1. The European XFEL Experience and Lessons Learned
(Being Learned)
Nick Walker, DESY
On behalf of many talented and dedicated people
LCWS 2017 – Strasbourg –

2. The XFEL experience – with ILC in mind1
XFEL is now delivering beam to first users 👍
This talk will focus on the 1.3GHz SRF 17.5 GeV linac
a prototype of the ILC 😉
Focus on
Production – from cavities, to modules, to accelerator
all important Mega Volts per Metre
Recent commissioning and operational experience
Many lessons learnt!
Mostly positive
This will be just the highlights
slide 1

3. The European XFEL Built by Research Institutes from 12 European Nations
Some specifications
Photon energy keV
Pulse duration ~ fs
Pulse energy few mJ
Superconducting linac 17.5 GeV
10 Hz ( b/s)
5 beam lines / 10 instruments
Start version with 3 beam lines and 6 instruments
Several extensions possible:
More undulators
More instruments
…….
Variable polarization
Self-Seeding
CW operation
SASE2
(= SASE1)
17.5 GeV
SASE1, lu= 40 mm
0.2 – 0.05 nm
SASE3, lu= 68 mm
1.7 – 0.4 nm

4. Accelerator Overview Injector L1 L2 L3 n.c.Gun 6 MeV LH, Dogleg, BC0
1.3 GHz module ≈ 150 MeV 3.9 GHz 3rd harm. ≈ 25 MeV L1
4 modules (1 RF station) L2
12 modules (3 RF stations) L3
80 modules (20 RF stations)
SASE2
300 kW
1 Module A2 A3 A4 A5 A6
A25 A1
AH1
SASE1
0.24 kW
5 kW
SASE3
9 kW
300 kW
n.c.Gun
1.3 GHz
6 MeV
LH, Dogleg, BC0
130 MeV
BC1
600 MeV
BC2
2.4 GeV
Collimation
6 to 17.5 GeV
300 kW
0 m
40 m
240 m
470 m
1460 m
2130 m
2440 m
3100 m
3300 m
s.c. linac with GHz superconducting modules + 1 third harmonic module
design gradient: 23.6 MV/m; pulsed with 1.4ms pulse length; 600 µs flat top
4 modules / 32 s.c. cavities are connected to one 10 MW klystron (“RF station”)
12 modules form a cryogenic string
Down to app. 50m behind the last module the complete beam vacuum is “particle free”
Slide 3

5. Parameters: XFEL and 500 GeV ILC compared4
XFEL
ILC
max electron beam energy
GeV
17.5
250
average Eacc
MV/m
23.6
31.5
macro pulse repetition rate Hz 10
5 (10†)
RF pulse length ms
1.4
1.6
# of bunches per pulse
2700
1312 (2625‡)
beam current mA 5
5 (9‡)
max. bunch charge nC
1.0
3.2
electron bunch length mm
600
300
norm. emittance (x / y) nm
400 / 400
500 / 30
average beam power MW
0.5
5 (10‡)
cavities per 10 MW klystron 32
39 (26‡)
Slide 4
† 10-Hz e+ production mode / low energy running
‡ luminosity upgrade

6. A 6% Production Prototype5
XFEL
TDR ILC
Number of cryomodules
100
~1800
Number of cavities
800
~17000
Number of Klystrons/Modulators/LLRF 26
~600
ILC-relevant lessons learnt across the board:
Industrial production of cavities
String assembly by supervised industrial labour
High production rate testing and QA/QC
Installation
Commissioning and beam operation
And of course COST $ (and managing IKC)

7. From Niobium to Functioning Linac6
Niobium material
Cavity fabrication
Cavity testing
Module assembly
Module testing
Installation
Operation
Performance Results
Performance Results
Performance Results
In addition:
Cold mass / insulation
HPC
Quadrupoles
BPMs
...
In addition:
Klystrons
Modulators
Pulse transformers
Vacuum
LLRF (controls)
Interlocks
...
Cryogenics!

8. How to mass produce 100 cryomodules7
Slide 7: Only to stress life is complicated in the IKC world!

9. Cavity Production (EZ, RI)8
Entirely produced by industry and delivered “ready to go”
Slide 8-1

10. Cavity Production (EZ, RI)8
Entirely produced by industry and delivered “ready to go”
Slide 8-2
Lesson learned #1: Yes you can do this and it worked really well

11. Cavity Production (EZ, RI)8
Entirely produced by industry and delivered “ready to go”
Slide 8-3
Lesson learned #1: Yes you can do this and it worked really well
Lesson learned #2: Be prepared to invest a lot of effort into making it work

12. Vertical tests at AMTF

13. As Received Maximum Gradient in the VT9
typical individual error: 10%
Slide 9

14. As Received Usable Gradient in the VT10
typical individual error: 10%
XFEL retreated
ILC
Include operations spec
Q0 ≥ 1×1010
FE threshold (X-ray)
 Usable Gradient
Slide 10: Slide Although average is excellent, large performance spread is indicative of a process that is not well enough controlled -> Important for ILC -> R&D
Cavities below 20MV/m mostly successfully recovered by additional High Pressure Rinse

15. Extrapolation to ILC – how close are we?11
see talk by D. Kostin “European-XFEL summary: Cavities/Modules performance” in SRF session, Tue 24th AM
No direct ‘correct’ comparison possible
Cut off for XFEL retreatment ≤20 MV/m
ILC is ≤28 MV/m
Can try to use retreatment MC model based in XFEL results ✔
Slide 11
More retreatments - but mostly only HPR
Number of average tests/cavity increases from 1.25 to 1.55 (1st+2nd) or 
20% over-production or additional retreat/test cycles

16. Cryomodule assembly at CEA Saclay
Opening up the vacuum is always a risk
Even in the clean room

17. The XFEL Village at IRFU / CEA Saclay12
Slide 12

18. Cryomodule Test at AMTF

19. Cryomodule production rates13
Slide 13

20. The First and the Last Module14
In total 96 modules in 103 working weeks
The initially projected rate was 1 acc. module per week.
Variation in coupler availability was compensated by additional efforts at CEA / Irfu wrt. assembly rate.
Gained experience with module testing was used to shorten test duration of module 40+ .
Slide 14

21. XFEL Module Gradient Performance15
Module performance well above XFEL specs. and visible improvement with time
Tunnel installation used sorting of modules based on AMTF performance
XFEL Spec
23.6 MV/m
see talk by D. Kostin “European-XFEL summary: Cavities/Modules performance” in SRF session, Tue 24th AM
vertical test (clipped at 31 MV/m)
module performance
Slide 15
Remarks:
Clipping at 31 MV/m is done due to max. available RF power; limit given by waveguide distribution.
XM98 is a scavenger module.
Ncavs
Average
RMS VT
815
28.3 MV/m
3.5 CM
27.5 MV/m
4.8

22. Lessons learned: Accelerator Modules at AMTF & WATF16
During the 2nd production year AMTF module testing was performed without any delay.
During the end of production the major non-conformity was overheating at the 70k coupler window; all respective warm coupler parts were exchanged.
Waveguide tailoring was done for all modules.
Successfully repaired modules were retested at AMTF when needed.
Not installed are
XM8 (leaky cryogenic line)
XM46 & XM50 (inacceptable cav. performance)
XM99 (leaky beam line; meanwhile repaired)
XM100 spare module & replaced by XM-2

23. Cryogenics is quite challenging17
Complexity of cryogenic system asked for sufficient commissioning time; experts had to establish / optimize operation and to gain experience with new machines, especially the used cold compressors.
How to deal with…
671 control valves
>3,800 sensors (temperature, pressure, flow, level)
433 regulation loops
>22,000 records and >220,000 properties
and last but not least … >300 tons of material to be cooled down
Required 2K pressure stability of 2% peak from LLRF requirements (cavity detuning)
Tedious adjustment of regulation loops
Inner-system heaters to counteract dynamic processes
Slide 17: Same as last slide, although this is interesting stuff. LHC is >>300 tons.

24. First Cooldown of XFEL Linac during Dec 201618
Start asymmetrical operation of two cold boxes to speed up cooldown
Fast cooldown at temperatures below liquid nitrogen (no more thermal stress)
Entry of cold return flows in cold boxes to enhance cryogenic capacity
No Cold Leaks!!!

25. XFEL RF System HV Modulators in surface hall19
CM1 (8 cav.)
CM2 (8 cav.)
CM3 (8 cav.)
CM4 (8cav.)
KLYSTRON
HV Modulators in surface hall
Connected to pulse transformer via up to 2km long pulse cables
Klystron and modulator
below module

26. L3 RF Stations on the Status Pannel20
in operation
dito. but shifted off beam
off
Slide 20
All RF stations including CS8 are commissioned at moderate gradients.
Operation automised and handed over by experts; energy goal for 2017/2018 reached.
Detailed measurements will show the path towards higher beam energies.
The last two stations (CS9) require still longer tunnel access.

27. Beamline Commissioning Progress21
0m m m m m m m L1 L2 L3
13/01*
130 MeV
600 MeV
* Beam permission on 13/01
600 MeV
2.5 GeV
27/04 Beam spot before dump
2.5 GeV
6 GeV
12 GeV
Slide 21
27/04*
* Beam permission on 26/04
keen on lasing…

28. SASE Operation First lasing (0.9 nm) reached on May 2nd/3rd.22
SASE spot on YAG screen
First lasing (0.9 nm) reached on May 2nd/3rd.
Commissioning of the photon beam diagnostics and transport was next.
Beam based alignment in the SASA1 undulator section followed. And gave good results.
First laser light at 2 Å on May 24th.
On May 27th we reached an energy of up to 1 mJ i.e. close to saturation.
Safety authorities handed out the operating permission for the SASE1 hutches on June 21st.
On June 23rd we lased at 1.5 Å .
GMD intensity signal (calibrated)
SASE spot on FEL imager

29. XFEL Energy Performance (to date)23 L1 L2 L3
21 GeV (AMTF max)
XFEL design 17.5 GeV
19 GeV (AMTF max, nom. BC)
18 GeV (as above w/o CS9)
16 GeV (current possible)
14 GeV (current operation)
slide 23-1
CS9 not commissioned
Operational spare

30. XFEL Energy Performance (to date)23 L1 L2 L3
21 GeV (AMTF max)
XFEL design 17.5 GeV
19 GeV (AMTF max, nom. BC)
18 GeV (as above w/o CS9)
16 GeV (current possible)
14 GeV (current operation)
Operational limits identified
(See SRF WG: M. Omet “Beam Commissioning for E-XFEL” Tue 24.10)
slide 23-2
CS9 not commissioned
Operational spare

31. XFEL Energy Performance (to date)23 L1 L2 L3
21 GeV (AMTF max)
XFEL design 17.5 GeV
19 GeV (AMTF max, nom. BC)
18 GeV (as above w/o CS9)
16 GeV (current possible)
14 GeV (current operation)
31.5 MV/m
Operational limits identified
(See SRF WG: M. Omet “Beam Commissioning for E-XFEL” Tue 24.10)
slide 23-3
CS9 not commissioned
Operational spare

32. Cavities: from VT to (current) operation24
ILC 35 MV/m
ILC 31.5 MV/m
ILC 31.5 MV/m
Single Cavity Vertical Test
Single cavity CW
<E> ~ 30 MV/m
<E> ~ 28 MV/m clipped at 31 MV/m
Cryomodule Test Single Cavity (pulsed)
<E> ~ 27 MV/m
Current XFEL operation (approximate)
<E> ~ 22 MV/m
Work in progress – expect to do better
slide 24
goal

33. Lessons Learnt – Summary 1 of 225
TESLA technology has been successfully industrialised and can be mass produced
No reasons why this cannot be extrapolated to ILC numbers
Success requires DILIGENCE (and attention to detail)
Close cooperation with industry
Constant feedback, QA and QC
Standard ‘TESLA’ recipe can almost achieve ILC specifications
But improvement still needed
30 MV/m average is great
But 7 MV/m RMS spread is too large (yield!)
Production-line string and module assembly by industry without performance degradation is possible
Again, requires DILIGENCE!
Auditing, QA/QC, feedback, etc.
Robots 🤖 !
slide 25
ILC will just need “more” of everything we did right for XFEL

34. Lessons Learnt – Summary 2 of 226
Tunnel installation rates successfully achieved
Need to check ILC TDR installation rates against this and revise
Early equipment commissioning (in tunnel) is mandatory
XFEL could have done with a little more or this
Better QA/QC would have helped during accelerator commissioning
ILC will require order(s) of magnitude more attention
Very good experience with cryo
Long contiguous ~1km cryo string with large dynamic load
Use of heaters etc – connection to LLRF
Quickly achieved 12 GeV and moving up
Slowly increasing to XFEL goal of 17.5 GeV
Hoping to eventually achieve ~26 MV/m average (currently at 22 MV/m)
ILC will need much more AUTOMATION to commissioning linacs in a finite time
Don’t expect max energy from day one. Or even day >300.
Carefully consider energy margins when making final design choices.
Push to high currents / long bunch trains to come
Controlling beam loss is the challenge!
slide 26

35. Guest Scientists during commissioning
General Assembly of the European XFEL Accelerator Consortium
Guest Scientists during commissioning
THANK YOU TO ALL CONTRIBUTORS TO THE
EUROPEAN XFEL