Search for Dirac monopoles with the CDF II detector

Nicholas Cothard, Daniel Goldin, Jonathan Lewis, and Homer Wolfe


This analysis builds on the previous CDF result by using the same dedicated trigger, the same analysis technique, and much of the same offline software to search for Dirac monopoles with mass less than 800 GeV/c2. The analyzed dataset represents more than a thirty-fold increase in the statistical power of the preceding CDF result.

Because of the monopole's unique signature, it is possible to eliminate background effectively while maintaining high efficiency. Unlike ordinary charge particles, the magnetic monopole is not accelerated in the plane perpendicular to CDF's solenodial magnetic field. Instead, monopoles are accelerated in the direction of magnetic field causing a relativistically stretched parabolic trajectory. The dedicated monopole trigger requires large light pulses in CDF's Time-of-Flight (TOF) detector caused by the monopole's ionization effects. We then search the data for a signature with no curvature in r-φ and abnormally high ionization in the COT drift chamber. The trigger was recalibrated in 2010.

We have searched 1.21 fb-1 of CDF Run 2 data (Jan. 12, 2011 - Sep. 29, 2011) for evidence of magnetic monopoles.

Complete details are in CDF Note 11102.

Online Selection

Fiducial Acceptance

We modified GEANT 3 to propagate monopoles along a parabolic path in a magnetic field with the correct energy loss and hard ionization (delta rays). We assume that every monopole will produce enough light in the TOF to satisfy the trigger, and we created a simple TOF trigger simulation to account for other effects.

We impose a cut of β > 0.15 (where β = v/c ) to insure that enough light is collected for the trigger. [J. Derkaoui, et al., Astroparticle Physics 10, 339 (1999)]

We plot the acceptance for monopole masses of 100 to 800 GeV. Link to acceptance eps file

The upper curve shows the geometric and kinematic acceptance only without the effects of the trigger. The band on the upper curve represents the systematic uncertainty resulting from the material model. The lower curve includes the effects of the trigger that are modeled in the simulation. The band on the lower curve includes the systematic uncertainty on the trigger effects.


The monopole trigger is based on large pulses from the PMTs attached to the TOF scintillator bars. It is possible that a particle may reach the TOF bar earlier than a monopole and complete the charge integration window before sufficient monopole charge can be collected. Therefore, we expect an inefficiency due to pileup.

Following the previous analysis, we chose to examine Z → e+e- events since we expect the events to be similar to monopole pair production. We use the z position of the electron when it reaches the TOF and the known speed of light in the bar to predict the propagation time of light from the electron to the PMTs on each end of the TOF bar. We subtract this from the measured time in the PMTs to get the corrected electron arrival time as predicted by each PMT. The difference between the corrected electron arrival times of the east and west PMTs will tell us the amount of time between the arrival of each particle. We can use the difference to determine if a monopole would be lost because of a particle arriving early and completing the PMT charge integration window before the monopole signal can be collected.

We plot e+e- mass in the electron sample. Link to Z mass plot eps and gif

We plot difference of east and west PMT predicted arrival times. Link to east-west time difference eps

We fit a double Gaussian to the distribution and take the fraction of events in the wide Gaussian to be the pileup fraction.

To determine the systematic uncertainty on this, we calculated the ratio of the integral of the tail distribution (Δt > 2ns) to the integral of the entire distribution. The difference between this ratio and the ratio computed by fitting the gaussians is our systematic.

The inefficiency due to pileup was found to be 35.6 ± 9.4%.

To examine the effect of luminosity on the pileup, we divide our luminosity range into five bins each with an equal amount of integrated luminosity. We then find the pileup inefficiency for each bin and plot it against the average value of the luminosity in that bin. We plot the pileup inefficiency as a fraction of instantaneous luminosity. Link to pileup vs luminosity eps

The lowest point corresponds to the previous analysis where the pileup inefficiency was much lower.

Offline event reconstruction and selection


We divide the r-φ plane into 100 equal regions of azimuth and search for straight paths for highly ionizing hits, using the outer 6 superlayers of the COT. A segment is formed when at least 8 out of 12 layers in a superlayer have highly ionized hits along a line. We make a cut on the effective curvature of segments based on the angle with respect to a radius. A monopole candidate track is formed if at least 5 of the 6 superlayers has segments in one azimuthal region or a combination of two neighboring azimuthal regions.

Background and width cut

We characterize the monopole background by examining jet data. We expect the monopole luminosity distribution to look similar to that of the electron events used in the pileup study described above while the jet-trigger sample is of lower purity, especially at high luminosity.

We plot the instantaneous luminosity profile of the electron and jet candidates for the monopole run list. Link to luminosity profiles eps

We vary the segment curvature and COT hit-width cuts while running the monopole track finding code over CDF jet data. To correct for the high trigger fake rate at high luminosity in the jet sample,Using the electron data, we weight the fake monopole candidates by the electron sample's instantaneous luminosity.

We plot the weighted fake candidates per event for various curvature and width cuts. Link to the fake candidates eps

Extrapolating to the same cuts as the previous analysis (curvature Cseg<0.001cm-1 and width w>140ns), we expect to see 3.4 ± 1.6 x 10-9 fake candidates per event. With 1.2 x 107 events in the monopole dataset, we expect to see 0.040 ± 0.019 events due to false positives.

Monte Carlo

We plot the tracking efficiency for monopole masses of 100 to 800 GeV. Link to tracking efficiency eps

The error bars are the statistical error on the simulation samples. At a monopole mass of 500 GeV/c2, the tracking efficiency is 0.985 + 0.005 - 0.016.

Limit calculation and results

We plot our cross section limit. Link to limit plot eps

This is the cross section limit at 95% CL versus magnetic monopole mass, for an n=1 monopole. The excluded region is above. The dashed curve shows the predicted cross section for Drell-Yan production. The arrow indicated the mass limit M > 476 GeV/c2.

We plot the ratio of our data versus prediction and the compare it to that of ATLAS (Link to ATLAS paper). We are more sensitive for masses less than about 300 GeV. Link to data/prediction ratio compared to ATLAS eps