[Paper Review] Fast Reconstruction and Data Scouting

J. Duarte|arXiv (Cornell University)|Aug 2, 2018

Advanced Data Storage Technologies2 references4 citations

TL;DR

This paper presents data scouting—a high-rate, low-overhead technique for LHC experiments that uses online reconstruction of trigger-level physics objects to record events at 2–6× higher rates than standard triggers. By saving only compact, reconstructed object data instead of full raw data, it enables sensitive searches for new physics, such as dijet resonances, with performance comparable to offline reconstruction, as demonstrated in CMS and ATLAS Run 2 analyses.

ABSTRACT

Data scouting, introduced by CMS in 2011, is the use of specialized data streams based on reduced event content, enabling LHC experiments to record unprecedented numbers of proton-proton collision events that would otherwise be rejected by the usual filters. These streams were created to maintain sensitivity to new light resonances decaying to jets or muons, while requiring minimal online and offline resources, and taking advantage of the fast and accurate online reconstruction algorithms of the high-level trigger. The viability of this technique was demonstrated by CMS in 2012, when 18.8 fb$^{-1}$ of collision data at $\sqrt{s} = 8$ TeV were collected and analyzed. For LHC Run 2, CMS, ATLAS, and LHCb implemented or expanded similar reduced-content data streams, promoting the concept to an essential and flexible discovery tool for the LHC.

Motivation & Objective

• Address the limitation of standard LHC trigger systems, which restrict data collection to ~1 kHz due to data volume and reconstruction load.
• Enable the recording of significantly more collision events—especially for rare or high-multiplicity final states—without increasing DAQ or storage burden.
• Maintain sufficient physics object quality for offline analysis by leveraging fast, accurate online reconstruction algorithms in the high-level trigger (HLT).
• Extend the reach of resonance searches to previously inaccessible regions, such as low-mass dijet resonances, by exploiting higher event statistics.
• Provide a scalable, flexible, and resource-efficient alternative to data parking and standard triggering, applicable across LHC experiments.

Proposed method

• Implement dedicated HLT data scouting streams that run in parallel with standard trigger paths, using looser selection criteria to accept more events.
• Perform online reconstruction of physics objects (e.g., jets, muons) using the same algorithms as the standard HLT, but with reduced output format.
• Save only the reconstructed object data (e.g., jet four-momenta, track parameters) in a compact, lossy format, reducing event size by 100–1000× compared to raw data.
• Apply calibration procedures to align trigger-level jet energy scales with offline standards using data-driven methods, such as dijet balance 'tag-and-probe' techniques.
• Use sliding-window fits to model and subtract Standard Model backgrounds in dijet mass spectra, enabling sensitivity to new resonances.
• Leverage the fact that scouting triggers can 'shadow' standard HLT paths, capturing physics objects from events rejected by primary triggers.

Experimental results

Research questions

• RQ1 Can data scouting significantly increase the number of recorded events for rare physics processes without degrading object quality?
• RQ2 To what extent can trigger-level physics objects be calibrated to match offline reconstruction performance for resonance searches?
• RQ3 What is the sensitivity gain of data scouting over standard triggering for low-mass dijet resonances?
• RQ4 How does the performance of scouting compare to data parking and standard triggers in terms of data rate, storage, and reconstruction load?
• RQ5 Can data scouting be generalized to support diverse final states beyond dijets and muons in high-luminosity LHC runs?

Key findings

• Data scouting enabled the collection of 18.8 fb⁻¹ of data at √s = 8 TeV in 2012, demonstrating its viability for high-rate physics.
• In Run 2, CMS and ATLAS used scouting streams to collect 27 fb⁻¹ and 29.3 fb⁻¹ of data, respectively, increasing event statistics by a factor of 2–6 over standard triggers.
• The 2016 CMS dijet search using calo-scouting data achieved sensitivity to resonances with masses between 0.5–1.6 TeV, including a previously inaccessible region below 0.6 TeV.
• ATLAS achieved 0.05% agreement in jet energy scale between trigger-level and offline jets after calibration, enabling high-precision resonance mass measurements.
• The sensitivity to universal quark couplings of a leptophobic Z′ boson was improved by a factor of two or more in the 0.45–1.0 TeV mass range compared to pre-LHC searches.
• The technique was validated across experiments: CMS, ATLAS, and LHCb implemented scouting in Run 2, confirming its role as a flexible and essential discovery tool.