Image: the High Granularity Timing Detector and where it will fit within ATLAS. (Credit: ATLAS collaboration)
Last week, I travelled to Lyon to give an overview of the status of the ATLAS High Granularity Timing Detector (HGTD) upgrade project for the leaders of the IN2P3 (the French particle and nuclear physics institute). For each major infrastructure project where the IN2P3 is involved, there is a yearly meeting to review progress, expected costs and person power needs. This scrutiny is necessary because building a new instrument can be a wild, rocky ride. I wanted to dedicate this post to the trials and tribulations of Instrumentation, and make a case for why it’s vital for every aspiring young physicist to get involved in the business of building detectors.
I spend most of my time these days project managing the analysis of the collision data collected by ATLAS: but we could not even contemplate exploring nature at the energy scales of the LHC without our unique, irreplaceable instruments. Of course, in outreach activities we talk about how large and impressive our detectors are, but we rarely put forward that it is nothing short of a miracle that we managed to design and assemble these colossal cameras. I was not around when the original ATLAS detector was first put together, but even my involvement in a comparatively modest upgrade programme has given me a taste of the challenges faced by those wishing to construct a detector.
HGTD, as the name implies, is a timing detector. Its role is to maintain ATLAS’s performance in the much more intense beam we will have for the last phase of the LHC, by adding additional information: time. Recent developments in silicon technology mean that there now exist sensors which can operate with the required time resolution. They involve thin layers of silicon and high bias voltages (usually several hundred volts) to get extremely quick response times. We can build individual timing sensors, hook them up to power supplies and an oscilloscope, and start measuring hits. So far so good!
The problem is all to do with the scaling. One sensor (typically at most 1mm by 1mm large) by itself will not help us to discover new physics. To be useful, we need to cover large areas. So, we create matrices of 15×15 of the sensors (i.e. 225 sensors in total). That covers about 2cm2 of area. That’s still not enough to be useful and already poses a problem: 225 sensors means 225 channels to read out. Even the best oscilloscopes on the market only have 16 input channels. So we need a custom readout card to extract the information from each sensor.
Thankfully many labs, including Clermont, have excellent micro-electronics engineers who can craft this kind of custom read-out circuitry, but it is delicate business. For instance, what if a stray particle from a collision happens to hit one of the circuit paths carrying the signals? That would scramble the information. It turns out that there are A LOT of particles in a detector like ATLAS (who knew?) and that this would happen intolerably often. So inside the readout card, each path needs to be triplicated, with the same information flowing along three different routes so that at the end, if they disagree (because of a stray particle impact), the errors can be corrected by taking a “majority vote”. To be most effective, the three routes need to be as far away as possible in space. And remember there are 225 channels, and the chip is only 2cm2 in surface area, so… it’s quite an optimisation challenge!
Ok, assuming we solve the problem of the readout card: how can one physically attach the sensors to them? 225 sensors, each needing a separate input/output wire and power supply: at this stage we are talking about hundreds of wires to attach on an area the size of your thumb. Probably you can imagine that a soldering iron won’t do the trick. Special “wire bonding” techniques are used for this, to somehow (it still seems incredible to me) achieve this level of density. And now that we have attached a pair of sensor matrices to a readout card, we have one single module.
Image: a microscope view of the wires attaching a readout card to a sensor array. It’s pretty dense. Credit: ATLAS collaboration.
Image: a single HGTD module, with a readout chip connected to two sensors (glued directly below it). Credit: ATLAS Collaboration.
Despite what already feels like a monumental achievement, so far we have still instrumented only a couple of square centimetres. We still need to instrument square meters. In total one needs 8000 modules to do the job, and the next challenge comes into sight: each sensor needs to have a voltage of at least a 300 volts applied, and at this stage we have 225 x 2 x 8000 = 3 600 000 sensors on at once with high voltage. That generates heat. Without care, you will very quickly cause a fire. To make matters worse, this sort of electronics works best at cold temperatures: -30C. So, now a whole new aspect comes into play: how to keep this device freezing cold when its natural tendency is to burst into flames. We need to design a hermetic vessel to hold the assembled instrument, and use liquid CO2 cooling to evacuate heat, with fine tubes snaking their way across the mechanical structure of the detector.
Image: a prototype support structure with cooling plate, modules (sensors+readout cards), mechanical supports, and electronics. Credit: ATLAS Collaboration.
Wait, mechanical structure? Of course, did you expect that the modules would hang there in mid-air while we take data? Something needs to hold them, their cables and pipes in place. A special structure is 3D printed, custom-made for each part of the instrument, to hold everything in place, like the skeleton of the device. One needs special robotic systems to glue the sensors on, as any slip of the hand could easily destroy the precious modules.
Going back to the cooling, of course you need almost perfect thermal contact to keep the temperature under control. Thankfully there are products like thermal paste to help distribute the heat… but did anyone check if that paste will last 15 years and still work?That’s how long HGTD is needed for, with almost continuous operation. Did anyone check that it is resistant to radiation? It will be operating in one of the most dense environments on the planet. And did anyone check that it will do both those things at once? If you think the manufacturer knows that information about their own product, then I have a bridge to sell you.
And it goes on: each step of the process of design of an instrument presents new problems, some of them seemingly trivial, but with far reaching consequences. Recently, a simple 12-volt DC-to-DC converter nearly put a stop to the entire LHC upgrade programme because of freak behaviour after irradiation and cooling, which not only broke the component, but fried anything it was attached to. More mundane problems also have the power to blow projects of course: what happens when you realise you need much more wire than your ordered to complete a project (we typically buy wire in lengths of kilometres) but the manufacturer has gone bust and the technology to make it is obsolete?
Instrumentation is risky business, it involves worldwide collaborations trying hard to build custom, unique tools against all the odds. And that is all before a single collision is recorded. HGTD is entering the production phase soon, which will be followed by assembly and, with a fair wind and a dose of good fortune, installation in 2028. Building the instruments for our science is a skill we cannot afford to lose. Major upgrade projects are few and far between: retaining the skills to project manage and deliver these types of project is a challenge for the field. The next generation of detectors (for the future circular collider) will probably be built by people who have not even done their PhDs yet. For the next hadron collider, instruments will be built by people not even born yet. So not only is it important for young scientists to work on instrumentation where they can, it is vital for the future of the field to pass on the skills needed to successfully complete such ambitious projects.
Louie D. Corpe 