GridPix Detector Technology

• GridPix detector is a gaseous pixel detection technology that combines CMOS ASICs with a precisely aligned amplification grid to achieve single-electron sensitivity and accurate imaging.
• Its design features one-to-one grid hole to pixel alignment, minimizing cross-talk and enabling detailed charge and timing measurements using Timepix/Timepix3 ASICs.
• Applications of GridPix span low-energy X-ray detection, TPC readout, neutron imaging, and X-ray polarimetry, showcasing its versatility and scalability.

Searching arXiv for recent and foundational GridPix papers to ground the article. The GridPix detector is a gaseous pixel detector architecture in which a CMOS pixel ASIC is combined with a photolithographically integrated Micromegas-like amplification stage, usually termed an InGrid. In reported implementations, one grid hole is aligned to one pixel, so that avalanche charge from a single amplification cell is collected by exactly one underlying pixel. This combination enables single-electron sensitivity, high-granularity imaging, and, in Timepix3-based systems, simultaneous per-pixel time-of-arrival and time-over-threshold measurement. GridPix has been developed across several experimental contexts, including low-energy X-ray detection at CAST, low-mass and collider-oriented TPC readout, neutron imaging, and next-generation photoelectric X-ray polarimetry (Manikantan et al., 2 Sep 2025, Ligtenberg et al., 2019).

1. Detector concept and physical construction

At the device level, GridPix denotes a family of detectors built around a pixel ASIC with a Micromegas-style grid post-processed directly on top of the chip. A common implementation uses a Timepix or Timepix3 ASIC with a 256 × 256 matrix of 55 μm × 55 μm pixels. Reported single-chip active areas include approximately 14.1 mm × 14.1 mm or, equivalently, about 199 mm². A 4 μm Si-rich Si₃N₄ or Si₃N₄ protection layer is placed on the ASIC surface for discharge protection, and the grid is held above the pixel plane by SU-8 pillars that define an amplification gap of 50 μm in many systems, or 80 μm in the polarimetric prototype (Krieger et al., 2014, Ligtenberg et al., 2018, Manikantan et al., 2 Sep 2025).

The mesh itself is typically aluminum or gold, with holes of about 35 μm to 38 μm diameter on a 55 μm pitch. The crucial design feature is lithographic registration of the hole pattern to the pixel matrix. In several descriptions, the hole-to-pixel alignment is stated to be better than a few microns, and in the polarimetric prototype the mesh hole pattern is aligned one-to-one with the 55×55 μm² pixels so that avalanche electrons generated in each hole are collected by exactly one underlying pixel. This geometry suppresses cross-talk and makes the detector effectively a single-electron-counting micropattern gas device with pixelized readout (Cavoto et al., 2023, Anastassopoulos et al., 2018).

Mechanical realizations range from single-chip modules to tiled arrays. The “quad” module places four Timepix3-GridPix chips on a common carrier plate and reaches dimensions of 39.6 mm × 28.38 mm with an active surface coverage of 68.9%. The CAST septem detector arranges seven GridPix chips as one central chip surrounded by six identical chips, yielding a roughly circular, 78 mm-diameter readout plane with a 30 mm gas-drift height and an active volume of approximately 150 cm³ (Ligtenberg et al., 2020, Altenmüller et al., 9 May 2025).

Field-shaping structures are application-specific. Examples include a surrounding copper anode ring that maintains a uniform drift field up to the mesh edge, a guard plane 1 mm above the grid to minimize distortion at the amplifier entrance, and multi-ring or multi-frame field cages for extended drift regions (Manikantan et al., 2 Sep 2025, Ligtenberg et al., 2018, Desch et al., 5 Dec 2025).

2. Signal formation, amplification, and readout

GridPix operates as a gaseous drift-and-amplification detector. An absorbed X-ray or a traversing charged particle ionizes the gas, after which the primary electrons drift under a uniform electric field toward the integrated grid. Electrons entering a grid hole initiate an avalanche in the high-field amplification gap. In several papers, the gain is written in the first Townsend approximation as

$G = \exp(\alpha \cdot d),$

where $\alpha$ is the first Townsend coefficient and $d$ is the amplification gap. Representative operating points include about 350–400 V across a 50 μm gap in Ar/isobutane mixtures, corresponding to gas gains from about $10^3$ to $10^4$, and $E_{\rm gain}\sim 40$–$50\,\mathrm{kV/cm}$ in the polarimetric prototype (Krieger et al., 2014, Krieger et al., 2017, Manikantan et al., 2 Sep 2025).

Readout mode depends strongly on the ASIC generation. In Timepix-based X-ray systems, the detector commonly operates in Time-over-Threshold mode for charge measurement or Time-of-Arrival mode for timing, often in frame-based acquisition. Reported examples include a 600 μs shutter with 40 Hz full-ASIC readout, and a typical shutter time of 2.2 s with approximately 10% dead time for a 175 ms readout in the seven-chip CAST detector (Krieger et al., 2017, Altenmüller et al., 9 May 2025).

Timepix3-based GridPix systems instead provide per-hit data in a sparse, data-driven mode. Each hit can carry pixel coordinates, a Time-of-Arrival value with 1.56 ns resolution from a 640 MHz TDC, and a Time-over-Threshold value clocked at 40 MHz. In the polarimetric development based on Timepix3, the recorded ToA values are converted to a drift coordinate using

$z_i = v_{\rm drift}( \mathrm{ToA}_i - T_0 ),$

which yields a true three-dimensional point cloud. In the same framework, $\mathrm{ToT}_i$ is proportional to the collected charge $Q_i$, enabling reconstruction of local energy deposition along the track (Manikantan et al., 2 Sep 2025).

A key Timepix3 issue is time walk, because smaller signals cross threshold later than larger ones. In TPC studies this effect is corrected using the measured ToT as a proxy for pulse size, with the empirical form

$\alpha$0

After correction, residual drift-coordinate bias is reduced to below a few microns in the test-beam study (Ligtenberg et al., 2018).

A common simplification is to treat GridPix as uniformly dead-time-free. The published record is more specific: frame-based Timepix systems do incur shutter and transfer dead time, whereas the Timepix3 polarimetric concept is described as providing essentially dead-time-free, digital read-out through fully data-driven sparse acquisition (Altenmüller et al., 9 May 2025, Manikantan et al., 2 Sep 2025).

3. Gas mixtures, transport parameters, and resolution

The gas choice in GridPix is application-driven. CAST X-ray detectors and several soft-X-ray calibration studies used Ar:isobutane 97.7:2.3 by volume at 1050 mbar with a 3 cm drift length and $\alpha$1 (Anastassopoulos et al., 2018, Krieger et al., 2017). Collider-oriented TPC studies used the T2K mixture Ar:CF₄:iC₄H₁₀ = 95:3:2, with reported drift fields of 280 V/cm or 400 V/cm and drift velocities of 78.86 μm/ns or approximately 54.6 μm/ns, depending on the setup (Ligtenberg et al., 2018, Ligtenberg et al., 2020). Low-mass TPC work characterized helium-rich He:iC₄H₁₀ mixtures at 95:5, 90:10, and 85:15, while the polarimetric prototype considers, for example, Ar–DME 80:20 at 1 bar (Cavoto et al., 2023, Manikantan et al., 2 Sep 2025).

The helium–isobutane study provides a detailed benchmark for transport properties. For the 90:10 mixture at $\alpha$2 and $\alpha$3, the measured drift velocity rises from 0.6 cm/μs at 300 V/cm to 1.6 cm/μs at 900 V/cm, with agreement with Garfield++ to better than 10%. The same work reports transverse diffusion constants $\alpha$4 of 200, 110, and 95 μm/$\alpha$5 at 300, 700, and 900 V/cm, respectively, and longitudinal diffusion constants $\alpha$6 of 280, 140, and 130 μm/$\alpha$7. The attachment coefficient is measured as $\alpha$8, implying less than 2% hit loss over 9 cm (Cavoto et al., 2023).

For spatial resolution, the literature consistently identifies diffusion as the dominant contribution beyond very short drift distances. In helium–isobutane, single-hit resolution in $\alpha$9 is of order 100 μm and $d$0 in $d$1 over a 9 cm drift (Cavoto et al., 2023). In a single-chip Timepix3 TPC prototype, the transverse resolution is modeled as

$d$2

with $d$3 at zero drift; the study reports systematic distortions of 7 μm in the pixel plane and 21 μm in the drift direction (Ligtenberg et al., 2018). The quad study similarly found that the error on the track position measurement in both coordinates is dominated by diffusion and achieved corrected pixel-plane residuals with an r.m.s. of 13 μm over the full plane and 9 μm in the central border region (Ligtenberg et al., 2020).

In soft-X-ray calibration with Ar/isobutane 97.7/2.3, the transverse cluster rms peaks near 0.8 mm for both 1.5 keV and 8 keV, with fitted transverse diffusion coefficients of 474 μm/$d$4 at 1.5 keV and 504 μm/$d$5 at 8 keV (Krieger et al., 2017). By contrast, the next-generation polarimetric concept exploits lower diffusion: in a 1 cm drift volume filled, for example, with Ar–DME 80:20 at 1 bar and operated at $d$6, the ionization electrons are described as diffusing only approximately 8 μm rms transversely, with measured $d$7 at 5 cm drift and $d$8–$d$9 (Manikantan et al., 2 Sep 2025).

4. Low-energy X-ray detection and rare-event searches

GridPix first became prominent in low-background soft-X-ray detection at CAST. The InGrid-based low-energy X-ray detector described by Krieger et al. used a 3 cm drift region, Ar:isobutane 97.7%:2.3%, and a 2 μm aluminized Mylar entrance window. It detected the C K$10^3$0 line at 277 eV and reported an average background rate of $10^3$1 between 2 and 10 keV with lead shielding. The same work emphasized topological background suppression using cluster size, shape, and total multiplicity, with a likelihood cut tuned to keep approximately 90% signal efficiency across 0.3–8 keV and background suppression factors of order 10–100 in the 2–10 keV band (Krieger et al., 2014).

Calibration studies later quantified the soft-X-ray response from 277 eV to 8.048 keV in the same gas. Two energy estimators were used: the number of hit pixels and the total measured charge. Both were found to be linear in energy and to intercept at zero. The charge-based energy resolution follows approximately

$10^3$2

while the pixel-based measure follows approximately $10^3$3. Numerically, this corresponds to better than 20% at 0.5 keV and better than 10% above 2 keV (Krieger et al., 2017).

In the 2014–2015 CAST solar-chameleon search, a single-chip GridPix operated with Ar:isobutane 97.7:2.3 at 1050 mbar and $10^3$4, covering 0.2–10 keV. The total detection efficiency shown in the paper rises from approximately 10% at 200 eV to approximately 60% at 1 keV and to at least 90% above 2 keV. Event-shape rejection used three variables—cluster eccentricity, track-likeness, and the fraction of pixels within one $10^3$5 of the centroid—with an 80% software signal efficiency and rejection of at least 90% of the background. In the central 5 × 5 mm² “gold region,” the reported background in 0.2–2 keV was below $10^3$6 (Anastassopoulos et al., 2018).

The seven-chip CAST detector extended the concept with a roughly circular multi-chip plane, ultra-thin silicon nitride windows, and layered vetoes. One version employed a 300 nm Si$10^3$7N$10^3$8 membrane plus about 50 nm Al; another description specifies a 300 nm Si₃N₄ membrane coated with 20 nm Al and supported by a strongback. These windows were reported to increase low-energy transmission by more than a factor of five relative to the approximately 2 μm Mylar windows formerly used at CAST, with more than 50% transmission at 0.5 keV and more than 90% above 1 keV in the later characterization paper (Altenmüller et al., 9 May 2025, Desch et al., 5 Dec 2025).

Background rejection in the seven-chip system combines passive shielding, scintillator vetoes, an induced-grid FADC pulse-shape veto, an outer-chip or “septem” veto, and software classification. The 2025 CAST axion search used a fully connected Multi-Layer Perceptron with 14 features and reported a remaining background rate of approximately $10^3$9 in the central 5 × 5 mm² spot over 0.2–8 keV at a total efficiency of about 80% (Altenmüller et al., 9 May 2025). The later detector-construction paper reports that, in a 3500 h background campaign, the septem veto removes approximately 50% of the low-energy background, the FADC veto approximately 40% across 0–8 keV, and the scintillator veto approximately 20% above 2 keV, leading to a final background level of $10^4$0 (Desch et al., 5 Dec 2025).

5. TPC readout, tracking, and neutron detection

A major branch of GridPix development targets TPC readout. In this context, the principal advantages are single-electron sensitivity, very fine granularity, and direct access to per-hit timing and charge. The helium–isobutane study was motivated by tracking low-momentum muons and pions under severe multiple scattering constraints. It reports that a low-mass design using He-based gas, 25 μm ETFE windows, PMMA enclosure, and thin PCB interfaces yields an overall material budget below 0.1% $10^4$1 per readout plane. Fully efficient operation, approximately 100% hit detection, was reached at gains of about $10^4$2–$10^4$3 for He:iC₄H₁₀ = 90:10 and 85:15 at $10^4$4–420 V (Cavoto et al., 2023).

Timepix3-based TPC studies emphasize precision tracking. In the single-chip prototype, single-electron detection efficiency reaches nearly 100% at a grid voltage of 350 V, with no evidence of crosstalk or spurious double hits. The detector achieved a truncated-sum $10^4$5 resolution of 4.1% for an effective track length of 1 m, and the same paper states that simulation studies show that a pixel readout can improve the momentum resolution of a TPC at the ILC by about 20% (Ligtenberg et al., 2019). The underlying mechanism is that hit uncertainty is reduced and the number of effective measurements scales with track length rather than pad-row count.

The quad module addresses scalability to large readout planes. It consists of four Timepix3-GridPix chips with services routed to the back, a central guard electrode, and cooling integrated under the active area. In test-beam operation with 2.5 GeV electrons, the setup reached a total position resolution of 41 μm, with a total systematic error of 24 μm. The study concludes that the GridPix concept can achieve sub-50 μm spatial resolution over pad-row lengths of a few millimeters and that scaling through tilable quad units is straightforward (Ligtenberg et al., 2020).

GridPix has also been adapted to neutron detection via the time projection method. The 2017 neutron prototype used eight TimePix ASICs with post-processed InGrid meshes mounted on an “Octoboard,” a 3.8 cm drift region, and a 96% $10^4$6B$10^4$7C converter layer 1.04 μm thick on the cathode. Thermal neutrons captured in the boron layer produce back-to-back $10^4$8 and Li ions, whose dense Bragg-peak tracks are reconstructed in 3D using a combination of TOT and TOA information. The reported spatial resolution is $10^4$9, and the readout architecture based on the RD51 Scalable Readout System is described as conceptually straightforward to upscale to much larger active areas (Köhli et al., 2017).

This range of implementations suggests that “GridPix detector” is better understood as a technology platform than as a single detector geometry. The common element is the integrated hole-per-pixel amplification stage; drift depth, gas, window technology, veto structure, and acquisition mode are adapted to the target measurement.

6. X-ray polarimetry, radiation hardness, and prospective developments

The most recent development direction connects GridPix to photoelectric X-ray polarimetry. The IXPE legacy paper frames GridPix as a radical departure from the two-dimensional, analog-only gas pixel detectors flown on IXPE, combining true three-dimensional imaging capability with essentially dead-time-free, digital read-out. The motivation is the need for detector upgrades in future polarimetric missions, particularly in the read-out ASIC and possibly the multiplication stage (Manikantan et al., 2 Sep 2025).

In the reported polarimetric concept, a 1 cm drift volume is coupled to a Timepix3-based InGrid readout. Each of the 65,536 pixels contains its own pre-amplifier and discriminator, giving an input capacitance of only approximately 10 fF per pixel, electronic noise of approximately 60 $E_{\rm gain}\sim 40$0 rms, and threshold dispersion of approximately 35 $E_{\rm gain}\sim 40$1 after per-pixel equalization. The dynamic range extends up to approximately 14,000 electrons, stated to comfortably cover the 2–20 keV X-ray band. The measured azimuthal modulation is characterized by

$E_{\rm gain}\sim 40$2

and early tests reach $E_{\rm gain}\sim 40$3–$E_{\rm gain}\sim 40$4 at 4.5 keV, already comparable to or exceeding GEM-based GPDs (Manikantan et al., 2 Sep 2025).

Because a space-borne polarimeter requires radiation tolerance, the same work reports proton irradiation of a complete GridPix assembly at the Bonn Isochronous Cyclotron with 13.6 MeV protons at rates adjusted to below 100 cps on the detector. By combining a Geant4 model of the test stand with cosmic-ray spectra from SPENVIS, the authors determined that approximately $E_{\rm gain}\sim 40$5 protons deliver the equivalent 20 year LEO total ionizing dose plus displacement damage in the 1 cm gas cell. The actual exposure was $E_{\rm gain}\sim 40$6 protons, approximately 100 times the flight dose, with no observed single-event latch-ups or destructive sparking. Post-irradiation calibration with an $E_{\rm gain}\sim 40$7Fe source showed a modest gain shift of approximately 8%, stable energy resolution with FWHM changing from 21% to 17%, and unchanged noise, threshold dispersion, and 3D imaging fidelity (Manikantan et al., 2 Sep 2025).

The forward path is described in unusually concrete terms. Proposed detector-side upgrades include reducing the mesh–pixel distance to 50 μm or optimizing the hole aspect ratio, using gas mixtures richer in DME or Ne–DME, and implementing a fully sealed module with an integrated thin Be window and bake-out-qualified InGrid. On the electronics side, the next-generation “XPOL-III” ASIC is said to feature deeper ToT counters, a faster 80 MHz timestamp, and on-chip clustering logic to reduce downstream data volume, boosting throughput by another factor of 5–10. Preliminary Garfield++ simulations predict that these changes could push the modulation factor to $E_{\rm gain}\sim 40$8 at 4 keV, lower the energy threshold to approximately 500 eV, and sustain source count rates of $E_{\rm gain}\sim 40$9 with less than 1% dead time (Manikantan et al., 2 Sep 2025).

A plausible implication is that GridPix now occupies a distinctive position among gaseous detectors: in CAST it has been optimized for low-threshold, ultra-low-background X-ray detection; in TPC programs it has been optimized for diffusion-limited hit resolution and digital $50\,\mathrm{kV/cm}$0; and in polarimetry it is being reconfigured for three-dimensional photoelectron-track imaging under space-radiation constraints. The shared fabrication principle remains stable across these domains, but the detector’s scientific role is increasingly defined by the surrounding system architecture—window technology, veto instrumentation, field uniformity, readout mode, and calibration strategy—rather than by the mesh-on-pixel concept alone.