CLEO Tracking Week In Review

May 23, 2006

I’ve wasted a couple of weeks of my life tracking down a rather odd problem with the CLEO detector, and I feel a certain need to offload. So here follows a rather detailed discussion of the CLEO tracking system — actually, drift chambers in general, so if you’ve ever wondered how exactly we follow all the particles that we create in collisions, here you are.

Say you have a particle whose path you wish to trace. You can’t really look directly at it; you’d need very intense light, which would disrupt the particle’s path, and at any rate isn’t easy to make. We instead put instrumented material in front of the particle. The aim is to use the interactions of the particle with the material to figure out where it was, while having those interactions be so gentle that the particle isn’t greatly affected (usually by using some sort of amplification of a weak signal). This is the principle behind detectors all the way from the cloud and bubble chambers of yesteryear to the silicon (and diamond?) detectors developed today.

CLEO uses a technology called the “drift chamber.” These consist of a volume of gas with lots of wires, running parallel to each other, spread out through the region (the guts of one can be seen in this photo, and the sort of wire arrangement we use is depicted here). The wires are brought to high voltage, with some positive and some negative; this sets up precisely shaped electric fields. As a particle traverses the chamber, it runs into atoms and molecules of gas and knocks electrons off them. The energy required to do this isn’t terribly large, so the particle’s momentum is not affected much (although the energy loss is measurable and important for other reasons).

The liberated electrons start drifting along the electric field lines towards the positively charged (“sense”) wires. Because they keep bumping into more gas, for most of the drift period, their speed is limited by the time between collisions; this “drift velocity,” which depends on the electric field and the type of gas, occupies a lot of some people’s time.

As the electrons approach the sense wires, the electric field increases as the field lines get closer together. At some point near the wire, the electrons accelerate enough between collisions to liberate more electrons when they crash into the gas. These new electrons go on to free more electrons, and so on in what’s called an “avalanche” — this is the main amplification step, where a few tens of primary electrons from the particle are turned into a measurable signal.

So what do we do with this information? The positions of the wires that were hit gives us very rough information on where the particle went, up to a couple of centimeters (a typical separation between wires): if we stop here we have a Multiwire Proportional Chamber. But we can do much better. As mentioned earlier, for most of their drift, the electrons travel at a determined velocity. You can design the field and choose the voltages such that the drift speed is pretty much constant over large regions. This means that the time the electrons take to drift can be translated to a distance measurement — you now know, to a precision more like 100 micrometers, how far away from the wire the particle passed.

Of course, you only know the particle position to within a cylinder of some distance from the wire. At CLEO, the drift chambers are cylindrical, and we know the particles are supposed to come out from the center of the cylinder, so the ambiguity is essentially one of “left” or “right” as the point of closest approach is at practically the same radius as the wire. To break this ambiguity, we stagger the wire layers at different radii, so going “left” or “right” at a certain radius means you hit different wires at the next stage. (Incidentally, the way you do the staggering can cause interesting effects, since positive and negative particles are bent in opposite directions by the experiment’s magnetic field and so produce different stagger patterns.)

One final word: you may have noticed that none of this information tells you where along the length of the wire the particle passed (in the usual experimental setup, this is defined to be the z-axis). We can get this information — at the price of breaking the cylindrical symmetry. Essentially, you give a layer of wires an angle to the z-axis, and tilt another layer in the opposite direction (these are “stereo” layers, as opposed to “axial” layers where the wires are strictly parallel to z). The resulting geometry is a hyperboloid. You can now intersect the two lines you get from the distance measurement, and their intersection gives information on the position along z. This is a really tricky operation, incidentally.

So, on to my obsession for the last week. CLEO uses two drift chambers: an older one called the DR — technically DR3, it’s the third in its lineage — which, with 47 cylindrical layers, provides most of the track information, and a relatively new one, the ZD, which is inside the DR and has six layers. The outermost layers of the DR are stereo, but to get a good measurement of z near the particle production point (and to get a good lever arm to measure angles), the ZD was designed to be all stereo layers.

The timing measurement is all-critical. Unfortunately, there isn’t a little readout telling us when, exactly, a given collision occurred. Collisions are separated by as little as 14 nanoseconds, and our electronics are nowhere near fast enough to react in that time period. We run all events through a “trigger” system to decide whether to take them or not; the trigger makes its choice quite a bit after the actual collision time, and with 42 nanosecond granularity. 42 nanoseconds is over a millimeter of drift distance — much worse than the intrinsic 0.1 millimeter of the detection method.

How do we deal with this? At first, we take the trigger time as the truth. For every event, we then run a program which tries to find little “tracklets” with the DR information. Knowing the timing structure of the collisions from the accelerator, it tries different possible collision times (on the 14 nanosecond clock) near the time the trigger reported, and finds the one where the hits are most consistent with the tracklets; we then take this as the Event Time. In short, the actual time the event occurred, the zero time from which all the drift distances are computed, is obtained entirely from DR information — and this time is used for the ZD as well.

This is fine unless the ZD gets, for some reason, a different time from the electronics than the DR does. As configured when the ZD was installed, the two detectors get their “record your times now” signals from two different circuit boards, so the possibility for a problem existed: if one of the boards broke subtly, the two might not always be in sync. If the two were always out of sync by a constant amount, we would calibrate it away and never see it, so we are only sensitive to one jittering with respect to the other, an even more subtle effect…

And that’s precisely what we saw (and fixed this morning). A few percent of events from the last data run seemed to have much worse tracking in the ZD than they should have. After days of tracing the problem back (made worse because I had just taken over the procedure that raised the red flag, and so I thought I just didn’t understand what I was doing), we decided the timing system was probably to blame. Physical investigation revealed a broken board which output timing commands which jittered, in a random few percent of events, 42 nanoseconds away from where they should have been, in exactly the manner to produce the observed problem. We replaced the board, and the new one seems to be doing OK, but I think I’m going to be eternally paranoid about this.

In the short term, I’m trying to repair the data that we’ve already taken to remove this effect, which seems quite doable. We’re also going to put monitoring in place to catch this quickly if it happens again. Hopefully I can go back to extracting physics results soon, though.

Many thanks to D.P. and K.E. for teaching me what I know about drift chambers…


One Response to “CLEO Tracking Week In Review”

  1. Michael Says:

    This is an excellent description of how drift chambers work – and also what it really means to be responsible for one.

    CDF have a very good drift chamber, too, a new one built for the higher event rates in Run II. In order to avoid the overlapping of events, the designers selected a thin and fast gas so that the drift velocity is a factor of two or three higher than typical drift velocities, and made the drift cells quite small. It works great! But we did have a scare when, after about a year of good physics running, some of the inner layers started to show clearly degraded performance, which was ultimately traced to contaminated argon. The problem was miraculously cured by injecting a very small amount of oxygen for a short time. The physicists who discovered and solved this problem are certainly heros of CDF.

    Your problem with the timing was certainly very serious, too, and I’m glad that you and your colleagues found a good solution – and that you will monitor this for the future.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s