Chips, the Higgs Particle, and Beyond
February 10, 2014, ISSCC, San Francisco—Erik Heijne from CERN talked about how chips pave the road to the Higgs particle and the atto world beyond. Fundamental research on elementary constituents of matter need more sophisticated tools with greater performance and resolution.
Researchers working in high energy physics are taking advantage of microelectronics trends for smaller, faster, and lower power to instrument facilities for investigating higher energy particle physics. The large hadron collider (LHC) has raised the energy levels in order of magnitude above any previous particle accelerator. Data capture instrumentation for the LHC requires half a billion sensor elements which act as giants cameras each capable of 40 million 3-D pictures per second.
Although the LHC covers many kilometers, space and power constraints are very severe. As a result, the researchers are taking advantage of trends electronics that move devices and circuits into the micro-dimensional spaces. To go along with these fine line processes, they also have a great need for design tools and a fab that can produce radiation hardened CMOS devices.
The LHC has been in design since 1984 and required the cooperation and coordination of many governments, countries, researchers, and supplier companies. Some of the earlier work looked into matter and antimatter constituents like quarks and leptons. With the higher energy is available in the LHC, physicists believed they could find the Higgs particle.
The problem with this type of work is very few particles in the intersecting streams actually collide, and of those collisions only a very small number of constituent sub particles are of interest. Most of the experiments differentiate the mass of a particle, and this new particle they were searching for is many many times greater than the mass of a proton.
Quantum physics attributes both wave and particle characteristics to the subatomic particles, and wavelengths are in the atto-meter range. The particles themselves have extremely short lifetimes and these shorter wavelengths need larger images for detection. To get to these small-scale devices requires that the colliding particles have an energy of at least 7 T E-V, and detectors need to find one event in 10-13 total events.
Back in the ’70’s, detectors were bubble chambers and throughput was about one picture per second. The LHC is now taking about 40 million images per second and the positioning of the sensors allows 3-D tracking. These sensors are built with custom radiation hardened devices and data is stored within 2 to 3µs. The data is moved to external FPGAs that provide a stepped acceleration cascade to differentiate and select images.
The sensor arrays are arranged in a hierarchy and the signals of interest are about the same as the particle energies, around 70 electron volts. The particle collisions generate 20k electron-hole pairs which the large diode detectors can detect. The layers are pixel, strip and an electro-magnetic colorimeter. The hadronic colorimeter detects muons and are arranged in a circle of 28k chips in a 80m segment of the collider. The strip detector is 198m2 and has time-matched cables for high-resolution and accuracy.
The detector diodes are integrated with a readout circuit with 128 channels on each chip. The readout includes an analog amplifier that is multiplexed into a 128 sample buffer. The thousands of parallel elements are all matched for gain and delay. The large number of chips and their relatively close spacing requires the circuits be very low power and low noise. The samples are time stamped at 20ns intervals.
Another consideration is that the detectors and other circuitry has to be working in an area with very high energy particles. As a result, all of the circuits are rad hard and single event upset (SEU) tolerant. The SRAM uses a thin oxide and an enclosed n-type transistor. The sensor design changed over time. In ’85, the detectors were CCDs. These detectors changed to analog storage in ’92, and in ’02 the architecture changed again to compute first, then store. The chips are mounted on flex circuit. The pixel detectors are 50 X 400 microns in 0.25 micron CMOS. The acquisition chips demonstrated their performance and accuracy with electrons in early trials.
The LHC generates lots of data, about 1MB every 25ns ( a 40 MHz rate). They move this data to the cloud and in ’00 generated 25 PB per year. Now the data volume exceeds an exabytes a year. All of this data is now producing amazing results. The data analysis indicated a likely Higgs particle, and last summer and these data confirmed the discovery. The Higgs particle is a fermion. The availability and ability to make silicon detectors and readout circuitry in a massive array was essential to the project and the particle discovery.
The applications and communications functions that were integrated into the detectors increased the ability of the sensors to detect and track the billions of particles and the combination of custom circuits and high-performance computing increased the filtering capabilities that enabled the needed system performance.
