• John
  • Felde
  • University of Maryland
  • USA

Latest Posts

  • USA

  • James
  • Doherty
  • Open University
  • United Kingdom

Latest Posts

  • Andrea
  • Signori
  • Nikhef
  • Netherlands

Latest Posts

  • CERN
  • Geneva
  • Switzerland

Latest Posts

  • Aidan
  • Randle-Conde
  • Université Libre de Bruxelles
  • Belgium

Latest Posts

  • Vancouver, BC
  • Canada

Latest Posts

  • Laura
  • Gladstone
  • MIT
  • USA

Latest Posts

  • Steven
  • Goldfarb
  • University of Michigan

Latest Posts

  • Fermilab
  • Batavia, IL
  • USA

Latest Posts

  • Seth
  • Zenz
  • Imperial College London
  • UK

Latest Posts

  • Nhan
  • Tran
  • Fermilab
  • USA

Latest Posts

  • Alex
  • Millar
  • University of Melbourne
  • Australia

Latest Posts

  • Ken
  • Bloom
  • USA

Latest Posts

Posts Tagged ‘pixels’

A Change of Pace

Monday, February 4th, 2013

Some physicists and engineers from Purdue and DESY, and me, at the beamline we used to test new pixel designs

Every so often, a physicist needs a vacation from doing data analysis for the Higgs boson search. A working vacation, something that gets you a little closer to the actual detector you work on. So last week, I was at the DESY laboratory in Hamburg, Germany, helping a group of physicists and engineers study possible changes to the design of individual pixels in the CMS Pixel Detector. (I’ve written before about how a pixel detector works.) We were at DESY because they had an electron beam we could use, and we wanted to study how the new designs performed with actual particles passing through them. Of course, the new designs can’t be produced in large scale for a few years — but we do plan to run CMS for many, many years to come, and eventually we will need to upgrade and replace its pixel detector.

What do you actually do at a testbeam? You sit there as close to 24 hours a day as you can — in shifts, of course. You take data. You change which new design is in the beam, or you change the angle, or you change the conditions under which it’s running. Then you take more data. And you repeat for the entire week.

So do any of the new designs work better? We don’t know yet. It’s my job to install the software to analyze the data we took, and to help study the results, and I haven’t finished yet. And yes, even “working on the detector” involves analyzing data — so maybe it wasn’t so much of a vacation after all!


Curveballs are Fun

Friday, December 19th, 2008

We’re not big fans of rigid hierarchy in academia, not even on big experiments like ATLAS with multifarious coordinators and project leaders.  On the one hand, this means that nobody ever gives me orders — but on the other hand, it does mean that there are a lot of people who can give me “strong suggestions.”  And sometimes one of those people decides to throw me a curveball…

Friday was a day of two work days.  First I worked a pretty normal eight hours debugging code, then spent the evening at a few holiday parties before heading to the ATLAS Control Room at 11 PM for an eight hour shift.  After I arrived, while waiting for the expert running things to let me do my shift so he could go home and get some sleep, I found an email in my inbox which had been sent only that evening.  It asked me to give a talk at the ATLAS Inner Detector-wide meeting about the activities of the Pixel group over the previous week.  All of the work to be discussed had done by others rather than me, and some of it I hadn’t even been aware of — and the talk was on Monday.

I had never received a request like that before, but believe it or not, I’m not complaining.  Yes, it was rather short notice, but it wasn’t even a strong suggestion, really — I was allowed to opt out if I didn’t have time.  But more importantly, after I thought about it, I decided that giving the talk was entirely a good thing for me.  There are a couple of reasons I can think of to give an inexperienced person the responsibility of summarizing the work of the whole Pixel Collaboration.  One is to give everyone who’s done work on the Pixel Detector a turn to make their participation visible to the wider Inner Detector community, even if their work contributed only indirectly to the material being presented.  (In my case, the contributions were taking shifts and writing tools for analyzing calibration scans.)  Another is to give the person giving the talk the opportunity to learn more about the broader work on the detector.

In my case, it was an opportunity I had to take quickly, so I sprung into action: I checked the agenda for Monday, found that the meeting wasn’t until 3 PM, and decided I could delay the writing of the talk itself until Monday morning.  I did look at the list of topics to cover during my shift, and asked a few questions; then I printed out all the supporting material on Sunday night.  But otherwise I continued with my weekend as scheduled.  This required Monday to be a very productive day: I got up at 6:30 AM to start reading everything I had printed out, then got intto work by 8:30 am and started writing.

Most of the slides were summarized from elsewhere, or even provided for me.  The most important part of what I had to do was to understand what was on them, so that I could provide context for the work and avoid sounding like an idiot if I had to go “off script.”  The way I think about it was that the people who had done the studies had given me intermediate-level information to present, and nobody would expect me to answer really hard stuff during a summary talk, but that I absolutely had to have a command of the basic way in which the material I was presenting fit into the broader picture.  I needed some help with that, and got plenty of it, from the experts who did the original work as well as from the person who asked me to give the talk.

By 3PM, I was ready, but also nervous about talking in a new venue and in front of new people.  I hadn’t given myself time to be nervous up until that point, but I had plenty of it while watching the other four talks ahead of mine.  My strategy during the talk itself was to try to sound confident that I understood everything, unless I actually didn’t know something and had to punt questions to the other pixel people in the room — which it turned out I never did.  In the end, in fact, I was told the talk was clear and went well.   So I suppose I managed to hit the curveball, and it definitely made for a more exciting Monday than usual!


How Tracking Works

Tuesday, November 25th, 2008

Author’s note: I didn’t mean for this to end up so complicated that it had equations, figures, and footnotes, but that’s how it turned out. I do apologize for the inconvenience, and if it’s any compensation I can assure you that about half the footnotes are funny.

I’ve written before about how a pixel detector works, but at the time I left as a “topic for another day” the broader question of what a pixel detector is for.  I’m going to answer one part of that question today, and discuss the tracking system, of which a pixel detector is one possible component.1 I’ll have to leave the question of the specific advantages of using pixels, as opposed to other tracking technologies, for another other day.

Regardless of the technology used, the basic idea of a tracker is to put together a bunch of stuff that measures the path a charged particle has taken.   The “stuff” could be silicon, in which electron-hole pairs are separated as the charged particle passes through, and can be used to produce a current, as I explained in my pixel detector entry.  It could also be gas, in which case electron-ion pairs are separated and produce a current in wires; this is the technology used in the ATLAS Transition Radiation Tracker.  If you want to “track” a baseball through the stands, the “stuff” is people: even if you can’t see the baseball in the crowd on other side of the stadium, you can see where it’s gone by who stands up or jumps down and starts grabbing under the seats.  An individual jumping person, or silicon pixel producing a current, is what we call a hit.

Our primary interest actually isn’t in how particles move through the detector, even though that’s what we directly measure.   So let me take a step back now and describe what we are measuring, first and foremost: momentum.

Momentum: What It’s Really All About

The best way I can think of to describe momentum in a few words is to quote Newton and call it the “quantity of motion.”2 It reflects not just the speed and direction (i.e. velocity) of an object, but also the amount of stuff (i.e. mass) that makes up that object.  In ordinary life, if you double the mass then you double the momentum, and if you double the velocity you get double the momentum too; in other words:

  • p = mv

where m is the mass, v is the velocity, and p is the momentum.3 Unfortunately, things get a little more complicated when the particle goes really fast, which they usually do in our detectors; then the equation doesn’t work anymore.  We’ll get to one that does in a minute.

Momentum intuitively seems the same as energy of motion, but technically the ideas aren’t exactly the same, and it just so happens that the difference is important to how the LHC detectors work.  One way to think of the energy of a particle is as follows: if you slammed the particle into a big block of metal and then extracted all the ensuing vibrations of the metal’s atoms4 and put them in a usable form, it’s the amount of mechanical work you could do.  In fact, that’s exactly what a detector’s calorimeter does, up to a point.  It’s made of big blocks of metal that absorb the particle’s energy, and then it samples that energy and turns it into an electrical current — not so we can do any kind of work with it, but just so we know how much energy there was in the first place.  So the calorimeter is the piece of ATLAS or CMS that measures the energy of particles and absorbs them; the tracker, by contrast, measures the momentum of particles and lets them pass through.   These two pieces of information are related by the following equation:

  • E2 = p2c2 + m2c4

where p and m are still momentum and mass, E is the energy, and c is the speed of light.  The intuitive understanding of this equation is that the energy of a particle is partially due to its motion and partially due to the intrinsic energy of its mass.  The application to particle detectors is that if you know the mass of a particular particle, or if it’s going so fast that its energy and momentum are both huge so that the mass can be roughly ignored, then knowing the energy tells you the momentum and vice versa — and knowing at least one of the two is critical for analyzing where a particle might have come from and understanding the collision as a whole.  We have both kinds of systems because they have different strengths — for example, some kinds of particles don’t get absorbed by the calorimeter, and some kinds of particles (the uncharged ones) can’t be seen in the tracker — and together, they cover almost everything.

(By the way, the second equation is relativistic; that is, it’s compatible with Einstein’s Theory of Relativity.  That means it always works for any particle at any speed — it might assume that space is reasonably flat or that time really exists, but these are very reasonable assumptions for experimental physicists working on Earth.  For those who haven’t seen the equation before and enjoy algebra problems: what famous equation do you get if you take the special case of a particle that isn’t moving, i.e. with a momentum of zero?)

Particle Motion and Momentum

The next ingredient you need to understand what a tracker does is something I haven’t mentioned yet: the whole thing is enclosed in a huge solenoid magnet, which produces a more-or-less uniform magnetic field pointing along the direction of the LHC beam.  As a charged particle moves through a magnetic field, the force exerted on it by the field works at a right angle to both the direction of motion and the field — I tried to illustrate this in figure 1, where the magnetic field is pointing into your screen if you assume the particle is positively charged.5 This means that as the charged particle flies from the center of the detector, it curves (figure 2).  The amount it curves by is inversely proportional to the momentum, which means that higher-momentum particles curve less.  Along its path, it leaves hits in the detecting material, as I discussed above (red dots, figure 3).  Finally, in a process called track reconstruction, our software “connects the dots” and produces a track — which is just our name for “where we think the particle went” (figure 4).

You’ll notice that figure 2 looks a lot like figure 4, but the conceptual difference is a very important one.  The red line in figure 2 is the actual path followed by the particle, which we don’t see directly, while the black line in figure 4 is our track as determined by detector hits.  If we do our job right, the red line and black line should be almost exactly the same, but that job is complex indeed — literally thousands of person-years have been put into it, including two or three Seth-years6 spent on detector calibration and writing automated tools for making sure the tracking software works properly.

The detector is shown here with only three layers.  Although this would be enough to find a particle’s path in ideal circumstances, we actually have many more: this allows us to still make good measurements even when one layer somehow doesn’t see the particle, and to get a final result for the path that’s more accurate.  And don’t forget that there will actually be many particles passing through the detector at the same time — so we need lots of measurements to be sure that we’re seeing real tracks and not just a bunch of “dots” that happen to “line up”…!

More Than Just Momentum

If you measure the path of a particle, you can do more than just find its momentum; you can also see where it came from, or at least whether it could have come from the same place as another particle.  Pixel detectors excel at making accurate measurements to figure out this kind of thing, but as I said already, to do that subject justice will require another entry.

So there you have it.  In a very broad sense, that’s what I’m working toward when I talk about calibrating the pixel detector.  Tracking provides critical basic information about every charged particle that passes through our detector; combined with data from the calorimeter and the muon systems, this information is what will let ATLAS and CMS measure the properties of the new particles that we hope the LHC will produce.

1 Both ATLAS and CMS have one, but many other detectors at colliders do not, because the technology is complex, relatively new, and expensive.
2 See Corollary III here for what he says about it, if you like your science extra-opaque.
3 I’m really not sure why we always use p for momentum, although a good guess seems to be that it’s related to impetus or impulse.
4 A friend of mine, who has the mysterious superpower of understanding how bulk matter works rather than just mucking about with individual particles, looked at a draft of this and was very concerned that I’m implying that all the energy from such a happening would end up as atomic vibrations. So let the record show that this probably isn’t true. And now, if you’d be so kind, can we pretend it is true? It will make illustrating my point very much easier. Thanks!
5 The particle is definitely not actual size, and don’t ask me why it’s green.
6 A Seth-year doesn’t make nearly as big a contribution as a year of work by any of our real experts, but they do happen to be of particular interest to me.

(Back to top)


Currently I am sitting at Geneva airport waiting for my plane to finally leave for Amsterdam. Looking east I see something the average cernoise is always happy to see: First snow on the Jura.


Pixeling along, 24-7

Tuesday, August 12th, 2008

The LHC startup is getting closer and closer. A few previous blog entries already informed you that there was a successful insertion of beam into the LHC. This is of course great news, but means that the testing and final preparations of the detectors has now become serious business. As the CMS pixel detector was planned to be installed as one of the final components before the first beam was delivered, we are very much under pressure to be ready in time. The initial performance of the CMS forward pixel detector You can read that as ‘continuously on shift until things are stable enough to be run by non-experts’. This also explains the lack of blog entries by me and some of the other people working closely on the detector, at the moment the pressure is really on and the detector comes first!


How a Pixel Detector Works

Friday, July 25th, 2008

If you read about the LHC detectors a lot, then you’ve heard about pixel detectors from time to time. Both ATLAS and CMS have one. But what is a pixel detector? Ken explained briefly here—and mentioned, in the process, that when he was my age, their trackers only had 26,000 channels, dinosaurs roamed the earth, they had to walk ten miles to the lab in the snow, and it was uphill both ways—but, anyway, I thought it might be worth expanding on the explanation a bit.

A schematic of the CMS pixel detectorI’ve heard several times that a pixel detector is like a digital camera. Indeed, they both have pixels, but how far does the parallel really go? The biggest, most obvious difference is the shape, which follows from different purposes. Your camera is designed to collect all the light that reaches a particular point, namely the camera lens, and so there’s a flat rectangle of pixels at the spot where the light is focused. A pixel detector at a collider is designed to collect information on particles coming from a particular point, namely the place where the particles collide, so it’s built to surround that point as completely as possible. The schematic of the CMS detector in the upper right illustrates this point nicely; obviously, on a large scale, it looks nothing like the inside of your camera. (A pixel detector has more pixels too, but it’s about the same scale—ATLAS has about 80 million compared with 6 million in my digital camera.)

But the real question is how the individual pixels differ, and the answer turns out to be: not by as much as you might guess. I’m no expert on digital cameras, so I had to look it up, but it turns out that digital camera sensors are actually little arrays of silicon detectors. When a visible photon hits the layer of silicon that makes up a pixel, it knocks an electron out of its place in the silicon. The electron is pulled in one direction by an electric field, while the hole (the empty space where the electron should be) is pulled in the other. The charge thus collected by each pixel is proportional to the number of photons that hit it, and hence the intensity of the visible light; this charge is eventually read out by (for example) a charge-coupled device, and the picture can be assembled.

How a silicon detector worksThe basic idea of the silicon detectors we use in particle physics is the same, but we’re looking at fewer particles with much higher energy. Whereas a visible photon has an energy of a few electron volts (eV), the interesting particles passing through our silicon detectors at the LHC will have an energy somewhere from several hundred million eV to several hundred billion eV. Thus, as illustrated at left, when a particle passes through our silicon detector, it knocks loose a bunch (thousands or tens of thousands) of electron-hole pairs and doesn’t stop at all. That’s exactly what we want, actually; this type of detector is for telling us where a particle went, not for absorbing it. (This is called tracking, and it’s a topic for another day.)

That’s step one. In step two, the electron-hole pairs are pulled in opposite directions by an electric field, and pulled into “contacts.” (Actually, specially-doped regions of silicon, if you’re curious.) In step three, the charge built up on those contacts produces a current that flows into our electronics—another topic for another day.

So far I’ve discussed silicon detectors in general; I could just as well be talking about the “silicon strip” detectors that are also used in ATLAS and CMS, for example. The key feature of a pixel detector is that the individual contacts are two-dimensional; for every 0.05 by 0.4 millimeter pixel, there’s a separate circuit and separate electronics. This gives us a very precise measurement of where, exactly, the particle passed through the detector.

A little piece of pixel detector

Of course, those pixels aren’t so very small—I’m pretty sure they’re actually larger than the ones in your camera. But they have to read out much faster than your camera does, since the LHC produces collisions forty million times a second. They also have to withstand the intense radiation found right next to the collisions, which can damage the silicon structure, for years and still work. It’s challenges like this that make pixel detectors such a complex and expensive job, but they’re vitally important for our physics program—but that, yet again, is a topic for another day.

Update (November 25): “Another day” has arrived, or at least one of them: How Tracking Works


Pixel cosmics

Friday, June 20th, 2008

While most of my time is spent on computing issues, there is (just a little) more to my research life.  Our group is also involved in building, installing and commissioning CMS’s forward pixel detector.  This is a silicon-based detector which records the position of charged particles.  Wafers of silicon are segmented into little squares (the pixels) that are 150 microns on a side.  When charged particles pass through a pixel, a little bit of charge is liberated in the material.  This charge can be picked up on a tiny amplifier and recorded into the data stream.  By looking at patterns of pixels that have charge in them, we can reconstruct the paths of the particles that passed through.  One particularly striking aspect of these detectors (to me, at least) is the density of readout channels.  The forward pixel detector is a few wheels that are about 10 cm in radius, but the whole detector has millions of readout channels!  (When I was in grad school, I worked on a similar device that had about 26,000 channels.  It seemed like a lot at the time.)

Even though this detector hasn’t seen any beam yet, it is already time to think about replacing it; in a few years, its performance will be impaired by through radiation damage (it sits right next to the beam), and we can apply lessons that have been learned from building the current detector to the new one.

This summer, we set up a little lab down the hall from my office where we can start to study prototype detectors and readout chips for the next round.  We have a few spare pieces from the current detector and a rudimentary readout system for the electronics.  A couple of our students set up a couple of scintillators and phototubes so that we can trigger the readout on cosmic rays that pass through the silicon.  Here is a plot of the pulseheights that they read out of the system during a cosmic-ray run:

The curve that is superimposed on the distribution is a Landau function, which is what you expect to see.  Not bad!  Our guys pretty much got this on the first attempt — good for them.  I’m told that no one has actually seen an output spectrum from real incident particles from these devices for some time.  (We’ll be seeing lots of them in a few months, when the LHC starts up.)  Now we can start to think about looking how the performance varies as we vary the operating parameters, and about testing out prototype detectors when they become available.