Two magnetic spectrometers
MAGIX scatters the energy-recovering MESA electron beam on a windowless gas-jet target. Two high-resolution magnetic spectrometers, which rotate around the interaction point, analyse the scattered electrons.
A quadrupole – dipole – dipole layout turns momentum into a position in the focal plane and keeps the angle information. The TPC sits in this focal plane.
Below, a cut through the target chamber and both spectrometers. In the left one, the glowing dots are electrons with five different momenta coming from the target. The magnets bend them down onto the detectors, and each momentum is focused onto its own spot in the focal plane, which is exactly where the TPC sits.
What the TPC must deliver
- Track position in the focal plane to ~100 µm
- Track angle to ~3.5 mrad
- Area of about 140 × 760 mm², with as little material in the electron path as possible
Tap any picture to enlarge it.
A TPC lying in the focal plane
Scattered electrons leave the spectrometer vacuum through a 75 µm polyimide window and cross the gas volume of the TPC.
On their way they ionise the counting gas. The freed electrons drift in a homogeneous field towards the readout PCB, where an amplification stage of four GEM foils multiplies them. The 9216 pads then record the projection of the track and the time of arrival of its charge. Together this gives a 3D track.
Built in Mainz
- Three TPCs: one per spectrometer, plus one spare for tests
- First active area finished January 2026, first cosmics recorded
- Trigger/veto system below the TPC: its 22 fast scintillator bars provide the external start time T0. This is the anchor in the drift direction: without it, only the track's angle in drift direction could be measured, not its absolute position. Lead and further scintillator layers add particle ID
Four GEMs, not three
A single drifting electron is far too small a signal. A Gas Electron Multiplier is a 50 µm polyimide foil with 5 µm copper on both sides, pierced by tens of thousands of holes per cm². A few hundred volts across the foil create strong fields in the holes, and these fields start electron avalanches.
MAGIX stacks four foils instead of the usual three. Each foil can then run at a lower voltage, which lowers the risk of discharges and reduces ion backflow.
The voltage divider
High-voltage divider board: eight resistors set the eight voltage steps; spring-loaded pogo contacts touch the foils.
Geometry
- Active area of 768 × 192 mm² (384 × 24 pads), foils stretched with CERN's NS2 system
- Conical holes: 50 µm inner, 70 µm outer diameter
- Pitch 140 µm (outer foils) and 280 µm (middle foils)
- Each foil's pattern is rotated by 90°, so there is no straight line of sight from cathode to anode
9216 pads, 8 copper layers
The anode is the gas-tight back wall of the detector. The pads face the gas; on the outside, connectors take the front-end hybrids. In between, every one of 3072 pads per board is routed through the inner layers to its connector pin.
Neighbouring pads always go to the same connector, on alternating contact rows. They reach neighbouring VMM channels, so the VMM neighbour logic still works: a channel that fires also reads out the pads to its left and right.
Pads are 2 × 8 mm²: long in the direction the electrons travel, narrow across the focal plane. Each row is shifted by ⅓ pad, so a straight track never lines up with the grid.
Two kinds of vias
Every signal takes the same path, read straight from the board file: pad on L16 → through via → trace on L3 or L4 → blind via L1–L14 → connector pin on L1. The blind vias stop at L14, so they never pierce the pad side.
VMM hybrids
Every pad needs its own amplifier, shaper and digitiser. MAGIX uses the VMM ASIC, originally developed for the ATLAS New Small Wheel. It has 64 channels per chip, two chips sit on each hybrid, and the data are read out through the RD51 Scalable Readout System (SRS).
- Self-triggering with zero suppression, up to 4 MHz per channel
- Time resolution below 1 ns
- Neighbour logic: channels next to a hit are read out too
- 8 gain settings, 0.5–16 mV/fC
- Data, configuration and power (2.9 V and 3.9 V) share one HDMI cable per hybrid
Cooling in 24 mm
Each hybrid gives off up to 3.5 W. Standard cooling blocks are 45 mm tall, but only 24 mm are available. Custom 2.3 mm aluminium profiles and high-pressure fans keep the ASICs below 60 °C.
An open field cage
A TPC needs a field cage: conductors around the gas volume that step the voltage down evenly, so the drift field stays homogeneous. Normally it closes all sides.
Here, though, the electrons enter through one side. Any material in their path means multiple scattering and lost resolution. So MAGIX leaves this side open, and the only material in the way is the 75 µm window.
Instead, field extension plates reach about 23 cm up into the vacuum. Their potential lines have a saddle point, a shape derived analytically (see Sources). It keeps the drift field homogeneous in the active volume even though the cage is open there.
The exit side, toward the trigger/veto system, is not a PCB either: it is also a polyimide foil, with the field-cage strips printed on it.
Voltages reach the vacuum side through spring-loaded contacts instead of HV vacuum feedthroughs.
Starry Night
At high rates, charge on the window and ions in the gas can distort the drift field and with it the measured tracks. To monitor this, MAGIX creates electrons at known places and known times.
UV LEDs behind the cathode shine through fused-quartz plates onto thin aluminium spots. Through the photoelectric effect these release electrons, which drift across the whole chamber like real ionisation. Where and when they arrive shows the actual drift field.
Tap Distort the field to add a local disturbance, such as charge collecting on the window. The drift lines bend, the spots on the readout move (pink arrows), and the regular pattern shows exactly where the field is wrong.
One complete module has been tested and works. The full set of 14 is being built. The LED pattern and the disturbance in the 3D view are schematic.
Track
Each square is a pad that fired: its position gives two coordinates, the drift time the third.
- Events built and clustered (DBSCAN) as in the group's analysis script; the 10 cleanest long tracks without gaps per campaign
Trigger & veto system
Below the TPC sits a sandwich of scintillators and lead. The trigger layer gives the start time for the drift measurement and starts the data acquisition. The veto layers tell real electrons from cosmic muons and neutrons.
Why rhombic?
The bars' cross-section is a parallelogram (base 36 mm, height 26 mm, 60°). Seen from above, neighbouring bars overlap, so electrons arriving at up to 16° never slip through a gap. Light guides turn the slanted end face into the round 1-inch PMT window. The veto segments use the same 60° shape.
Strictly a parallelogram, not a rhombus, but the name stuck. From S. Stengel, doctoral thesis, JGU Mainz (2024).
Sources & links
Scan a code with your phone to open the page.
Made from real design data
- Readout board, field-cage boards and GEM voltage divider: the original Eagle board files
- Open field cage: analytic solution and field map by H. Merkel, "Field cage and construction", MAGIX collaboration meeting (ref. [68] in D. Markus' thesis)
- Cosmic tracks: MAGIX TPC runs of 27 March 2026, clustered with the group's analysis script
References
- D. Markus, Construction of a Time Projection Chamber as Focal Plane Detector for MAGIX and Preparation of the Hoyle State Measurement at MAGIX, doctoral thesis, JGU Mainz (2026)
- S. Stengel, Design and Development of the MAGIX Trigger Veto System, doctoral thesis, JGU Mainz (2024)
- Development of large area focal plane detectors for MAGIX, arXiv:1906.05900
- The MAGIX focal plane time projection chamber, J. Phys.: Conf. Ser. 1498 (2020) 012022
Interactive poster created by Pepe Gülker in collaboration with AI.