CEPC Reference Detector Overview
- The CEPC Reference Detector is a baseline design that defines performance metrics and guides subsystem R&D for the Circular Electron Positron Collider.
- It features a particle-flow oriented architecture with TPC-based tracking, silicon vertex and calorimeter systems, ensuring high-resolution measurements.
- Ongoing R&D and prototype tests, from MAPS sensor development to TPC optimization, validate the detector’s ability to meet demanding operational targets.
The CEPC Reference Detector denotes the baseline detector definitions used to quantify detector performance, guide subsystem R&D, and support project-level design decisions for the Circular Electron Positron Collider. In the 2018 Conceptual Design Report, the practical reference detector is the particle-flow baseline with a silicon vertex detector, silicon inner and outer tracking, a large TPC, high-granularity ECAL and HCAL, a solenoid, and a muon detector in the return yoke; the CDR also presents an IDEA alternative and a full-silicon tracking variant, but the TPC option is the one used for detailed full-simulation physics performance studies (Group, 2018). CDR-era performance studies refer to the baseline geometry as APODIS, or CEPC_v4, optimized from CEPC_v1 (Zhao et al., 2018). In the Engineering Design Report phase entered in January 2024, CEPC developed a “reference TDR design” or TDR_ref, intended mainly for project review and approval in China rather than as the final detector TDR of an experiment collaboration (Gao, 7 May 2025).
1. Conceptual evolution and institutional role
The CEPC detector program did not begin from a single immutable hardware definition. A 2016 study proposed a SiD-derived CEPC detector concept, denoted SiDcc1, and explicitly characterized it as a “promising detector approach” within conceptual design studies rather than as a finalized reference detector; in that study, the CEPC-specific simplifications were a solenoid-field reduction from to and HCAL reductions from barrel layers and endcap layers, with benchmark distributions agreeing with the original SiD geometry within statistical uncertainties (Chekanov et al., 2016).
The 2018 CEPC CDR then established a more definite reference framework. It presents two primary detector concepts plus a tracking variant, but states that the baseline particle-flow detector with the TPC option is the concept studied in detail through realistic simulation and used for the physics performance studies. In that sense, the TPC-based particle-flow detector is the de facto CDR reference detector, whereas IDEA and the full-silicon tracker remain alternative or variant concepts within the same design report (Group, 2018).
By the EDR stage, the meaning of “reference detector” became more administrative and project-defining. The “reference TDR design” was reviewed by the CEPC Detector Review Committee on April 14–16, 2025, was expected for release by mid-2025, and was intended for project review and approval in China, to be included in the 2025 proposal to the Chinese government. The same source is explicit that this document is not yet the final detector TDR of an experiment collaboration; formal detector TDRs are deferred until after project approval and the formation of genuine international detector collaborations (Gao, 7 May 2025).
2. CDR baseline detector architecture
In the CDR baseline, the detector is organized from inside outward as beam pipe, silicon pixel vertex detector, silicon inner tracker, TPC central tracker, silicon external tracker, forward silicon tracking disks, high-granularity ECAL, high-granularity HCAL, superconducting solenoid, instrumented iron return yoke and muon detector, and forward luminosity and machine–detector-interface instrumentation. The vertex detector has 6 cylindrical layers with radii from to , more explicitly , with target single-point resolutions of for layers 1 and 2 and for layers 3–6, material budget of about 0 per layer, and pixel-sensor readout time shorter than 1. The benchmark impact-parameter target is
2
The default tracking system then combines silicon layers at 3, 4, and 5 with a TPC of inner radius 6, outer radius 7, and length 8, segmented into 220 radial layers with 9 step and operated with 0; the quoted TPC single-hit resolutions are 1 in 2 and 3 in 4, and the combined tracker target is
5
The calorimetry follows the particle-flow logic: a silicon–tungsten ECAL with 30 layers, total tungsten thickness 6, and 7 cells; and a hadronic calorimeter with 40 layers, about 8, and 9 cells. The solenoid is 0, and the muon system is integrated in the return yoke (Group, 2018).
Several subsystem studies sharpened that baseline. A photon-reconstruction study of the CEPC baseline detector describes the same particle-flow-oriented layout in fuller subsystem language: silicon pixel vertex detector, silicon inner tracker, TPC, silicon external tracker, silicon–tungsten ECAL, steel–glass RPC HCAL, a 1 solenoid, and a return yoke with muon detectors. In that study the ECAL barrel radius is 2, the endcaps are placed at 3, the calorimeter has 30 longitudinal layers with total tungsten thickness 4, and the silicon cell size is 5 (Shen et al., 2019).
An earlier ECAL optimization, starting from the CEPC_v1 reference geometry, recommended a cost/performance compromise of 25 layers, 6 silicon, 7 tungsten per layer, and 8–9 transverse cells, while keeping the same total tungsten thickness of 0 (Zhao et al., 2017).
3. EDR reference design and subsystem targets
The EDR reference detector preserves the particle-flow orientation but updates the subsystem technology choices. The layout shown for the reference TDR design includes a 1 superconducting solenoid, a precision vertex detector, a combined gaseous and silicon tracking system, a particle-flow electromagnetic calorimeter, a particle-flow hadron calorimeter, a muon detector in the yoke, and forward calorimeters including LumiCal. The figure labels are explicit: “3T Magnet (SC Solenoid),” “Yoke + MU (PS + SiPM),” “PFA HCAL (Glass Scintillator),” “Crystal PFA ECAL (Transverse bar),” “LumiCal (SiDet + LYSO),” “TPC (Pixelated Micromegas),” “VTX (MAPS SiPixel + Stitching + Bending),” “ITK,” and “OTK (AC-LGAD strip)” (Gao, 7 May 2025).
The EDR table of baseline technologies and specifications gives a 6-layer CMOS silicon pixel vertex detector with spatial resolution of about 2 and material budget 3 per layer. The tracking system combines the vertex detector and TPC with CMOS SPD inner tracking and AC-LGAD-strip outer tracking; the momentum-resolution target is printed in the approximate form
4
For particle identification, the table gives 5 measurements by the TPC with relative uncertainty of about 6, and time of flight by AC-LGAD SSD with 7. The ECAL target is
8
with effective granularity of about 9. The HCAL target is a single-hadron resolution of about
0
and a jet-energy resolution target of about
1
The same paper also reports subsystem-development status: stitched-CMOS R&D with the ALICE team for the vertex detector, completed TPC mechanical and cooling design with a prototype pixelated Micromegas module ready for test beam, development of long crystal bars for ECAL, demonstrated mass production of full-size 2 glass-scintillator HCAL samples, production of 3 plastic-scintillator strips for the muon detector, and development of an aluminum-stabilized NbTi Rutherford cable for the solenoid (Gao, 7 May 2025).
This suggests that the EDR “reference detector” is not merely a restatement of the CDR baseline. It is a project-baseline detector definition with broader engineering scope, including electronics, TDAQ, software/computing, mechanical integration, detector cost estimation, preliminary hall design, and machine–detector interface, while still remaining upstream of collaboration-owned final detector TDRs (Gao, 7 May 2025).
4. Detector performance in full and fast simulation
Full-simulation Higgs benchmark studies using APODIS provide the most compact CDR-era statement of reference-detector performance. In 4 events, the 5 channel is reconstructed with a mass resolution of 6 7 after fitting the dimuon mass spectrum to a Crystal Ball function. The 8 channel is characterized by 9, where 0 is the half-width of the narrowest interval containing 1 of the distribution, corresponding to 2 3. For jet final states, after the truth-level cleaning
4
the reconstructed Higgs mass resolutions are 5 for 6, 7 for 8, 9 for 0, 1 for 2, and 3 for 4. The same study also reports a clear separation of 5 and 6 decay cascades in visible-mass space, including four visible-mass peaks for the different visible and invisible 7 combinations in 8 (Zhao et al., 2018).
Photon performance has been studied in full simulation with the CEPC baseline detector and Arbor reconstruction. For unconverted isolated photons above 9, the photon identification efficiency is above 0 and the neutral-hadron misidentification rate is below 1. The fitted photon-energy resolutions are
2
for the realistic baseline detector and
3
for a simplified geometry without geometry defects and without material before the ECAL. In 4, the relative Higgs mass resolution is about 5 in the baseline detector before correction, about 6 after a position-based correction, and about 7 in the simplified geometry. The same study reports photon conversion rates of roughly 8–9 in the central region and about 0 in the forward region, and a 1 photon effective resolution improvement from 2 to 3 after correction (Shen et al., 2019).
Particle-identification requirements have also been mapped onto the baseline detector. A full-simulation requirement study concludes that, for incident particles in the barrel region with relevant momentum larger than about 4, if the detector achieves a TOF benchmark of 5, then the TPC 6 resolution should be better than 7. Under that condition, the study finds 8 identification efficiency and purity of about 9 in the realistic 00-degradation scenario, 01 reconstruction of 02, and 03 reconstruction of 04 (Zhu et al., 2022).
The reference detector also serves as a fast-simulation benchmark in detector-level physics analyses. A study of Higgs 05 properties in 06, 07, at 08 and 09 uses WHIZARD 3.1.6, Pythia6, and DELPHES with the CEPC detector card. After the preselection 10, the effective Gaussian width of the recoil-based event-energy variable is 11, and the expected 12 CL limit on the 13-odd coupling parameter improves from 14 to 15 when 16 information is included. The paper is explicit that this is not a full GEANT-style detector optimization; the detector enters as the CEPC DELPHES card, so the conclusions are tied to that simplified CEPC reference configuration (Drutskoy et al., 9 Nov 2025).
5. Subsystem R&D, prototypes, and optimization studies
The vertex-detector program has progressed from chip prototypes to integrated mechanical demonstrators. A beam test of a baseline CEPC vertex-detector prototype reports a cylindrical barrel structure with six double-sided detector modules based on TaichuPix-3 MAPS sensors. In a 17 electron beam at DESY II TB21, the offline analysis indicates spatial resolution of about 18, detection efficiency exceeding 19, and a beam-test-defined impact parameter resolution of about 20. The same paper notes that the prototype ladder material is about 21, well above the CEPC final per-layer target of 22, so the result is a system-level milestone rather than a final detector realization (Li et al., 2024).
CEPC MAPS R&D has also addressed readout architecture and sensor operating point. TaichuPix1 demonstrated a 23 monolithic pixel with data-driven, FE-I3-like column-drain readout, 24 system clock, analog peaking time 25, measured time walk around 26, and serializer operation up to about 27 (Wu et al., 2021). TaichuPix-3 studies then focused on the long-barrel geometry with no endcap, where forward tracks enter barrel sensors at large angles; at ITHR 28, the measured mean cluster size rises from 29 at 30 to 31 at 32, 33 at 34, and 35 at 36, with a phenomenological dependence 37 used to guide digitization and occupancy estimates in the forward barrel (Lu et al., 7 Mar 2025). JadePix-3 characterization, in turn, studies substrate reverse bias and concludes that reverse bias expands the depletion region, reduces input capacitance, enhances charge collection efficiency, and lowers fake-hit rate, with the paper identifying an optimal operational working condition around 38 for the sensor (Hu et al., 28 May 2025).
TPC R&D has concentrated on the circular-collider challenge of continuous operation. A feasibility study of a hybrid GEM–Micromegas module reports ion-backflow suppression to about 39 at gain about 40, with no obvious discharge behavior and X-ray energy resolution below 41 at 42 (Yuan et al., 2019). A later space-charge study for the high-luminosity 43 pole uses Mokka/Geant4 ionization distributions and COMSOL field calculations and finds maximum distortions of 44 for Higgs operation but 45, reaching under 46, at the updated high-luminosity 47 pole. That study therefore derives a much stronger TPC requirement for 48-pole operation: gain needs to be about 49 with 50, and it identifies the pixel TPC as a potential option to replace traditional MPGD readout (Yuan et al., 2021).
Optimization studies around the baseline geometry have begun to isolate which design parameters dominate CEPC flavor performance. Starting from the CDR-like six-layer vertex design with first layer at 51, per-layer material budget 52, and hit resolutions of 53–54, a scan of inner radius and spatial resolution finds that halving the inner radius and spatial resolution improves transverse and longitudinal impact-parameter resolution approximately by a factor of two, while improving the 55 accuracy by 56 and the 57 significance by 58; doubling these parameters causes comparable degradation, and variations in inner radius are identified as the dominant factor (Li et al., 6 Aug 2025).
On the calorimeter side, CEPC AHCAL prototyping has reached system level. The analogue hadron calorimeter option comprises a 40-layer steel–scintillator sandwich with 59 tiles, 60 sensitive area, 324 channels per layer, and 12,960 channels in total. Cosmic tests of the full 40-layer system report a MIP most probable value of 61, corresponding roughly to 62, and layer efficiencies of about 63; the paper further cites an optimized CEPC AHCAL design giving a Higgs boson mass resolution of 64 in 65, and preliminary beam-test hadron energy resolution of approximately
66
These results are presented as evidence that a highly granular scintillator–steel AHCAL is a feasible hadron-calorimeter option for the CEPC detector (Shi et al., 13 Nov 2025).
6. Background environment, timing architecture, and interpretive limits
The reference detector is shaped not only by reconstruction goals but also by machine backgrounds and timing constraints. A dedicated beamstrahlung study for CEPC Higgs-factory operation at 67 and 68 finds an average beamstrahlung parameter 69 and concludes that incoherent 70 pairs are the dominant beamstrahlung-induced background. Under the assumed interaction-region geometry, the hit density at the first vertex layer at 71 is 72, corresponding to occupancy well below 73 for 74 pixels and 75 integration time, while annual radiation levels are about 76 and 77 at the first layer (Xiu et al., 2015).
Timing coordination between accelerator and detector has likewise become part of the reference-detector definition. A joint accelerator–detector study sets the rule that the spacings between adjacent bunches in any CEPC operation mode are integer numbers of 78, and that the accelerator will provide the detector systems with a beam-synchronous master clock of 79. The same paper proposes a fine-tuned ring circumference of 80, which makes the orbit length closer to 81 and benefits the detector in the first 82-year operation. For the vertex detector CMOS sensor, the paper compares two bunch-interval cases and states that, with the modified bunching structure, the overall power consumption of the vertex detector can be reduced to 83 (Wang et al., 23 Sep 2025).
At the same time, the term “reference detector” should not be conflated with a fully validated, subsystem-by-subsystem engineering design. Some CEPC detector studies are full-simulation analyses, while others use fast simulation or DELPHES cards as common detector benchmarks. The Higgs 84 study cited above is explicit that the CEPC reference detector enters as smearing, acceptance, and reconstructed-object response in a simplified DELPHES configuration rather than as a transparent breakdown of subsystem performance (Drutskoy et al., 9 Nov 2025). A plausible implication is that the CEPC Reference Detector functions in two distinct but related ways: as a project-baseline detector definition for design and approval, and as a standardized simulation benchmark for detector-level physics projections.