The Large Hadron Collider (LHC) is a 27 km circumference hadron collider, built at CERN to explore the energy frontier of particle physics. Approved in 1994, it was commissioned and began operation for data taking in 2009. The design and construction of the LHC presented many design, engineering and logistical challenges which involved pushing a number of technologies well beyond their level at the time. Since the start-up of the machine, there has been a very successful 3-year run with an impressive amount of data delivered to the LHC experiments. With an increasingly large stored energy in the beam, the operation of the machine itself presented many challenges and some of these will be discussed. Finally, the planning for the next 20 years has been outlined with progressive upgrades of the machine, first to nominal energy, then to progressively higher collision rates. At each stage the technical challenges are illustrated with a few examples.
During the early 1980s, the high-energy physics community was already beginning to look at what would be needed as a future facility at the energy frontier. At CERN, the construction of the Large Electron Positron Collider (LEP) had just been approved and in the USA, the Tevatron had just begun operation with collisions of hadrons at around 2 TeV in the centre of mass (CM). The consensus was that the next machine needed to push into a new energy domain to explore the region between the LEP mass limit and 1 TeV. For this purpose, a hadron–hadron collider with the highest possible primary beam energy and collision rate was considered to be ideally suited. The requirement of high primary beam energy comes from the fact that hadrons are composite particles and the energy is shared between the quarks and gluons. The high collision rate is needed to search for the very rare events that would be associated with new particles. These two factors therefore became the main design challenges for the new machine.
In the USA, plans were drawn up for the superconducting super collider (SSC) . This was designed to be a machine with a CM energy of 40 TeV and would be constructed on a ‘green-field’ site. An alternative design was developed at CERN to make use of the existing infrastructure of the laboratory, including the 27 km LEP tunnel, then under construction . This machine became the Large Hadron Collider (LHC). Eventually, cost issues led to the cancellation of the SSC project and the relatively cheaper LHC machine was approved for construction in the mid-1990s.
2. Design challenges
In a circular accelerator dipole, magnets are used to bend the trajectory of the particles around the circumference of the machine. There is a relatively simple relationship between the energy of the beam (in GeV), the dipole field needed (in tesla) and the bending radius of the ring (in metre) 2.1 In the LHC case, R was fixed at 2804 m by the existing tunnel built for LEP. To reach a beam energy of 7 TeV (allowing 14 TeV CM collisions), a dipole field of 8.33 T would, therefore, be required with a total of 17.6 km of active dipole field. In addition, given the constraints of the existing tunnel cross section, a two-in-one dipole construction would be needed. The collision energy available in LHC represents a factor 7 increase over previous machines.
In a collider, the event rate for producing a particular type of event is the product of the collision rate (or luminosity) and the cross section for that event type. The cross section is expressed in units of area, called barns (1 barn=10−24 cm2). The luminosity, L, is given by the equation (2.2)  2.2 where the beam takes the form of nb bunches each having Nb particles making frev revolutions of the accelerator per second. γ is the relativistic factor and ɛnβ* is related to the beam size at the interaction point. The factor F is related to the geometry of the collision, with a value of 1 for head on collisions and reduces as the crossing angle increases 2.3 Here, θc is the crossing angle, σx is the beam size in the crossing angle plane and σz is the longitudinal bunch length.
From equation (2.2), it can be seen that to achieve high luminosity, one has to make lots of high population bunches collide at high frequency in locations where the beam optics provides as low values of the beam size as possible. Optimization of these parameters is an important part of the design process. The parameter set for the LHC promised a luminosity of 10+34 cm−2 s−1, which represents a factor 100 increase over previous equivalent colliders.
3. Engineering challenges
Realizing a paper design of a large facility like the LHC presents engineering challenges across virtually every engineering discipline. In this paper, I consider only two: magnets and cryogenics. More details on the overall design of the machine as well as the main technical systems can be found in [4–6].
In an electromagnet used in particle accelerators, a current flowing through a set of coils generates the field. In a normal-conducting electromagnet, this field is dominated (and limited) by the magnetization of an iron yoke through which the coils pass. The quality of the field in the beam aperture is determined by the exact geometry of the yoke close to the beam pipe. This type of magnet is limited by saturation of the yoke to a central field of around 2 T . In a superconducting magnet, the field is directly generated by a suitable distribution of the current, properly arranged around the beam aperture. In this case, the field is limited by the current density that can be achieved .
Figure 1 shows the standard cross section of the LHC superconducting dipole magnet. This illustrates the mechanical design with the two-in-one design of the magnets as well as the mechanical assembly of the cold mass and the associated cryostat design. The two-in-one design complicates the magnet structure because the separation of the two beam channels (194 mm) is such that they are coupled both magnetically and mechanically.
The magnetic field in the LHC dipoles represents close to a factor 2 higher strength than magnets used in previous superconducting accelerators. In addition to a high field, the field quality within the beam aperture represents a considerable engineering challenge. The field is determined by the geometrical arrangement of the coils around the aperture. Any deviation in this arrangement will give rise to multipole fields in addition to the simple dipole field. Control of these higher harmonic field components is crucial to the good performance of the accelerator as they can give rise to perturbations and instabilities in the beam. Figure 2 shows the coil distribution around the two beam apertures (a) as well as the resulting field distribution in the cold mass and the beam apertures (b).
Each individual dipole has a magnetic length of 14.4 m when cold, an overall length of 16 m and a weight of approximately 36 000 kg when mounted into its vacuum cryostat. In order to reach the very high current density, the temperature of the magnets must be cooled with superfluid helium (He II) and operated at a temperature of 1.9 K. Below 2.17 K (the lambda point), helium is superfluid with the following characteristics:
— very low viscosity approximately 100 times less than water at normal boiling point;
— very high specific heat capacity approximately 100 000 times higher than the cable conductor in the magnet by unit mass; and
— very high thermal conductivity, approximately 3000 times higher than pure copper.
These characteristics have been fully exploited in the magnet design and in the cooling scheme to enable the dipole field to reach close to 9 T .
The majority of the magnets in the LHC are superconducting and must be cryogenically cooled. In the arcs, the magnet cold mass is immersed in a pressurized bath of superfluid helium at a pressure of 1.3 bar and operated at a temperature of 1.9 K. A few magnets in the straight sections of the machine require lower fields and are therefore operated at 4.5 K. In total, the LHC cryogenic system has to cool approximately 37 000 t of material to 1.9 K. This requires a total helium inventory of approximately 130 t. The production and distribution of such a large amount of He II at pressures above atmospheric pressure is therefore a significant engineering challenge for the system.
The secret lies in the use of a two-stage cooling system . The primary cooling plants comes from eight, 4.5 K warm compressor plants, each capable of producing 18 kW of cooling capacity at 4.5 K. Most of the liquid helium produced is then distributed around the eight cryogenic sectors of the machine and used to fill the magnets in a static pressurized bath. Figure 3 shows the layout of the cryogenic plants around the LHC.
The second cooling stage uses cold compressors installed in each underground location. These produce He II by reduction of the pressure in the circuit to 15 mbar. The low-pressure saturated He II is then transported around the sectors of the machine and used to feed a low-pressure heat exchanger tube, which runs through each magnet. This line is situated at the top of the magnet cold mass in figure 1. The exchange of heat between this line and the main helium bath allows cooling of the whole cold mass to 1.9 K. The arrangement is shown schematically in figure 4.
The resulting cryogenic system is very large and complex. Careful control of the cooling process is needed, especially during cool down and warm up of the sectors to avoid large temperature differences and hence mechanical stress. A complete cool down of a sector from room temperature to 1.9 K can be achieved in around three weeks. In addition to the static losses coming from the heat loss to the environment, the system must cope with dynamic losses from the powering of the magnets and from beam induced heating. Finally, the reliability of the system as a whole must be very high as any breakdown can lead to long recovery times due to the large thermal inertia of the system. In practice reliability in excess of 95% has been achieved during routine operation of the LHC.
4. Logistical challenges
The operation of the LEP machine was terminated in November 2000. The first job was to remove the LEP machine and much of the existing infrastructure in order to prepare the tunnel for the LHC. The LEP dismantling was in itself a huge project with some 35 000 t of material to be removed. Large sections of the tunnel had to be stripped down almost to the bare concrete in order to enable the civil engineering teams to move in. The most critical two-thirds of the tunnel circumference were emptied in a very intense three-month period .
While it is true that the tunnel already existed, in many areas the underground structures needed to be adapted to the needs of the LHC and major civil engineering works undertaken. The major areas included the construction of the ATLAS and CMS caverns, the extraction lines for the beam dumps and the boring of two, 3 km transfer lines between the LHC and the injector machine, the Super Proton Synchrotron (SPS). As an example, figure 5 shows the CMS cavern during construction (a) and completed (b). The existing machine tunnel can be clearly seen in the photographs.
In addition to modifications to the underground structures, a total of 30 new buildings totalling a footprint area of 28 000 m2 had to be built on the surface at various sites around the machine.
Before the installation of the machine could begin, a complete integration study had to be done in order to make sure that everything would fit into the confined space. As well as the machine elements themselves, the cryogenic distribution line (QRL), cables, pipework and ventilation ducts had to be incorporated. In addition, the needs of the installation itself required a certain amount of space to be left free. Figure 6 shows a typical three-dimensional integration of equipment into one of the machine caverns.
(a) Industrial procurement
The main components of the LHC were first developed in house, and then the technology was transferred to industry for the series production. As an example, we will discuss the production of the main dipoles.
The first critical element for the magnets is the superconducting cable. In the case of the main LHC magnets, this was based on relatively mature technology using niobium–titanium (Nb–Ti) as the superconductor. The filaments of Nb–Ti are embedded into a matrix of copper to make the individual strands. With a diameter of 7 μm, around 8000 filaments are used for each 1 mm diameter strand. Either 28 or 36 of these strands are then braided together to form the classical flat ‘Rutherford’ cable. In figure 7, the structure of the individual strands is shown (b) together with the final braided cable (a). A total of 7000 km of cable were manufactured for the LHC.
The assembly of the magnet cold masses was performed by three European firms, based on a recipe specified by CERN . CERN also acted as the main supplier of components. A total of 1232 dipoles plus spares were ordered necessitating major production facilities at the contractors' sites. Rigorous quality assurance procedures were put in place to ensure that the magnets produced met the specifications in terms of geometry, field quality and powering performance. Each complete dipole was delivered to CERN for final assembly into the vacuum cryostat and cold testing. Every magnet was cooled and trained to reach a field of 9 T . As the current in the magnet increases, so do the electromagnetic forces, which can cause minute movements of the pre-stressed coils. The energy liberated by such movements can be enough to quench the magnet. In most cases, two or three quenches were necessary to reach nominal field, however, in some cases it was significantly more.
The original planning of the LHC called for just-in-time delivery of the magnets for immediate testing and installation in the tunnel. In practice, this was hard to achieve. Delays in the civil engineering and in the installation of the infrastructure, especially the QRL meant that large numbers of completed dipoles built up in buffers on the surface. In the end, this backlog of magnets for installation was turned into an advantage as it allowed the sorting of magnets into families where the measured field defects in one magnet could be compensated by the selection of suitable partner magnets in the machine lattice .
The QRL is worth mentioning further, as the difficulties encountered here are a good illustration of the engineering and installation challenges of the large LHC systems. Conceptually, the QRL is rather simple. It contains a series of pipes within a continuous cryostat that deliver the cryogenic fluids at a variety of temperatures to the magnets in the accelerator and collects the warm return gasses for re-cooling. It stretches the length of each LHC arc and amounts to a total length of 20 km. At regular intervals, a jumper module allows connection of the magnets to the QRL. One of the major engineering challenges for this system is the longitudinal thermal contraction of the pipework during cool down to its operating temperature. All parts must therefore be free to move as cool down proceeds while compensators make up for the shrinkage of the pipes. The jumper modules themselves represent a significant additional complexity, as they must remain fixed with respect to the adjacent magnets. They also contain a large number of control valves for optimizing the cooling of the machine elements. Problems during the industrialization of the manufacture of the QRL components as well as difficulties installing, connecting and aligning the elements in the tunnel lead to significant delays in completing this part of the infrastructure.
Only one shaft was made large enough to lower a complete dipole assembly. The dipoles were lowered and then transported in the tunnel to their chosen location. The transport around the tunnel was done with optically guided electrical tractors moving at a maximum speed of 3 km h−1. In total, these tractors travelled approximately 20 000 km to deliver all magnets to their designated locations.
The arcs of the LHC consist of a 3 km continuous cryostat. Each magnet therefore had to be connected electrically, hydraulically and mechanically to its neighbours. As the magnets would be operated in a He II environment all hydraulic and vacuum connections were welded, rather than clamped. The electrical connections were spliced together, soldered and tested to ensure that connection resistance was sufficiently low. In total, around 250 000 high-quality welds and over 10 000 high-current electrical splices were made in the difficult tunnel environment .
5. Operational challenges
On 10 September 2008, everything was ready for first injection; the machine had been cooled down over the previous months and all circuits tested to the level needed for injection of beam. The machine was started with a blaze of publicity and within a few hours circulating beam was established. Over the next 9 days, rapid progress was made in commissioning the various systems with low-intensity beams.
Interleaved with beam commissioning, periods were set aside for commissioning the magnets to high energy. This work had been progressing very well and seven of the eight sectors of the machine had been commissioned to the equivalent field of 5 TeV (6 T in the main dipoles). However, during the commissioning of the last sector, a resistive zone developed in a faulty inter-magnet bus-bar splice. The resulting arc punctured the helium enclosure and a total of 275 MJ of energy were dumped into the joint and the liquid helium causing spectacular damage. The resulting repairs took almost 12 months to complete and involved exchanging a total of 54 magnets and cleaning around 4 km of beam vacuum pipe. An extensive inquiry was held to understand exactly what had happened and to put in place measures to ensure that this kind of event could never happen again . An extensive modification of the magnet protection system was undertaken to allow this system to monitor the inter-magnet connections and passive protection, such as larger pressure relief valves and more robust magnet anchoring to the ground were installed.
This event was a lesson in the energies involved when powering the LHC. As a result of this incident, the protection of the accelerator from the energy in the beam, as well as from its own stored energy, was to become a critical aspect in the operation of the LHC.
(a) Stored energy
When at operating temperature, the magnets of the LHC have essentially zero resistance to electrical current flow. They do, however, have an inductance and this is quite high, at 100 mH per dipole. As all dipoles in a given octant of the machine are connected in series, the inductance (L) of the overall circuit is around 15 H. The energy stored in an inductor is given by E=1/2⋅L⋅I2, where I is the current. This amounts to 1.1 GJ of stored magnetic energy per sector and 10 GJ for the whole machine when operated at the nominal current of 13 000 A. This energy must be controlled very carefully. In case of a powering failure, or other abnormality, it is necessary to extract this energy to ground as quickly as possible in order to avoid damage. In the LHC, the magnet protection systems are designed to extract this energy within approximately 105 s of detecting a fault. The magnets and powering systems themselves are designed to accommodate this time constant.
In the case of the beam, the stored energy is 362 MJ when the LHC is running at nominal energy and beam current. This, however is in the form of a series of bunches moving at essentially the speed of light around the machine. In this case any anomaly must be detected very rapidly and the beam dumped, before the onset of damage. The machine protection system is designed to detect and dump the bean within a few hundred microseconds (three turns of the accelerator) .
The operational mode of an accelerator is by definition one of the most exciting and challenging periods. In this mode, the different elements of the LHC have to act as a unified whole, so that the beam can be injected, accumulated, accelerated and brought into collision. The control system and applications software are crucial elements that allow the operators to run through the different processes and prepare for a data taking run. Beams will only survive the process if all systems are perfectly synchronized. Second-order corrections are carried out by dedicated feedback systems, which measure small deviations from the required conditions and apply corrections. All these systems are, of course, totally dependent on the beam instrumentation system, which must provide reliable observational data on various beam parameters at all times. In addition, all safety systems must be fully validated because stored energy of the beam as well as the magnetic stored energy has huge destructive capability. As an example, one system will be mentioned in more detail. This is the collimation system, which helps to protect the LHC from the stored energy in the beam.
The beam in an accelerator is not a rigid object; instead it is a loose collection of particles, which has a Gaussian distribution in its transverse dimensions. Although the vast majority of the particles are within the core of the beam there will always be some at higher amplitudes in the tails of the distribution. The energy of each particle in the beam is so high that even a very small fraction hitting the surface of the superconducting magnets of the machine can cause a quench, and worse still damage. The collimation system is designed to protect the cold aperture of the machine by removing the outlying particles cleanly and absorbing their energy. Each collimator consists of pairs of jaws, 1.2 m long that can be moved close to the beam. Depending on the amount of energy that is will be deposited in the jaws during normal operation, these can be made of carbon, tungsten or copper. Some act in the horizontal plane, some in the vertical plane and others in the skew directions (at 45° to the horizontal). In total, 108 collimators are presently installed in the LHC [18,19].
In order to function properly, the collimators must be adjusted precisely with respect to the position of the beam, while the jaw gaps themselves are changed during the different parts of the operational cycle of the machine. The collimators closest to the beam are known as the primaries, where initial scattering of the beam particles to be removed occurs. Secondary and tertiary collimators then absorb the showers generated by the initial interaction. To avoid damage a very strict hierarchy of collimator positions must be observed and the position of each collimator must be maintained accurate to around 10 μm with respect to the centre of the beam. The set-up and qualification of the collimation system is established during special low-intensity runs and then operation relies on excellent reproducibility of the beam conditions from run to run, and on sophisticated beam feedback and control systems.
Run 1 of the LHC lasted 3 years seeing the machine advance from initial commissioning then through a series of intensity increases to reach a peak instantaneous luminosity of 7×10+33 cm−2 s−1. Figure 8 shows the integrated luminosity evolution over the 3 years.
This remarkable performance was not only due to the quality of the design of the machine and its sub-systems but also due to the reliability of the equipment making up the various sub-systems. Many of these sub-systems are huge installations in their own right. As an example the operational efficiency of the LHC during the proton run is given in figure 9. The time is split into six modes as follows.
— ‘Set-up’ corresponds to the preparatory period where the machine is being made ready for taking beam.
— During ‘Injection’, bunches are progressively accumulated in the LHC by transfer from the injector complex. During 2012, up to 144 bunches could be transferred in a single injection.
— ‘Ramp’ corresponds to the period where the energy of the beam is progressively increased from the injection level (0.45 TeV) to the collision energy. During Run 1, this energy was limited to 3.5 or 4 TeV per beam.
— ‘Squeeze’ corresponds to the period where the beams are adjusted prior to collisions. This consists of adjusting the focusing of the beams around the interaction points to minimize the beam size and hence maximize the luminosity.
— ‘Physics’ is the period during which the beams collide in a stable way, the experimental detectors are turned on and are gathering collision data. The luminosity delivered to the experiments during this period is integrated and over many physics fills forms the plots shown in figure 8.
A figure of 36.5% for the fraction of the scheduled time spent with stable colliding beams is quite remarkable for a machine of the size and complexity of the LHC.
6. Future challenges
The LHC is presently undergoing major works during a long shutdown, lasting almost 2 years. This shutdown is to prepare all systems for operation at nominal energy and involves a number of measures to consolidate the machine in the wake of the 2009 incident. In addition, after 3 years of almost continuous operation the maintenance and upgrade of many technical systems is needed. At the same time, the injector complex for the LHC is being upgraded and maintained ready to deliver the high-intensity, high-quality beams needed to increase the performance of the LHC further.
The LHC will begin re-commissioning with beam early in 2015 and will then run until the middle of 2018, with only short technical stops over the new-year period. During this ‘Run 2’, the operation of the machine will be established at 6.5–7.0 TeV per beam and the performance pushed beyond the nominal design luminosity of 1×10+34 cm−2 s−1. As well as almost doubling the energy of collisions, the number of bunches circulating in the machine will also be increased by a factor 2. This will be an exciting period as the machine will again move into unknown territory and systematic searches for physics beyond the standard model will be high on the priority list. In addition, the increased performance will provide significantly more statistics on the Higgs boson discovered in 2012, allowing its nature to be investigated in more detail.
In 2018–2019, a second long shutdown is planned. Lasting around 18 months, this will allow major upgrades to the LHC detectors as well as the machine elements. The aim here is to increase the performance of the machine again to around three times the nominal design value. By the end of 2022, it is forecast that the LHC will have delivered approximately 300 fb−1 to CMS and ATLAS. This value represents the initial design goal for the LHC.
In 2013, the update of the European Strategy for Particle physics was published . This process, undertaken every 5 years brings together the community and funding agencies to set out the major strategic goals for the discipline. In the recent update, the highest priority was established to maximize the potential of the LHC by launching an upgrade programme for the machine and detectors in order to collect a factor of 10 more data by the early 2030s . The ambitious programme to deliver 3000 fb−1 to ATLAS and CMS is now launched as a major international project called HL-LHC (high-luminosity LHC).
The HL-LHC aims to maximize the luminosity production of the LHC by re-optimizing the parameters of equations (2.2) and (2.3). The key parameters here are the brightness of the beam from the injectors (defined as Nb/ɛn), reduction of the beam size at the IP, via lower β* and a mechanism for controlling the luminosity, via the factor F (equation (2.3)). The higher brightness beam concerns upgrades to the injector complex, the other two refer to upgrades in the LHC itself. A full description of the upgrade plans for the LHC is outside the scope of this paper. However, a couple of examples will be given here of the technical challenges concerning this luminosity upgrade.
The modifications to the LHC for HL-LHC will involve essentially re-building around 1.2 km of the machine with new equipment. The major elements will be new final focus quadrupoles around CMS and ATLAS. These wide-aperture quadrupoles will allow the β* (and hence the beam size at the IP) to be reduced by a factor of 5. The technology of these new high-field magnet quadrupoles has to be pushed considerably and their development and construction will form part of an international collaboration . The cable for these magnets must be constructed from Nb3Sn, which is a superconductor capable of generating the higher magnetic fields required in this configuration. Intensive R&D is taking place at CERN, in the USA and in Japan to develop them. Unfortunately, the full benefit of the beam size reduction from the potentially lower β* will not be directly available in the form of increased luminosity, as the reduction in beam size at the IP will also reduce the factor F, in equations (2.2) and (2.3). In order to overcome this and to provide a mechanism for levelling the luminosity (i.e. keeping it constant through the greater part of a physics coast), a second technical innovation is required, namely crab cavities.
Crab cavities, installed at either side of the IP, can be used to kick the head and tail of each bunch in such a way that they effectively cross head-on, in spite of the crossing angle. A schematic is shown in figure 10.
Crab cavities are a relatively new technology that has so far only been experimentally verified in the KEKB accelerator in Japan, and then only with leptons . In addition, for the LHC case compact cavities must be developed that will fit within the constraints of the machine, most notably the 194 mm distance between the two beams. Figure 11 shows schematic drawings of several crab cavity designs being developed by international partners in the USA, UK and Japan. Each superconducting cavity type will be tested at CERN before a final decision on the installation in the LHC can be taken.
After 25 years of design and development, the LHC has been commissioned at intermediate energy and during its first 3-year run has performed well and beyond expectations. Already, at this early stage, it has delivered large amounts of collision data to the experiments and allowed them to discover and observe the properties of a Higgs boson at 126 GeV. A 2-year shutdown of the machine is now underway with the aim of preparing the machine for ‘nominal’ operation in terms of energy and luminosity. LHC will begin operation in 2015 at 6.5–7 TeV per beam and run for 3.5 years before the next long shutdown. This period will be very exciting in terms of the possible observation of physics beyond the standard model. To follow from this, an upgrade plan has been elaborated and launched to provide a factor of 10 more data than was foreseen in the original design. The larger statistics will allow more precise investigations of the Higgs boson as well as searching for very rare new processes.
Throughout the various stages, the LHC has had to confront a number of technical, engineering and operational challenges. A few of these have been mentioned here to give a flavour of the complexity of the machine and its detectors.
I have profited enormously with material and advice from numerous colleagues at CERN without which this paper would not have been possible. I sincerely thank them.
One contribution of 12 to a Discussion Meeting Issue ‘Before, behind and beyond the discovery of the Higgs boson’.
- © 2014 The Author(s) Published by the Royal Society. All rights reserved.