HADES Multiwire Drift Chambers

• HADES Multiwire Drift Chambers are precision detectors for charged-particle tracking in high-energy experiments, featuring an ultra-low mass design and high spatial resolution.
• They employ 24 trapezoidal drift chambers with optimized stereo wire angles and graded cell sizes to achieve sub-150 μm position precision and efficient momentum reconstruction.
• Recent developments include upgraded front-end electronics, refined calibration techniques, and stabilization strategies to mitigate gas-related instabilities for FAIR-era operations.

HADES Multiwire Drift Chambers (MDCs) are the central charged-particle tracking detectors of the High-Acceptance Dielectron Spectrometer. They are implemented as 24 trapezoidal planar drift chambers arranged as four tracking planes in six azimuthal sectors, with two planes located in front of the toroidal magnetic field and two behind it. In this configuration the MDCs measure charged-particle trajectories before and after magnetic deflection, enabling momentum reconstruction, while also contributing supplementary particle identification through specific energy-loss information extracted from time-over-threshold readout. Their design is closely tied to the requirements of dielectron spectroscopy: high spatial precision, high efficiency, large acceptance, and an exceptionally low material budget (0902.3478).

1. Role in the HADES spectrometer

Within HADES, the MDC system provides the geometrical basis for reconstructing an “inner” track segment and an “outer” track segment, whose change in direction yields the bending in the toroidal magnetic field and therefore the particle momentum. The spectrometer has six identical sectors defined by six superconducting coils, and the MDC system follows this six-fold symmetry. The overall HADES acceptance is characterized by about 85% azimuthal coverage over a polar-angle interval from $18^\circ$ to $85^\circ$, and the MDCs cover this range as part of the tracking chain that interfaces upstream with the RICH and downstream with TOF/TOFINO and Pre-Shower detectors (0902.3478).

The design constraints were set by dielectron measurements. HADES targeted a vector-meson mass resolution around $2.5\%$, and the magnet geometry implied that the tracking detectors had to provide position resolution better than $150~\mu\mathrm{m}$ in order to keep the detector contribution to the momentum resolution below $1\%$ for a $p=1$ GeV/$c$ particle at $\theta = 20^\circ$, assuming two detector pairs spaced by $d=0.3$ m. This requirement, together with the need to suppress multiple scattering and external conversions, drove the choice of low-mass multiwire drift chambers rather than more massive tracking technologies (0902.3478).

The MDCs were also designed to operate in heavy-ion events. For central Au+Au collisions at 1 AGeV, the design estimate for the maximum cell occupancy was 30%. This high-occupancy environment shaped both the stereo-wire layout and the calibration strategy, because ambiguity resolution, efficiency, and timing robustness had to remain acceptable in multi-track topologies (0902.3478).

2. Chamber geometry, materials, and low-mass design

Each MDC contains six sense/field wire layers arranged at stereo angles $\pm 0^\circ$, $85^\circ$0, and $85^\circ$1. This stereo configuration was chosen to reduce ambiguities in high-multiplicity events and to optimize the resolution in the polar direction, which is also the direction of magnetic deflection. Each chamber type contains about 1100 drift cells. The cell size increases from the innermost to the outermost planes, from about $85^\circ$2 mm$85^\circ$3 in Plane I to about $85^\circ$4 mm$85^\circ$5 in Plane IV, matching the larger detector dimensions and lower occupancies at larger radii (0902.3478).

The low-mass concept is implemented aggressively. Sense wires are gold-plated tungsten, with diameters of $85^\circ$6 in planes I–III and $85^\circ$7 in Plane IV. Cathode and field wires are annealed aluminum, with diameters of $85^\circ$8 and $85^\circ$9; planes I–III use bare aluminum, while Plane IV uses gold-plated aluminum. The chamber entrance windows are $2.5\%$0 aluminized Mylar. Active areas range from about $2.5\%$1 up to $2.5\%$2, and the innermost chambers were subject to especially strong mechanical constraints because the available frame width was limited to 3 cm by the coil shadows (0902.3478).

The original counting gas was $2.5\%$3. This helium-based mixture was chosen because it minimized multiple scattering while still allowing efficient tracking at moderate gain. Material minimization was a primary design theme: the detector thickness per chamber was about $2.5\%$4, so that all four MDC planes together contributed about $2.5\%$5, while the air in the tracking system added about $2.5\%$6. The total material from MDCI to MDCIV thus remained typically below about $2.5\%$7 radiation length. The good signal-to-background ratio in dielectron measurements was explicitly linked to the low conversion probability arising from the use of He-based gas and Al wires in the tracking system (0902.3478).

Long-term operation was an explicit design target. Systematic tests showed a 10% wire-tension loss in five years, and accelerated ageing tests with $2.5\%$8Fe on prototype chambers showed no noticeable gain drop larger than 5% over an equivalent of two years of continuous running. The expected maximum accumulated charge was of order 10 mC per year and cm of sense wire. These results initially suggested acceptable long-term stability, but later operating experience showed that the relevant failure modes of the upstream chambers were not exhausted by classical ageing tests alone (0902.3478, Wendisch et al., 2024).

3. Readout architecture, calibration, and reconstruction performance

The original readout was based on custom front-end electronics mounted directly on the chamber frames, outside the active area. Four sense wires were connected via flexible printed circuits to analog boards. The front-end ASIC was ASD8-B, with 8 channels per chip, intrinsic noise of 1 fC, power consumption of 30 mW/channel, and an adjustable threshold. The ASIC performed amplification, shaping, and discrimination, and produced a logical output pulse whose width equaled the time the shaped signal remained above threshold. This provided both a drift-time measurement and a time-over-threshold observable (0902.3478).

Custom TDC chips digitized these logical signals with 0.5 ns/channel, common-stop operation, and a $2.5\%$9 full range. The ToT information was subsequently used for noise rejection and for $150~\mu\mathrm{m}$0-like information. For each MDC type, the ToT–energy-loss relation was calibrated in bins of impact angle and distance to the wire using

$150~\mu\mathrm{m}$1

with $150~\mu\mathrm{m}$2 as fit parameters. After normalization of the cell-by-cell measurements to a common reference, a truncated mean was formed, keeping on average 18 out of 24 cell measurements, corresponding to a 20% truncation cut. The resulting MDC energy-loss measurement enabled $150~\mu\mathrm{m}$3/p separation up to about $150~\mu\mathrm{m}$4 GeV/$150~\mu\mathrm{m}$5, improved the signal-to-background ratio in $150~\mu\mathrm{m}$6 measurements, and achieved a resolution of about 7% for minimum-ionizing particles and about 4% for more strongly ionizing particles (0902.3478).

The distance–time relation was obtained from GARFIELD simulations for each drift-cell geometry and cross-checked against measurements. The self-tracking studies showed typical differences below the simulated time dispersion of 2 ns. In a prototype Plane-II chamber tested with a $150~\mu\mathrm{m}$7 GeV/$150~\mu\mathrm{m}$8 proton beam and an external silicon tracker, a spatial resolution of $150~\mu\mathrm{m}$9 was achieved over 70–80% of the drift cell at the optimal high voltage of $1\%$0 V. Under real in-beam conditions, time resolutions ranged from 2.3 ns to 3.6 ns depending on chamber type, gas, and cell geometry, corresponding to spatial resolutions from approximately $1\%$1 up to $1\%$2 for MDCI–IV (0902.3478).

Measured layer efficiencies for minimum-ionizing particles were about 90% and 97% in the inner planes and almost 100% in the outer planes. Using $1\%$3 elastic scattering at 1.25 GeV, the candidate-search efficiency was close to 100%; with a fitted inner MDC segment it was about 92%; with both inner and outer fitted segments it was 87%; and full track reconstruction including Runge–Kutta momentum determination reached 86%. Event-based alignment used straight tracks without magnetic field, cosmic-ray tracks, and kinematically constrained $1\%$4 elastic scattering, yielding a precision on global alignment parameters of order 0.1 mm or better. In a $1\%$5 elastic-scattering analysis at 3.5 GeV, the observed integrated momentum resolution was about 4%, whereas a full GEANT simulation with nominal GARFIELD drift-time resolutions predicted about 1.5%; the discrepancy was attributed mainly to misalignment and calibration offsets (0902.3478).

4. Front-end electronics replacement for FAIR-era running

Operation at the future FAIR SIS-100 accelerator motivated a major MDC readout upgrade. Two distinct pressures were identified: the higher reaction rates expected at SIS-100 would stress the detector electronics and DAQ, and the ASD-8 analog readout ASIC previously used for the MDCs could no longer be procured. In response, the collaboration evaluated PASTTREC, an ASIC originally developed for PANDA straw-tube readout, as a replacement front-end for the HADES MDC system (Wiebusch et al., 2018).

The comparison of ASD-8 and PASTTREC was performed in beam tests, laboratory drift-chamber tests with cosmic muons, $1\%$6-based charge tests, and 3D GARFIELD simulations. The central timing difference was linked to the shaping time: ASD-8 had a peaking time of 7 ns, whereas PASTTREC had a peaking time of 15 ns. Timing extraction used walk correction based on time-above-threshold information. After this correction, PASTTREC fell short of ASD-8 timing precision by only 20%. In the beam test at COSY/Jülich, which used a minimum-ionizing proton beam and a diamond detector as trigger and reference time, the beam slice selected by the diamond detector was $1\%$7, and the arrival-time precision was extracted from the standard deviation of an asymmetric Gaussian fit to the arrival-time distribution (Wiebusch et al., 2018).

The measured drift-time precision as a function of distance to the sense wire exhibited the characteristic “W” shape. In the 3D GARFIELD modeling, this behavior was reproduced by adding a Gaussian error as a proxy for input noise and by imitating discriminator-threshold effects through waiting for the $1\%$8-th fastest electron to arrive at the sensing wire. The effective response inferred from the comparison was that ASD-8 behaved as if it were sensitive to the 3rd arriving electron, whereas PASTTREC behaved as if it were sensitive to the 4th to 5th arriving electron. This provided a physically transparent explanation of why the shorter-peaking ASD-8 remained better for pure timing (Wiebusch et al., 2018).

The trade-off favored PASTTREC in other respects. Charge information was inferred from ToT through a calibration function, and in $1\%$9 tests PASTTREC clearly separated the $p=1$0 K$p=1$1 peak from the Ar escape peak while varying the gas gain by a factor of 15, whereas ASD-8 showed no clear separation. PASTTREC was also far less susceptible to pickup noise and far less prone to self-oscillation in both beam-test and laboratory conditions. The upgrade study therefore characterized PASTTREC as a viable replacement candidate: it retained timing performance within about 20% of ASD-8 after walk correction, while improving charge-related observables and practical operating stability (Wiebusch et al., 2018).

5. Malter-like instabilities, diagnosis, and recovery

A distinct chapter in the history of the MDC system concerns the unstable operation of the chambers in front of the magnetic field, closest to the interaction point. The affected population comprised the 12 upstream chambers, notably all six original type I chambers and later type II chambers; rebuilt type I.1 chambers also eventually showed related but less severe instabilities. The observed pathologies included self-sustained currents, sudden discharges, HV trips, and current instabilities developing on a minutes time scale. In some cases self-sustained currents could re-establish after HV cycling even without radiation (Wendisch et al., 2024).

Two operational episodes were distinguished. In the first roughly five years of operation, the original type I chambers developed localized HV instabilities. Opened chambers revealed deposits on aluminum wires, and energy-dispersive X-ray spectroscopy identified silicon in these deposits. The contamination was traced to an O-ring near one window frame that had apparently been treated during fabrication with standard vacuum grease containing silicon compounds. Because conditioning and increased gas flow did not improve the behavior, all six type I chambers plus one spare were rebuilt, producing type I.1 chambers (Wendisch et al., 2024).

The later and more general problem affected rebuilt type I.1 and especially type II chambers after prolonged exposure to isobutane-based gases. Visual inspection of type II wires after rupture showed dark spots on cathode wire planes, distributed over the active area with approximately one dark spot per 1 to 2 cm$p=1$2. EDX indicated that these spots were mainly hydrocarbon compounds, and no silicon was found in this later case. The authors classified the effect as Malter-like rather than classical wire-chamber ageing. Their argument rested on the location and chemistry of the deposits, the self-sustained and off-spill currents, the explicit statement that a loss of gain was not observed, and the modest charge load: the estimated accumulated lifetime charge close to the interaction point was about 15 mC/cm on anode wires, about 10 mC/cm on cathode wires, and about 5 mC/cm on field wires, with operating current density typically between below 1 and 5 nA/cm. These values were far below the $p=1$3 mC/cm scale cited in the literature for classical hydrocarbon ageing (Wendisch et al., 2024).

Stable recovery required both removal of isobutane from the gas system and addition of controlled water vapor. By 2014, isobutane was fully banned from the MDC gas system, and the replacement working gas became $p=1$4. A prior test with Ar/isobutane did not solve the problem. The successful stabilization procedure added deionized water vapor at 1000 to 3500 ppmv, individually optimized for a given chamber. For type I.1 chambers the concentration was below 1500 ppmv. Because gas purifiers would alter the water content, the affected front chambers were switched from recirculation to an open gas system. Water concentration, oxygen concentration, and overpressure were monitored continuously, with overpressure maintained at 30–50 Pa. Before physics runs, optimization was performed chamber by chamber using an X-ray tube that reproduced the polar rate distribution and duty cycle of SIS18 heavy-ion operation; a chamber was considered acceptable if it showed no measurable residual current during spill breaks (Wendisch et al., 2024).

The recovery protocol enabled stable operation in several production runs, including high-intensity heavy-ion induced reactions and, specifically, Au+Au at $p=1$5 GeV. The trade-off was non-negligible. After switching from He/isobutane to Ar/CO$p=1$6, the increased electron drift velocity in the new operating regime caused a loss in time and spatial precision of about 10–15%, and GARFIELD simulations showed significantly lower drift velocity in low-field cell corners for CO$p=1$7-based gases than for isobutane-quenched gases, leading to reduced charge collection and a detection-efficiency drop in cell corners. A common misconception would be to interpret the episode as ordinary classical ageing; the evidence presented in the recovery study instead points to cathode-deposit-induced Malter-like effects, with water acting as an operational stabilizer rather than a permanent cure, since removing the water caused the instabilities to return (Wendisch et al., 2024).

6. FAIR-oriented calibration and control developments

In later FAIR-oriented work, the HADES MDCs were also used as a test case for real-time detector calibration and control. The central idea was to predict ionization-loss calibration parameters from environmental and operational data rather than from computationally expensive track-based offline calibration. The response variables were tied to ToT distributions, and the updated offline targets were extracted with a Landau–Gaussian convolution rather than a pure Landau fit. The input space explicitly included gas concentrations $p=1$8 and $p=1$9, detector count rates, gas composition, pressure, and high voltage (Kladov et al., 22 Sep 2025).

Several architectures were compared: fully connected neural networks, graph-convolutional LSTMs with spatial convolutions or Chebyshev convolutions, and transformers. The updated method also introduced a trainable exponential normalization applied to both input and output values inside the model after fixed-range rescaling. Because the February 2022 beam-time data had nearly constant HV, an additional dense layer was inserted in the middle of the network to combine HV directly with intermediate features and preserve HV sensitivity. The HV dependence was then recovered by fine-tuning on cosmic-ray data with manually varied HV settings, while freezing the majority of the network weights before the inserted HV layer (Kladov et al., 22 Sep 2025).

The reported performance for a period of five days after the training dataset ended was RMSE $c$0, precision $c$1, and robustness $c$2. Figure 1 in that study compared offline calibration results from cosmic data, network predictions after fine-tuning on cosmic data, and beam-time data with artificially varied HV values, and the three were reported to be in good agreement. This established a practical framework in which the same learned detector-response model could be used not only for real-time calibration prediction but also to infer optimal HV settings for stable operation during data acquisition (Kladov et al., 22 Sep 2025).

These developments connect directly to the earlier MDC history. The original system emphasized ultra-low mass, precise drift-time measurement, and ToT-based supplementary $c$3; later work confronted component obsolescence, rate-capability demands, gas-dependent operational pathologies, and the need for chamber-by-chamber stabilization. The recent calibration-and-control studies suggest that the MDCs have also become a methodological platform for FAIR-era online calibration and detector steering, while retaining their original role as the tracking core of HADES (Wiebusch et al., 2018, Wendisch et al., 2024, Kladov et al., 22 Sep 2025).