A high-precision track reconstruction method for a GeV-level proton beam telescope
By employing a high-current-intensity mode and a layer-by-layer state prediction intelligent search method in the beam telescope, the problems of multiple scattering and multi-particle accumulation under low-energy proton beams were solved, achieving high-precision track reconstruction and high-efficiency detector performance calibration.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA SPALLATION NEUTRON SOURCE SCI CENT
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-10
AI Technical Summary
Existing beam telescopes struggle to simultaneously address the issues of multiple scattering interference and multi-particle accumulation identification under low-energy, high-current conditions, leading to deterioration or failure of reconstruction capabilities.
An intelligent search method based on layer-by-layer state prediction under high current intensity mode is adopted. Through multiple Coulomb scattering models, the track is searched and pruned layer by layer to optimize the track. Combined with the data acquisition system, the location information of all hit clusters is read synchronously to reconstruct the track with high precision.
High-precision track reconstruction was achieved under GeV level proton beams, improving position resolution and detection efficiency, solving the bottlenecks of multiple scattering and multi-particle accumulation, and improving computational resource efficiency by 5 times.
Smart Images
Figure CN122362469A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to high-energy physics and nuclear detection technology, and in particular to a high-precision track reconstruction technology for a GeV-level proton beam telescope. Background Technology
[0002] Beam telescopes are core equipment for calibrating the position measurement capabilities of advanced particle detectors (such as silicon pixel detectors, microstrip detectors, time projection chambers, drift chambers, etc.). They reconstruct the precise tracks of incident particles through multiple position-sensitive reference detectors, and then measure key parameters such as the position resolution and detection efficiency of the detector under test (DUT).
[0003] The structure of a beam telescope is as follows: Figure 1 As shown, the key components and their functions are as follows:
[0004] Beam: A particle beam composed of high-energy protons, used as the incident source for calibrating the detector under test, possessing a specific initial angular dispersion and momentum distribution. The proton test beam under construction in China for HPES has an energy of 1.6 GeV.
[0005] The beam telescope (reference system) consists of six high spatial resolution silicon pixel detectors arranged sequentially along the beam propagation direction. Each detector is an independent position-sensitive unit; when a particle passes through it, it generates a hit signal, which represents the particle's two-dimensional coordinates (x, y, y) on the detector plane. p ,y p The detector indexes p=1,2,…,6 provide a reference baseline for subsequent track reconstruction.
[0006] The detector under test (DUT) is a new type of particle detector located in the detection path of the beam telescope reference system (usually between or at the end of the six silicon pixel detectors). Its performance (such as position resolution and detection efficiency) needs to be calibrated. Its working principle is the same as the reference detector, synchronously recording the hit position (x-axis) of the particle as it passes through. DUT ,y DUT ), used for comparative analysis with reconstructed tracks.
[0007] Overall working logic: Protons in the incident beam pass sequentially through the six reference detectors of the beam telescope and the DUT. The multiple sets of two-dimensional coordinates recorded by the reference detectors are the raw data. The motion trajectory of the particles is reconstructed by mathematical fitting algorithms (such as linear fitting and polyline fitting) to form a reconstructed track. Finally, based on the difference between the reconstructed track and the hit position recorded by the DUT, the performance of the DUT is calibrated.
[0008] like Figure 2 As shown, the principle of track reconstruction is as follows:
[0009] By exaggerating the particle scattering effect, the relationship between the real track and the reconstructed track is clearly presented:
[0010] Actual Track: The solid line represents the actual trajectory of the high-energy proton as it passes through the entire system. Due to the Coulomb interaction between the proton and the atomic nuclei of the detector material, multiple scattering occurs (especially at low-to-medium energy beams such as 1.6 GeV, where the scattering effect is more significant), resulting in the actual track not being an ideal straight line, but exhibiting a slight directional deflection (this effect is highlighted in the figure by magnifying the deflection angle).
[0011] Reconstructed Track: The dashed line represents the trajectory estimate obtained through mathematical fitting based on the hit positions of six reference detectors. The fitting process comprehensively considers the initial motion direction of the particles and the scattering characteristics of the detector materials, striving to offset the deviations caused by multiple scatterings as much as possible, ultimately forming an optimal approximation of the true track. Figure 2 It is evident that scattering causes the actual track to deviate from the ideal straight line.
[0012] The effectiveness of traditional beam telescopes is closely related to the beam energy: under high-energy beam conditions (typically above 5 GeV), the probability of particles undergoing multiple scattering is low, and sufficient statistical data can be obtained without extremely high current intensity, thus achieving telescope position resolution at the micrometer level. However, in beam scenarios with relatively low energy (1.6 GeV) such as HPES, traditional technologies face significant bottlenecks, specifically the following contradictions and shortcomings:
[0013] The contradiction between multiple scattering and statistics: Under low-energy beams, the multiple scattering effect is significantly enhanced when particles pass through the detector, severely interfering with the accuracy of track reconstruction algorithms. Traditional optimization approaches to address this issue involve directly discarding instances with significant multiple scattering. While this improves the reconstruction quality of individual tracks, it drastically reduces the utilization efficiency of available instances, leading to insufficient effective statistics and ultimately failing to improve the overall position resolution of the telescope.
[0014] The contradiction between high flux intensity and multi-particle accumulation: Theoretically, this can be compensated for by increasing the proton beam intensity, but high flux intensity introduces the problem of multi-particle event accumulation. Existing beam telescope data analysis frameworks (such as Corryvreckan) only have track screening functions based on goodness of fit (χ²) or single-point residual thresholds, which are only applicable to low flux intensity and single-event conditions. In scenarios with severe accumulation caused by high flux intensity, their discrimination ability is severely insufficient, and they cannot effectively separate and identify multiple particle tracks in the accumulation, resulting in a significant deterioration or even complete failure of reconstruction capabilities.
[0015] In summary, traditional track reconstruction techniques cannot simultaneously solve the dual problems of multiple scattering interference and multi-particle accumulation identification in low-energy, high-current application scenarios. Summary of the Invention
[0016] The technical problem to be solved by this invention is to provide a track reconstruction and screening method that overcomes the technical bottleneck of high-precision calibration under a fixed low-energy beam by effectively utilizing the adjustable current characteristics of existing track reconstruction technology to meet the accuracy and efficiency requirements of detector performance calibration.
[0017] The technical problem to be solved by this invention is to provide a high-precision track reconstruction method for a GeV-level proton beam telescope, comprising the following steps:
[0018] S1. Set the proton beam intensity to a high-current mode so that M particle events are generated simultaneously within a single trigger time window. The data acquisition system synchronously reads and stores the position information of all hit clusters generated on all reference detectors. The reference detectors are several position-sensitive detectors arranged sequentially along the beam direction.
[0019] S2. Employing a layer-by-layer state prediction intelligent search method based on a multiple Coulomb scattering model, starting from the first layer of reference detectors, track reconstruction search is performed on the hit clusters, specifically including:
[0020] (1) Initialization and first layer scattering: Starting from a cluster of hits on the first layer reference detector, the initial angular dispersion standard deviation of the incident beam is obtained, and the standard deviation of the scattering angle caused by the particles passing through the material of the first layer reference detector is calculated; the total angular dispersion standard deviation after passing through the first layer is calculated based on the initial angular dispersion standard deviation and the standard deviation of the scattering angle after passing through the first layer, and a circular candidate region is determined on the plane of the second layer reference detector accordingly.
[0021] (2) Second layer matching and branch generation: Search for hit clusters on the second layer reference detector within the circular candidate region, and continue, split or terminate track branches according to the matching results;
[0022] (3) Progressive prediction based on the current branch: For a confirmed track segment, its direction is updated to the direction of the line connecting the last two hit clusters that constitute the track segment. Based on this direction and the standard deviation of the scattering angle caused by the current layer reference detector material that the particle is about to pass through, the total angular dispersion standard deviation after passing through the current layer is used to calculate the new candidate region on the next layer reference detector plane. Starting from the second layer reference detector, the total angular dispersion standard deviation is determined only by the standard deviation of the scattering angle caused by the current layer material.
[0023] (4) Iterative search and pruning: Repeat step (3) and advance layer by layer. In each layer, search for clusters only in their corresponding candidate regions and determine the continuation, splitting or termination of track branches based on the matching results until the last layer reference detector, and output the candidate track set.
[0024] This invention proactively leverages high current intensity to obtain high statistical values, employing intelligent algorithms to ensure high-quality datasets. It transforms the traditional sequential approach of independent reconstruction followed by line-by-line filtering into an intelligent approach of layer-by-layer search to find the optimal track. While maintaining reconstruction accuracy, it fully considers the feasibility of engineering applications. Through optimized algorithm design, typical M=4 or M=5 stacks can be solved precisely within milliseconds. In beam telescope applications, limited computing resources are used to achieve orders-of-magnitude improvements in reconstruction accuracy and efficiency; efficiency can be increased to N times that of traditional reconstruction algorithms. 5 times.
[0025] The beneficial effects of this invention are:
[0026] To address the problem of severe multiple scattering of protons at the GeV level, a strategy combining high current intensity and efficient track reconstruction algorithm was designed to reduce the beam energy suitable for detector testing of the proton beam telescope to 1.6 GeV.
[0027] A scheme is proposed to use the pruning algorithm to process track reconstruction under high packing conditions, thereby improving the accuracy of track reconstruction under severe multiple scattering conditions. Attached Figure Description
[0028] Figure 1 Beam telescope structure;
[0029] Figure 2 The principle of track reconstruction (exaggerated scattering angle demonstration);
[0030] Figure 3 A schematic diagram of intelligent search with layer-by-layer state prediction;
[0031] Figure 4 This is a flowchart. Detailed Implementation
[0032] The core of the method of this invention lies in setting the beam intensity to a preset high current intensity mode, so that M particle events are generated simultaneously within a single trigger time window, thereby using a data acquisition system to synchronously read and store the positions of all hit clusters generated on all reference detectors.
[0033] Within a single time window, assuming each detector records N clusters, selecting one cluster from each of the six detector layers and connecting them can form a potential track. The total number of possible track combinations will reach N. 6 The simplest data processing method is to reconstruct each of these track combinations and select the optimal solution, but this method obviously wastes a huge amount of computational resources.
[0034] In the specific search process of track reconstruction, this invention employs an intelligent search method based on a multi-coulomb scattering model for layer-by-layer state prediction. Because the change in particle orientation only occurs when passing through the detector material, in each search step, only the orientation of the currently determined track segment is used, superimposed with the scattering angle broadening introduced by the next layer of material the particle is about to pass through, to dynamically define the reasonable search area for the next layer's impact point. This achieves fast and accurate track tracking and invalid branch removal.
[0035] Suppose the set of hit clusters recorded on the p-th layer detector. , where N p Let p be the total number of clusters hit in the p-th layer, with detector indices p = 1, 2, ..., 6. For the cluster index of the p-th layer, For the p-th layer A cluster, =1,2,…, .
[0036] A candidate track can be represented as an ordered sequence of cluster hits. ,in ∈H p , where i is the candidate track index.
[0037] Taking the search process starting from the cluster hit by the first-layer detector D1 as an example, the specific process of intelligent search is as follows: Figure 4 As shown:
[0038] (1) Initialization and first-layer scattering: Track reconstruction of the hit clusters identified from the first-layer detector D1, for example Figure 3 (a) from h1 (1) Beginning. The beam itself has an initial incident direction with angular divergence. Initial direction of the track. There is some angular dispersion (scattering angle broadening), and the initial standard deviation of the angular dispersion of the incident particles in the beam is... This describes the initial angular uncertainty of the beam before it enters the detector system. Based on this, the distribution of multiple Coulomb scattering angles caused by particles passing through the material of the first layer of detector D1 is superimposed. It is the standard deviation of the scattering angle θ of the multiple Coulomb scatterings that occur when a particle passes through the first detector. The standard deviation of the scattering angle θ is the same as that of the multiple Coulomb scatterings that occur when the p-th detector passes through it. , ,in, Let be the momentum of the incident particle. Let be the speed of light, β be the ratio of the particle's velocity to the speed of light, and βc be the velocity of the incident particle. It is megaelectron volts. It is an empirical constant. The thickness of the detector material, The radiation length of the detector material.
[0039] The standard deviation of the beam divergence after passing through the first layer detector D1 A circular candidate region is formed by geometric projection onto the D2 plane of the second-layer detector.
[0040] (2) Second-level matching and branch generation: On the D2 plane, check whether there is a hit cluster in the candidate region calculated in the previous step: If there is a unique hit cluster, such as h2 (1) Then h2 (1) With h1 (1) Confirmed as a single track segment; if multiple hit clusters exist, such as Figure 3 As shown in (a), there is a hit on cluster h2. (1) and h2 (2) This will generate two branches, such as Figure 3 As shown in (b) and (c), each branch represents a possible local track, h2 (1) With h1 (1) h2 is a track. (2) With h1 (1) This is a trajectory; if no cluster is hit within the region, this initial directional branch is pruned, and the particle is considered to be at h1. (1) If large-angle scattering occurs, the hit is discarded.
[0041] (3) Progressive prediction based on the current branch: For any path confirmed in the previous step (e.g., by h1) (1) and h2 (1) The track segment formed is then updated with the direction of the line connecting the two measured points. Then, based on this new direction, the angular dispersion standard deviation caused by the material of the D2 detector that the particle will subsequently pass through is superimposed. At this point, the angular dispersion standard deviation is only related to the second layer detector, and the calculation formula is... Therefore, new candidate regions on the D3 plane of the third-layer detector can be calculated. That is, starting from the second-layer detector, the region traversed by the D... p Angular deviation caused by detector material .
[0042] (4) Iterative search and pruning: Repeat step (3) layer by layer. At each layer, the algorithm searches for hit points only within the physically reasonable area predicted based on the latest measured state, and decides whether to continue, split, or terminate the track branch based on the matching results, such as unique match, multiple selections, or no match. This process continues until the last layer of detectors, and finally outputs a dataset of all possible high-quality tracks.
[0043] (5) The track dataset output in step (4) is directly input into the subsequent standard analysis process (such as residual calculation and DUT resolution fitting) to calibrate the position resolution and detection efficiency of the detector under test. Thus, the present invention has achieved the goal of stably outputting high-precision track data under the conditions of GeV level proton beams, significant multiple scattering and multi-particle accumulation, overcoming the fundamental contradiction of the traditional method's reconstruction capability deteriorating or even failing under high current intensity.
[0044] The final output of step (4) is a physically plausible and high-quality set of candidate tracks. This set has the following characteristics:
[0045] For accumulation: By using multi-cluster association and branch pruning within the time window, multiple particle tracks in accumulation cases are effectively separated.
[0046] For multiple scattering: By using dynamic search region constraints based on the scattering model, spurious tracks caused by large-angle scattering are suppressed.
[0047] The output track set satisfies the following: each track has a unique cluster match on the six-layer detector; the track direction variation conforms to the physical expectations of multilayer Coulomb scattering; and it can still maintain high-purity track reconstruction efficiency under high current intensity (M=4−5) stacking conditions.
[0048] This layer-by-layer state prediction intelligent search method effectively achieves the elimination of false positives and retention of high-quality ones: abnormal and inferior tracks caused by multiple scattering effects are effectively removed during the pruning process; while high-quality tracks with good penetration and no multiple scattering are retained as a set of high-quality track examples for DUT performance calibration, thereby improving the quality of track reconstruction while ensuring data statistics.
Claims
1. A high-precision track reconstruction method for a GeV-level proton beam telescope, characterized in that, Includes the following steps: S1. Set the proton beam intensity to a high-current mode so that M particle events are generated simultaneously within a single trigger time window. The data acquisition system synchronously reads and stores the position information of all hit clusters generated on all reference detectors. The reference detectors are several position-sensitive detectors arranged sequentially along the beam direction. S2. Employing a layer-by-layer state prediction intelligent search method based on a multiple Coulomb scattering model, starting from the first layer of reference detectors, track reconstruction search is performed on the hit clusters, specifically including: (1) Initialization and first-layer scattering: Starting with a hit cluster on the first-layer reference detector, the initial angular dispersion standard deviation of the incident beam is obtained. And calculate the standard deviation of the scattering angle caused by the particle passing through the first layer of reference detector material. ;according to and Calculate the standard deviation of total angular dispersion after passing through the first layer. Based on this, a circular candidate region is determined on the plane of the second-layer reference detector; (2) Second layer matching and branch generation: Search for hit clusters on the second layer reference detector within the circular candidate region, and continue, split or terminate track branches according to the matching results; (3) Progressive prediction based on the current branch: For a confirmed track segment, its direction is updated to the direction of the line connecting the last two hit clusters that make up the track segment, based on this direction and the standard deviation of the scattering angle caused by the current layer reference detector material through which the particle is about to pass. The standard deviation of total dispersion after passing through the current layer p is the detector index, and the new candidate region is calculated on the plane of the next layer reference detector; starting from the second layer reference detector, the total angular dispersion standard deviation is determined only by the scattering angle standard deviation caused by the current layer material; (4) Iterative search and pruning: Repeat step (3) and advance layer by layer. In each layer, search for clusters only in their corresponding candidate regions and determine the continuation, splitting or termination of track branches based on the matching results until the last layer reference detector, and output the candidate track set.
2. The method as described in claim 1, characterized in that, The standard deviation of the multiple Coulomb scattering angles The calculation formula is: ； in, Let be the momentum of the incident particle. Let be the speed of light, β be the ratio of the particle's velocity to the speed of light, and βc be the velocity of the incident particle. It is megaelectron volts. It is an empirical constant. The thickness of the detector material, The radiation length of the detector material.
3. The method as described in claim 1, characterized in that, The hit cluster is a cluster formed by clustering the signals generated on several adjacent pixels when a particle passes through the detector, and its position is represented by the centroid coordinates of the cluster.
4. The method as described in claim 1, characterized in that, The range of values for M is determined by the current intensity.
5. The method as described in claim 1, characterized in that, The number of reference detectors is 6.
6. The method as described in claim 1, characterized in that, The output set of candidate tracks is used to calibrate the position resolution of the detector under test.