The construction of HERA - the Hadron Elektron Ring Anlage - at DESY between 1984 and 1992 provided a working example of the way forward for international and inter-regional collaboration. Germany announced its firm intention to build the new electron-proton collider and invited other countries who wished to do so to join in the construction. In the event a number did by constructing elements in their own countries that formed part of the machine, most notably the construction of 3 km of superconducting magnets by Italy. Other substantial components were contributed by Canada, France, Israel, the Netherlands and the USA. Major expert assistance was supplied by Poland and China, and Czechoslovakia: the DDR and the United Kingdom also supplied expertise. Countries who did not contribute in a major way to building the machine could contribute to the experiments, but subsequently pay certain running costs.
CERN is following a similar model for the Large Hadron Collider. The full project for an energy of 14 TeV was approved by CERN Council in December 1994. Based on the member-state funding foreseen at that time it will operate at reduced energy from 2004, and at full energy from 2008. Non-member states who wish to use the accelerator are invited to contribute in cash or kind in a way that will bring this timetable forward. The invitation is open, but there is an expectation this time that such countries will indeed contribute to the machine.
Japan responded with an immediate gift of 5BYen, perhaps with more to follow, and agreements have been reached, or are well advanced, with Canada, Israel, India and Russia. Negotiations with the USA are underway, with a view to reaching an agreement before the progress of the LHC project is scheduled for review in 1997. The somewhat harsher financial climate of the 1990's is part of the explanation of this stiffening of attitude, together with the perception that it is some years since there was a broad balance in the two-way flow of scientists across the Atlantic, at least within the field. The USA continues to offer entry to Babar on the basis of contributing to detector and associated running costs but not to the capital cost of the accelerator and European scientists are certainly keen to work there, though by no means as many as the Americans who want to use the LHC.
But HERA provides a pointer to the future in another and an important way. Table 1 illustrates this. It shows the time intervals at which beam-beam crossings occur at different storage rings, and the expected number of interactions per crossing (HeraB is a high-rate b-physics experiment being constructed at HERA to run in 1997). Every crossing must be interrogated to see if it contains a useful interaction. One can record only a few events per second. HERA puts us for the first time into a situation where the interval between crossings that must be interrogated is less than the time it takes to make a decision, and in fact less than the time it takes to read the raw data off the detector into a buffer.
| Facility | Year | Crossing Interval | Events/crossing |
| PETRA | 1978 | 3.6 ms | rare |
| LEP | 1989 | 22 ms | rare |
| HERA | 1992 | 96 ns | rare |
| HeraB | 1997 | 96 ns | 4 |
| LHC | 2004/8 | 25 ns | 25 |
Real-time event selection and processing is therefore needed. Unwanted crossings must be rejected, and wanted ones identified. These form only one part in 104 or 107 of the total number of candidate crossings, for a range of relatively common or rare processes. The ``trigger" needs high and known efficiency, and the ability to develop to counteract backgrounds or to accept processes not considered at the original design phase. One sits on a learning curve, first from HERA to HeraB to LHC on the use of ``pipelined logic" and secondly in the early part of the running of the experiment, on developing triggers of increasing sophistication.
Figure 3 indicates the logic structure proposed for the ATLAS experiment for the large hadron collider[10]. The data flows are tremendous. The event sizes are such that the front-end data flow exceeds that of present day television sets in the whole of Europe. Successive trigger levels have more time to think, and allow increasingly complex logical decisions to be made on the data. Level 3 involves full event reconstruction.
Figure 4 shows the `learning curve' for the ZEUS experiment. Data flow restrictions force the first level trigger rate (FLT) to be below a few hundred Hz and the output to tape (TLT) to be below about 10Hz. At the start of running in 1992 the rates for all trigger levels were very similar and independent of the (very low) luminosity (and so were background-dominated). By 1994 the techniques had become much more sophisticated. Rates are far below linear extrapolations with intensity from 1992, whilst a broader range of physics processes is addressed. The lessons for LHC start-up, when a variety of presently foreseen or unforeseen processes will need to be studied that are excluded by simply selecting on the highest transverse energy, are clear. The HeraB experiment will provide the next test of these triggering ideas[11]. From 1997 four events every 96 ns will have to be tested to see if they are the 1 in 105 in which b-quarks are produced.
To handle the energy, the rate, and the complexity of the events the LHC experiments will also be large and complex. Figure 5 illustrates the structure of the CMS experiment [12], and figure 6 shows a typical simulation of an event reconstructed in the ATLAS inner tracker. The technical demands imposed by the LHC on detector technology have stimulated work on detector Research and Development across the whole of the CERN user community, whether on radiation hard semiconductor pixel detectors, high resolution electromagnetic calorimetry, fibre-optic data links, object-oriented programming, cooling and alignment or parallel computing [13]. The `technology-foresight' and industrial links provide a valuable spin-off from the field.