The European Spallation Source Design

The ESS linac requires strict loss control to stay below the 1 W m⁻¹ of loss limit [20], and this sets a strong limit on the amount of halo produced in the linac. There are already several mathematical definitions of the beam halo, e.g. [21, 22], and even more publications on the processes in which halo is generated, e.g. [23, 24]. The main sources of halo generation are mismatches, high space charge and tune depression, nonlinear fields, and small longitudinal acceptance which results in particles falling off the accelerating bucket. Such particles will either be left unaccelerated or accelerated to a different energy, in either case resulting in transverse oscillations and eventual loss.

To avoid producing particles that will populate the beam halo this well-known beam physics rules are applied to the ESS linac design:

• The transverse phase advance per period figure 7 is limited to 87° to reduce the percentage of the beam that, due to their phase, would have a phase advance exceeding 90° per period.
• The phase advance per metre (average phase advance) variation should be smooth and continuous. For the ESS design the average phase advance variation is chosen to be monotonic, figure 8.
• The tune depression, $\eta ={\sigma }_{{\rm{sc}}}/{\sigma }_{0}ok$, must stay above 0.4 in all the planes during acceleration, see figure 9. This limits the number of mismatch resonances to only two (figure 10) from which one is always present irrespective of the tune depression. In order to avoid the second resonance the tune depression should be kept above ∼0.4, [25].

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

On top of these, both the transverse and longitudinal apertures should be properly chosen, a too large aperture will increase the cost and a too tight aperture will increase the risk of losses. To avoid particles escaping the accelerating bucket the synchronous phase is kept below $-15^\circ$ at all times, with bigger values at lower energies, figure 11.

Zoom In Zoom Out Reset image size

The superconducting section of the ESS linac is designed to have equal tune-depressions in the three planes [26]. An equipartition design would discriminate one or several planes in favour of the other ones if the emittance ratios are not equal to 1, see figure 12. As the emittance ratios at the ESS are $\sim 30 \%$ higher in the longitudinal plane compared to transverse, equipartitioning increases halo growth in at least one plane.

Zoom In Zoom Out Reset image size

Though every section of the accelerator is designed to match the average phase advance of neighbouring structures, a further phase advance smoothness matching is performed during the integration to assure a premium beam quality. As the linac is designed in a collaboration, the design of the linac has also been influenced by the focus on improving the integrity of the accelerator as a single structure and improving the real estate gradient. Matching between the structures is done by using a pair of quadrupoles, and a maximum of four cavities in each side of intersection. In order to have a current-independent linac, all matchings are done by smoothing the transverse and longitudinal phase advance, not just adjusting the Twiss parameters.

The longitudinal aperture and acceptance are as important as in the transverse plane as any longitudinal losses result in increased longitudinal halo that will cause a loss in transverse plane somewhere downstream in the linac. The longitudinal aperture is defined by the synchronous phase, ${\phi }_{s}$, and the amplitude of the field in the cavity, E₀ [27]. The phase acceptance is $3\times {\phi }_{s}$ and energy acceptance is:

$$\begin{eqnarray}&&w=\pm \sqrt{\displaystyle \frac{2{{qE}}_{0}T{\beta }_{s}^{3}{\gamma }_{s}^{3}\lambda }{\pi {m}_{0}{c}^{2}}({\phi }_{s}\cos {\phi }_{s}-\sin {\phi }_{s})}.\end{eqnarray} \tag{ 1 }$$

There is a frequency doubling at the transition from spoke to medium β in the ESS linac However, such a change in the frequency causes an abrupt change in the average focusing forces in the longitudinal plane at the frequency jump according to equation (2), which if not handled properly could degrade beam quality and increase losses.

$$\begin{eqnarray}&&{k}_{l0}^{2}=\displaystyle \frac{2\pi {{qE}}_{0}T\sin (-{\phi }_{s})}{{{mc}}^{2}{\beta }_{s}^{3}{\gamma }_{s}^{3}\lambda }=\displaystyle \frac{2{f}_{{rf}}\pi {qT}\sin {\phi }_{s}}{{{mc}}^{3}{\beta }_{s}^{3}{\gamma }_{s}^{3}}.\end{eqnarray} \tag{ 2 }$$

There are studies [28, 29] proposing solutions to avoid these unwanted effects. One of the drawbacks of these methods is that they find solutions by only adjusting the structure which comes downstream of the frequency jump.

At the ESS linac the average phase advance in the spoke section is increased by lowering the synchronous phase gradually. This keeps the phase advance per metre equal to the phase advance per metre at the entrance of the medium-β section. Downstream of the frequency jump also the phases are bigger than what it would have been without the frequency jump, and the accelerating gradient is lowered to limit the phase advance per period to less than 90°.

This section discusses the design criteria of the transport lines in the ESS linac.

LEBT: The LEBT of the ESS linac consists of two solenoids, with a 330 mm length, 50 mm aperture radius, and ∼25 T maximum strength. Its main function is to focus the diverging beam from the ion source to the small aperture of the RFQ. It is estimated that the current out of the ion source can take up to ∼3 ms to be stabilised. Thus, including the required 2.86 ms, the length of the pulse out of the ion source can be up to ∼6 ms. Between the two solenoids, the LEBT houses an electric deflector (slow chopper), with a rise and fall time on the order of 100 ns. The slow chopper removes this initial transient part of ∼3 ms, as well as the short (on the order of 100 μs) transient part in the tail of the pulse, by deflecting this part of the pulse on the conical shape dump mounted on the entrance wall of the RFQ. The other device housed between the two solenoids is the iris with six blades. Taking into account potential losses within the LEBT and RFQ, the ion source is designed to produce a maximum proton current of ∼74 mA.

A MDIS generates a proton pulse up to 6 ms long with a maximum flattop of ∼3ms [30]. The proton intensity (H₁⁺) exceeds 74 mA after the extraction electrodes at an energy of 75 keV.

The source is followed by the LEBT [31], which is composed of two magnetic solenoids (330 mm in length, 50 mm beam aperture radius and a maximum field of ∼0.25 T). The LEBT matches the beam to the RFQ. It incorporates a chopper system that removes low quality head and tail of the beam, a multi-blade iris to reduce the current and a suite of beam diagnostics that measure the beam properties. The gas pressure, distribution and composition in the LEBT is adjusted to compensate 95% of the space charge, this compensation is built within $20\,\mu {\rm{s}}$ when the proton beam passes through the LEBT. The ion source and LEBT is designed and built by the INFN-LNS in Catania, Italy.

MEBT: The medium energy beam transport (MEBT) system [32] measures, cleans and matches the beam out of the RFQ structure and transports it to the DTL. The MEBT is equipped with a suite of beam diagnostics to measure the current, transverse and longitudinal properties of the beam and provides means to collimate the beam in transverse using a 3 sets of adjustable collimating blades [33]. The pulse has a head with $20\,\mu {\rm{s}}$ uncompensated space charge (and therefore mismatched), a fast stripline chopper device (with a repetition rate of up to 200 kHz) cleans this head and provides short pulses needed for commissioning (in combination with the LEBT iris that creates lower current beams). The MEBT has 11 quadrupoles (80 mm magnetic length, 18.4 mm beam aperture, and ∼2.6 T maximum integrated gradient) equipped with trajectory correcting coils (Maximum integrated field 2.8 mT m) and three copper plated buncher cavities (14.5 mm beam aperture with maximum voltage of 150 kV and maximum power of 22.5 kW) resonating at the beam frequency. The MEBT is designed and built by the ESS-Bilbao, in Bilbao, Spain.

DTL: The DTL accelerates the proton beam energy to 89.6 MeV in five independently powered tanks [34, 35]. Each tank is powered by a 2.8 MW klystron, keeping 30% margin for LLRF, tuning and waveguide losses. 2.2 MW of power is used to excite the cavity and accelerate the beam. At full beam current $\sim 50 \%$ of this power is transferred to the beam and the rest is lost as ohmic losses. Higher energy at the DTL entrance results in longer low-energy cells with several positive consequences: longer cells can house bigger quadrupoles reducing the magnetic gradient for the same integrated gradient, longer cells and gaps reduce the magnetic field and the electric field, respectively and higher energy enhances the effective shunt impedance (ZTT). The transverse focusing is provided by Halbach permanent magnet quadrupoles (PMQs), with $45\,\mbox{--}80\,\,\,\mathrm{mm}$ in length, beam aperture of $10\,\mbox{--}\,12\,\mathrm{mm}$ (Tank1–5) and a maximum integrated gradient of ∼3.1 T. These are housed in every other drift tube. The constraints present in a DTL require an optimisation process to decide where to put the corrector dipoles and BPMs and how to minimise its quantity. The DTL is designed and built by the INFN-LNL in Legnaro, Italy.

Spoke: The rest of the acceleration in the ESS linac is provided by superconducting cavities. A low energy differential pumping section (LEDP), separates the DTL from the Spokes. Double spoke cavities [36] are used to accelerate the beam from 89.6 to 221 MeV. The choice of spoke cavities enhances the availability and reliability of the linac These cavities can be re-tuned for different beam energies. Such a flexibility permits operation of the spoke section while one or more cavities [37] are not operational. Spoke cavities have a larger transverse aperture compared to conventional normal conducting structures. The 352.21 MHz spoke cavities (maximum gradient of 9 MV m⁻¹ and beam aperture of 28 mm) with an optimum β of 0.50 are housed in pairs in 13 cryomodules [38] and are separated by LWUs. Every LWU is composed of a pair of DC quadrupoles [39] (magnetic length 250 mm, beam aperture 30 mm, and maximum integrated gradient 1.9 T), a dual plane corrector (maximum integrated field 1.2 mT m) and a BPM plus a central slot allocated to beam diagnostics. Depending on the required diagnostics per LWU, there are different LWU across the linac [40]. The spoke cavities and cryomodules are designed and built by the IPN in Orsay, France, and all the corrector magnets and quadrupoles downstream of the DTL by Elettra in Trieste, Italy.

Ellipticals: The RF frequency doubles to 704.42 MHz at the beginning of the following structure in the linac, the medium-β elliptical cavities. There are two families of elliptical cavities [41, 42], that accelerate the beam from the spoke output energy to 571 MeV using 36 medium-β cavities (maximum gradient of 16.7 MV m⁻¹ and beam aperture of 50 mm) and further to 2.0 GeV by 84 high-β cavities (maximum gradient 19.95 MV m⁻¹ and beam aperture ∼68 mm). In both sections, four cavities are housed in cryomodules of identical length [43, 44]. Having different geometric β of 0.67 and 0.86 respectively, the medium-β cavities are given an extra cell (6-cell) with respect to high-β cavities (5-cell) to make the lengths of the two cavity types as close as possible. There are identical LWUs for both elliptical sections. Thus an equal period length in the medium and high-β cryomodules, making them swap-able in case the required gradient in medium-β is not achieved. The elliptical LWU have the same functionality as spoke LWU, with stronger quadrupoles (maximum integrated gradient 2.3T) bigger beam apertures (50 mm), and longer magnetic length (350 mm) plus stronger dual-plane magnetic correctors (integrated field 2.4 mT m). To have the same flexibility at the spoke to medium-β transition the period lengths in elliptical section is chosen to be exactly twice the one of the spoke section. The cryomodules are designed and built by IPN in Orsay and CEA in Saclay, France, the medium-β cavities are designed by CEA Saclay and INFN-LASA in Milan, Italy, finally, the high-β cavities are provided by CEA Saclay and Daresbury Lab in UK.

HEBT: For contingency purposes, the same periodicity, in transverse plane, is maintained for 15 periods after the high-β section in the HEBT. After this contingency area, there is one more LWU which is followed by a vertical dipole with a bending angle of 4° that also works as a switch magnet between the beam dump and the target. In the path to target, after 6 periods of longer doublet focused sections that are adjusted to create an achromat dogleg, the beam is bent to the horizontal plane using a second vertical dipole. This beam is transported to the target using a set of 6 quadrupoles and 8 raster magnets that paint the target surface in horizontal and vertical directions at different frequencies [45]. To reduce the beam centre movement on target due to energy jitter, the phase advance between the second dipole and the target surface is set to be a multiple of 180°.

Dumpline: When the beam is directed to the 12.5 kW beam dump (by switching off the first dipole) the beam is magnified on to the dump entrance face using three quadrupoles. This line is also equipped with an non-invasive profile monitor (NPM), close to the dump to assure that the beam is not focused on the dump surface.

Raster system: The beam density on the surface of the target is desired to be as uniform as possible to achieve the lowest peak density. One can either use a nonlinear magnetic system to achieve this goal, which is very distribution dependent, or paint the area with the beam using dynamic dipolar fields. The ESS HEBT uses a fast horizontal–vertical sweeping dubbed raster system [46]. The raster system is composed of four dithering dipole magnets per plane which sweeps the beam on the target surface within the 2.86 ms pulse, shaping a rectangle of 180 × 60 mm² (H × V) with almost a uniform density within the footprint that drops to zero at the same rate as the initial beam out of the linac does [45]. These magnets are driven by two triangular waveform currents with first harmonic frequencies of up to 40 kHz and up to fifth harmonics. The raster magnets are designed and built by the Aarhus University.

In order to define the tolerances on the active components, large statistical studies with multiple linacs are simulated, each with distributed errors on alignment and field vectors of quadrupoles and cavities. These studies are broken in two steps, initially to investigate the effect of tolerances on each variable, and secondly applying all the errors (normalised to have comparable effects) to simulate the effect. These studies are performed using ${10}^{5}\,\mbox{--}\,{10}^{6}$ macro particles, with ${10}^{3}\,\mbox{--}\,{10}^{4}$ linacs, depending on the purpose of the study. On each of the produced linacs the error on the variable under study has a uniform distribution limited to maximum error × step of error. In the second set of studies dedicated to simulate the effect and the losses, all the errors are applied: alignment errors with a uniform distribution and jitters with a Gaussian distribution.

As the first step for starting the statistical error studies, it is needed to define a strategy on how to proceed and where to set the limits or, in other words, how large errors to accept. This process can be re-iterated as some of the resulted tolerances, which are not very demanding be tightened in favour of those that are more costly to achieve or maintain. For this first step, the limit on losses is set at a 99% confidence level to be within the 1 W m⁻¹ limit. There is a limit on the emittance growth due to errors too. An emittance two times larger than the nominal case without errors is set as the limit for the final emittance in the presence of errors. As explained above, the ESS linac consists of five accelerating structures, RFQ, DTL, Spoke, medium-β, and high-β, plus two transport lines, MEBT, and HEBT. These sections have different lengths, but it takes almost the same time for the beam to travel through each section. Therefore, the additional emittance growth per section due to errors has been limited to $\sqrt[7]{2}-1$, or 10% per section in each plane (for 99% of the cases), setting a second limit for defining tolerances.

Having set a strategy in place to define the static tolerances, it is also important to set the limit on dynamic errors and jitters. A dynamic error includes all the errors which cannot be corrected or happen in between different linac tunings. To make sure that the dynamic errors do not cause unwanted effects, the criteria to set them is to have an effect which is less than $1\times$ rms spread of the beam in the plane of study. The applied errors on the quadrupoles are the alignment and gradient. The cavities have a rotational symmetry, hence alignment errors did not include the roll angle, accelerating field and synchronous phase errors.

To study the transverse errors, a functional beam trajectory correction system is put in place. This has been done by optimising the positions of the steerers and their corresponding BPMs, specially in the DTL, where the number of choices is limited [52]. The main objective in the other parts of the linacis to reduce the number of active BPM, without loosing the control over the beam trajectory. Enabling the correction system, with the right limit on the steerer strengths, the transverse tolerances can be set.

The errors investigated in this study belong to one of the following types: machining, welding, positioning, powering, and input beam errors. The errors are uniformly distributed around the defined tolerance.

RFQ: To set the tolerances on the beam parameters at the RFQ entrance, the effects of the position, angle, and Twiss parameters of the beam are studied. The machining errors in the RFQ include errors in the longitudinal and transverse profile of the vanes. The welding errors investigated are the vane welding errors that can be vertical and horizontal shift of the vertical vanes and the effect on the horizontal vanes. These errors can have a different value on the two ends of the vane, still within the limit, which causes a tilt. The RFQ is being built in several sections, and the alignment of these units with respect to each other (shifts and tilts) are also included. The last error that has been covered in the RFQ is the voltage jitter. Tolerances of the RFQ are listed in table 6. The parameters of the beam, their average, and extents are kept at the end of the RFQ. The result of errors on the final beam out of the RFQ is presented in table 7. Finally, the tolerances along the whole linac are presented in table 8.

Table 7.  Beam error included in the error studies at the MEBT input (max values).

	dx/dy	dx/dy'	dϕ	dE	d	Mismatch	dI
	mm	mrad	deg	keV	%		mA
Beam	0.3	1	0	36.2	5	5	0.625

MEBT, SCL, and HEBT: In the MEBT, HEBT and the superconducting linac (spoke, medium-β and high-β sections), the errors applied are the following:

Table 8.  Tolerances for the ESS linac (max values).

	${{\rm{\Delta }}}_{{xy}}$	${\phi }_{{xy}}$	${\phi }_{z}$	${\rm{\Delta }}G$	${\rm{\Delta }}{\phi }_{{\rm{rf}}}$	${{\rm{\Delta }}}_{{\rm{rf}}}$
	mm	deg	deg	%	deg	%
			mebt
Quad	0.2	—	0.06	0.5	—	—
Cav (S*)	0.5	0.115	—	—	1	1
Cav (D)	—	—	—	—	0.2	0.2
			dtl
Quad	0.15	0.5	0.24	0.9	—	—
Cell (S)	—	—	—	—	0.5	1
Tank (S)	—	—	—	—	1	1
Tank (D)	—	—	—	—	0.1	0.1
			scl
Quad	0.2	—	0.06	0.5	—	—
Cav (S)	1.5	0.129	—	—	1	1
Cav (D)	—	—	—	—	0.1	0.1
			hebt
Quad	0.2	—	0.06	0.5	—	—
Dipole	0.2	—	0.06	0.05	—	—

Note. *S: Static, D: Dynamic.

Quadrupoles: alignment of the quadrupoles including their transverse position, rotations around the beam axis, and gradient errors.

Cavities: alignment of the cavities including their transverse position, rotations perpendicular to the beam axis, static and dynamic accelerating field amplitude and RF phase errors.

Dipoles: alignment of the dipoles including their transverse position, rotations perpendicular to the beam axis, and field error.

Input beam: variations of the beam parameters at the end of the RFQ, which represent the beam errors at the input of the MEBT, including the position and angle errors, emittances, mismatch in Twiss parameters, energy jitter and current variations. The nominal distribution out of the RFQ is used as input distribution.

DTL: The DTL uses PMQs for transverse focusing. Each tank is fed by one klystron and therefore all the errors caused at that level are coupled. Synchronous phase and field amplitude in each cell is determined by the shape and position of drift tubes and are mostly random. The applied errors here are:

PMQ: the same type of errors as powered quadrupoles, with different amplitudes, are applied here.

RF: uncoupled field and phase errors are applied to RF cells. These errors are due to machining or installation errors and are static in time. Coupled field and phase errors are applied to all cells of each tank. These errors are due to RF set-points and are both static and dynamic, with different amplitudes.

In the presence of these errors the beam behaves differently than with nominal conditions: mismatches in both planes due to different focal lengths of quadrupoles and cavities, beam centre jitter due to misalignments of such elements and halo growth caused both by mismatch and by longitudinal tails. The positions of the tracked particles are stored during error studies and different post processing methods are performed to get, the losses, confidence levels, or the proximity of the halo to apertures [53].

For the ESS lattice, it is important to have an excellent knowledge of the expected loss levels in the machine. The accelerator will deliver 5 MW proton beam power, while the losses are required to stay below 1 W m⁻¹ for the entire linac. That means that the relative amount of losses that can be accepted are on the order of 10⁻⁴–10⁻⁷ per metre, depending on the beam energy where the losses happen.

Losses are seldom distributed evenly. Certain weaknesses in the lattice might produce more losses and the misalignments are more unfavourable in certain regions. It is somewhat easier to predict loss distributions from a lattice than it is to predict their quantitative levels.

The study shown here is done for 20 000 machines with 600 000 macro-particles in each machine. For each machine the nominal tolerances for static and dynamic imperfections defined above have been applied [53, 54]. In order to provoke more losses for the beam loss and activation studies, a dedicated study is performed with doubled dynamic RF amplitude and phase jitters. In the simulation the loss levels are quoted in arbitrary units, as a normalisation would represent losses that are significantly higher than what is actually expected in the machine. The increased dynamic error used does not change the distribution of losses significantly, which makes this a useful way to increase statistics. Simulations are done from the exit of the RFQ until the beam reaches the target.

Figure 17 shows the losses along the linac, where the count is equal to the number of particles lost (i.e. no weight on the energy of the lost particles). There is a dense distribution of losses in the warm linac, a very clean spoke region and then losses from the medium-onwards. The last peak at around 480 m is the beginning of the dogleg, where the off momentum particles are over steered into beam pipe.

Zoom In Zoom Out Reset image size

There is an interesting regularity of the losses in the superconducting region which is shown better in figure 18. The figure shows the loss distributions in each period of the medium-overlapped. It can be seen that the majority of the losses are in the warm linac (first 2 m of the figure), but there is a non-negligible amount of losses also inside the cryomodule. The relative amount of losses in the cryomodule may reduce with a more realistic aperture model, and would then be expected to be lost in the beginning of the downstream LWU.

Zoom In Zoom Out Reset image size

In figure 19, the energy distribution of the losses in the medium-and start of high-is shown. There are very few losses being initiated from the normal conducting linac and the spoke section. The majority of losses have an energy equal to the medium-input energy of 216 MeV or more. This is expected to originate from the challenging frequency jump between the spoke and medium-section. It is also shown that there are almost no losses from the second half of the region where this data is taken from, indicating that these are slow (longitudinal) losses.

Zoom In Zoom Out Reset image size

The RF design parameters that need to be specified for the 2D cross section are frequency, length, mean value of electrode tip-to-axis distance r₀ versus abscissa z along the RFQ (varying between 3.530 and 5.332 mm), radius of curvature of electrode tips ($\rho =3.0$ mm), and inter-electrode voltage V_p versus z (from 80 kV at RFQ input to 120 kV at RFQ output). The RFQ cross section is designed to achieve the ideal voltage profile V_p(z) by modulating the lateral surface geometry of the electrodes.

Stability analysis shows that, with the proper design of the RFQ end-circuits, the RFQ dipole stabilising rods are not necessary. For this purpose, the vane undercuts are defined to be 41 mm and 45 mm respectively, at the input and at the exit of the RFQ. The quadrupole rods, located close to the pole tips in each quadrant, are located at 50.1 and 50.7 mm, with a translation range from 40 to 60 mm that gives a comfortable tuning interval from −0.1 to $+0.1$ (V m⁻¹) V⁻¹ during the validation process (before the final length is fixed). As a result, the closest quadrupole mode Q₁ is located at $+1.65\,\mathrm{MHz}$ from the fundamental mode Q₀. The closest dipole modes are shifted from Q₀ by −5.25 MHz for D₂ and $+2.67\,\mathrm{MHz}$ for D₃.

Slug tuners compensate the inevitable construction errors. The requirements on the voltage errors are given by:

$$\begin{eqnarray}&&\displaystyle \frac{{U}_{Q,S,T}-{V}_{p}}{{V}_{p}}\ \leqslant \ 0.02,\end{eqnarray} \tag{ 3 }$$

where U_Q is the quadrupolar component and U_S and U_T are the dipolar components of the RFQ voltage V_p indicated in figure 24. A perfect RFQ has U_Q = V_p and ${U}_{S}={U}_{T}=0$. There are 15 tuners per quadrant—60 in total—equally spaced with a diameter of 80 mm. Error analysis shows that with a maximum penetration range of 26 mm in the quadrant, the tuning scheme compensates for construction tolerances of $t+\delta \leqslant 70\,\mu {\rm{m}}$ with 30% safety margins, assuming that the centre of curvature of each electrode tip is located in a square with side $2t$ (centred at its theoretical location), and the difference between actual and theoretical electro tip profiles is bounded by δ. At position 0 the resonant frequency is 349 MHz.

Zoom In Zoom Out Reset image size

The voltage profile and the frequency is adjusted after an initial dedicated bead-pull measurement in the magnetic field region of the RFQ. The RFQ goes through its final tuning once installed at its final position in the tunnel. The position of the bead trajectory is close to the quadrant bottom walls, with a total of 30 field sampling-points per quadrant. In the absence of experimental errors the voltage profile can be reconstructed with a relative accuracy of 10⁻⁴. A unique feature developed for the ESS RFQ is the use of adjustable tuners that allow higher efficiency, time and risk reduction, including the possibility to re-tune the RFQ during its lifetime. Voltage profile monitoring is performed by a total of 20 pick-up loops that are located in the mid-plane of each module, calibrated against a reference bead-pull measurement.

The RF power is coupled to the RFQ through two coupler loops with a design optimised to minimise voltage perturbations. Both loops are located close to the RFQ mid-point, where modes Q₁ and D₃, most sensitive to perturbations, present a voltage node. The loops are placed in opposite quadrants, as shown in figure 25, in order to avoid inducing dipole-like perturbations. The power coupled into the RFQ cavity is adjusted by rotating the loop, giving the maximum coupling when the loop is aligned to the transverse plane. Each loop is specified to handle a coupled power of 1 MW, although it is expected that the maximum power transmitted during operation will nonetheless be limited to 0.8 MW, equally balanced between the couplers, in order to meet the RF budget detailed in table 10.

Zoom In Zoom Out Reset image size

Table 10.  RFQ power budget.

Component	Power
	kW
Copper power dissipation
2D design	524
Input and output end circuits	5
Vacuum ports	62
Tuners at +26 mm max position	408
Wall roughness	100
Total dissipation, theoretical 3D design	1099
Maximum operational power ($+25 \%$)	1375
Beam power	225
Maximum total coupled power	1,00

Active water cooling controls the RFQ deformations due to power losses on the cavity walls that cause frequency shifts. The cooling channels are located in the cavity body, close to the electrode tips. The temperature of the water is kept constant at $30\pm 0.1\,^\circ {\rm{C}}$, enabling the electrode temperature (close to the body temperature) to be set with a precision of $0.1\,^\circ {\rm{C}}$. The cooling water circulation is alternated from one section to the next.

The five sections of the RFQ are assembled using positioning pins, RF seals and Helicoflex seals (or gaskets). The RFQ is also equipped with 60 tuners, 2 couplers (4 coupler ports in total), from 8 to 10 turbomolecular pumps (TMPs) (36 vacuum ports in total), 22 pick-up ports including 2 for LLRF (28 ports in total), and 80 cooling connectors on 40 cooling plates.

Each section is made of four poles (one per vane) in pure copper, two minors and two majors as shown in figure 26. Pure copper (Cu-OFE, $99.99 \%$ Cu) is the main material used for the bulk of the RFQ, components and ports (except for stainless steel flanges) chosen for its high electrical and thermal conductivity as well as for its brazing possibilities. Every pole comes from the same casting in order to assure identical properties. Poles undergo several annealing and stress relieving heat treatments at different steps of the manufacturing process to eliminate stress or constraint in the material. A high isostatic pressing treatment during RFQ manufacture avoids any milimetric shrinkage defect or porosity in the copper. The ports are pure copper tubes with stainless steel flanges. Poles are machined with a precision of $20\,\mu {\rm{m}}$ and then positioned and brazed with a $30\,\mu {\rm{m}}$ precision, according to beam dynamics and RF studies and design.

Zoom In Zoom Out Reset image size

Pole and port assembly is performed in two brazing steps under vacuum. The first is a bi-metal brazing between the copper tube and the stainless steel flange, at high temperature, to obtain a port (i.e. vacuum, tuner, pick up or coupler port). The second step is a copper–copper brazing, at $1000\,^\circ {\rm{C}}$, to precisely assemble the poles and ports on the poles. The manufacturing process includes different machining steps and cleaning processes, consistent with brazing, vacuum and RF conditions.

The requirements of table 11 imply that the quality factor Q₀ and the shunt impedance r_sh must be laeger than 18 000 and 1.26 MΩ, respectively. Careful cavity shape optimisation generates the results presented in table 12, showing comfortable design margins compared to the requirements, and operational margins to adjust the proton beam energy if the RFQ output energy is lower than expected.

Power is coupled into the cavity via an inductive loop. Its prototype is shown in figure 27. The input coupler uses a standard flanged rigid coaxial interface, while the output coupler interface is a standard ConFlat vacuum flange inserted into the buncher cavity. The optimisation goal is to obtain ${S}_{11}\lt 40$ dBs in a wide frequency band from 300 to 400 MHz.

Zoom In Zoom Out Reset image size

A fixed slug tuner on the cavity is used to compensate the frequency shifts due to machining and assembly errors. A movable tuner is available for the low level RF (LLRF) system to tune the cavity frequency during operation, with a frequency correction of up to +2 MHz. A magnetic loop probe samples the field in the cavity, with minimal effect on the axial field. Active water cooling with an inlet temperature of $25\,^\circ {\rm{C}}$ minimises deformations during operation, paying particular attention to the nose region where the heat deposition is the highest and where the accelerating field is the most sensitive to changes in shape.

The cavity mechanical design comprises two main pieces: one includes the cavity barrel and a cover, while the other includes only one cover [57]. Each piece is made of stainless steel (AISI304). Commercial flanges are welded to the body after machining. Once all the components of the cavity are assembled, the interior of both pieces is plated with $30\mu {\rm{m}}$ of copper, creating a layer thick enough for the RF requirements. Both parts will be assembled using an aluminium seal with raw stainless steel surfaces, to guarantee the vacuum requirements.

The model is fabricated with all the ports opened and the tuners (one fixed and one movable) in the correct positions. The assembled cavity is then measured to verify the design and to check the resonant frequency. The cover piece edge is machined later, shrinking the gap length when the full cavity is assembled, and resulting in a lower resonant frequency. This process continues iteratively until the desired frequency is obtained. A buncher prototype built and tested at ESS Bilbao (Spain) is shown in figure 27.

The drift tubes bodies are made from OFE copper, with copper electro-plated AISI 304L stems. After inserting the internal components, such as the BPMs, the electro-magnetic dipoles, and the PMQs, the drift tube bodies are vacuum brazed and sealed by electron-beam-welding. Prototypes of a drift tube and a PMQ are shown in figure 29.

Zoom In Zoom Out Reset image size

The post couplers needed for RF stabilisation do not need to be bent, thanks to the choice of constant E₀ in each tank [59]. The linear density of the post coupler distribution along each DTL tank varies like

$$\begin{eqnarray}&&{N}_{{\rm{PC}}}\ ({{\rm{m}}}^{-1})\ =\ \displaystyle \frac{| {E}_{{\rm{first}}}-{E}_{{\rm{last}}}| }{{E}_{0}}\displaystyle \frac{1}{{L}_{{\rm{tank}}}}\end{eqnarray} \tag{ 4 }$$

with the increasing cell length and with the 2 m tank modulation. The detuning induced by the post couplers is +20 kHz.

One 2.9 MW klystron feeds each DTL tank. A 30% fraction of that power goes to waveguide losses and LLRF regulation. The remaining 2.2 MW enters the cavity through two iris couplers located at 1/3 and 2/3 of the tank length to minimise the induced field perturbation. The coupling strength is optimised in order to critically couple the waveguide and the beam loaded cavity. The iris height and aperture allow a coupling strength $\beta =1.2$, with a 20% margin that can be adjusted by shifting the short circuit at the end of the waveguide.

Tuners compensate for static frequency errors as large as ±750 kHz caused by construction errors, taking into account realistic tolerances on the principal DTL dimensions, such as tank diameter, and drift tube lengths, diameters, and face angles. The tuning sensitivity is 7.0 (kHz mm⁻¹)× m, with a diameter of 100 mm. Frequency errors due to thermal deformations in operation are corrected by three movable tuners in each tank, with a tuning range from +120 kHz at the maximum penetration depth of 80 mm, to −90 kHz at zero penetration. The nominal water cooling temperature is optimised during the RF conditioning phase, to benefit from a thermal tuning range from +40 kHz at $26\,^\circ {\rm{C}}$, to −100 kHz at $40\,^\circ {\rm{C}}$.

The principal DTL RF parameters are summarised in table 14.

The electromagnetic design of the spoke cavity is guided by the resonant frequency, the optimum beta and the optimisation of the peak surface fields. Whereas the most important parameter for the beam is the accelerating field (or voltage) seen by the beam, two other important parameters are the ratios of the peak surface fields to the accelerating gradient; the peak electric and magnetic surface fields should be minimised for a given accelerating field. Another important parameter is the overall length of the cavity. A spoke cavity has re-entrant end-cups that can be lengthened to give more volume per unit of stored energy, thus decreasing the peak surface fields. Such lengthening has the drawback of decreasing the real-estate gradient, with more longitudinal space taken to produce the same accelerating voltage.

The spoke cavity is designed and optimised using the 3D CAD tools of the CST Microwave Studio software package. First the fundamental TM₀₁₀ accelerating mode is calculated with reduced CPU-time, by applying a magnetic boundary condition (${H}_{{\rm{tangential}}}=0$) to both symmetry planes of a geometric model of 1/4 of the cavity. The main goal is then to reduce the peak surface field ratios to

$$\begin{eqnarray}\begin{array}{rcl}{E}_{{\rm{pk}}}/{E}_{{\rm{acc}}} & \ \lt \ & 4.38\\ {B}_{{\rm{pk}}}/{E}_{{\rm{acc}}} & \ \lt \ & 8.75\,\,\,(\mathrm{mT}\,{(\mathrm{MV}{{\rm{m}}}^{-1})}^{-1})\end{array}\end{eqnarray} \tag{ 5 }$$

using an optimisation procedure that is divided into three parts:

• (i)  
  Optimise the geometry without any cavity body ports.
• (ii)  
  Integrate the RF coupler port with the cavity body and perform Q_ext calculations.
• (iii)  
  Include all ports: RF coupler, pick-up probes and ports dedicated to cavity preparation (chemistry and rinsing).

Benchmark tests show that at least 10⁵ hexahedral mesh cells are required in order to calculate accurate values of both ${E}_{{\rm{pk}}}/{E}_{{\rm{acc}}}$ and ${B}_{{\rm{pk}}}/{E}_{{\rm{acc}}}$ ratios. Table 15 lists the cavity design parameters, after the optimisation is complete.

Table 15.  Simulated spoke cavity design parameters.

Parameter	Units	Mesh type
		Hexahedral	Tetrahedral
Number of mesh cells	106	1.2	0.65
Optimum β		0.50	0.50
Peak electric field, ${E}_{{\rm{pk}}}/{E}_{{\rm{acc}}}$		4.96	4.47
Peak magnetic field, ${B}_{{\rm{pk}}}/{E}_{{\rm{acc}}}$	mT (MV m−1)−1	7.03	6.74
Shunt impedance, r/Q	Ω	428	427

The power coupler has the double role of supplying RF power to the cavity while insulating the cavity vacuum from the atmospheric pressure inside the RF line. The design is based on the coupler developed for the superconducting spoke cavities in the framework of the EURISOL design study. To adapt that design to the ESS power coupler requirements, a water-cooling system is integrated in the inner antenna and the water-cooling system of the 100 mm diameter ceramic window has been modified to direct the water flow more effectively. The outer conductor is cooled with supercritical helium. The ESS spoke power coupler has been designed for a peak power of 400 kW (335 kW nominal) and its doorknob transition from coaxial to half height WR2300 waveguide. The power coupler prototype can be seen in figure 31.

Zoom In Zoom Out Reset image size

Each ESS double-spoke cavity is tuned with a double-level arm type tuner with eccentric shaft actuated by a cold motor and equipped with 2 piezo stacks (figure 32). The coarse tuning range is about +170 kHz with a maximum stroke of 1.3 mm and the fast tuning range shall be +675 Hz minimum to compensate the dynamic Lorentz force detuning. Two versions of CTS have been designed in order to test three different lengths of piezos and two sets of each version have been fabricated. The CTS version 1 (nominal configuration) can host 36 mm and 50 mm long piezos whereas the version 2 (optional design) can host 50 mm and 90 mm long ones. Several configurations of CTS version 1 and 2 have been investigated during vertical tests of the cavities.

Zoom In Zoom Out Reset image size

Spoke cavities operate in a boiling bath of superfluid helium at a temperature of 2 K and a stable pressure of 31 mbar. The cryogenic transfer line (CTL) distributes 4.5 K, 3 bar helium for cool-down and for nominal operations of the equipped cavities and the RF power couplers, from the cryoplant, which also serves the medium and high beta sections of the linac. Two helium return lines operating at 1.05 bar and 31 mbar close the cold mass cooling circuit. A valve box between the CTL and each cryomodule distributes cryofluids via a jumper. A vacuum barrier placed on the tunnel side of the valve box permits each cryomodule to be individually disconnected without warming-up its neighbours.

The CTL also provides 19.5 bar helium from an additional circuit to operate a TS at an intermediate temperature range of 40–50 K. This coolant is efficiently re-used at both ends of each cryomodule to intercept heat on the cold to warm transitions that link the cold beam pipe to the warm ultra-high vacuum gate valves.

Superfluid helium is produced at each cryomodule by a subcooler heat exchanger and an isoenthalpic expansion Joule–Thomson valve, placed within the jumper. Superfluid helium is collected into a biphasic pipe with an internal diameter of 63 mm, and is distributed through the top port of the cavity helium tank for filling. Complete cavity tank filling is regulated by measuring the superfluid helium level with a double wired superconductive gauge inside the biphasic pipe collector. In normal operation the biphasic pipe ensures the collection of the pumped saturated vapours. In the worst accident case of a beam vacuum break associated with an air entry into the beam pipe, it ensures the brutal evacuation of warmer gas towards the burst-disk.

The single-window power coupler is connected to the cavity by a double-wall tube. Heat conducted from ambient temperatures and from RF dissipation is intercepted by supercritical helium circulating in three parallel channels inserted helicoidally along the double-wall tube. Laminar flow in these channels, each 1.5 mm $\times \,2\,\mathrm{mm}$, allows a mass flow rate of about 0.015 g s⁻¹ with a limited pressure drop. The ceramic window and the antenna of the coupler are cooled at room temperature by two demineralised water circuits that can also be connected into a single circuit for the series of cryomodules. The magnetic shield of the cryomodule is actively cooled by the helium cooling line, before the cavities make the transition into the Meissner state.

The cryogenic distribution is dimensioned to handle the estimated static heat loads given in table 16. Dynamic loading induced by RF and beam dissipation brings an additional 4 W into the superfluid helium bath of each cryomodule, and requires about 0.015 g s⁻¹ of liquefaction power to cool each power coupler. The resultant cool-down times are four hours for the thermal and magnetic shields and eight hours for the cold mass. This gives sufficient margin for a full cooling time of one day.

The ESS prototype power-coupler is composed of three main parts:

• A window with its antenna to allow RF power coupling to the cavity and to isolate the cavity vacuum from atmosphere thanks to an alumina disk.
• A cooled double-wall tube to keep a coaxial configuration between window and cavity and allow thermal transition between ambient and cold cavity temperatures.
• A doorknob transition to allow RF matching between the coupler and the RF power source.

The antenna is cooled by a water flow and the external conductor is a double wall with an inner chicane cooled by supercritical helium gas entering the tube at the cavity side flange at 3 bar and 4.5 K. The ceramic windows and antennas are the same for both types of cavities and their Q_ext is adjusted by using external conductors with two different lengths, one for the medium and one for the high-beta cavities. Three diagnostics are used for protection of the ceramic window: a pressure gauge, an electron pickup antenna and an arc detector. Figure 33 shows the power-coupler system and its cooling circuits. The coupler conditioning is being performed first in travelling wave (TW) with low power, short pulses and low repetition rate. Depending on the couplers vacuum multipactor effect and arc breakdown diagnosis, the average power is increased up to 1.2 MW using $3500\,\mu {\rm{s}}$ pulses with a repetition rate of 14 Hz (nominal duty cycle). Then the power is maintained at nominal value for at least 15 minutes before decreased to check for multipacting zones. In standing wave (SW) RF conditioning, the measurement is similar to the TW one. However, the RF is totally reflected by a movable short circuit allowing the maximum field position sweeping along the coupler and particularly near the ceramic.

Zoom In Zoom Out Reset image size

Since these cavities are aimed to work in pulsed mode, it is important to minimise their sensitivity to Lorentz detuning. To do this, stiffening rings are placed between the cavity cells. The static Lorentz coefficient, ${{\rm{L}}}_{{\rm{L}}}={Df}/{E}_{{\rm{acc}}}^{2}$, measures the resonance frequency shift produced by the mechanical deformation induced by the electromagnetic field in continuous wave. The positions of the stiffening rings and the cavity thickness have been optimised in order to minimise $| {KL}|$ when the cavity has fixed extremities. The result is a niobium cell sheet of 3.6 mm in thickness and a stiffening ring radius of 84 mm for the high beta cavity. With these values, $| {KL}|$ is equal to $0.36\,{\rm{Hz}}\,{({\rm{MV}}{{\rm{m}}}^{-{\rm{1}}})}^{-2}$ for cavites with fixed ends. An exhaustive list of the corresponding RF and mechanical parameters is given in table 17.

The CTS is based on the proven mechanical design principles of the CEA tuners (France) that can be seen in figure 34. This type of tuner is working at cryogenic temperature inside the insulating vacuum tank. All the surfaces submitted to friction are treated with a dry lubricant coating efficient at cryogenic temperatures. It combines a slow and a fast tuner. The slow tuner is a double lever system with eccentric shafts actuated by a motor gear box and a screw. This mechanism is running under vacuum at cryogenic temperatures. The fast tuner is made by two piezo actuators that can act simultaneously or not to compensate the Lorentz force detuning. The tuning range of this slow tuner is $\pm 600\,{\rm{kHz}}$.

Zoom In Zoom Out Reset image size

Figures 35 and 36 illustrate the cryomodule design. The cavity package includes a cavity welded to a titanium helium vessel, a vertical power coupler equipped with a single coaxial ceramic window, a WR1150 doorknob transition, and a cold magnetic shield. The CTS includes two piezo stacks allowing fast and slow adjustment of the resonant frequency. Four cavities are connected by three 140 mm diameter hydroformed bellows, with 100 mm diameter cold-to-warm transitions at each end. Neighbouring beamline elements are connected (and isolated) by two DM 100 ultra high vacuum gate valves.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The cavity string is supported within the cryomodule by a network of tie rods made of a TA6V titanium alloy, connected to a stiff spaceframe. The vacuum vessel itself is a 304L stainless steel cylinder with a 1344 mm diameter, capped by two removable lids. Table 18 shows the expected cryogenic heat loads from each elliptical cryomodule. The interface with the linac cryogenic system consists of four helium circuits (inlet and outlet for the cavity string and cooling for the thermal shielding) via a jumper connection.

Solid state amplifiers (SSA) are expected to be more reliable, and usually lighter than tube amplifiers. The SSA in the klystron gallery amplify the RF energy from the LLRF to provide the right power to the next level of the chain, or directly to the cavity, as listed in table 22.

Table 22.  Solid state amplifiers in the klystron gallery. Output power is quoted at the 1 dB compression point.

	RFQ	Bunchers	DTL	Spokes	Medium-β	High-β
Number of SSAs	1	3	5	26	36	84
SSA input to	Klyst.	Cavity	Klyst.	Tetr.	Klyst.	IOT
Frequency (MHz)	352.21	352.21	352.21	352.21	704.42	704.42
Output power (kW)	0.2	$\geqslant 30$ peak	0.2	7.4	0.2	15
Min. gain (dB)	53	76	53	69	53	73
Goal efficiency (%)	$\geqslant 40$	46				50
PS AC voltage (V)	230	400	230	400	230	400
PS phases	1	3	1	3	1	3
Operation class	A	A or AB	A		A	AB
Cooling (Air/Water)	A/W	Water	A/W	Water	A/W	Water
LDMOS technology	Yes	(Open)	Yes	(Open)	Yes	Yes

In addition to the gain and output power (quoted at the 1 dB compression point) the main points of the specifications are:

• RF Pulse and rise time
• Power stability during the pulse
• Phase stability during the pulse
• Power stability between pulses <3% (or 0.12 dB)
• Phase stability between pulses <2°
• Gain and phase variation
• Harmonics
• Voltage SW ratio

The normal-conducting section of the linac operates at 352.21 MHz and contains the RF quadrupole and five DTL tanks, each fed through two couplers by an individual power source. The power requirement for this section is about 1.6 MW for the RFQ and 2.2 MW for the DTL tanks, plus 30% to compensate for losses and overhead required for regulation. This brings the power required from the source up to 2.1 MW for the RFQ and 2.9 MW for the DTL.

ESS uses six 352.21 MHz, 3 MW klystrons to feed the RFQ and DTL tanks. Klystrons at this frequency and power level are usually very long devices (more than 5 m) and they will be horizontally orientated in the gallery. Klystrons satisfying these specifications are currently available on the market from two different suppliers, as shown in figure 37. The main specifications are shown in table 23. Each modulator feeds two klystrons, with the RFQ klystron sharing the power supply with the first DTL tank.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 23.  Main specifications for the RFQ and DTL klystrons.

Parameter	Units	Value
Nominal output power	MW	3
Frequency	MHz	352.21
Bandwidth	MHZ	$\geqslant \pm 1$
Pulse width	ms	3.5
Repetition rate	Hz	14
Perveance		$1.3\times {10}^{-6}$
Efficiency	%	$\gt 52$
Voltage standing wave ratio		$\leqslant 1.2$
Power gain	dB	$\geqslant 40$
Group delay	ns	$\leqslant 250$
Harmonic spectral content	dBc	$\leqslant -30$
Spurious spectral content	dBc	$\leqslant -60$

The linac spoke section is formed by 26 superconducting accelerating cavities increasing the proton beam energy from 90 to 220 MeV. The RF peak power for this energy gain ranges from 260 to 330 kW at the input coupler of each superconducting spoke cavity.

The maximum required RF peak power of 400 kW meets the request for the cavity accelerating voltage level and takes into account some margin for RF distribution losses and regulation. The tetrodes are the simplest RF power source available to achieve such a power level by means of a combination of two tube transmitters. The choice of this power size allows matching each cavity to one RF power station (RFPS), ensuring an independent control of each system to cope with any phase offset, amplitude level and adjustment required. 26 RFPSs supply the linac spoke section.

The RFPS is the complete system that generates the required RF power levels for acceleration of the proton beam from the electrical grid connection. Driven and regulated by the LLRF signal and triggered at the repetition frequency, the RFPS amplifies the RF power according to the main parameters listed in table 24.

Table 24.  Spoke RF power station parameters. The RF efficiency ${\eta }_{{\rm{RF}}}^{}$ is defined as the ratio of the RF output power to the DC power input to the power supplies, during the time of the pulse.

Parameter	Units	Value
Central frequency	MHz	352.21
Nominal power, peak PN	kW	400
Nominal Power, average	kW	20
Bandwidth at −1 dB	MHz	$\gt \pm 1$
Gain at PN	dB	86
RF drive maximum	dBm	0
Operational mode		Periodic pulsed
Nominal repetition rrequency	Hz	14
Nominal pulse width	ms	3.5
Minimum pulse width	μs	140
RF output line		6-1/ ${8}^{{\prime\prime} }$ EIA coaxial line
RF efficiency ${\eta }_{{\rm{RF}}}^{}$	%	$\geqslant 50$

The RF amplification is based on a combination of a solid state pre-amplification and a tetrode amplification stages. Two RF branches are foreseen, with two transmitters. The final output power is achieved by means of high power hybrid combiner. The RF power level is monitored at several consecutive points. Gain and phase adjustment at the beginning of each branch allows the final power combination to be fine-tuned.

The solid state amplification stage has a nominal gain of 71.3 dB with a dynamic range that is optimised for the last 10 dB of the input drive, in order to maximise its linearity in the acceleration power range.

The modulators that supply the tetrode anodes with electrical power are pulsed high voltage power electronic converters. They are physically separated from the other RF power source cabinets. Each modulator has its own internal control and interlock systems, and is fully integrated in the RFPS as a part of the complete system. Each modulator supplies two tetrodes in parallel, hence 26 modulators are required. The tetrode modulators are based on switched mode power converter technology with high voltage stability regulation, high reproducibility and voltage droop compensation.

The main parameters of the modulators are listed in table 25. Key factors are a pulse stability of 0.3% of U_A over 2 h, and a reproducibility of 0.1% of U_A. AC grid power quality of the modulator follows the best available performances and IEC standard with power factor $\gt 0.9$ and an AC relative voltage flicker of less than 1% within one minute at the nominal repetition rate of 14 Hz.

Table 25.  Tetrode modulator rating and parameters. The pulse type is constant voltage, with a pulsed load current.

Parameter	Units	Value
Nominal output voltage, UA	kV	18
Nominal current amplitude, both outputs, pulsed IA	A	44
Nominal power, both outputs, pulsed PA	kW	720
Maximum droop or slow oscillation during flat-top	%	5
Minimum efficiency	%	90
Voltage ripple, maximum		0.1
0.3 kHz $\lt \,f\lt$ 1 kHz	%	0.3
1 kHz $\lt \,f\lt$ 100 kHz	%	0.1
100 kHz $\lt \,f\lt$ 300 kHz	%	0.3
Voltage regulation	%	0.05

A compact design of the RFPS is mandatory to fit all the equipment inside the available service area room. The optimal footprint for the RFPS including the modulator unit is 3600 mm × 900 mm. With this footprint, up to eight complete RFPS can be hosted in a 14 m × 10 m RF gallery area, providing enough space for operation and maintenance. Rear doors or movable panels allow easy access to the main RFPS components. The RFPS is equipped with knobs, switches, display and indicators that permits operation in local mode during commissioning and testing periods.

The medium-β and high-β sections contain respectively 36 and 84 superconducting elliptical cavities at 704.42 MHz. For regulation purposes each cavity is supplied by a single RF source through one coupler, with RF power levels varying from 225 kW to 800 kW for medium-β, and from 740 kW to 1.1 MW for high-β, depending on the location of the cavity along the linac. The RF source must be able to reliably supply the amount of power required at the cavity coupler, plus a 5% margin to compensate for the losses in the RF distribution system (RFDS), and an overhead for LLRF regulation purposes, currently estimated to be 25% of the coupled power.

Klystrons are used in the medium-β section, and multibeam IOTs in the high-β section. As the MB-IOT is a new technology, klystrons of the same type used in the medium-β section can also be used in the high-β section, as a backup solution. As a result, the medium-β klystrons have to be able to produce up to 1.5 MW of RF peak power at 704.42 MHz, with a pulse length of 3.5 ms at a repetition rate of 14 Hz.

Because of space constraints in the RF gallery (see figure 38), the medium-β klystrons operate in a vertical position, with the collectors at the highest part of the tube. Each single modulator (660 kVA) is used to supply four klystrons. The main klystron specifications are summarised in table 26.

Table 26.  Main specification for the medium-β klystrons.

Parameter	Units	Value
Nominal output power	MW	1.5
Frequency	MHz	704.42
Gun type		Diode
Bandwidth	MHZ	$\geqslant \pm 1$
Pulse width	ms	3.5
Repetition rate	Hz	14
Perveance		$0.6\times {10}^{-6}$
Efficiency	%	$\gt 60$
Voltage standing wave ratio		$\leqslant 1.2$
Power gain	dB	$\geqslant 40$
Group delay	ns	$\leqslant 250$
Harmonic spectral content	dBc	$\leqslant -30$
Spurious spectral content	dBc	$\leqslant -60$

The power levels required from the klystrons in the medium-β section vary from less than 300 kW to about 1.1 MW. Operating all the klystrons at the same design beam power (112 kV and 22 A) will lead to an efficiency as low as 11% for the first amplifiers of the section (including 30% overhead). A possible solution is to operate some klystrons at reduced beam voltages, which will also increase their lifetime.

When reducing the beam voltage and current, the space charge forces between the electrons in the klystron vary and the reduced plasma wavelength changes. In this situation, the cavity position, strictly related to the reduced plasma length, is no longer optimal. Furthermore, the impedance of the beam increases in such a way that also the external Q of the output gap is sub-optimal. As a consequence, the saturated efficiency of the tube drops. The external Q of the output cavity can be changed to increase the efficiency at reduced beam voltages by introducing a mismatch at the klystron output.

The high-β part of the linac accounts for 80% of the accelerator total beam power and consists of 84 high power sources each operating at approximately 1.2 MW.

Figure 39 shows the basic layout of how the RFDS transmits power from the HPA to the FPC, by means of either waveguides or coaxial lines depending on the power level of the linac module. A dummy load facilitates amplifier testing by dissipating the amplifier power. The high power loads are water cooled, connected in series with the klystron collector cooling circuit, simplifying energy recovery and cooling system design.

Zoom In Zoom Out Reset image size

When closed, a shutter switch after the circulator connects the amplifier to the load via the circulator, for testing. This scenario depends on the conditions to access to the tunnel and hence on permissions from the PSS. The circulator protects the amplifier from unwanted reflections, using an anisotropic ferrite material that transmits power only in one direction, and directs reflected power to the load.

Dual directional couplers (DDCs) are used to measure the forward and reflected power. The directivity of the DDC decides the error in the measured signal. The LLRF system requires a directivity of more than 40 dB for the DDC. The signal from the DDC is split and used by the LLRF and local protections system. The signal from the DDC is used to protect the amplifier and the cavity with the help of the local protection system. Signals from the DDC will be used to maintain the required gradient in the cavity, with the help of the LLRF system.

Flexible waveguide bellows take care of thermal stresses on critical components, like ceramic windows, in the klystron and in the FPC. Bellows in the stub take care of thermal expansion due to heat generation, and due to relative movement between the gallery and the tunnel. Bellows at the stub exit take care of differential thermal expansion of the tunnel wall and the tunnel flooring.

Microwave devices and systems working in high-power area may quickly build up extremely high electric field strengths that exceed the breakdown strength of the air, causing arcs. Once ignited, these arcs grow and travel towards the RF source. Hence arc detection and switching off the RF drive within 10 μs are vital. Locations prone to arcing are the ceramic window at the amplifier and the circulator and the load. Arc detectors will also be placed on the couplers.

The main requirements on the klystron modulators, in order of importance, are:

• (i)  
  Safety, health and environmental sustainability: safe and reliable energy storage; biodegradable dielectric materials; efficiency >92%.
• (ii)  
  Pulse power ratings: pulse voltage 115 kV; pulse current 100 A, driving 4 klystrons in parallel; pulse width 3.5 ms; pulse repetition rate 14 Hz.
• (iii)  
  Pulse quality: rise time $\lt 120\,\mu$s; high flat-top accuracy better than 0.15%; pulse-to-pulse stability $\lt 0.1$%; reverse voltage $\lt 15$%; energy in case of klystron arc $\lt 10$ J.
• (iv)  
  AC grid power quality: flicker $\lt 0.3$%; THD: $\lt 5$%; PF >0.95.
• (v)  
  Reliability: mean time between failures greater than 70 000 h.
• (vi)  
  Accessibility for maintenance and module replacement: mean time to repair less than 3 h.
• (vii)  
  Footprint: width $\lt 4.5$ m; depth $\lt 1.2$ m; height $\lt 2.4$ m; weight total $\lt 10$ ton; weight per component $\lt 2$ ton.
• (viii)  
  Cost: as effective as possible.
• (ix)  
  Standard components: use to the maximum extent, with at least two vendors.

Safety is the most important requirement for the modulators, so polypropylene self-healing capacitors are adopted, rated for 1 kV. Their extra cost is largely compensated by their longer lifetime—typically 100 000 h. The total volume of polypropylene capacitor banks represents about 7% of the total modulator volume—not a driving factor for size, weight or footprint. Modulators are heavy and bulky pieces of equipment. There is no crane available in the RF gallery for maintenance reasons, and so the weight of individual components does not exceed a few tons. A modular configuration is used, in which the total power is split amongst several identical components.

The proton beam is extracted from the plasma source of the MDIS type (see above). While the extraction voltage is kept at 75 kV, the RF power is pulsed at the nominal beam frequency to form and maintain the plasma. It takes up to 3 ms to reach the stable current once the RF has been switched on and the plasma decays in less than 100 μs when no RF power is available. The undesired transients are deviated to the chopper dump, located in the RFQ cone, by the electrostatic field created in the LEBT chopper to keep only 2.88 ms of stable beam pulse.

Because the rise time of the LEBT chopper is 100 ns chopper, it takes 100 ns to deviate the beam from its nominal axis to the dump (and vice versa). When the LEBT chopper is turned off and the beam is on axis again, it takes between 1 and 20 μs for the space charge compensation, which was annihilated due to the presence of the electric field, to recover and thus producing the matched Twiss parameters into the RFQ. Because of this phenomenon, a second chopper is required in the MEBT. The latter needs to be very fast with a rise and decay time of less than 10 ns to prevent potential hazardous losses at high energy of non fully kicked bunches [60].

The linac has two choppers and their associated dumps: one in the LEBT and a second one in the MEBT. The main mission of the choppers is to remove the transients from the beam pulse thus keeping only the stable 2.86 ms of the pulse required on the neutron production target. The source of transients are of two kinds: from the plasma generation in the ion source and from the space charge compensation in the LEBT. This is why it is mandatory set a chopping strategy in two steps. In addition to the their main feature, the choppers are also used for machine protection (MP) purposes. The chopper beam dump in the LEBT can handle the full beam current at nominal duty cycle while the one located in the chopper can withstand a 200 μs beam pulse without damage.

The normal conducting part of the linac hosts two types of collimators: an iris between the two solenoids in the LEBT and three pairs (i.e. vertical and horizontal) of scrappers in the MEBT at optimised locations [60] along the beam line. The iris permits to reduce the beam current at a desired level while the scrapers remove the unwanted beam halo that may cause losses at high energy.

The power of the beam during production mode is not fixed. Moreover, it is important to have the capability to limit the power losses at high energy to an acceptable level if required, while keeping the beam power as high as possible. Because the beam frequency and pulse length are fixed values, the only way to modulate the power is by modulating the intensity of the beam. The ion source alone can only produce stable beams with beam currents not lower than 90% of the nominal. In order to produce very low intensity beams of ∼6 mA with a controllability within 2 mA, an iris with movable blades is foreseen in the LEBT.

As experienced in other facilities such as SNS, the only way to reduce hazardous losses at high energy is to reduce the halo formation in the MEBT. For this purpose three movable scrapers are installed in the MEBT that can remove the beam halo above 3 sigmas.

LWUs located between the cryomodules in the superconducting linac (SCL) as well as in the HEBT and dogleg part of the accelerator-to-target (A2T) section of the ESS accelerator each contain two quadrupole magnets and one dual-plane corrector magnet for beam focusing and trajectory correction. All magnets are normal-conducting and operate in DC mode. There are three different types of quadrupole magnets and two different types of corrector magnets used in LWU. The LWU located between the spoke cryomodules contain two quadrupoles of type Q5 and a combined horizontal and vertical plane corrector of type C5. The LWU located between the elliptical cryomodules and in the HEBT contain two quadrupoles of type Q6 and a combined horizontal and vertical plane corrector of type C6. The LWU located in the dogleg part of the A2T section contain two quadrupoles of type Q7 and a combined horizontal and vertical plane corrector of type C6. Different types of the quadrupole and corrector magnets have different aperture, length and strength.

All three quadrupoles have a laminated yoke and water-cooled coils. Since Q6 and Q7 have the same bore diameter and the same good field region radius, these two quadrupoles can have the same 2D magnetic design with the same lamination geometry, the same ampere-turns and two different yoke/magnetic lengths. Furthermore, in order to use only one type of power converter, the design of all the quadrupole coils allows adopting the same conductor cross section and the same maximum current, or rather the same maximum current density. In order to minimise as much as possible the power consumptions, the conductor must have a cross section area which keeps the maximum current density lower than 4.5 A mm⁻². At the same time, the conductor must have a cooling channel with a diameter sufficiently large to reduce as much as possible the required liquid velocity in order to limit the pressure drop in the cooling circuit.

The magnetic field values and the artistic views of the quadrupoles for the LWUs are shown in figures 41 and 42. The main design parameters are summarised in tables 28 and 29.

Zoom In Zoom Out Reset image size

Table 28.  Magnet parameters and performance for quadrupoles type Q5, Q6 and Q7.

Parameter	Q5	Q6	Q7	Unit
Quantity	26	95	12
Aperture radius	34	56	56	mm
Yoke overall width and height	440	650	650	mm
Good field region radius r0	22	35	35	mm
Yoke length	170	230	290	mm
Coils overall length	250	340	400	mm
Magnetic length Leff at nominal Ic	200	277	337	mm
Maximum integrated gradient	2.21	2.47	3.01	T
Harmonic contents at r0	$\lt 0.01$	$\lt 0.008$	$\lt 0.008$	%
Inductance	8.2	36.5	45.0	mH

The overall dimensions of the correctors C5 and C6 are mainly driven by the LWU layout. In particular, the foreseen available space for corrector installation limits the C5 overall length to 70 mm and the C6 overall length to 100 mm. The full apertures of both correctors are equal to those of the quadrupoles in the corresponding LWU. The correctors are designed as window-frame magnets with the yokes made of laminated steel sheets to minimise as much as possible the eddy currents in the iron in order to allow possible feedback operations. The required field quality has to be better than 4%. Since correctors C5 and C6 are powered by only one type of power converter, the coils use the same conductor cross-section and the same maximum current, or rather the same maximum current density. The correctors are designed so that they can be split into two halves allowing installation around the vacuum chamber.

Artistic views of the correctors for the LWUs are shown in figure 43. The main design parameters are summarised in table 30.

Zoom In Zoom Out Reset image size

Table 30.  Magnet parameters and performace for correctors type C5 and C6.

Parameter	C5	C6	Unit
Quantity	13	55
Yoke width and height	138	206	mm
Yoke length	35	65	mm
Yoke mass	1.6	7.6	kg
Coil cooling	Air	Air
Maximum current density	1.5	1.5	${\rm{A}}\,{\mathrm{mm}}^{-2}$
Resistance of one coil	16.1	40.4	mΩ
Coil overall length	68	98	mm
Coil mass	1.4	3.6	kg
PC Maximum current IMax	16	16	A
Maximum voltage	0.5	1.3	V
Maximum power dissipation	8	22	7 W

Two vertical dipole magnets D1 bring the beam from the accelerator tunnel (AT) level up to the level of the target where the beam is focused and transported to the target in a straight section using six quadrupole magnets of type Q8. The trajectory correction is done by four combined horizontal and vertical plane corrector magnets type C8. These dipoles, quadrupoles and correctors are all normal-conducting and operate in DC mode. The dipoles and the quadrupoles coils are water-cooled. The main design requirements for these magnets are summarised in table 31.

Table 31.  Main design requirements of the accelerator-to-target magnets.

Parameter	C8 corrector	D1 dipole	Q8 quadrupole	unit
Min. aperture diameter	137	112	137	mm
Effective length	n.a.	1800	800	mm
Max.overall length	399	2200	1000	mm
Nominal bending angle	0.052	4	n.a.	o
Nominal bending radius	n.a.	25.783	n.a.	m
Nominal magnetic field	n.a.	0.36	n.a.	T
Nominal integrated field	$8.4\times {10}^{-3}$	0.648	n.a.	T m
Nominal integrated gradient	n.a.	n.a.	7.8	T
Integrated field uniformity	$\lt 5\times {10}^{-3}$	$\lt {10}^{-3}$	n.a.
Magnetic field harmonics	n.a.	n.a.	10−3
Quantity	4	2	6

The dipole magnet is designed with an opening in the yoke to accommodate a special vacuum chamber with a straight-out beam port, allowing the beam transport towards the tuning beam dump, when the dipole is powered-off. Due to an expected remnant field when the dipole is powered-off a correction coil may be added to nullify that field and to allow the beam to pass straight to the tuning beam dump. Two dipoles are powered in series by a single power converter.

To reduce the time-averaged beam intensity, the ESS will use a fast AC transverse (RSM) system that introduces beam centroid displacements across the target and PBW. The total raster system consists of 8 colinear RSM, two sets of four units each acting in the respective transverse planes. Even though the RSM in a set should ideally be synchronised and share the same field amplitude, each RSM is powered by a dedicated modulator based on a capacitor-charging supply and an H-bridge. Not only does this modular design reduces the magnetic load on the RSM and the peak output power of each modulator, but it is also a straightforward approach to reduce the impact of element failures and implement serial redundancy. An isolated RSM or modulator failure within each set can thus be compensated by a 33% amplitude increase of the unaffected subsystems. To uphold a high reliability, active cooling of magnets and modulators is avoided by operating the system at a duty cycle of only 5%, appropriately more than the 4% beam pulse duty cycle, 2.86 ms at 14 Hz.

To avoid burn-in near pattern edges and corners, the sweep speeds should ideally be constant, while the direction alternates as dictated by the frequency. Each RSM subsystem is thus to generate a pulsed triangular magnetic field waveform, acting in either the horizontal (H) or vertical (V) direction.

Maintaining a sweep frequency ratio differing from the unity in the two planes, ${f}_{x}/{f}_{y}$, the RSMs produces a fine-meshed Lissajous-like, diagonal crosshatch beam centroid displacement pattern with a rectangular outline (see figure 44). Such an approach, combined with a $\simeq 0.68\,{{\rm{cm}}}^{2}$ beamlet, generates a time-averaged intensity distribution with a large uniform central region and less than 1% beam deposited outside the 160 × 60 mm² nominal beam footprint.

Zoom In Zoom Out Reset image size

Substantial fundamental raster frequencies are required, tens of kHz with 40 kHz as the specified maximum, to minimise raster-correlated modulation of the moderated neutron pulse.

The same RSM and modulator design is applicable to both the horizontal and vertical set of RSMs, differing only by magnet orientation. The RSM are placed in pairs of identical field direction, $({B}_{y}{B}_{y})({B}_{x}{B}_{x})| | ({B}_{x}{B}_{x})({B}_{y}{B}_{y})$. With the required magnitudes of dB/dt, a ceramic vacuum chamber is required, and a RSM pair shares a single 850 mm ceramic vacuum tube with metallic flange connections. A cross section and isometric view of a horizontal RSM (with a vertical field component) can be seen in figure 45. The magnet is based on a window frame yoke that gives good field uniformity in a large region of the magnet aperture. The considerable operating frequency calls for a yoke material with low eddy current losses and a high frequency response. A NiZn ferrite yoke, possibly CMD5005 from Ceramic Magnetics, is an obvious choice. Due to the skin depth in Cu, the coils are cut from thin, 1 mm, OFHC Cu plates and bent into 2-turn bedstead coils.

Zoom In Zoom Out Reset image size

It is planned to install in total about 100 BPMs on the ESS linac. The BPM measure the beam position and phase, and provide input to the beam interlock system (BIS) to stop the beam in case of fault. The phase measurement is mainly intended for cavity tuning and ToF energy measurements.

The ESS BPM system uses BPM sensors of different sizes and types. Small-aperture stripline BPM are used in the low-energy linac including the MEBT and DTL sections. The BPM of the downstream sections are of electrostatic button type.

Most of the BPM belong to the LWUs installed in between each two successive cryomodules. The current plan is to have one BPM per LWU with staggered BPM location (i.e. BPM location being alternatively closer to the upstream and downstream ends in the successive LWU). This creates two baselines for TOF measurements, and the phase difference between the two BPM including the number of periods can then be calculated by comparing the readout phase from the short and long baselines.

In order to minimise potential disturbances from nearby RF sources, the BPM signal processing is done opposite to the RF. This means that the second harmonic (i.e. 704.42 MHz) of the BPM signal is processed in the spoke and upstream sections, and the fundamental harmonic (i.e. 352.21 MHz) in the sections downstream to the spokes. A direct consequence of this is that the BPM and the LLRF systems shall use phase references with opposite frequencies.

Most of the BPM need to have an overall accuracy of $\pm 200$ μm and a resolution of 20 μm with the nominal beam current of 62.5 mA and pulse width of 2.86 ms. The BPM also need to successfully measure the beam position under off-optimal conditions such as with a de-bunched and low-current beam of 6.25 mA and pulse-width of 5 μm that is foreseen for linac commissioning. It is therefore important to ensure that even under such extreme conditions, the BPMs still provide a rough position so that the beam can be safely steered all the way to the target or the tuning dumps.

The BPM button voltage is expected to decrease to less than a half from the beginning to the end of the linac [61]. This is due to the beam velocity increase and the changes in the processing frequency and beam pipe size. Despite these variations, the same electronics is used for most of the BPM. Since the LLRF and BPM systems have similar requirements, considerable technical effort has been put in place to maximise synergies by using same or similar electronics and firmware for both systems.

The chosen platform for the BPM electronics is μTCA.4. Its design includes a rear transition module, where the BPM signals are level-adjusted, down-mixed to IF and filtered in the analogue domain, as well as an advanced mezzanine card for analogue-to-digital conversion and FPGA processing.

The BCM system are used to measure the pulsed beam current as well as the per-pulse and cumulative beam charge. It consists of 15 AC-current-transformers (ACCTs) and one fast current transformer along the 600 m linac with a higher concentration of the sensors in the low-energy sections. This is mainly to address issues regarding fast detection of high beam losses and initiating a beam abort request especially in areas where the BLMs cannot be successfully used.

The MPS requirements include differential beam current measurement using several ACCT pairs. The readout current from the two ACCT are compared with each other and in case the difference exceeds a certain threshold, a beam abort request is sent to the BIS. The pulse width, amplitude and frequency are also measured by the ACCT in absolute mode and a beam abort signal is sent to the BIS in case a mismatch is detected between the real beam and the selected beam mode.

It is planned to use commercial off-the-shelf ACCT sensors. The associated electronics is installed in the klystron gallery and consists on a stand-alone analogue front-end box and customised readout electronics based μTCA.4 technology.

The primary goal of the BLM system is to detect abnormal beam behaviour and trigger a beam abort in case of beam failures in order to keep the machine safe from beam-induced damage. In addition to its protection functionality, the system is expected to provide the means to monitor the beam losses during normal operation with the aim to avoid excessive machine activation.

The protection functionality gives a constraint on the system response time and sets the upper limit of its dynamic range. Similarly the lower limit of the dynamic range is set by the monitoring functionality.

Regarding the response time, the BLM system is required to react within ∼10μ in the superconducting and within ∼1 μs in the normal conducting parts of the ESS linac. These numbers were set based on a simplified calculation of the time needed to melt a block of material under a perpendicular incidence for the beam parameters expected in the time. The numbers have recently been recheck with updated values for beam current and beam size, while assuming a Gaussian beam instead of a uniform circular one. The results raise a concern for the 1 μs limit expected for the NC linac (see figure 46). However, these calculations give a conservative result on the melting time as no cooling process was included. On the other hand, the results agree well with the 10 μs limit for the SC parts of the linac. Though experience at SNS raises a question for this limit, since there a degradation of the superconducting cavities is observed already after losing $\lt 15\,\mu {\rm{s}}$ pulse of a 26 mA beam about 10 times per day [62].

Zoom In Zoom Out Reset image size

In the normal-conducting parts of the ESS linac, between one and two BLM devices per metre are foreseen. On the other hand, three to four devices per double lattice cell are planned in the superconducting parts, four where there is a cryomodule (three surrounding the quadrupole doublet and one more per cryomodule) and three in the transport section. The number of devices in each of the linac sections is listed in table 32.

Parallel plate gas ionisation chambers (IC) filled with N₂, which were developed for the LHC BLM system, are selected for the SC parts of the linac as a BLM detectors, mainly due to their their fast response performance.

What must be taken into account when using ICs is the photon background due to the RF cavities. This is mainly due to field emission by the electrons from the cavity walls, resulting in bremsstrahlung photons created on the cavities or the beam pipe materials [63]. Estimations of the energy spectra show that photons with energies up to few tens of MeV can be expected [64]. The LHC ICs have a cut-off for transverse incidence for photons below ∼ 2 MeV [65], thus in order to know the importance of this background, its level needs to be determined. However these levels are difficult to predict numerically as they depend on the quality of the cavities, beam loading, operation conditions and time.

In order to address this issue during the operation, baseline correction of the recorded signal is anticipated. Ideally the data for the baseline calculation is foreseen to be sampled in the time window after the RF power is turned on and before the beam pulse arrives in order to correct the BLM raw data in the pulse accordingly.

The expected particle fields outside the tanks in the normal-conducting parts of the linac are dominated by neutrons and photons. Therefore IC are not a good option for a BLM device in these parts of the linac. Instead micromegas detectors are currently being considered as BLM detector choice in the low energy parts of the linac Micromegas detector sensitive to fast neutrons and blind to photons based on signal discrimination may be used.

Current concept for the electronics of the IC based part of the BLM system located in the superconducting parts of the linac consists of two separate units (see figure 47). The BLEDP card [66], which was developed for the new BLM system at the CERN injector complex, is used as acquisition unit, serving as an analogue front end and a digitiser. This card has a wide dynamic range (10 pA—200 mA) and is likely to fit the ESS requirements for the dynamic range also after the revision.

Zoom In Zoom Out Reset image size

The acquisition unit can provide information on the integrated loss over a fixed time of 2 μs, so called running sum 0 (RS0). Processing unit is in turn planned to provide additional RS giving information on losses integrated over longer time scales. The unit acquires the baseline data and subtract them from the raw data. Depending on the beam mode, each RS for each channel can be masked if needed.

It is essential to have the linac equipped with a certain set of BI during its commissioning phase and start-up in order to tune it for optimal beam transmission and minimal beam losses during the operation periods. An example of such an instrument is a LBM, which provides information about an average time structure of the bunches. The ESS accelerator hosts four LBM devices. Their positions along the linac are indicated on figure 48. The one-sided rms longitudinal bunch size is expected to shrink from ∼150 to 3 ps during the acceleration process. The intrinsic limit for the LBM methods that are based on detection of the bunch electric field at the beam pipe boundary can be estimated [67] as

$$\begin{eqnarray}&&{\rm{\Delta }}t=\displaystyle \frac{{R}_{{\rm{bp}}}}{\sqrt{2}v\gamma },\end{eqnarray} \tag{ 6 }$$

where ${\rm{\Delta }}t={\rm{\Delta }}d/v$ represents the rms value of the longitudinal charge distribution on the inner wall of the beam pipe with radius R_bp, which is produced by a charged particle moving with speed v. Due to the rather low Lorentz factors γ of the ESS beam the one-sided rms bunch lengths will be far below this limit (see figure 49). Therefore the available options to measure the bunch profiles are rather limited.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The most common device used for measuring the longitudinal bunch profile in proton machines like ESS is the bunch shape monitor (BSM) proposed by A. Feschenko [68]. Feschenko's BSM is based on collecting the low energy secondary electrons (SE) emitted from a thin wire target placed in the beam. The time structure of the primary bunch is transformed into spatial distribution of SE through a RF modulation generated by a RF deflector. High negative voltage potential is applied to the wire to accelerate the SE towards the RF deflector with a set of slits and an electron detector at the exit.

A typical phase resolution of this device is ∼1^o [69], which is accurate enough only for one of the four planned LBM devices at the ESS linac³ , namely LBM1 located in the MEBT section. However it has been recently proposed by A. Feschenko that a resolution of 0.5^o or less could be achieved with a modernised version of the RF deflector, which should provide more symmetric electro-magnetic fields with less fringing fields at the edges [69]. As this is accurate enough for both LBM1 and LBM2 device, it has been decided to use Feschenko's BSM with proposed modernised RF deflector for both. Depending on how well the modernised BSM resolution can be improved the LBM3 could potentially also be a BSM device. However the expected bunch lengths at the LBM4 location are shorter than 3 ps, thus to short to be resolved even with the modernised BSM. Since a standard solution with sub-picosecond resolution does not exist for proton machines like ESS, LBM4, and potentially LBM3 as well, require development of a new device type. Additional argument for new development in the case of LBM3 follows from the study of the space charge effect on the performance of a BSM, which is described in [70].

For the case of LBM3 and LBM4, an alternative based on Cherenkov or transition radiation together with a streak camera is considered. The primary motivation for focusing on photons as the secondary particles is to avoid the space charge effects. Additionally both Cherenkov and transition radiation interactions are considered to occur fast in contrast to SE emission, where it was experimentally shown, that the emission time delay does not exceed (4 ± 2) ps [71]. Due to the rather low beam energy at the LBM3 location, there is a concern regarding a sufficient photon yield from transition or Cherenkov radiation. Therefore it is foreseen to use the new method proposed in [72] for this location, either as a supplementary option or as a back-up solution.

Wire scanners have been deployed successfully for decades in accelerators. This represents a conservative choice for beam profile measurement. Their principle is rather simple and consists of moving a wire across the beam while monitoring a signal proportional to the number of particles interacting with the beam.

Eleven wire scanners are installed in the linac at different locations. The wire cannot withstand the beam power during the production mode (2.86 ms, 62.5 mA, 14 Hz), thus, the beam power shall be reduced in order to preserve the wire integrity. Two beam modes are dedicated to commissioning and specific beam studies when the insertions of interceptive devices are allowed:

• Slow tuning mode (i.e. 100 $\mu {\rm{s}}$, up to 62.5 mA, 1 Hz).
• Fast tuning mode (i.e. 10 $\mu {\rm{s}}$, up to 62.5 mA, 14 Hz).

At low energy (i.e. below 200 MeV), the primary source of the signal is SEM current from the wire, while at higher energy the primary source is the light generated in a scintillator by high energy secondaries. All the wire scanners in the linac shall be able to reconstruct the beam profile in SEM mode.

The numbers of WS station per section, the numbers of actuator and acquisition channel are summarised in table 33.

Two types of wires have been considered a $33\,\mu {\rm{m}}$ carbon wire for the warm linac and a $40\,\mu {\rm{m}}$ tungsten wire for the cold linac

Due to the low beam energy, the MEBT represents the worst case for the wire scanner, in addition beam density is relatively high. The maximum temperature reached by each wire scanner during a scan at 1 Hz are summarised in table 34 as well as the expected beam sizes at the wire location.

Table 34.  Maximum temperature with a scan at 1 Hz.

Beam sizes (mm)		Tmax (K)	Tmax (K)
${\sigma }_{x}$	${\sigma }_{y}$	50 $\mu {\rm{s}}$	100 $\mu {\rm{s}}$
1.85	1.45	1900	—
3	1.5	1430	2110
2.5	2.4	1240	1780
3.2	3.2	975	1350

As shown in table 34, the temperature increase for the first wire scanner is above the mechanical limit of the wire even with a reduced beam power, the wire survives only if the pulse length is reduced to 50 $\mu {\rm{s}}$. In this case, the maximum temperature is around 1900 K, below threshold of thermoionic emission and the limit of a carbon wire, in fast mode, the temperature is below 1400 K.

In the spoke section, a wire scanner is installed after the first four cryomodules, the beam sizes at the wire scanner location and the maximum temperature reached during the scan are summarised in table 35.

Table 35.  Beam parameters at the spoke wire scanner location.

Beam sizes [mm]		Tmax [K]	Tmax [K]
${\sigma }_{x}$	${\sigma }_{y}$	slow mode	fast mode
2.5	1.9	1790	1325
2.9	1.6	1730	1260
2.6	2.1	1610	1240
2.8	2.2	1500	1170

In the elliptical section, the average beam sizes (${\sigma }_{x}={\sigma }_{y}=2$ mm) have been considered for a beam energy of 200, 500, 1000 and 2000 MeV, the maximum temperature on the wire are shown in table 36.

Table 36.  Maximum temperature on a tungsten wire in the elliptical section.

	${T}_{\max }[K]$
Energy [MeV]	slow mode	fast mode
200	1590	1215
500	1340	978
1000	1250	890
2000	1210	860

In the spoke section, the temperature might be too high during the slow tuning mode and a smaller wire diameter (20 μm) is used. In this case the temperature drops by almost 200 K after spoke 1 during the slow tuning mode and 100 K during the fast tuning mode after spoke 4.

In the elliptical section, The temperature is below 1600 K for all the cases and should not be an issue for the wire integrity. More details can be found in [64, 73].

In the MEBT, at full current (62.5 mA), the maximum peak signal expected is around 400 $\mu {\rm{A}}$ for the more focused beam and 180 $\mu {\rm{A}}$ for the less focused one.

In the spoke section, the maximum expected signal will be 120 $\mu {\rm{A}}$ after the first spoke cryomodule and 60 $\mu {\rm{A}}$ after the fourth cryomodule.

The wire scanner shall be also able to measure the beam profile with the lowest current foreseen in the linac (i.e. 6.5 mA). The level of the signal increases linearly with the beam current at constant beam sizes, in first approximation, it can be assumed that the signal is divided by a factor 10 compared to the previous estimation for the lowest current. This means that the peak signal will be around 40 $\mu {\rm{A}}$ in the MEBT and will drop to 6 $\mu {\rm{A}}$ in the spoke section.

In the cold linac, the BI is positioned in the warm sections, between the quadrupoles doublet. Due to the low energy of the beam, the shower created in the wire is stopped at the quadrupole, in order to keep a sufficient signal, both wire scanner actuator and scintillator shall be positioned between the magnetic elements, in consequence the full system has to fit in less than 45 cm.

The signal produced by the detector shall be independent on the beam position, the geometry of the detector is done in order to minimise this effect. The detector consists in 4 scintillators positioned around the beam pipe, the size of each active element is $5\times 5\times 25\,\mathrm{cm}$, arrange in order to have a square with a dimension equal to $30\times 30\,\mathrm{cm}$, each scintillator is surrounded by a 1 mm thick aluminium foil.

The homogeneity of the detector has been checked at different beam energies by moving a 1D gaussian beam ($\sigma =2\,\mathrm{mm}$) across the beam pipe aperture. For each energy, 49 points have been simulated to cover an square aperture from −30 to 30 mm in steps of 10 mm in both transverse directions. The signal on each scintillator shows a strong dependency on the wire position. In the worst case, a variation about a factor 2 can be measured along the full beam pipe aperture. By summing the signal from the four scintillators, the variation on the deposited energy can be lowered to just some units per cent.

At low energy, the detector is less homogenous. For the 220 MeV case, the error compared to the reference is less than 10% if the beam is kept in a square of $40\times 40\,\mathrm{mm}$ and less than 5% for a square of 10 × 10 mm (see figure 50). At 2 GeV, the error compared to reference is less than 5% over the range considered in the simulations. From these results, and assuming a beam size less than 3 mm (i.e. 1 rms) along the cold linac, it can be concluded that the homogeneity of the detector is sufficient for transverse profile measurement.

Zoom In Zoom Out Reset image size

A plastic scintillator might not be suitable for the ESS linac in terms of radiation hardness and light production. Other type of scintillators have been studied. The evolution of the signal as function of the beam energy has been estimated for three commons scintillators type [74]:

• BC 412 plastic scintillator
• NaI crystal
• BGO (bismuth germinate)

Beam energies from 200 to 2100 MeV have been considered with steps of 100 MeV, assuming the beam in the centre of the beam pipe, and with the same geometry as the one above.

The number of photons produced per primary crossing the wire in each scintillator has been calculated from the average energy deposition in the four scintillators and the photon yield. These values have been used to estimate the light power generated in each scintillator for typical beam sizes in the superconducting linac and for the nominal beam current. For these estimations, the scintillation photons have been considered as mono energetic, with an energy equal to the luminescence peak. The minimum power is then about 3.5 mW at 300 MeV for a plastic scintillator and up to almost 50 mW at 2 GeV for a NaI crystal.

At these levels of power, a photodiode can be used a light detector. Assuming a light collection and transmission from the scintillator to a photodiode of 20% and a typical spectra response of 0.2 A W⁻¹ for the NaI and plastic scintillator and 0.3 A W⁻¹ for the BGO scintillator, the signal level at the output of the photodiode is in the mA range for the inorganic scintillators [75].

The accelerator has two EMUs: one in the LEBT and the second one in the MEBT.

The LEBT EMU is used intensively during the commissioning phase to characterise the ion source, the beam transport in the LEBT and measure the Twiss parameters at the RFQ matching plane, the results of this measurement campaign is also used to provide input for the end to end beam dynamics simulations of the ESS linac In order to fulfil these requirements, the EMU shall be able to reconstruct the transverse emittance with an error lower than 10%.

The specifications of the EMU are:

• The maximum peak power absorbed by the thermal screen shall be 7.5 kW⁴ .
• The EMU shall be used at 8% of duty cycle (i.e. 6 ms, 14 Hz).
• The minimum beam sizes at the Allison scanner position shall be ${\sigma }_{x}={\sigma }_{y}=3\,\mathrm{mm}\,(1\,{\rm{rms}})$.
• The maximum beam sizes at the Allison scanner position shall be ${\sigma }_{x}={\sigma }_{y}=10\,\mathrm{mm}\,(1\,\mathrm{rms})$.
• The resolution in position shall be better than 500 μm.
• The EMU shall be able to measure a maximum beam divergence of $\pm 100\,\mathrm{mrad}$.
• The angular resolution shall be better than 0.5 mrad.
• The user of the EMU shall be able to choose the signal sampling in a range between 100 kHz and up to 2 MHz.
• The head of the Allison scanner shall fit in a DN250 CF flange or less.
• The length of the movement shall be design in order to scan a beam pipe aperture of 160 mm.

In order to withstand the beam power, a full finite element analysis has been performed [76] showing that only few materials are able to meet the specification on beam power density. The final design of the EMU is shown in figure 51, its consist of a beam stopper in copper plated with tungsten and entrance slit in TZM to withstand the beam power.

Zoom In Zoom Out Reset image size

MEBT EMU. One of the functions of the MEBT is to fully characterise the beam at the exit of the RFQ [60]. For this, a slit and grid system are installed in the reserved drift space for beam diagnostic downstream the chopper.

The space available for the beam diagnostics after the second triplet is less than 350 mm (flange to flange) and might not be sufficient to meet the performance requirement for EM with the slit and grid system. The distance between the slit and grid has to be increased to provide enough angular resolution, therefore the slit has to be positioned between two quadrupoles, in this case the distance between the slit and the grid is increased to 400 mm.

The mechanical design of the slit is constraint by the space available between the two quadrupoles and the function to absorb the beam power. The slit is not able to absorb the nominal beam power, during emittance scan the pulse length as well as the repetition rate has to be reduced to 50 $\mu {\rm{s}}$ and 1 Hz respectively. From the studies done for the LINAC4 emittance metre [77], graphite is the best candidate for the slit material. The slit has to be tilted with respect to the beam axis in order to spread the energy deposition on a larger surface, the angle can not be less than 45° due to space limitation.

Error on emittance reconstruction. The slit and grid method for measuring the beam transverse phase space might induced some errors in the emittance reconstruction, the major errors contributions are:

• Error on the slit/grid position
• Multiple scattering on the slit edges
• Space charge effect in the drift between the slit and the grid
• Cross talk between the wires

An optimised design of the slit and grid system reduces the contribution of these errors to an acceptable level. The influence error on slit/grid position can be reduced by increasing the distance between the slit and the grid, the effect of multiple scattering can be reduced by increasing the slit thickness above the penetration depth of the proton beam and increasing the slit aperture while the influence of space charge can be reduced by decreasing this aperture, a more complete study of this various effect can be found in [78, 79].

In order to avoid cross talk between the SEM grid wires, a second plan of wire shall be positioned downstream the grid and be polarised to attract the SEs [77].

In order to preserve a good accuracy for the measurement the slit and grid system has to fulfil these specifications:

• For the slit:
  • An aperture of 100 μm
  • A thickness of 200 μm
• For the SEM grid:
  • A minimum number of wire equal to 24.
  • The grid pitch shall be better than 500 μm
  • The grids shall be equipped with 20 μm tungsten wire
• The minimum distance between the slit and the grid shall be at least 350 mm

The overall accuracy on the emittance reconstruction depends on the knowledge of the relative longitudinal and transverse distances between the slits and the grids. A constant offset between their positions does not affect the emittance reconstruction accuracy, The slit and grid system are designed in order to insure mechanical stability of their position with respect to each other. In addition, the actuator system of is able to measure the position of each unit with an accuracy better than ±50 μm.

In total, four FCs are installed in the ESS linac: one in the LEBT, one in the MEBT an two in the DTL.

LEBT: FC The LEBT FC is the primary diagnostic used for the ion source and LEBT commissioning [80]. This FC is designed in order to withstand the full beam power at the exit of the ion source, the relevant beam parameters are:

• Beam current equal to 100 mA.
• Beam pulse length equal to 6 ms.
• Repetition rate equal to 14 Hz.

For a duty cycle above 8% and an average power above 600 W, a cooling circuit is necessary. A representation of this FC is shown in figure 52.

Zoom In Zoom Out Reset image size

The accuracy of the current measurement shall be better than 0.5%, a biased electrode is needed to achieve this specification.

In the MEBT, a FC is installed in order to avoid beam losses in the downstream section during the tuning of the first part of the MEBT, specially for emittance scan with a slit and grid system. The FC is located as close a possible to the end of the vacuum chamber foreseen for the installation of the various beam diagnostics in order to optimise the distance between the slit and the grid.

The cup is designed to absorb a beam with an energy equal to 3.62 MeV, with a peak power equal to 230 kW, at a maximum pulse length of 50 $\mu {\rm{s}}$ with a repetition rate equal to 1 Hz. In addition, the FC is able to withstand the same peak power, at a beam pulse length of 10 $\mu {\rm{s}}$ and a repetition rate of 14 Hz. At the MEBT FC location, assuming a beam with Gaussian distribution, the beam sizes are 2.5 mm (1 rms) in both transverse planes.

The MEBT FC is able to measure the full beam current without limitation on the beam position, the aperture of the FC shall be at least equal to 50 mm. To preserve space, the length of the FC is less than 50 mm, including the repeller electrode.The MEBT FC is able to measure the beam current with accuracy better than 1% and a resolution better than 0.1%, for all beam intensity foreseen in the ESS linac (6.3–62.5 mA). The signal is sampled at a frequency of at least 2 MHz.

In the DTL, two FC are installed in order to avoid beam losses in the downstream part of the ESS warm linac during tune up of the first DTL. The FC is installed in the inter-tank area, the FC positioned after the second DTL tank is named here FC2, the one after the fourth DTL tank is named FC4.

The FC2 and FC4 are designed to absorb a beam energy range from 20 MeV to 39 MeV and 39 MeV to 74 MeV respectively. From these values, the maximum peak power absorbed by each FC is calculated. Assuming a beam current equal to 62.5 mA, the peak power absorbed by the FC2 and FC4 is 2.43 MW and 4.62 MW respectively.

The FC is primary used during the RF tuning of the upstream tank(s) with a reduced beam duty cycle. For this particular studies it is foreseen to use a 5 $\mu {\rm{s}}$ pulse length at a repetition rate of 14 Hz. In this case, the maximum average beam power absorbed by the FC is 170 Watt for the FC2 and 324 Watt for the FC4.

In addition, the two FCs shall be able to absorb a 100 $\mu {\rm{s}}$ pulse length at a repetition rate of 1 Hz, at this duty cycle, the maximum average beam power shall be 243 Watt for the FC2 and 462 Watt for the FC4.Thermo mechanical analysis are mandatory to estimate the maximum power which can be absorbed by the FC and define the beam modes/duty cycle which allow a safe operation of the FCs. At the their locations, assuming Gaussian distribution of the beam, the beam sizes are 1.3 mm by 2 mm (1 rms) at the FC2 location and 1.6 mm by 1.9 mm (1 rms) at the FC4 location.

NPMs measure the profile of long high power pulses by imaging the interaction between the proton beam and the residual vacuum in the beam chamber. They complement lower intensity wire scanners by measuring transverse beam profiles of macropulses with more than 10¹² protons, without any upper limit. They also provide longitudinal profile information, making them important diagnostics for high-power commissioning and during routine operation. There are two NPM designs: the beam induced fluorescence (BIF) design, and the ionisation profile monitor (IPM) design.

The simpler BIF design is the first NPM installed, imaging the fluorescence of the residual gas interacting with the proton beam. It consists basically of an imaging system: a source, an optical system, and an image readout system. The BIF system is simple, with the ability to deliver 2D transverse profiles, initially providing two orthogonal projected 1D profiles of the beam. The source exists whenever protons are present, hence it is passive and robust. However, the residual gas pressure in the cold section of the linac is low, in the 10⁻⁹ mbar range. Consequently the source may not be bright enough to be displayed within the required performance. Also, there are technical challenges, in particular due to the hard-to-predict radiation environment in which the optical system will work.

The IPM design does not suffer from these challenges, but it is more complex. It relies on a high voltage applied across the ionised residual gas to project ions onto a screen. The detector observes the projected particles (ions and electrons) in the vacuum chamber. Again, the radiation environment is a critical issue for both operation and reliability. Maintenance is not trivial, in the high vacuum in the proximity of the cold accelerating cavities.

Doppler shift measurement diagnostics provide an accurate measure of the intensities of the different ion species produced by the source. It is based on the interaction of accelerated particles from the source (mainly H⁺, ${{\rm{H}}}_{2}^{+}$, and ${{\rm{H}}}_{3}^{+}$ ions) with the vacuum chamber residual gas, producing a proportional fraction of neutral excited atoms that emit de-excitation radiation at specific wavelengths. Doppler shifted spectra produced by H, H₂ and H₃ are observed at a given angle to the moving beam, allowing the fraction of H⁺, ${{\rm{H}}}_{2}^{+}$, and ${{\rm{H}}}_{3}^{+}$ to be measured. Doppler measurements are available from the first operation of the ion source, installed in the LEBT, and able to characterise the PS during commissioning.

The target proton BI systems monitors the beam on the target and on the tuning dump. This instrumentation supports commissioning, startup, and operations by providing measurement data through the EPICS control system. In addition, the systems detects errant beam conditions that could be damaging to the dump or target components and informs operators via the alarm system. If the errant conditions meet a predefined severity threshold, the instrumentation mitigates it by suppressing beam production via a direct and low-latency interface to the BIS. As part of the MPS, this interlock can suppress beam production within 10 μs of receiving a signal from an instrumentation system. Most of the systems described below measure beam parameters and detect errant conditions within the pulse period of 1/14 Hz. With this latency, each pulse is measured and qualified before the upcoming pulse is allowed. Some systems should make measurements many times during a pulse and could therefore detect and mitigate errant conditions within a single pulse.

To fulfil its primary mission of protecting target and tuning dump components, this instrumentation suite measures the proton beam current density in the two transverse dimensions, and the beam current outside of a desired footprint. The harsh environment and the requirements for MP about redundancy and diversity. In other words, each of these primary beam parameters has to be measured by at least two different systems. Each instrumentation system provides additional measurements for diagnostic purposes.

Components located within the target monolith and, to lesser extent, components located in the tuning dump are subjected to extreme radiation. For this reason, all of these beamline devices incorporate remote handling features. In addition, cost and radioactive waste associated with maintenance and replacement activities are minimised.

Imaging systems: Three imaging systems are present in the ESS accelerator: one for the target wheel, one for the PBW and one for the tuning dump. These measure the two-dimensional distribution of the beam current at or near the target and dump components. The density distribution is measured for each pulse and based on the result, the next pulse is enabled or disabled by the MPS. For the imaging systems, the most important parameter to measure is the maximum current density within the beam footprint. This should be measured with a precision of about 10% of the nominal peak density and with a spatial resolution of about 1 mm. If required, a measurement from an upstream beam current monitor can be used to normalise the results from the imaging system. To support imaging and calibration without beam, an illumination system should cover the entire field of view. Near the edges of the field of view, fiducials allow alignment, determination of scale factors and, optionally, correction of distortion. The imaging system that views the target wheel is also capable of measuring the relative position of the wheel and the moderator assembly. To support background suppression and diagnostic studies, the imaging systems provide adjustable spectral selectivity over the visible and near IR range.

For diagnostic purposes, it is desirable to measure the beam density distribution as a function of time during the pulse. This can be the result of data collected over many pulses in order to achieve an adequate signal-to-noise ratio. If errant conditions can be detected within a single pulse, the imaging systems and the MP interface is designed to interrupt beam production during the pulse. To verify configuration of the beam delivery system, a short pulse of about 10 μs is transported and measured.

The beamline device includes the source of the light to be imaged, plus a complete optical system to bring the image out of the shielding to a location suitable for electronics. Optical components that support illumination are also included.

The current concept for the source of light is based on the previous experience from the SNS at ORNL. Cooled surfaces in the proton beam path receive a coating of chromium doped alumina. This coating is applied in a thermal spray operation. For ESS, the coated surfaces are: the beam window directly upstream of the tuning dump, the PBW within the target monolith, and the rim of the target wheel. Coatings must remain functional for the lifetime of the component onto which they are applied. In the case of the beam dump window, the goal is life of the facility, as this component receives very little fluence. During full power operations, the PBW must last about one year, and the target wheel must last about five years.

The optical system must provide a field of view determined by the full size of the beam aperture, including fiducial elements, and transport the image through the shielding to the electronics. For the PBW and the target wheel, this is of the order of 260 × 110 mm². For the tuning dump, the field of view is about 100 × 100 mm². The object resolution must be of the order of 1 mm. The thermo-mechanical design of the optical components near the beam aperture accommodates the energy deposition from the proton beam and from secondary particles.

All three systems incorporate an alignment and illumination system for in situ testing and calibration covering the entire field of view.

The grid system consists on one beamline device located in the target monolith along with the supporting electronics. The device measures the horizontal and vertical projections of the beam current density such that changes of 20% with respect to nominal peak density can be accurately determined. The goal is to make this determination in less than 100 μs. The system interfaces with the BIS so that the beam can be interrupted within the pulse if the current density exceeds a programmable threshold. The measurements are made over beam widths that range from about 160 mm horizontally and 60 mm vertically down to several millimetres in each dimension.

The scope of the beamline device includes a multiwire grid assembly consisting of two planes of wires (horizontal and vertical), and additional planes as needed to provide electric fields. This replaceable assembly is integrated into a shield plug that provides cooling and positioning and is part of the target work package. All radiation tolerant cabling and feedthroughs between the multiwire grid assembly and a location just outside of the monolith is also included.

The aperture monitoring system consists on three devices for the target (one integrated into PBW assembly, one in proton BI plug, and one in the neutron shield wall far upstream of the monolith), and one for tuning dump (integrated into the beam window just upstream of the dump surface). Requirements for the beam on the target and dump specify the amount of beam current that can reside outside of a nominal footprint. For example, in the case of beam within the target monolith, this current is about 0.1% of the total. If the current near the aperture and outside of the nominal footprint exceeds a predetermined value, then the aperture monitoring system suppresses the beam production via its interface to the interlock system. The aperture monitors must provide a measurement precision that supports this function.

Fixed sensors surround the apertures of the PBW, the proton BI plug, and the tuning dump aperture. At each of these locations, the fixed assemblies provide many measurement points and will be integrated into bulk steel assemblies provided by the target work package. In addition, moveable sensors reside close to the upstream aperture of the neutron shield wall. Sensors on the moveable unit are set far outside of the beam core, but close enough to detect the result of deviations from nominal DC optics. Each beam line device also includes radiation hard cable assemblies and feedthroughs that carry signals several metres through the bulk shielding.

Two types of sensor are used in each location. One type is a low mass thermocouple that can detect errant conditions after many pulses. To achieve faster response, a complementary technique that measures current induced in metal blade (primarily by the charged particle shower) will also be employed. These metal blades are arranged in a segmented array surrounding the aperture and located about 1 cm upstream of the thermocouples.

Three insertable beam stops enable commissioning, tuning and startup of the linac while minimising beam loss in the downstream sections. One is located in the first spoke section LWU and allows tuning through the DTL. The second is located in the middle of the spoke section and the third is located in the middle of the medium beta section. This latter stopper provides the final destination before beam is sent to the tuning dump or to the target. Because they function primarily as beam stops and not as measurement devices, the accuracy requirements of the upstream FCs do not apply to these beam stops. All should handle probe, slow tuning and fast tuning beam modes.

Each of them is positioned between two quadrupoles, leaving a length available for the dump and the longitudinal shielding of about 0.46 m. The main concern with these devices is their activation. Shielding keeps the received dose rate after a 4 h cooldown below 25 microSieverts h⁻¹ at 30 cm from the shielding surface. Because they are located in close proximity to the superconducting cavities, these stops are designed for easy cleaning and incorporate formed bellows. An interface from the motion control and instrumentation systems to the BIS assures that damaging beam conditions are mitigated before harm comes to the beam stop or the cavities.

The activated target wheel backshines gamma radiation through the proton beam pipe, following beam exposure during periods of operation. The line of sight to the target wheel is blocked by a gamma blocker inserted in the beam pipe upstream of the target during maintenance periods, allowing personnel to perform their activities on components located in the A2T section of the linac. There is a similar situation with backshine from the tuning beam dump in the dump line section of the linac, although the irradiation times and beam powers (5 MW to the target wheel versus 12.5 kW to the tuning beam dump) are very different.

A gamma blocker reduces radiation levels to be compatible with hands-on maintenance conditions. Additionally, the radiation hazard is minimised by reducing the strength of the source (given by the irradiation and cooling times of the target wheel and tuning beam dump), by reducing exposure times to personnel (determined by maintenance procedures), and by increasing the separation between source and components (set by the lattice design).

The gamma blocker design is inspired by the SNS gamma blocker. A cylindrical core of high-Z material that resides in vacuum to one side of the beam line is horizontally movable and able to block the whole aperture of the beampipe. The gamma blocker servicing the accelerator to the target section is located just upstream of the neutron shield wall, while the one servicing the tuning dump line is just upstream of the external wall of the dump vault.

Along the accelerator beam, halo particles can fall outside the beam envelope and interact with accelerator structures and components. These primary protons generate secondary particle showers via nuclear interactions and produce prompt ionising radiation. The maximum average normal operational beam loss of 1 W m⁻¹ has been adopted at ESS [82] to allow hands-on maintenance of accelerator structures [84]. This loss of 1 W m⁻¹, together with the beam spill limits defined in [84] is used as requirement for establishing the design of the prompt radiation shielding. Note that performed error study analysis show that the expected beam loss levels at the ESS accelerator are much lower than 1 W m⁻¹ [85]. Additionally, operating experience at other facilities in Europe and the US demonstrates that losses below 1 W m⁻¹ are routinely achieved. Thus a limit of 1 W m⁻¹ is conservatively established.

A dedicated model of the ESS AT and surrounding structures has been built in the MARS15 code [81]. A detailed geometry model of the DTL and superconducting linac has been created. As-built AT construction drawings and existing 3D models for planned surrounding structures are also used to create the corresponding geometry in MARS15. With a simulated average 1 W m⁻¹ beam loss along the entire linac, the expected maximum total prompt dose rate in the vicinity of accelerator components, as shown in figure 53, it is estimated to be around 10 Sv h⁻¹ [81].

Zoom In Zoom Out Reset image size

The ESS accelerator shielding is designed to guarantee an external prompt dose rate of less than 3 μSv h⁻¹, corresponding to the limit for a supervised radiation area [86] in all adjacent accessible areas. Bulk shielding has been derived and special care was taken to correctly model and calculate the radiation transport and resulting local doses associated with over forty penetrations that are required for personnel and equipment access to the tunnel [81]. The ESS procedure for designing shielding for safety [87] has been followed. Details of the study can be found in [81].

Figure 53 shows that the total prompt dose rate changes depending on beam energy along the linac. The ratio of the highest and lowest prompt dose in the immediate vicinity of the beamline (at the Linac end and beginning of DTL, respectively) is approximately 100.

Accidental worst full point beam loss has been simulated at 2 GeV and the prompt dose rates calculated outside of the accelerator shielding, with a maximum estimated dose rate of 400 mSv h⁻¹ (including a safety factor of 2) on top of the accelerator shielding [81] and a maximum value of 120 mSv h⁻¹ in the klystron gallery [81]. Figure 54 shows a prompt dose rate map for accidental beam loss with a shallow angle directed vertically up.

Zoom In Zoom Out Reset image size

Any lost proton beam leads to activation of accelerator components and structures through nuclear reactions. These activated structures and components generate ionising radiation, which is called residual radiation. The residual radiation is dominated by gamma particles and has been calculated using the same MARS15 model as for shielding calculations. The ESS procedure for activation calculation [88] has been followed and the on contact residual dose rates calculated for the quadrupole and bending magnets and concrete tunnel walls [81]. Obtained on-contact peak residual dose rate can be up to 4 mSv h⁻¹ , for 100 d of irradiation with a beam loss of 1 W m⁻¹ , with 4 h of cooling time. Results are consistent with hands-on maintenance conditions and a design criteria of 1 mSv h⁻¹ residual dose rate at 30 cm from a surface of any given accelerator components, after 100 d of irradiation and 4 h of cooling time [82].

Activation of accelerator systems and components: The same MARS15 geometry model has been used for radioactive inventory calculations. The ESS procedure for activation calculation [88] has been followed. Isotope production rates have been calculated using the MARS-DeTra tool to determine nuclide inventories [89]. The quadrupole yoke and coil materials, beam pipe, accelerating niobium cavities, liquid helium, and cooling water have been also studied [90].

Activation of tunnel air: For 1 W m⁻¹ beam loss model, the produced unstable isotopes and total production rates in the entire tunnel air have been calculated and tabulated [83]. No ventilation was assumed. These results are used as the source term to calculate on–and off-site doses from normal operations.

The tunnel has an internal volume of $1.26\times {10}^{4}$ m³, housing 43 cryomodules connected to a distribution system consisting of a cryogenic line, valve boxes and jumper connections. The entire cryogenic system enters its steady-state phase of normal operation once all cryomodules are cooled down to 2 K. Access to the tunnel is allowed after cooldown, in order to perform maintenance activities such as leak test, visual inspection, and repairs.

Table 37 shows the estimated helium inventory that is stored in the cryogenic systems components.

An accidental release of helium in the tunnel during the maintenance phase might lead to personnel exposure to oxygen deficiency and cold burns. For that reason, the need has been expressed to carry out a CFD in order to provide a complementary study to the preliminary ODH assessment carried out using the Fermilab and CERN approach and to estimate the temperature, the oxygen concentration as well as the pressure at different locations and at different times in case of a release of helium inside the tunnel in order to quantify cold burn hazard, oxygen deficiency and pressure rise The simulated failure scenarios are listed in table 38.

The first failure scenario chosen for the CFD is an air inlet into the beam pipe and affecting a high-beta elliptical cryomodule. If such event occurs, it will be characterised by a high helium discharge dynamic. This event will lead to condensation and solidification of the air on the inner surface of the cavities which is not protected by any insulation layer. Heat from the air will then be transferred to the helium bath leading to an overpressure in the 2 K LHe circuit and bursting of the disks causing helium vapour to be discharged in the tunnel, figure 55.

Zoom In Zoom Out Reset image size

A scenario with the characteristics indicated in figure 57 has been considered for the cryodistribution simulations.

It shall be noted that the simultaneous release of helium from several cryomodules has been discarded due to the fact that the gate valves used to allow the vacuum isolation of the different sections and located in between the cryomodules, are closed during the maintenance periods of the accelerator. Thus preventing any propagation of a pressure increase in the beam vacuum to the adjacent cryomodules.

Figure 56 shows an estimation of the temperature and O₂ concentration near the faulty cryomodule 10 s after a helium discharge. Near the area of interest, O₂ concentration can drop down to 6% and the temperature can reach −135 °C.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Additional scenarios have been investigated in order to assess the level of safety for operators potentially located in critical areas, such as the dead-end of the A2T area, including helium pressure levels that might compromise the mechanical integrity of the exit doors, as occurred during the LHC accident in 2008. Moreover, the ventilation system of the AT plays a significant role in the helium dissipation in case of an accidental release and thus the efficiency of such system is of importance and needs to be analysed via a CFD.

A summary of the different failure scenarios simulated via CFDs can be found in figure 58.

Zoom In Zoom Out Reset image size

In order to prevent and minimise the exposure of the personnel to helium discharge in the tunnel, several control measures are implemented during the design and operation phases.

The implementation of a helium collection header allows venting the helium outside of the tunnel in case of a significant overpressure of the 2K helium vessel of one of the cryomodules (figure 59).

Zoom In Zoom Out Reset image size

In addition, the AT is equipped with an air management system based on the ISO 17873 standard. This system facilitates the dissipation of the vented helium in case of a discharge/leak in the tunnel.

During maintenance phases, all control measures are implemented as defined in the ESS ODH Safety Process and Implementation:

• Fixed and personal oxygen monitoring systems with a set alarm at 19.5% O₂
• ODH training.
• Operation, emergency and access procedures.
• Warning signs.

A dedicated ACCP is used to serve the accelerator superconducting section. The ACCP is the largest and most complicated of the ESS cryoplants. The cooling requirements and the initial concept design have been developed and improved over the years [107–111].

Architecture. The ACCP consists of a warm compressor, oil handling and gas management station, located in the ACCP compressor building, and one single coldbox, located in the cold box building. The simplified block diagram in figure 60 shows the main components of the ACCP, consisting of:

• The warm compressor station circulating helium at the flow rates and pressures required by the ACCP cold section
• Helium gas treatment comprising bulk and final oil removal and a gas dryer
• Gas management panel for process control
• One cold box mainly containing six turbines, three cold compressors in series, two 80 K adsorbers, one 20 K adsorber, several heat exchanger blocks and acceptance test equipment to provide the specified cooling and tests
• An ambient heater to assist warm-up of the system or the cool-down of single cryomodules
• Liquid helium storage tank

This equipment forms part of a single procurement by ESS. Other related equipment required to run the ACCP, particularly warm helium storage tanks, warm interconnecting piping, the CDS or the recovery system are separate procurements or in-kind contributions to ESS.

Zoom In Zoom Out Reset image size

Cooling requirements and steady state operation modes.Detailed thermal performance studies and analyses have been performed to estimate the heat load imposed on the CM and the CDS including heat load due to beam losses. The 2 K heat loads consist on the isothermal load, which is applied to the liquid helium bath surrounding the cavities and the non-isothermal load from the CDS. The isothermal load, including the statics and dynamics from RF loses in the cavity walls and beam losses, determines the helium mass flow through the 2 K circuit handled by the cold compressors. The non-isothermal load will strongly affect the 2 K return temperature at the inlet of the first cold compressor. The installed cooling capacities of the ACCP are 3050 W at the 2 K circuit, 9.0 g s⁻¹ liquefaction rate at 4.5 K for power coupler cooling and 11380 W at approximately 43 K for the TS.

The main steady state operation modes include the nominal design comprising both beam and RF operational and the nominal turndown with both beam and RF non operational. The former mode features the highest cooling capacity needed in both stages. In the latter, the total heat deposited into the 2 K circuit comprises only the static load. The 2 K refrigeration cycle is able to efficiently handle the dynamic heat load range. Two standby operation modes include 4.5 K standby for several-days interruptions of the RF power and beam and TS standby for the scheduled machine maintenance lasting 90 d per year (approximate 60 d in summer and 30 d in winter). In addition, the maximum 4.5 K liquefaction capacity of the designed system is used for cool-down and CM re-filling operation.

Project stages. The different groups of cryomodules, spoke, medium beta and high beta cryomodules, are installed and commissioned stepwise in the tunnel.

In order to achieve stable and efficient operation the cryoplant operates in two configurations, namely stage 1 and stage 2. This is realised by means of two sets of flow parts for cold rotating equipment, turbine expanders and cold turbo compressors, and variable frequency drives in the warm compressor system. The stage 1 turndown 2 K load is 48% of the stage 2 installed 2 K load. In the case that not all safety factors are required, The stage 1 turndown 2 K load is 30% of the stage 2 installed 2 K load. As this turndown ratio can not be handled by the cryoplant in an efficient manner; the relatively simple configuration with two sets of flow parts provides a good additional load adjustment possibility and has advantages in the spare strategy. The variable frequency control for the low pressure (LP) compressor, the sub-atmospheric pressure (SP) compressor and the cold compressors also help to provide efficient plant adaptation in all intermediate modes.

Design choices. The cryogenic system provides 4.5 K helium at a pressure of 0.3 MPa to each cryomodule. The actual production of 2 K helium occurs in the tunnel by means of a combination of a 2 K heat exchanger and a subsequent Joule-Thomson valve, found in each of the cryomodule-valve box assemblies. This kind of set-up permits independent warm-up/maintenance/cool-down of single cryomodules while the rest of the system is maintained in cold condition. The helium vapour with the corresponding saturation pressure of 3.1 kPa is warmed in the 2 K heat exchanger and by static heat load in the CDS before returning to the cryoplant. There, a combination of cold and warm compression stages ensures high flexibility regarding load adaption at optimal efficiency. Industry studies have shown that this concept and the decision to use one integrated coldbox for all cold ACCP components provide highest space and investment cost saving.

The ACCP operates without using liquid nitrogen pre-cooling as the load is dominated by 2 K refrigeration. The other two cryogenic loads that the cryoplant has to supply—TS cooling and liquefaction load for cooling the RF main power couplers—represent only 11% and 10% respectively of the plants total energy. This load combination makes the use of liquid nitrogen pre-cooling less attractive as performance boosting with LN2 pre-cooling is moderate in this load regime. The very thin walled cavities result in a relatively low cold mass of only 20 tons for the whole accelerator [108]. Hence cool-down is not a significant issue and LN2 support not required. Only expansion turbines will provide the cryogenic refrigeration in the ACCP, increasing reliability, capital cost saving and traffic reduction.

The TS temperatures are chosen to optimally fit not only the CM designer needs but also the ACCP process layout. This applies particularly on the temperature step across shield feed and return which should match the temperature step the related turbine string in the ACCP does. Turbine 3 in the ACCP takes care of the shield cooling and operates between HP inlet and MP outlet, resulting in a temperature step of about 20 K. The selected TS temperatures are >33 K in and <53 K out.

The ACCP control system is based on local PLC for internal, time critical and deterministic control loops and safety functions. Supervisory controls and high level batch operations run in the EPICS process controller. The HMI on local screens as well as in the control room, alarm handling and data archiving are all based on EPICS. This functional split between local PLCs and EPICS IOC provides safe operation even in case the EPICS IOC shuts down. The ACCP control system is compatible with the other linac control, providing all advantages of an open control system.

The ACCP compressor system is one of the heavy energy users at ESS. Apart from the focus on matching the cryoplants refrigeration capacity optimally with the linacs load requirements a second approach for using energy efficiently is to recover heat from the cooling water. For this purpose the compressors oil coolers are designed to minimise the temperature differences between cooling and warming flows. The cooling water flow to the single coolers is controlled primarily to achieve the required oil outlet temperatures but secondarily also to achieve a preferably high temperature spread between inlet and outlet of the cooling water. Besides reducing the required cooling water flow the water returns at elevated temperatures, enabling efficient heat recovery to serve Lund's district heating system and greenhouse applications.

The ESS CDS connects the ESS superconducting linac cryomodules with the ACCP. Its main function is to transfer the required cooling power from the ACCP to all the elliptical and spoke cavity cryomodules. The nominal operating temperature for the linac cavities is 2 K. However, the helium is delivered in two temperature levels: 4.5 and 40 K [109]. Then, the 2 K temperature level is produced at each cryomodule by isoenthalpic expansion of the 4.5 K supercritical helium, previously subcooled to 2.2 K in a counterflow heat exchanger.

The CDS consists of a CTL, a cryogenic distribution line and four auxiliary process lines, as shown in a schematic layout of the ESS linac cryogenic system in figure 61. The CTL runs from the ACCP cold box located in the cold box hall into the dedicated CTL gallery with a chicane leading directly to the linac tunnel. The CTL ends with a splitting box equipped with two cryoline terminals. The first terminal is connected to the cryogenic distribution line designed for serving all the 13 spoke and 30 elliptical cavity cryomodules of the Optimus + linac [112]. If the ESS linac requires any future extensions an additional CTL will connect all the contingency or/and upgrade cryomodules to the other terminal.

Zoom In Zoom Out Reset image size

The cryogenic distribution line of the Optimus + linac is located in the linac tunnel alongside the cryomodules. It consists of 43 valve boxes and an end box. Each valve box is equipped with cryogenic control valves and some other process control elements, such as check valves, warm control valves, safety valves as well as pressure transmitters and temperature sensors. Figure 62 shows the conceptual design and flow diagram of the valve box for the elliptical cavity cryomodules.

Zoom In Zoom Out Reset image size

The valve box design results from a number of main functional and technical requirements for the CDS that fundamentally come from the top-level requirement of 95% availability of the ESS linac itself. The CDS must be extremely reliable throughout the expected lifetime of 40 years. It must facilitate easy repairs or even easy replacements of faulty components, since the linac is operated remotely and continuously 24 h per day, for about 250 d per year. As a result, the valve box and entire CDS designs allow for warm-up and cool-down of a single cryomodule while keeping the rest of the system at cryogenic temperatures. They also keep very high thermal and mechanical properties within the operation lifetime and tolerate rapid cool-downs and warm-ups [113].

To meet these requirements the valve box contains six cryogenic valves (CV03, CV04, CV05, CV60, CV61 and CV63) and 2 warm valves (CV05 and CV62) as shown at the bottom of figure 62. In the nominal operation conditions CV03 and CV04 are fully open. Supercritical helium at 4.5 K and 3.0 bar flows from the Helium supply line to the cryomodule cold circuit. The major portion of this helium is throttled below 31.3 mbar and the liquid fraction of the produced 2 K helium cools down the cavities. The required subatmospheric pressure in the cavity helium vessels is produced by the constant evacuation of the helium vapour via the BB line, CV04 and further via the VLP line to the ACCP cold compressor system. The other portion of the supercritical helium is used for cooling the cavity power couplers, and returns to the ACCP compressor station as gaseous helium at 300 K and 1.1 bar.

The CDS also distributes 40 K helium to the cryomodules. This helium flows from the TS supply line to the cryomodule TS circuit, via the BE line and CV60, and return to the TS return line via the BF line and CV61. Since the BF line and TS return line are thermally linked to the CDS TSs the lines also work as thermal sinks for the 40–50 K heat loads of the cryogenic transfer and distribution lines.

The two remaining cryogenic valves (CV06 and CV63) and the warm control valves (CV05 and CV62) are dedicated to facilitating and speeding up the warm-ups and cool-downs of a single cryomodule. If one or a few cryomodules need some maintenance or repair work at room temperature the cooling loops in the particular cryomodules are shut off, whilst the other cryomodules are brought to the so-called 4.5 K stand-by mode. Then, just after the evaporation of the liquid helium from the cavity helium vessels in the serviced cryomodules, warm helium from the HP line is supplied to the cryomodule circuits via the BH line, CV05 and CV62. The supplied helium warms up the cold parts of the cryomodules and flows to the ESS helium recovery system via CV06, CV63, the BR line and the helium recovery line. Later, after maintenance or repair, valves CV05 and CV62 are closed and valves CV03 and CV04 are opened in order to bring the serviced cryomodules back to the cryogenic conditions. As soon as the cryomodule cooling loops are sufficiently cold, valves CV06 and CV63 are closed and valves CV04 and CV61 are opened, and the entire linac cryogenic system is brought back to the nominal operation conditions.

The valve boxes for the spoke cryomodules differ slightly from those for elliptical cryomodules. In order to optimise the design and assembly of the spoke cryomodules some items of the cold circuit are moved from the spoke cryomodules to the related valve boxes. Therefore each of these valve boxes additionally houses a counterflow heat exchanger and two additional cryogenic control valves, i.e. a throttling valve and a filling valve of the cavity helium tank.

One key feature of the target system is the hydrogen moderators, which use supercritical hydrogen at 17 K and 1.5 MPa to reduce the energy of the neutrons before they reach the instrument lines. Neutrons, protons and photons deposit significant energy into the hydrogen, which must be removed to maintain the hydrogen at its nominal operating temperature of 17 K. The TMCP [114] cools this hydrogen cryogenic moderator system (CMS). The heat deposited into the hydrogen is removed via a heat exchanger in the CMS coldbox that transfers the heat from the hydrogen circuit to a gaseous He circuit operating between 15 and 20 K which is connected to the TMCP cold box via a CTL.

Target cryogenic system overview. The target cryogenic system consists of three subsystems: the CMS, the TMCP, and the CTL. The CMS [115] comprises a vacuum insulated cryostat that contains hydrogen circulation pumps, a helium/hydrogen heat exchanger, an ortho-to-para hydrogen conversion catalyst bed, a pressure control accumulator, interconnecting piping, valves, and instruments. Supercritical cryogenic hydrogen circulates through vacuum insulated piping connecting the CMS cryostat with the cold moderators, which are located at the target wheel. The TMCP is comprised of a cold box containing turbo-expanders, heat exchangers, and associated valves and piping; a warm compressor station containing helium compressors, oil removal system, gas management panel, and associated piping and controls; and an automatic control system. The CTL consists of a set of vacuum insulated piping (one supply line and one return line) that circulates cold helium between the TMCP and CMS. Figure 63 shows a simplified schematic of the target cryogenic system.

Zoom In Zoom Out Reset image size

TMCP design. The total heat load handled by the TMCP is the sum of the CMS heat load plus the additional static ambient heat load into the helium from the CTL. The TMCP must operate efficiently over a wide range of heat loads due to the following factors:

• (i)  
  The neutronics heat load can change significantly depending on the proton beam power.
• (ii)  
  Beam trips will also result in large changes in heat load for short durations.
• (iii)  
  The ESS linac will be commissioned in phases, with proton beam power increasing incrementally as additional cryomodules are installed, over several years.

Table 40 outlines the required cooling capacity of the TMCP with no beam, initial beam commissioning, and at each phase of accelerator commissioning.

The TMCP minimum allowable helium temperature is set considering the H2 freezing point of 13.8 K. A margin of 1.2 K is added, setting the minimum TMCP supply temperature at 15 K. The maximum temperature is determined considering the maximum CMS hydrogen temperature of 20.5 K. Therefore, the design of the TMCP is based on a heat load range of approximately 4–32 kW, and operating temperature range of 15–20 K.

TMCP operation. The TMCP accommodates ESS overall operating modes and the CMS requirements by defining three sets of operational modes:

• (i)  
  Steady state modes—Nominal design, nominal low power, and nominal turndown.
• (ii)  
  Transient operating modes—Cool down and warm up.
• (iii)  
  Switching modes—Long term switch from nominal to turndown, long term switch from turndown to nominal, and beam trip.

Accommodating the large variation in heat load and narrow operating temperature range of 15–20 K requires a sophisticated control scheme to operate efficiently. The proposed TMCP process design features two parallel warm compressors and two parallel turbo-expanders to accommodate the high load variation, as well as using the so-called floating pressure cycle [116] for a significant portion of the operating range.

The TICP primarily provides refrigeration for the cryomodule test stand in a closed loop [117]. Its secondary function is to provide liquid helium for the neutron science instruments in an open loop. The TICP is the smallest of all three cryoplants. It is essentially a customised standard plant, twenty times smaller than the ACCP [118]. The sub-atmospheric pressure (SP), required for the test stand operation temperature of 2 K, is created by means of a warm process vacuum pump station, comprising a set of two parallel roots pumps and rotary vane pumps. The flow is too small for efficient cold compression.

Like the ACCP, the TICP delivers supercritical 4.5 K helium for the 2 K circuit, 40 K refrigeration for TS cooling and helium liquefaction for the coupler cooling. With very low 2 K heat loads, cold sub-atmospheric compression is not feasible. Instead, the returning vapour flow is heated up and compressed in process vacuum pumps to the low cycle pressure. Hence for the TICP the supercritical 4.5 K helium flow translates to constant level liquefaction of about 4 g s⁻¹. Furthermore, the TICP provides refrigeration of about 390 W between 33 and 53 K for TS cooling. The TICP size is not impacted by the static load on the transfer line, interconnecting TICP coldbox and cryomodule test stand, apart from the TS. In fact the load on the transfer line is allowed to be quite high and is only limited by the operational restrictions of the economiser in the cryomodule.

The test stand load translates into energetically equivalent rising level liquefaction of about 130 l h⁻¹. This is considerably more than required later, when serving the liquid helium demands of the neutron instruments and their sample environments. When operating to supply the cryomodule test stand, the characteristic plant load is ideal for liquid nitrogen pre-cooling. This reduces the plant size significantly and halves the electrical power and cooling water consumption. For the normal TICP liquefaction operation in an open loop, there is the choice to pre-cool with liquid nitrogen or not, depending on the specific needs at the time of operation. A variable frequency drive for the recycle compressor helps to cope with the different load scenarios and use the plant most efficiently.

One of the worst case scenarios in view of the cryogenic system at ESS is a long power shut down, causing the helium in the cryomodules to evaporate due to the static heat load. In this case the evaporating helium flows eventually into the main LP line of the ACCP that is connected to the equivalent main LP line of the TICP. The TICP recycle compressor, that receives power from a backup system, is able to compress the evaporating helium and shift it to the warm medium pressure storage tanks, avoiding a major loss of helium.

Apart from the three cryoplants and the vast cryodistribution system, the CDS comprises auxiliary equipment including an external, stand-alone purifier, an impure helium recovery system, warm high and medium pressure gas storage, liquid helium and nitrogen gas storage.

The cryogenic system contains substantial amounts of helium. During nominal operation about 3 Tm of cold helium circulates between the cryoplants and their loads, not counting the helium in the 20 m³ s⁻¹ fill liquid helium storage tank attached to the ACCP [118]. Up to 19 tanks, each with a geometrical capacity of 70 m³, store the helium in warm condition at about 20 bar.

About two thirds of this helium is contained in the cryomodules and distribution system of the cold linac Before the pump-down to 2 K, when being filled with liquid helium, the linac and cryodistribution systems require about 130 kg more helium than during nominal operation. This is due to the large 2 K vapour return line, which is filled with 25 times denser vapour at 4.5 K compared to the gas at nominal conditions. The density increase of the liquid helium during pump-down has a slightly lower impact and affects only the cryomodules, not the CDS.

A significant portion of the helium is deposited in the long transfer lines connecting the TMCP and its load (the CDS) in the target building. The cryoplant loads and cools down about 350 kg of helium from warm to operating conditions [119].

When the TICP is serving the cryomodule test stand, it operates in a closed loop. The helium to fill the connected test cryomodule is stored in a warm buffer tank. The amount needed is insignificant compared to the amount required for the bigger cryoplants.

The expected monthly consumption of the neutron instruments, sample environments and related facilities is about 7500 l of liquid helium. Helium is liquefied by the TICP into a dedicated 5000 l liquid helium storage tank, transferred to mobile helium dewars and transported to the neutron science users, with three satellite gas collectors and gas bags in the neutron instrument halls. From there the helium returns warm at a more or less constant flow rate to the central helium recovery system where it is pressurised and stored in twelve 1 m³ high pressure storage vessels at 200 bar. Connected to the central helium recovery system are also the flash gas return connections form the mobile dewar filling station, the neck cooling discharge connections from the liquid helium storage tanks, the discharge connections of the safety relief valves protecting the sub-atmospheric circuits, discharge from helium quality detectors and diverse purge and regeneration connections from the cryoplants and helium purifiers.

Pressurised 'dirty' helium is mainly purified in the internal purifier of the TICP cold box when the liquefier is switched on. An external purifier is available when the TICP cold box is not running or helium has to be replenished for the closed loop systems. This purifier is a fully automatic stand-alone unit, based on liquid nitrogen cooled absorber beds and a drier connected upstream. The purified helium from this stand-alone purifier is distributed to the clean medium pressure storage tanks or can feed directly the ACCP.

Accelerator vacuum systems service the entire accelerator from the PS to the target station.

The front-end systems comprise the PS, LEBT, RFQ, the MEBT and the DTL. With the exception of the MEBT and DTL, these front-end systems handle the high hydrogen and nitrogen gas load needed for PS operation. As such, the vacuum pumping systems are sized to handle the gas loads needed for ionisation within the PS and exhaust from these pumps which must eliminate the potential for the formation of explosive mixtures. Mass flow controllers provide the independent flow of hydrogen and nitrogen into the PS with pressure measurements made using a capacitance manometer to eliminate errors due to changes in gas composition inherent with a Pirani style of gauge.

Injected gas pumping is provided by four TMP, nominally rated at 600 l s⁻¹, located in the LEBT. No pumping is provided in the PS. Two dry pumps back the four TMP's through a common manifold to provide redundancy. While the typical operating pressure of the PS and LEBT is in the 10⁻² Pa range, these systems are directly coupled to the RFQ, which operates at a pressure of less than 10⁻⁴ Pa to minimise beam-gas and RF interactions. An orifice 14 mm in diameter, located on the beam path, enables a pressure gradient from the LEBT to the entrance of the RFQ where up to ten 200 l s⁻¹ TMPs satisfy the pressure constraints at the entrance to the MEBT, where the pressure is 10⁻⁵ Pa or better. As with the LEBT, the TMPs are backed by two dry pumps connected to a common manifold to provide redundancy.

The MEBT operates in the range 10⁻⁵–10⁻⁶ Pa. It is pumped by a series of noble diode ion pumps, nominally 100 l s⁻¹, distributed along its length. Initial pump down uses a TMP directly connected to the MEBT beam pipe and backed by a dry pump. Vacuum measurements are made by a combination of Pirani and Penning vacuum gauges. Gate valves located at the outlet of the LEBT, RFQ and MEBT allow these sections to be isolated for vacuum pump down and maintenance. All metal sealing is used to the maximum extent possible and all systems are vented with dry nitrogen.

On exiting the MEBT, the beam enters the five DTL tanks, which also operate in the range 10⁻⁵–10⁻⁶ Pa. These tanks are 7.5 m in length and 0.5 m in diameter, fabricated from copper and driven by RF power entering each tank though ceramic windows. Each tank is pumped by three noble diode ion pumps rated at 300 l s⁻¹ distributed along the length of the tank, pumping through a screen that limits the RF power deposited into the ion pump to 4 mW. Initial pump down to the ion pump initiation pressure is accomplished from a common pumping manifold servicing all tanks, backed by dry pumps. This also provides the capability for DTL tank leak testing. Vacuum measurements are made using both Pirani and Penning vacuum gauges. Gate valves are located between each tank, except that space constraints force tanks 1 and 2 to share a common vacuum.

In-between the warm and cold sections of the linac is the LEDP unit, designed to protect and limit the flow of gas into the cold section of the accelerator, with a pressure reduction ratio of 100:1. Pumping is performed by three noble ion pumps distributed along the length of the LEDP. In addition, cold cavity protection is provided by a fast valve with a closure time of less than 10 ms. This valve closes on rising pressure and limits the flow of gas in an unplanned air in-rush event. Initial pump down is achieved using a mobile pump cart incorporating a TMP and dry backing pump. Again, Pirani and Penning vacuum gauges and all metal sealing are used to the maximum extent possible. The system is vented with dry nitrogen.

Downstream of the LEDP is the first spoke cavity. At full beam power of 5 MW, the cold section incorporate 26 spoke, 36 medium-β and up to 84 high-β cryogenically cooled elliptical cavities. These cavities are cryogenically pumped at between 2 and 4 K. It is imperative to limit the build up of the monolayers, so as not to degrade cavity performance. The limit is about 20 or 30 monolayers at the end of operations that extend over 30 years. The vacuum interface with the cryomodules must also take into account the limited capacity for intervention.

The cryomodules are also susceptible to particulate contamination at submicron levels that may degrade their performance due to field emission and will lead to cavity quenches. The cavities pump cryogenically when cold, and so do not require auxiliary pumping. Each cryomodule incorporates a metal seal gate valve at each end that remains open during operations, and is only closed during accelerator maintenance periods. Pressure within the cavity is measured by a means of a Penning gauge. A separate vacuum jacket is cryogenically pumped following initial pump down, providing part of the thermal insulation of each cryomodule. Active pumping is available to pump potential helium leaks into the jacket. This system is a TMP based system servicing a number of cryomodules.

LWUs as shown in figure 64 are located between cryomodules and at regular intervals from the last cryomodule to the PBW. Each LWU comprises a beam pipe with centrally located pumping system and accelerator beam diagnostics. Bellows at each end of the beampipe act as interface between cryomodules. The LWUs are cleaned and assembled under ISO class 3 conditions, and installation is performed in portable clean rooms providing ISO class 3 conditions. Each LWU is evacuated after installation, using a mobile pump cart before starting the low profile NEG-ion pump used in normal operation that provides an emergency egress route under the LWU beam pipe. The LWU is pumped to a pressure below 10⁻⁶ Pa before opening the gate valves and cooling the cryomodules to operating temperature. When cold, cavity pressures will be less than 10⁻⁸ Pa.

Zoom In Zoom Out Reset image size

At the exit from the last cryomodule the high energy differential pumping unit, like the LEDP, protects and limits the flow of gas into the cold section of the accelerator.

The major target vacuum system is in the monolith, which operates under either helium or a high vacuum environment. The PBW is eliminated In the vacuum mode, providing the beam with a clear path to the target wheel, with monolith pressures of less than 10⁻³ Pa. The monolith operates under high vacuum and incorporates both roughing and high vacuum systems. Condensing coils are used in conjunction with a turbo-molecular pump during initial pump down, to handle the high water vapour load that is anticipated.

The PBW is installed in the helium mode, isolating the monolith and HEBT vacuums. The PBW incorporates inflatable metal seals that mate with metal flanges in the accelerator beam pipe. The window is designed for routine remote replacement, and allows remote leak testing. A service valve located on the upstream side of the PBW allows vacuum isolation during PBW replacement.

The vacuum systems for the first instruments incorporate choppers, neutron guides and detector tanks. Both choppers and neutron guides operate in the 10⁻¹ Pa range, incorporating pumping systems using TMP.

The neutron filter materials used in some neutron guide tubes pose specific vacuum issues due to outgassing and, in some cases, undesirable constituents resulting in the need for preconditioning prior to installation. The Material Test Facility is used for material evaluation and to assist in the design and strategy needed during installation, commissioning and installation.

Detector tanks have volumes from around 1 m³ to several hundred m³, operating in the pressure range from 10⁻¹ Pa to 10⁻⁴ Pa, with some tanks requiring fast cycle times. This variety in design parameters leads to a spectrum of vacuum system designs. A central venting system provides a controlled environment to the various vacuum systems. A supply of clean, dry and filtered air prevents contamination that could degrade equipment, materials, etc, resulting in a loss of performance. Dry air enables fast system pump down, a critically important feature for instruments that requiring a fast cycle time.

Thin windows on the entrance and exit sides of the neutron guides and choppers are usually diaphragm plates, in which the deflection may be greater than half of the diaphragm thickness. These windows have limited exposure to cyclic loading (typically less than 1000 cycles) and are not generally subject to negative load cycling. The Thin Window Test Facility enables loading and unloading cycle tests, which are performed on each window assembly to ensure structural integrity prior to installation.

ESS is supplied by two transformers (130/24 kV, 63 MVA) and uses 130 kV switchgears. The transformers and switchgear are outdoors, installed within protective concrete walls that contain fire and explosion. The connection points to the ESS electrical grid are at the 24 kV incoming cubicles inside the Primary Substation building, as shown in figure 65 [121].

Zoom In Zoom Out Reset image size

Primary substation (H05) directly feeds the target station, instrument halls, and site offices, as well as the distribution substation (H06) which, in turn, supplies the cryoplants (G04), the accelerator (G02), the central utility building (H01) and houses the backup generators. The ground point for the 24 kV power systems is provided by zero-point transformers and grounding resistors (zero-point equipment), limiting the single-phase fault current to approximately 5 A. Under these conditions, the mandatory required separation between the MV cables and other services is reduced to 50 cm.

Secondary substations convert MV from the distribution substation to the voltage levels required by plant and equipment: 400, 600, 690 V and 6.6 kV. The secondary substations are equipped with four transformers, one MV and one LV switchgear, both including two separate groups of three phase bus bars that can be connected through transfer switches in case one incoming feeder trips. The installed power per substation is 40% more than the actual maximum load, allowing the substation to operate with one transformer off.

The 6.6 kV switchgear is of the same type and construction as the 24 kV systems, to simplify service, operation and maintenance of the HV equipment. The 400 V and 690 V switchgears comply with applicable standards for enclosure isolation class form 4 A/B, and are equipped with arc guard and over current protection.

Power grid status and event lists are monitored and operated by a supervisory control and data acquisition system. The SCADA system communicates this information to the ICS), enabling machine operators to monitor the status of the power supplies, and to assist them with diagnosing failures.

A cluster of redundant diesel generators, providing up to 3 MVA, is installed in the secondary distribution station to back up the 24 kV grid. This power is provisioned by the SCADA system. For specific, highly safety-critical purposes, local backup or UPS-solutions are required. Remote backup and UPS-power is foreseen for conventional safety-related systems, such as ventilation, fire, and security. Smaller amounts of UPS power, around 40 kVA, are provided by UPS systems installed in the low voltage switchgear room of the secondary substations. Small UPS systems are designed for five minutes of endurance, backed up by generators.

The grounding system satisfies several general requirements: electrical safety requirements; EMC/EMI compatibility; and practicality for the installers and users. It is based on a standardised meshed bonding network (MESH-BN) topology [122, 123]. The mesh includes all metallic civil elements (such as pillars, reinforcements) as parts of the protective bonding (PB). Galvanised strip steel conductors are installed and clamped to the reinforcement bars at regular intervals. Conductors entering the buildings are connected to the potential equalisation (PE) network at the penetration points. The PB and PE networks are bonded together and connected to physical earth at the lightning down-conductors, and the PE-bars, inside the electrical substations, as shown in figure 66.

Zoom In Zoom Out Reset image size

Additional grounding meshes are installed in the floors of the klystron gallery, the AT and the front end building (FEB), due to the expected level of EMI in these areas. This EMC-mesh spacing is sized to approximate a ground plane for the expected EMI from the RF-power supplies, resulting in a mesh size of 2 m by 2 m in the tunnel and the gallery, and 1 m by 1 m in the FEB. A 4 m by 4 m grid of floor-pads provides direct connections to the EMC-mesh in the klystron gallery and FEB. The EMC mesh is also connected to the other grounding systems [124].

Neutron instruments have different grounding requirements, and have adopted a star-ground topology that potentially allows better control of EMI, which requires a careful cable routing and reliable connections to the BN. This is managed by making of each instrument a ground zone with only one connection to the buildings PB-network.

Based on the requirements for flow, temperatures, operation pressure and total pressure drop in the system, an industrial process gas blower or turbo compressor has been identified as the two preferred machine types. The helium operation pressure is set to 3.5 bar with corresponding volume flow of up to 6 m³ s⁻¹, thus available blowers or compressors on the market is highly limited.

To remove 3 MW of heat and keep the tungsten temperature well below the limit of 500 °C, a flow of 2–5 kg s⁻¹ is needed. The removed heat power is given by the following equation:

$$\begin{eqnarray}&&P\ =\ Q\,{C}_{p}\,({T}_{{\rm{out}}}-{T}_{{\rm{in}}}),\end{eqnarray} \tag{ 7 }$$

where Q is the mass flow of helium, C_p is the heat capacity for helium, T_out temperature of helium leaving the target and T_in temperature of helium entering the target. If Q is set to 2 kg s⁻¹ and T_in = 40 °C, The outlet temperature is be too high to secure a tungsten temperature below the critical temperature. Whereas, if Q is set to 3 kg s⁻¹ and T_in = 40 °C, The outlet temperature is around 200 °C which is a good compromise. Note also that the inlet temperature cannot be much lower since it would not be cost effective.

The ESS moderators maximise the neutron brightness—the average flux per unit solid angle at the surface of the moderator—by using the distinctive and novel concept of low-dimensional moderators [129, 133]. Low-dimensional moderators increase the brightness by a factor of two or three by matching neutron cross-sections and mean free path (MFP) lengths in the liquid para-H2 moderator and pre-moderator to the moderator geometry.

Flat moderators are quasi-2D. The height of a flat moderator is reduced with respect to a conventional volume moderator, such as the coupled moderator used at J-PARC, ideally reducing it to a two-dimensional shape. Tube moderators are quasi-1D, removing one more dimension from flat moderators. Physics studies suggest that tube moderators are fundamentally the brightest moderators. Some examples of the application of tube moderators in an ILL-type reactor are discussed in [129]. This flat and tube classification applies to cryogenically cold moderators. However, ESS delivers bi-spectral neutron beams, also implementing a thermal neutron source. It is also advantageous to reduce the dimensions of the thermal moderator, since less beryllium reflector is removed for beam extraction, resulting in a lower brightness penalty.

The brightness distribution from the surface of a volume moderator filled with pure para-H2 is non-uniform [134]. Figure 76 shows the calculated brightness distribution over the emitting surface of a 10 cm tall moderator. Two areas of the moderator have increased brightness—the bottom (target) side of the moderator, and the top (reflector) side—because of the non-uniform source of neutrons, and because of the MFP difference of cold and thermal neutrons in para-H2.

Zoom In Zoom Out Reset image size

The MFP of thermal neutrons in para-H2 is about 1 cm, while the MFP of cold (5 meV) neutrons is about 11 cm. The thermal MFP of 1 cm causes the two hot spots in the volume moderator, at the target and reflector sides: thermal neutrons within about 1 cm of the moderator edges are slowed down to cold energies. The total cross section is reduced by one order of magnitude for cold neutrons in para-H2, making it easier to extract neutrons from the whole volume of the moderator.

A simple model based on this MFP difference reproduces quite well the angular emission of cold neutrons from a pure para-H2 disc-shaped moderator [134]. Thermal neutrons from the water pre-moderator are primarily slowed down to the cold regime by the first collision after entering the para-H2 close to the moderator wall, escaping the moderator volume as cold neutrons with a high probability of no further collision. In practice, the two hot spots become a single hotspot in a flat moderator, with a thickness comparable to the MFP.

Several other design criteria and requirements are incorporated in the flat moderator. The viewed surface of at least 3 cm × 6 cm (height by width) is seen everywhere in the two 120${}^{\circ }$ suites of beam ports. Bi-spectral extraction is possible from all beam ports.

Pre-moderator. Pre-moderators are in wide used at cold neutron sources, plays a crucial role for both cold and thermal moderators by shaping the neutron spectrum from the spallation target and reducing neutron energies from about 1 MeV to thermal values. An extended pre-moderator (water disk) with transverse dimensions practically equal to the beryllium reflector, and a thickness of about 3 cm, gives a substantial gain in both thermal and cold brightness.

Vertical position. Placing the moderators close to the neutron production hot spot is quite important for brightness optimisation, even though many of the neutrons reaching the moderator are scattered by the inner reflector. Proximity is important both vertically (the distance from target to moderator) and horizontally (in the plane parallel to the target), with a brightness slope of as much as 5% per cm in both directions. It is important to take into account that proximity to the neutron source is important but it implies the disadvantage of a larger energy deposition into the moderator that needs to be removed with additional cooling power.

Horizontal position. While the moderator closest to the hot spot is the brightest, bi-spectral extraction is needed for a significant number of instruments. The pancake design incorporates a butterfly thermal moderator above the hot spot, for two reasons:

• (i)  
  the thermal moderator is more compact than the para-H2 moderator, so that both moderators can be placed close enough;
• (ii)  
  the thermal moderator works as a pre-moderator for the cold moderator, compensating for the brightness lost by distancing the cold moderator.

Figure 77 shows the target-moderator-reflector assembly, with thermal layers of water forming X-shaped pre-moderators, placed near the hot spot, viewed by the full ${120}^{\circ }$ arc of each beam port suite. Two cold moderators are in the shape of a butterfly. The upper cold moderator is 3 cm high, while the lower, which is 6 cm high, is less bright but has a divergence spectrum that is more convenient for some instruments, such as imaging.

Zoom In Zoom Out Reset image size

Figure 78 shows how the cold and thermal brightnesses vary as a function of the moderator height. A 3 cm tall cold moderator is 2.5 times brighter than a 10 cm moderator, viewing $2^\circ \times {120}^{\circ }$ (equivalent to a 12 cm cold moderator viewing $2^\circ \times {60}^{\circ }$). This cold brightness is consistent with the performance of a pancake moderator, while the thermal brightness is a factor of two larger, probably due to the large amount of water on the sides of the tall butterfly moderator.

Zoom In Zoom Out Reset image size

Figure 79 shows the thermal and cold brightness performance of the butterfly 3 cm top moderator. Spectra are shown for a beam line placed at the maximum cold brightness position, at ${50}^{\circ }$ with respect to the proton beam. The thermal brightness is maximum at an angle of ${90}^{\circ }$ with respect to the proton beam. Calculations are performed with the model of figure 77. Cold and thermal brightnesses for the bottom moderator are factors of 1.60 and 1.45 lower than for the top moderator, respectively. The top moderator has a peak thermal brightness that is, on average, about 10 times brighter than the ILL time average thermal brightness [136]. The cold brightness at 4Å is more than 100 times the ILL brightness. Engineering reality reduces the brightness of the moderators from the MCNPX model values, with a penalty that is independent of the moderator height—both moderators are expected to suffer a loss of about 20%.

Zoom In Zoom Out Reset image size

The heat load in the moderator-reflector is critical, in particular for the cryogenic system with liquid hydrogen in the cold moderator vessel and pipes. Calculations with MCNPX generate a total heat load of 7.0 and 10.6 kW in the top and bottom cold moderators, respectively.

The average angular separation of the beam ports is ${6}^{\circ }$, alternating between the two values of ${5.7}^{\circ }$ and ${6.3}^{\circ }$. Ports are divided equally between two ${120}^{\circ }$ sectors: one is directed towards experimental hall number 1, and the other towards experimental halls 2 and 3. Ports in both sectors are in turn divided into two groups, in order to optimise the use of the butterfly moderators. The beam ports in each group diverge from one of four focal points that are located where the cold moderator butterfly wings meet the central, X-shaped, thermal pre-moderators, as shown in figure 81. This arrangement allows each instrument line to be fully cold, fully thermal, or bi-spectral.

Zoom In Zoom Out Reset image size

Temperature stability in the cold moderator design avoids changes in neutronic performance. Single phase operation avoids flow instability. Mechanical stresses are sustained under pressure, and during pressure pulses. Heat is removed into the cryogenic cooling system.

The hydrogen phase diagram includes a still undefined area, besides the usual solid, gas and liquid phases [137]. This mixed phase area in $(p,T)$ space, called supercritical because it exceeds the critical point defined by $({p}_{c},{T}_{c})$, contains two sub-areas:

• (i)  
  subcritical pressures $p\lt {p}_{c}$ with overcritical temperatures $T\gt {T}_{c}$
• (ii)  
  overcritical pressures $p\gt {p}_{c}$ with overcritical temperatures $T\gt {T}_{c}$.

Only the second case is truly supercritical, in the sense that no more phase changes are observed as the temperature is increased. Nonetheless, overcritical pressure operation can result in normal phase change phenomena such as blisters and film boiling, which must be avoided. A pseudo-critical (PS) line describes a kind of boiling curve that is interpreted as an extension of the normal boiling curve, leading to a pressure-dependent PS boiling temperature ${T}_{{\rm{PS}}}(p)$, with ${T}_{{\rm{PS}}}\approx 34$ K at p = 15 bar.

The high nominal operating pressure of 15 bar avoids pseudo-phase changes, and increases the low temperature range of the clear liquid phase, but increases the structural loads. The nominal temperature range, from 17 K at the inlet to 21 K at the outlet, is well below the PS boiling temperature. The mass flow derived from the total heat input $\sum {Q}_{i}$ and the temperature rise ${\rm{\Delta }}T$ is

$$\begin{eqnarray}&&{m}_{{\rm{flow}}}^{}\ =\ \displaystyle \frac{\sum {Q}_{i}}{{C}_{{\rm{pm}}}^{}{\rm{\Delta }}T}\ =\ 1000\,({\rm{g}}\,{{\rm{s}}}^{-1})\end{eqnarray} \tag{ 8 }$$

including static heat flows in the cooling circuit, pump et cetera.

The worst case model time structure of the neutron heat input follows the proton pulse on target, with a rectangular pulse of length of 2.86 ms arriving at a 14 Hz rate, resulting in a transient heat load that is 25 times the time-average load. The local heat load depends strongly on location, decreasing almost quadratically with increasing distance from the spallation centre. In addition, the local heat load depends on the material that acts as shielding between the point of interest and the source. In general higher density materials have increased interaction probabilities, receiving more heat than lower density materials. Figure 82 show the neutron heat load distribution that is used in simulation. Integrated over the volume of the four cold moderators, the total heat load adds up to $\sum {Q}_{i}=17$ kW.

Zoom In Zoom Out Reset image size

Figure 83 shows the fluid dynamic behaviour of the cold collimator system, over five beam pulses [137, 141]. The absence of an upward trend in temperature shows that there is sufficient time between beam pulses to remove the neutronic heat. Even the maximum temperature—in the aluminium on the outside of the moderator vessel—does not exceed the allowable critical temperature of 34 K. More important is the hydrogen-aluminium interface temperature, with a maximum temperature of about 30 K, well short of the criticality value. Pressure peaks reach up to a maximum of about 17 bar.

Zoom In Zoom Out Reset image size

Strength and fatigue analysis simulations according to RCC-MRx are ongoing at the time of writing, using the computation fluid dynamics simulation results as input, to ensure that the maximum allowable stress of 55 MPa is not exceeded [138]. Preliminary results show that the local stress exceeds this maximum only in small areas. This is mitigated by small changes in geometry, increasing the radius of curvature in these critical areas.

Manufacturing the flat cold moderator is particularly challenging. Forming the moderator from upper and lower halves, as shown in figure 84, enables the welds to be placed outside high stress areas, at locations that are accessible and testable. The halves are high-speed milled from two solid work pieces. Electron-beam welding filler grooves are 0.5 mm wide and 2 mm deep, with a tolerance of 0.05 mm. An undercut of 5 mm allows the weld seams to be placed out of the high stress area. The maximum clearance between parts is 0.1 mm during the welding process. Stiffness is maintained by following a particular milling sequence, in order to achieve this high precision, particularly during the undercut milling. Work pieces are prepared by face milling to decrease the inner tension equally on all sides, and to make them planar. Alignment holes are drilled, and then the 30 mm deep undercut is made with top down movement to reduce the forces on the remaining wall, only 3 mm wall thick.

Figure 84 shows how the surrounding material is kept for support during milling, and to be able to precisely mate the lower and the upper halves, using the fixture holes. The two pieces are connected after the upper and lower halves have been machined separately. The remaining material is milled down to 3 mm wall thickness, and the welding filler groove is made. The surrounding material is kept to guarantee stiffness during the welding process.

The 600 mm diameter thermal pre-moderator disc is shown in figure 85. Both the 30 mm high inner flow channel water cross and the 40 mm high vacuum surround are high-speed milled from a solid AL6061-T6 work piece. The operating pressure is only 5 bar, so the welding is less demanding than for the cold moderator. Through-welding the closing plate to the flow channels eliminates the need for any undercut milling. A sheet filler is used to meet the through-welding challenge, with good dimensional stability and faster installation, avoiding oxidation. An example of the weld quality and the penetration depth, using an AlSi12 filler material, is shown on the right of figure 85. The weld seam is slightly off-centre to the left, indication that the electron beam welding beam power could be higher to increase the weld thickness. Compliance of the tolerances is very important.

Transmutation of aluminium to silicon introduces significant hardening and embrittlement, and introduces additional SiMg₂ precipitation hardening, in addition to the displacement damage. Unlike 5-series aluminium alloys, the reduction of ductility of AL6061-T6 alloy saturates at high irradiation doses. Its total elongation is more than 4% at a thermal neutron dose of 10²³ neutrons cm⁻² that corresponds to about 1.5% of silicon yield in aluminium. This indicates that AL6061-T6 has a longer lifetime than 5-series alloys.

The integrated thermal neutron flux in the cold moderator is $2.4\times {10}^{22}$ neutrons cm⁻² for a year of 5 MW beam operation, leading to a maximum silicon yield rate of about 0.9% per year. The cold moderator should retain more than 5% of total elongation after a year, due to silicon embrittlement. Cold moderator lifetimes would be longer than a year if silicon embrittlement were the only significant mechanism. Unfortunately this is not the case.

Helium is mainly produced in aluminium by high energy neutrons and protons with kinetic energies greater than 1 MeV. Helium production may enhance both irradiation hardening and embrittlement effects, although information and experience with AL6061-T6 in this high energy regime is lacking. Instead, PIE data of the aluminium alloy AlMg₃ that was used as the material for safety-hulls of the SINQ Target-III is used as a guideline in determining the lifetime of AL6061-T6 due to helium embrittlement [143]. Tensile test results on the SINQ window, which was exposed to 6.8 A h⁻¹ of 570 MeV protons and the resulting spallation neutron irradiation, indicate strong irradiation hardening and embrittlement effects. The total elongation decreased from 22% in non-irradiated condition to about 8% after the maximum dose of 3.6 dpa, with further decreases expected with increasing dose. The reduction of the uniform and total elongations of the AlMg${}_{3}^{}$ specimens at doses above 2.7 dpa is due to high density helium bubbles at grain boundaries, with a maximum helium yield of 1125 appm for the total proton charge of 6.8 A h⁻¹.

The SINQ Target-9 has been irradiated with a total proton charge of 13.2 A h⁻¹ during 2011 and 2012, without any structural failure in its safety-hulls. Assuming the same beam footprint, the Target-9 window received a dose of 2200 appm helium and 7 dpa radiation damage. The maximum helium yield in the ESS thermal moderator is 310 appm per 5 MW year, far below this value. Thus, helium bubble induced material degradation seems relatively unimportant in determining the moderator-reflector assembly lifetime.

The pre-moderator, next to the target, suffers a maximum displacement damage rate of 29 dpa per 5 MW year, due to the fast neutron flux ($E\gt 0.1$ MeV). Operating experience at the high flux isotope reactor at ORNL shows that AL6061-T6 is able to withstand up to 40 dpa of irradiation, indicating that the pre-moderator is replaced every 7000 h. Displacements and point defects in aluminium self-anneal at a moderately elevated temperature, so there is room for further relaxation of the dpa-based lifetime limit. A long-term materials research programme will identify the displacement damage levels that limit the pre-moderator lifetime.

Current knowledge of 5 and 6 series aluminium alloy material properties suggests that the lifetime of the moderator vessels is limited by the displacement rate of 29 dpa yr⁻¹ at the pre-moderator vessel, which is predominantly exposed to the fast neutron flux. After taking a safety margin, the lifetime of the moderator system is determined to be one 5 MW year, at least for initial operations.

Helium bubbles produced in beryllium via (n, 2n) and (n, α) reactions may lead to swelling. Swelling increases monotonically with dose, bounded above by an estimate of 1% for beryllium containing 5000 appm of helium, after irradiation at temperature between room temperature and 350 °C. The maximum helium production rate in the beryllium reflector is 520 appm per 5 MW year, corresponding to a swelling ratio of less than 0.1% per year when the beryllium volume is water-cooled. The beryllium is not expected to swell enough to cause a structural failure of the reflector vessel, during the initially set 5 MW year moderator-reflector lifetime.

Neutron irradiation may lead to a decrease in thermal conductivity of beryllium. The maximum fluence in the beryllium reflector is about $0.2\times {10}^{22}$ n cm⁻² on one 5 MW year. The thermal conductivity of beryllium is conservatively estimated to be about 50% of the unirradiated value after a year of 5 MW beam operation, increasing the operational temperature with time, with an accelerated the swelling rate.

The exothermic beryllium-steam reaction could occur in a LOCA in the reflector. This is a safety concern, as it produces toxic beryllium oxide, flammable hydrogen, and heat. Studies of the chemical reactivity of irradiated beryllium and steam indicate that hydrogen generation is enhanced at a temperature of 700 °C, in a test in which the specimen experienced a temperature excursion above 1000 °C. The temperature in the beryllium reflector must be kept below 350 °C during nominal operations, to avoid a swelling rate larger than 0.1% per 5 MW year. In accidents, the temperature in the reflector must be kept below 600 °C, to avoid accelerating the beryllium-steam reaction.

The most concrete and novel concept in the green field construction is that the cells have no glass windows to look through from the technical galleries. Windows generally are a weakness in the static barrier of the civil construction, both in shielding and also structurally, especially during construction. Windows would require significant maintenance during the 60-year design lifetime of the ACF.

Window removal enhances operational flexibility, centralised operation and cell utilisation. Operational flexibility is improved by removing the constraint that operations must be planned near windows through which operators have limited visibility, with through-wall manipulators placed close by. Operations made via cameras and sensors are instead performed anywhere in the cell, using master-slave arms that are operated by motors instead of by classical manual force, and which are placed where needed. The design and layout of the facility is based on the logistical path for waste treatment, packaging, interim storage and shipment.

Spent monolith components are inserted in shielded casks with floor valves that keep the components encapsulated, before transportation. Figure 89 shows the target wheel cask, as an example. Component are hoisted by an internal lifting device within the cask, and the whole assembly is then transported by the high bay crane. The maximum capacity is 95 tonnes, using two 50 tonne cranes in a single crane mode. Different casks are docked on the roof of the ACF, on top of a floor valve that either has an interface to the process cell or the maintenance cell, or has direct access to a storage pit. This storage pit is deep enough to accommodate the tallest components from the monolith without any cutting operations. The storage pit is used if a component cannot be immediately lowered into the process cell.

Zoom In Zoom Out Reset image size

Inside the ACF, a component is moved from the cask to the process cell where it is dismantled and reduced in size. Waste is then packed into wastebaskets that are decontaminated and moved, first to the maintenance cell, and eventually into a storage pit for interim storage until offsite shipment criteria are fulfilled. Wastebaskets are shipped offsite via the transfer hatch, through an arrangement for waste disposal within the Swedish radioactive waste system. The waste is dry and metallic, so the main path is to use a standard tank and shipment container system. Figure 90 illustrates the step-by-step logistics:

• (1)  
  An internal transport cask is docked on top of the ACF.
• (2)  
  A monolith component is loaded from the cask, through a floor valve and into the process cell of the ACF.
• (3)  
  The component is dismantled in accordance with a pre-defined procedure, and the pieces are loaded into a wastebasket.
• (4)  
  The wastebasket is decontaminated in a storage pit to mitigate unnecessary contamination of surfaces in the facility and shipment containers.
• (5)  
  The full wastebasket is stored in a storage pit until further transport is possible, depending on public road transport regulations.
• (6)  
  Wastebaskets are lifted from the storage pit, through the transfer hatch and into a shipping container.
• (7)  
  When full, the shipping container is rolled out from the loading position, through an airlock door and into the transfer hall.
• (8)  
  The high bay crane picks up the container and loads it onto a transport truck that then drives the container to its offsite location.

Zoom In Zoom Out Reset image size