLO calculations has serious drawback: it lacks reusability and flexibility. If at a later time a new, and often similar calculation is required (we get a lot of these requests!), one has to perform this calculation from scratch. Such an approach is problematic, for several reasons:

  1. Public codes that can perform such NNLO calculations are scarce.
  2. Any NNLO calculation is very time- and computing-resource- demanding. Typical costs are about 10k-50k CPU hours but could go into the millions. For such calculations large cluster-sized computing infrastructure is needed. Not everyone has access to one, not even counting the associated expenses!
  3. Such computing codes, even if made public, are too complicated for most users to use.

The question then arises: is there a way to completely eliminate the above obstacles? The answer is yes!

As a result of many years of combined experience with developing and performing NNLO calculations our team has shown that it is possible to:

  1. Store the events computed in the course of a calculation.
  2. Not all computed events need to be stored thanks to a partial un-weighting technique we have employed. This way we can reduce the number of events that need to be stored which is essential for making the size of the database manageable.
  3. Simple things should be simple: the user do not really need access to the events themselves. We have created our own sophisticated internal infrastructure, which allows the users to run their own analysis over the database of events without ever needing to worry about computing, infrastructure and associated costs. They just get their results in (typically) minutes and that’s that!


How it works


Here is how it works:

  • The user submits their request. Here is the info about how this is done.
  • The request starts an analysis. It can take from few minutes to about half an hour depending on the requested options and how busy our servers are.
  • The results are returned to the user in the form of histograms, potentially accompanied by a nice plot.

That’s it!


Main features


Here are some of the features of the analysis:

  • It is completely generic: any InfraRed safe observable can be binned.
  • The binning is in terms of kinematic variables. A set of predefined standard variables specific to each process will be available to the user. The user can also specify their own kinematic variables; options are practically unlimited!
  • Arbitrary cuts can be implemented.
  • The user can either use the default predictions in terms of scales (factorization and/or renormalization) and parton distributions. A major new feature of our library is that it makes it possible to dynamically change the scale and pdf during the analysis. This will allow the user to obtain predictions based on any pdf set and scales!
  • The predictions return an estimate of the statistical error due to the finite size of the event sample.
  • Quality: the results derived from the library are not in any way inferior to published results. They are as good as a brand new calculation started from scratch.
  • Users can analyze any one of the existing datasets. The list of available processes will be made available so the user can select among them.
  • We will constantly be adding new processes. Please let us know if your favorite is not among them.


How to access the library


There are two ways to access the library:

  • Via ready-to-use Jupyter notebooks on Google Colab. Suitable for all users. No special software or hardware infrastructure is required (it can even be accessed via standard smart phone). No programmatic skills are needed to fully utilize all available options.


Our Team

Meet the team:

To contact us, please email: hightea@hep.phy.cam.ac.uk