Large-Area MCP-MPT and its Application at JUNO
Yifang Wang (right) gave a very different talk from the previous presentations. He discussed how they chose the photomultiplier tubes for JUNO.
A few years ago they studied the possible experiments to constrain the mass hierarchy. They understood that they needed a good energy resolution in their detector. Comparing with KamLAND, the largest liquid scintillator available at the time, they were asking for a gain of a factor of 5 in photoelectrons per MeV. This boiled down to a needed increase of a factor of two alone in the PMT collection efficiency and quantum efficiency.
There were a number of proposals for phototubes in 2009. A design with a reflective photocathode was by UC Davis; another idea was to replace dynodes by a Scintillator and APD. Large-area picosecond photodetectors were also proposed. Then in 2008 there was a proposed super-bialkali photocathode with high quantum efficiency. Hamamatsu had flyers claiming that their photocathodes could have the super-bialkali (SBA)-grade quantum efficiency. But there was no product to buy and test. The price was also a unknown, as well as the quantum efficiency that would be possible for very large PMTs.
At the end they came up with a new proposal: instead of a phototube with only a transmissive photocathode, you put a transmissive photocathode at the top and another reflective photocathode layer at the bottom. This could fit the bill. However, nobody knew how to do them this way.
The design started with a 8-inch prototype, moving then to a 20-inch one. The phototube is coupled with a micro-channel plate, whose parameters need to be optimized for the assembly. The photocathode is also very complex, with thin layers of K-Cs-Sb. 20-inch phototubes can only be done by artisanal glass-blowers.
The production of prototypes saw many issues: the gain of the MCPs is not constant. Also, the collection efficiency varies a lot over the incident angle. They had to come up with a new design, which only uses one set of MCP plate, with a new surface treatment, with a collection efficiency highly improved (from 60% to 100%!) by having electron hitting the surface and not the holes have a chance of getting back in the hole all the same. This has worse timing but better collection efficiency.
The performances indicate a 26% quantum efficiency, uniform, a small dark rate, and high efficiency. The down side is the 6 ns spread in the timing resolution. Decided to purchase the phototubes for JUNO based on these performances and risks. Finally they decided to purchase 15k (75%) of the MCP-PMTs from NNVT, and 5k Dynodes (25%) from Hamamatsu.
NNVT started the production line, planning to produce 7k such phototubes per year. Their pilot production is good. In summary, driven by the JUNO requirements, a new type of PMT has been designed. Potentially its performance is better than ones based on dynodes.