Highlights of The Structure Breakdown
Workshop
Held at SLAC, August 28-30. Roughly 40
people.
A web site has been made and scanned
copies of most of the transparencies are
available.
 Physics of breakdown and of damage.
 Conditioning experience.
 Experiments to determine what is
important and what can be achieved.
 Structure design.
 Fabrication, cleaning, surface analysis.
Physics of breakdown and
of damage
H. Padamsee
The origin of breakdown for DC is the
same as for RF and super-conductivity is
the same as warm copper… and it’s all
dust.
The super-conducting community has
made an extensive study of the location of
emission sites, corresponding damage sites
after breakdown and materials found in the
damage sites - there is always evidence of
impurities.
This has lead to a breakdown model of
breakdown.
 Dust particle gives enhanced field
emission.
 Dust particle heats a little and emits gas.
 Field emission current ionizes gas.
 RF fields sweep electrons away from
plasma but ions sit there creating a DC
potential.
 Ions move in the DC field - ion
bombardment begins really heating the
dust particle.
 More gas is emitted, the process runs
away and the dust particle explodes
leaving behind a crater.
The process has been simulated. The runaway takes just a few ns.
* Perhaps Higgs results could provide a
benchmark.
The process including damage is
entirely local and depends only on the
surface electric field. One (or a few)
pit(s) per breakdown event.
P. Wilson
Surfaces are eroded continuously by
'cathode spots'. These are well studied at
DC. Emitter sites ignite small selfsustaining
plasma
clouds.
Ion
bombardment melts copper and pressure
squirts copper away from the surface.
Clouds migrate around the surface.
An experimental signal is light.
The process including damage is
entirely local and depends only on the
surface electric field. Damage is not
correlated to breakdown events.
Conditioning experiences
W. Wuensch
Higgs, 11 GHz and 30 GHz accelerating
structures.
A. Vlieks
 Component testing.
 Bake in situ to 200 C.
 Start with narrow RF pulses!
 Set vacuum interlocks to low trip levels.
Processing is initially gas limited, later
not - this seems to be a general
feature. Perhaps glow-discharge could
accelerate the initial processing thus
reducing total damage.
 f changes during conditioning can help.
 Strong emphasis on diagnostics - RF, Xray intensity and energy and light.
 Components can breakdown-interact.
See comments below about couplers.
C. Adolphsen, R. Loewen
Tests in NLCTA
respectively.
and
in
ASTA
Tests of various constant gradient,
detuned, and DDS structures - result has
been disappointing gradients and damage
that has a noticeable effect on phase
advance.
Turning the structures around and 'reconditioning' gives less damage to low
group velocity cells and has lead to the
now popular idea that vg plays a dominant
role in damage caused by RF.
The evidence is very compelling and
SLAC will build 15(!) structures to test the
vg hypothesis and the test has a very high
priority. These tests look like they should
be definitive.
However, some ambiguity remains.
Since the subject is so important I will try
to summarize some discussion and
thoughts that followed from the
presentations.
 Not everyone at even SLAC is
convinced the vg hypothesis is correct.
 There seems to be a scale problem damage needs nJ and RF pulses have
about 10 J (D. Burke). How does a
proportionality work across so many
orders of magnitude?
 Low vg cells have less damage than high
vg cells but the coupler on the low side is
still (seriously) damaged. Did the
coupler protect the low vg cells? Do
couplers for low vg cells have higher
surface fields than couplers for high vg
cells? Is coupler design more important
than cell design? Pulsed surface
heating may be as bad as a TDS cell.
Our single feed structures are only
damaged in the input coupler…
 Breakdown
level
is
certainly
proportional to something like t½ and
damage seems to be a direct result of
breakdown
(Padamasee
and
Laurent/Sprehn but not P. Wilson) but
the famous equation doesn't have pulse
length in it.
 Dynamic changes of power flow,
reflections, during breakdown are not yet
included in the model.
 Damage and conditioning may be the
same thing.
 The structures were brazed in a hydrogen
furnace (see Laurent/Sprehn later).
 The structures are probably very dirty
(see Wang later).
 Structures were not systematically baked
out in-situ.
 The conditioning process was apparently
not well defined. Multiple shot
breakdowns were allowed for example.
Structures were removed and let up to air
and measured during the conditioning
process.
 The process was not very well
instrumented. Vacuum level was read off
ion pump currents for example.
(The last four points have most to do with
problems of absolute level)
Experiments to determine what is
important and what can be
achieved
L. Laurent/D. Sprehn
Windowtron - Highly instrumented
11 GHz cavity with removable high-field
'noses'. Designed to allow systematic study
of materials and processes. Not the whole
story on the level that can be achieved in a
TW
structure
but
essential
for
understanding what's important.
Hydrogen baked noses showed breakdown
damage
concentrated
along
grain
boundaries. Vacuum baked noses showed
much less damage which was evenly
distributed over the surface.
One pair of noses were conditioned with
92 breakdowns. About 90-100 pits were
found on the surface.
A 3 GHz Windowtron is under preparation
- to get frequency dependence!
A higher frequency windowtron is under
consideration.
S. Tantawi
Proposal to make high gradient tests by
using just WG's tapered to sufficiently
small gaps.
D. Pritzkau
Hydrogen brazed, vacuum baked copper
surface was cracked by 120º C pulsed
surface heating.
HIGGS
Structure design
X. Lin
Increasing disk thickness improves Es/Ea.
Go to 150º phase advance to reduce vg
without increasing the transverse wake.
May need to go to 'local' damping SLAC's name for what we do.
J. Wang
Described the 15-structure matrix that will
used for vg/breakdown/damage analysis.
Fabrication,
analysis
cleaning,
surface
J. Wang
Explained that cleanliness decreases as
fabrication steps progress. Culminates in a
brazing hall dirtier than the nearby parking
lot. > class 100,000.
The big issues
 f?
 t?
 What is the ultimate level?
 What is a breakdown?
 How does a breakdown produce
macroscopic
phenomena
like
reflected RF, vacuum rise, damage
etc.?
 How do you get rid of it?
 How do you attenuate its effects?
 Is something other than breakdown
occurring - erosion?
Where we stand
We have a proof of principle of gradient: A 26
cell (24 cm), 11.4 GHz was conditioned to a local
accelerating gradient of 154 MV/m, peak surface
electric
field
of
320 mV/m
(1.9
surface/accelerating, 1.1 over-voltage in the
coupler) for a pulse length of 150 ns. The structure
conditioned in 107 shots - fast. No damage (to the
extent we can verify it) but no lifetime test either.
Now we need to evolve to our 30 GHz structure.
 Physical length OK.
 Worse ratio of surface/accelerating fields in our
30 GHz structures, modification of geometry
may be necessary.
 If vg is a factor, modification of cell geometry
may be necessary.
 Frequency may yet help.
So far, life has been tough at 30 GHz. Testing
restricted to 16 ns pulses (and pulse length is very
important), 5 Hz. Damage to two structures has
been observed.
 Damage is confined to the 1.4 over-voltage
region of the input coupler (to the extent we can
verify it), which is single feed (always know to
be unacceptable in the long term).The two
structures have both been exposed to air (and
dust) for about six to seven years (the X-band
structure was exposed to air for probably less
than 48 hours). Both were operated in unusually
poor, perhaps 10-3 Torr, vacuum conditions.The
structures have not been systematically
conditioned.
30 GHz is so far a failure but with a lot of
excuses!
Observed 30 GHz fields
meaningful data points.
achieved
are
not
We do not yet know which issues are most
important - frequency, coupler geometry, cell
geometry, conditioning procedure, materials,
surface finish, oxides, dust, base vacuum pressure,
dynamic vacuum pressure, etc.
Our experimental and theoretical program is
expanding rapidly to identify the important issues
and then fix 'em.
Electrical breakdown doesn't look so bad, but
remember, the final limit may be 40ºC pulsed
surface heating.
WE NEED
30 GHz
POWER
SOURCES!