Run-7: Sometimes Less can be More
By Angelika Drees
Before the most recent Heavy Ion run of our RHIC accelerator we made – of course – plans and projected our goals for the 3rd running period with Au nuclei at full energy (100 GeV) to date. The last run of this kind took place in the year 2004. In particular, our goals were to increase the number of bunches in RHIC from 45 to 111, double the average luminosity per store from 4x1026 cm-2 s-1 to 8x1026 cm-2 s-1, double the peak luminosity from 15x1026 cm-2 s-1 to 30 and, most importantly, double the amount of integrated luminosity per week from 160 mb-1 to over 300 with a machine availability of 60% or more at store. Our predictions for the most optimistic scenario were 3500 µb-1 total integrated luminosity per low β* experiment, i.e. STAR and PHENIX.
A RHIC run is subdivided into “cool-down” of the superconducting magnets, “setup”, where the current machine configuration is commissioned and “physics”. Physics in turn contains a “ramp-up” period at the beginning, during which the machine performance is optimized by means of frequent machine development times with certain parameters being continuously pushed with the goal in mind to maximize the integrated luminosity. The effective set-up period this year, with both rings at operating temperatures, added up to 2 weeks. This year the physics period was interrupted for a low energy run at about half the typical injection energy. The setup of this special run took less than 24 hours and, at the end, provided colliding beams at a center of mass energy of about 9.18 GeV in two Interaction Regions (IR). You can read about the low energy run in the third story of this edition of RHIC News.
During the 100 GeV setup time, a ramp with a transmission efficiency of 90% and above was developed resulting in a first physics store with 51 bunches per ring and a peak luminosity of 14x1026 cm-1 s-1! After that we set out to increase the number of bunches step by step to the maximum of 111. This maximum was reached after 16 days as can be seen in Fig. 1. Fig. 1 also indicates the setup and ramp-up periods of approximately 2 weeks each as well as the beginning of physics.
At the same time, however, we were pushing for more intensity per bunch, beginning with about 0.9x109. The instantaneous luminosity L is proportional to:
where NB and NY correspond to the average bunch intensity in the Blue and Yellow ring, σx and σy are the horizontal and vertical beam sizes, assuming the two beams have the same size, and k is the number of bunch crossings. Therefore, while the luminosity increases linearly with the number of bunches, k, it increases quadratically with the bunch intensity and any effort that goes into maximizing the bunch intensity is awarded with an even bigger increase of luminosity.
When the bunch intensity approached 1.2x109 while the number of bunches was at its maximum of 111 we encountered two problems: (i) a loss of a varying but significant portion of the bunched beam at the time of rebucketing when the storage RF system with its 196 MHz frequency is turned on (as shown in Fig. 2), and (ii) an increasing loss of beam during the ramp, unevenly distributed over the bunches along the “train”. Both phenomena not only cause a reduction of the achievable luminosity, as can be seen in Fig. 2, they both could and would also lead to a loss of that store or ramp entirely, adding to our downtime rather than uptime.
The losses per bunch along the train during a ramp can be seen in Fig. 3. The top figure shows a ramp attempt (8936) with 103 bunches and 8 evenly distributed empty buckets (“103 standard”). The onset of the losses after bucket number 100, i.e. about a third of the ring, can clearly be seen. This ramp had a total loss of more than 20 109 particles and failed. The same loss distribution was observed with 111 bunches and is typical for an effect caused by electron cloud. The average intensity per bunch was 1.15x109. Right after this failed ramp we refilled immediately with another pattern that got later named “103 fancy gap”. With an average intensity per bunch of 1.17x109 this ramp, 8937, was very similar to the one before. However, the losses during this ramp start much later and drop after the 2nd gap while accumulating to about 5x109 total, a factor 4 less than in the failed ramp. We were able to reproduce this behavior whenever we tried again with 103 or 111 standard patterns. Consequently the resulting collision rate was significantly increased, even with fewer bunches than 111 and was maximized when we started using the 103 fancy gap pattern.
With fewer losses along the ramp, the ramp efficiency could be optimized as well as was the intensity per bunch that we could inject. The ramp efficiency for blue and yellow stores that lasted at least 2 hours increased as we started using less bunches in the 103 fancy gap pattern and reached an average of 95% in blue and 92% in yellow respectively. The resulting average store luminosity is shown in Fig. 4. Only physics stores that lasted more than 2 hours minimum are included. When we started using the 103 fancy gap pattern rather than the 111 standard pattern the luminosity could be maximized and we even reached 14x1026 cm-2 s-1 for a number of stores before the experimental magnet polarity was swapped! This exceeds our goal by 60%.
Thus, although we didn’t quite make our goal of 60% of calendar time at store and achieved only about 47%, we exceeded or met all other goals. This made it possible to deliver more than 90% of our most optimistic predictions for the integrated luminosity for this run as can be seen in Fig. 5. The delivered luminosity for the PHENIX experiment for instance totaled 3250 µb-1, 93% of our prediction. The two experiments, MONOp and LARP, were in large β* areas and STAR, with just 2 colliding bunches in IR6 less than PHENIX, could reach 100% or their goal for this run.