1. Dave McGinnis Chief Engineer – Accelerator Division
The ESS Linac
Dave McGinnis
Chief Engineer – Accelerator Division
15 October 2014

2. The ESS Linac
The European Spallation Source (ESS) will house the most powerful proton linac ever built.
Average beam power of 5 MW.
Peak beam power of 125 MW
Acceleration to 2 GeV
Peak proton beam current of 62.5 mA
Pulse length of 2.86 ms at a rate of 14 Hz (4% duty factor)
97% of the acceleration is provided by superconducting cavities.
The linac will require over 150 individual high power RF sources
with 80% of the RF power sources requiring over 1.1 MW of peak RF power
We expect to spend over 200 M€ on the RF system alone

3. ESS Schedule Full funding and ground-break in Fall 2014
1.25 MW of proton beam power by 2019
5 MW of proton beam power by 2022

4. Neutron Spallation Sources
Traditional neutron sources are reactor based
Neutron flux is limited by reactor cooling
Neutron energy spectrum is measured by time of flight using neutron choppers
Chopping throws away neutrons and limits neutron brightness
Spallation sources consist of a:
pulsed accelerator that shoots protons into:
a metal target to produce the neutrons
The pulsed nature of the accelerator makes the neutron brightness
much higher for a spallation source
for the same average neutron flux as a reactor
The accelerator complex of atypical spallation sources consist of a:
Linac to accelerator the protons
A storage ring to compress the linac beam pulse
Linac
Target and Instruments
Storage Ring

5. What is Different About ESS?
The average proton beam power will be 5 MW
Average neutron flux is proportional to average beam power
5 MW is five times greater than SNS beam power
The total proton energy per pulse will be 360 kJ
Beam brightness (neutrons per pulse) is proportional to total proton energy per pulse
360 kJ is over 20 times greater than SNS total proton energy per pulse

6. Short Pulse Neutron Spallation Sources
The neutrons are cooled by a moderator downstream of the target
The time constant of the moderation process is about 100 ms
Proton beam pulses shorter than 100 us serve only to stress the metal target and limit the beam power
Typical short pulse spallation sources have storage ring circumferences ~300 meters which produce 1 ms beam pulses
To build a storage ring with a 100 ms pulse would require a ring 30 km in circumference
The target stress from the short beam pulse places a limit on:
proton beam power
and ultimately neutron flux and brightness
The proton beam power of SNS (Oak Ridge Tennessee, USA) is limited to 1MW (17 MW peak)

7. Long Pulse Concept
360 kJ packed into a short pulse of 1 ms (360 GW peak) would destroy a target
ESS will not use a compressor ring
The linac will send the beam directly to the target over a period of 3 ms at a rate of 14 Hz.
Peak beam power on the target is less than 125 MW
The tradeoff is that ESS will
Have longer neutron guides between experiments and the target
Require a neutron choppers for precision energy measurements

8. The Long Pulse Concept Advantage - No compressor ring required
No space charge tune shift so peak beam current can be supplied at almost any energy
Relaxed constraints on beam emittance
This is especially true if the beam expansion system for the target is based on raster scanning of the beam on the target.
No H- and associated intra-beam stripping losses
Permits the implementation of target raster scanning
Disadvantage - Experiment requirements “imprint” Linac pulse structure
Duty factor is large for a copper linac
Duty factor is small for a superconducting linac

9. ESS Linac Evolution

10. Cost Targets
Although the 2008 design with 150 mA of beam current has higher technical risk, it has an inherently lower construction cost than the October 2012 baseline.
Large fraction of the 2008 linac consists of normal conducting structures which are significantly less expensive to build than superconducting structures
Lower energy (but higher beam current) requires a significantly shorter linac with less accelerating structures
However the current cost targets are based on the 2008 design even though the October 2012 design:
Has many more superconducting structures
But provides less technical risk
The only way to close the gap between the cost estimate and cost target is
to modify the October 2012 baseline by adding technical risk
or increasing the cost target

11. Cost Reduction Strategy
Keep constant
𝑃 𝑏 = 𝑃 𝑏𝑝𝑘 𝐷= 𝑃 𝑏𝑝𝑘 𝑓 𝑟 𝜏 𝑝
Reduce
𝑃 𝑏𝑝𝑘 = 𝐼 𝑏 𝐸 𝑝𝑘 𝑛=1 𝑁 𝑀 𝑐𝑒𝑙𝑙 𝐸 𝑎𝑐𝑐 𝑇 𝐸 𝑝𝑘 𝛽 𝑔 𝜆 2 cos⁡ 𝜙 𝑠 𝑛 + ℰ 𝐹𝐸 𝑞
The cost of the elliptical cryomodules and associated RF systems are the largest cost driver in the ESS Linac
Reducing the number of superconducting cavities will have the largest impact on cost and design contingency
each cavity that is removed from the design not only removes the cost of the cavity
but also removes the need (and cost) for the RF power sources that feed the cavity.
Therefore, the design contingency strategy will hold the average beam power constant while looking for avenues to minimize the number of superconducting cavities.

12. Cost Reduction Strategies
Keep constant
𝑃 = 𝑃 𝑝𝑘 𝐷= 𝑃 𝑝𝑘 𝑓 𝑟 𝜏 𝑝
Reduce
𝑃 𝑝𝑘 = 𝐼 𝑏 𝐸 𝑝𝑘 𝑛=1 𝑁 𝑀 𝑐𝑒𝑙𝑙 𝐸 𝑎𝑐𝑐 𝑇 𝐸 𝑝𝑘 𝛽 𝑔 𝜆 2 𝑛 + ℰ 𝐹𝐸 𝑞
Increase
duty factor, D
peak surface field, Epk
peak beam current, Ib
average value of EaccT sum by adjusting the power profile
ratio of EaccT/Epk by appropriate choice of bg
energy of the front end linac, EFE

13. Increasing The Duty Factor
The choice of a superconducting linac becomes obvious as the duty factor increases.
From an accelerator design point of view, increasing the duty factor has the least impact on the configuration of the accelerator.
As the duty factor is increased
by either increasing the pulse length or the repetition rate,
the final energy of the linac can be decreased and still provide the same average beam power.
However, increasing the duty factor will reduce the peak neutron flux

14. New Baseline New Baseline Headline Parameters
5 MW Linac
2.0 GeV Energy (30 elliptical cryomodules)
62.5 mA beam current
4% duty factor (2.86 mS pulse length, 14 Hz)
First beam by 2019 (1.0 MW at 570 MeV)
The new baseline was achieved by:
Increasing beam current by 25%
Increasing Peak Surface Field by 12%
Setting High Beta bg to 0.86
Adopting maximum voltage profile
Adopting a uniform lattice cell length in the elliptical section to permit
design flexibility
schedule flexibility.

15. Lattice Cell Length
For the October 2012 baseline design, the cell length along the linac changes substantially.
4.18 meters in the spokes,
7.12 meters in the medium beta section with one cryomodule per cell
15.19 meters in the high beta section with two cryomodules per cell.
For a maximized voltage profile, a high beta bg=0.86, and an Ib=62mA,
over half the medium beta cryomodules are eliminated
the beginning of the high beta region is now 520 MeV
At this energy, the current long high beta cells is too weak at to provide the desired phase advance per cell of 87 degrees with reasonable gradients in the quadrupoles.
Thus a fourth type of cell with one high beta cryomodule per cell would be needed in this region.

16. Uniform Lattice Cell Length
A tunnel design with many different cell lengths is very undesirable with the perspective of considering:
design contingency
future upgrades.
In the future, it might be advantageous to interchange
spoke cryomodules with medium beta cryomodules.
medium beta cryomodules with high beta cryomodules.
At the added expense of a longer linac, the new baseline has:
Spoke cell Length = 0.5 x Medium beta cell length
Medium beta cell length = High beta cell length
A uniform cell length provides the possibility that the medium and high beta cryomodules could be interchangeable and possibly identical.
6 cell medium beta cavities that would be close to the same length of the high beta cavities.
This would reduce the prototyping schedule (and cost) significantly because only one type cryomodule prototype would need to be constructed.
Also a 6 cell medium beta cryomodule requires one less high beta cryomodule to achieve 5 MW of beam power

17. 6 Cell Medium Beta Cavities
For a uniform lattice cell length,
the current 5 cell medium beta cavities need a drift of 0.2 meters after each cavity
Might require a specialized port on the cryomodule to access the tuner package for both species of geometrical beta
If 6 cell medium beta cavities (bg=0.67) are used,
the extra drift is reduced to 0.06 meters
One less high beta cryomodule required

18. New Baseline Layout
Done by a professional: M. Eshraqi

19. New Baseline Lattice

20. Design Risk Reduced the number of elliptical cryomodules from 45 to 30
Each cryomodule + RF to power the cryomodule costs ~6.5 M€
Elimination of 15 cryomodules yields 78 M€ savings (6.5 M€ x 15 x 80% (power factor) )
By accepting large technical risk
Power Couplers:
Maximum coupler power is 1200 kW
Went from 850 kW/coupler to 1100 kW/coupler
Reduced our design margin by 70%
Cavity Peak Surface Field
Maximum surface field is 50 MV/meter
Went from 40 MV/meter to 45 MV/meter
Reduced our design margin by 50%

21. Design Contingency ESS uses the Long Pulse concept
No compressor ring is required
Peak beam current can be supplied at almost any energy
If we fail to meet our goals on:
Beam current
Cavity gradient
Power coupler power
The accelerator complex will still function but at a reduced beam power
With the exception of reduced gradient on the spokes (energy acceptance in medium betas)
We can buy back the beam power in the future by adding high beta cryomodules to the end of the linac
As long as the additional space is reserved.
We proposed to mitigate these risks by reserving the tunnel space for 15 cryomodules (127.5 meters) as “design contingency”.

22. Conventional Facility Costs
The approximate costs for conventional facilities are:
Tunnel: 22,900 €/m (3270 k€ / m2) including berm, auxiliary costs
Gallery: 46,200 €/m (2800 k€ / m2)
The cost of accelerator equipment is:
6.5 M€ / cryomodule which includes the RF power
Average cost of superconducting RF accelerator equipment is:
790,000 €/m
35x more expensive than tunnel cost
11.4x more expensive than total CF cost
Average beam power cost for the accelerator equipment in a cryomodule cell is 18kW / M€.
The cost of the 127 meter contingency space without stubs and gallery is 2.9 M€
Equivalent to the cost of accelerator equipment needed to supply MW of average beam power (1% of 5 MW)

23. Linac Design Choices
User facilities demand high availability (>95%)
ESS will limit the peak beam current below 65 mA
Linac Energy > 2 GeV to accomplish 125 MW peak power.
The linac will be mostly (>97%) superconducting
Front end frequency is 352 MHz (CERN Standard)
High energy section is at 704 MHz

24. Front End Section
The RFQ and DTL will be similar to the CERN Linac 4 design.
The RFQ
will be 4.5 meters long
and reach an energy of 3.6 MeV
The DTL
Will consist of five tanks
Each tank ~7.5 meters in length
Final energy will be 88 MeV
Six klystrons
at 352 MHz
with a maximum saturated power of 2.8 MW
and a duty factor of 4% are required for the Front End

25. Spoke Cavities
ESS will transition to superconducting cavities at 88 MeV
ESS will be the first accelerator to use 352 MHz double spoke cavity resonators
Twenty-eight cavities with an accelerating gradient of 9 MV/m are required.
Each cavity will operate at a nominal peak power of 320 kW
What type of power source to choose?
Tetrode
Klystron
IOT
Solid State

26. Elliptical Cavities Universal Cryomodule Medium Beta bg = 0.67
Cryomodules are expensive and difficult to fabricate
Pick cavity bg and number of cells
Optimize power transfer
Optimize length
Power in couplers is limited to 1200 MW (peak)
Medium Beta bg = 0.67
6 cell cavities
Cavity length = 0.86 m
32 cavities packaged in 8 cryomodules
Maximum peak RF power = 800kW
High Beta bg = = 0. 86
5 cell cavities
Cavity length = 0.92 m
88 cavities packaged in 22 cryomodules
Maximum peak RF power = 1100kW

27. Lorentz De-tuning
Because of the enormous gradients in superconducting cavities,
the radiation pressure deforms the cavities
We expect over 400 Hz of detuning in the ESS cavities.
Unloaded cavity bandwidth = 0.07 Hz
Loaded cavity bandwidth = 1 kHz
The mechanical time constant of the cavities is about 1 ms compared to the pulse length of 3 ms
Static pre-detuning as done in SNS will not be sufficient
Dynamic de-tuning compensation using piezo-electric tuners is a must!
Or else pay for the extra RF power required

28. Cavity Power Configuration
Because of fabrication techniques,
superconducting cavity strings are usually much shorter (< 1 m) than copper cavity strings (> 5m).
The Lorentz de-tuning coefficient varies from cavity to cavity
Therefore, each superconducting cavity has its own RF power source

29. Transit Time Factor For proton linacs using copper RF cavities
the cavity cell structure is tuned to match the changing proton velocity as it accelerates.
The power profile is usually flat
Because of high fabrication costs and difficulty,
The cell structure of superconducting cavities is tuned for only one beam velocity.
Multiple families of cell velocities are chosen. ESS cell velocities:
Spoke: bg = 0.5
Medium beta: bg = 0.67
High beta: bg = = 0. 86
There is a limit on the surface field in a SCRF cavity (ESS 45 MV/m)
Since, the particle velocity does not match the geometrical velocity for the entire acceleration range,
The power profile is not flat

30. ESS Linac Cavity Power Profile
High Beta
Medium Beta
Spoke
Transitions

31. ESS Linac RF System

32. Spoke linac (352 MHz) RF System Layout
26 Double Spoke cavities
Power range kW
Combination of two tetrodes
Other options:
Solid State Amplifiers
Large power supply (330 kVA) to supply 8 stations (16 tetrodes)

33. Elliptical (704 MHz) RF System Layout
One cavity per klystron
4 klystrons per modulator
16 klystrons per tunnel penetration
Tunnel

34. Elliptical (704 MHz) RF System Layout
Klystrons
Modulator
Racks and Controls
WR1150 Distribution
4.5 Cells of 8 klystrons for Medium Beta
10,5 Cells of 8 klystrons (IOTs) for High Beta

35. Summary