The expected lead-lead collision rate in the LHC Run3 (2019 onwards) is 50 kHz. This corresponds, on average, to a collision every 20 μs. In the ALICE TPC, a 90 m³ gas volume where ionization electrons take 100 μs to drift the full 2.5 m distance to the Readout Chambers, the equivalent of 5 events will overlap, which suggest that in order to record the information of all charged particles of all collisions, a continuous readout technique should be used.

This poses two severe problems to the current Multi Wire Proportional Chambers used for the readout of the current TPC. These chambers are composed of an anode wire grid, where the electron amplification occurs, sandwiched between a cathode wire grid and a flat pad plane, where the signals are read out. But in addition, on top of this structure there is another wire grid, called the gating grid, which makes it possible for the detector to perform at high event rates and multiplicities. Introduced in the LEP era (ALEPH, DELPHI), the gating grid allows one to prevent electrons from non-triggered events to reach the amplification region, by applying an alternating voltage on its wires. Upon a trigger, the gating grid is quickly switched to a flat potential thus allowing ionization electrons of this particular event to reach the anode wires and induce a signal onto the pads. Now, the crucial role of the gating grid is to close itself just after all electrons from our event have reached the anode grid, such as to trap the positive ions produced in the avalanches and prevent that they invade the drift volume. In the same way, charge from non-triggered events is never amplified and thus no extra positive ions are produced. This mechanism allows one to keep the drift volume relatively clean of slow drifting ions (160 ms full drift time) which otherwise would build up a considerable space-charge density and lead to important distortions of the electric drift field. For example, if the current TPC would be run, without switching the gating grid, at 50 kHz Pb-Pb, tracks would appear distorted by as much as 1 m. But 1 m distortions are obviously too much to correct for. If we were to use the gating grid, the maximum trigger rate would be determined by the 100 μs for the drift of the electrons (gating grid open) plus another 180 μs for the corresponding ions to reach it (gating grid closed); this is about 3 kHz, much lower than the 50 kHz the LHC is expected to provide. So a gating technique is not possible.

Furthermore, the amount of charge reaching the anode wires would lead to the saturation of the amplification field in their vicinity, thus affecting the uniformity of the gas gain. Now, remember here that one of the functions of the ALICE TPC is particle identification through measurement of the specific energy loss of all charged particles; but if the gain is modified by fluctuating space-charge in the amplification region, the dE/dx determination will be seriously affected. So wire chambers altogether won’t do the job.

So we look to alternative solutions for the readout chambers, and an obvious choice are micro-pattern gaseous detectors, GEMs. These Gas Electron Multipliers, the famous foils with lots of tiny holes introduced by F. Sauli in the 90’s, are certainly capable to cope with the rates and multiplicities we expect and, it is said, provide ‘intrinsic ion blocking’, just what we need. However, standard configurations of triple GEMs do not provide sufficient ion blocking for us. We define the Ion Back-Flow (IBF) as the number of positive ions, after amplification, escaping back into the drift volume per initial primary electron. In order to keep the track reconstruction distortion to a bearable level, of order of 10 cm, the IBF from a GEM structure should be 1 % or below. Standard triple GEMs achieve about 5%. So there is some way to go. After an intensive R&D program, an IBF below 1% has indeed been achieved by using non-standard configurations of stacks of 4 GEMs, with hole pitches different from standard, and with voltages and fields different from standard. It turns out that the minimization of IBF enters in competition with the energy resolution, i.e. the precious dE/dx performance, and by careful optimization of the GEM structure a good compromise has been reached.

One GEM foil used in a full size prototype of the TPC inner chamber.

But what about the stability of such an arrangement? GEMs have been optimized for years in order to be robust against discharges. Although we have departed substantially from the standard configuration, it turns out that the sharing of the gain between four, rather than three, GEMs provide, at our odd fields, the same discharge probability (about 10^-8 for alpha particles) as the standard device. It should be noted here that, for various reasons, the operating gas of the upgraded TPC is Ne-CO₂-N₂ (90-10-5), where the addition of N₂ has proven to strengthen the stability.

So we think we have a concept that guarantees charged-particle momentum determination and excellent particle identification through dE/dx. And we keep the beautiful field cage of the TPC.

But there is one more change we have to undertake. The pad plane has now switched roles from being a cathode to, in the GEM case, an anode. This means that the polarity of the signal will be negative. This ‘little detail’, and the need to read out all pads continuously, leads to the necessity to redesign and build a new set of front-end electronics, of which the first test samples are now being tested.