DC breakdown measurements Sergio Calatroni Present team: Jan Kovermann, Chiara Pasquino, Rocio Santiago Kern, Helga Timko, Mauro Taborelli, Walter Wuensch

DC breakdown measurements

Sergio CalatroniPresent team: Jan Kovermann, Chiara Pasquino, Rocio Santiago Kern, Helga

Timko, Mauro Taborelli, Walter Wuensch

Outline• Experimental setup• Typical measurements• Materials and surface preparations• Time delays before breakdown• Gas released during breakdown• Evolution of and Eb• Effect of spark energy• Future

Helsinki Sergio Calatroni - 19.11.2010 2

Experimental set-up : ‘‘ the spark system ’’

Helsinki Sergio Calatroni - 19.11.2010

Vpower supply (up to 15 kV)

C (28 nF typical)

vacuum chamber (UHV 10-10 mbar)

anode(rounded tip,

Ø 2 mm)

cathode(plane)

HV switch HV switch

spark

m-displacementgap 10 - 50 mm

(±1 mm)20 mm typically

• Two similar systems are running in parallel

• Types of measurements : 1) Field Emission ( b)

2) Conditioning ( breakdown field Eb)

3) Breakdown Rate ( BDR vs E)

3

Experimental set-up : diagnostics


V V

HV probe

to scope

current probe

vacuum gauges

V

gas analyzer

optical fibre

photomultiplieror spectrometer

4

Field emission - measurement• An I-V scan is performed at limited current, fitting the data to

the classical Fowler-Nordheim formula, where [jFE] = A/m2, [E] = MV/m and [φ] = eV (usually 4.5 eV).


E

EjFE

2/332/1

26 1053.6exp)41.10exp(

)(1054.1

0.0065 0.007 0.0075 0.008 0.0085-35

-34

-33

-32

-31

-30

-29ln(I/E^2)

Linear (ln(I/E^2))

1/E

Log(I/E^2)

is extracted from the slope

Conditioning – average breakdown field

Molybdenum Copper

Conditioning phase: 40 sparks

Deconditioning 1-5 sparks or no conditioning

Helsinki 6Sergio Calatroni - 19.11.2010

Eb

Eb

Surface damage (Mo)


1 5 10

20 40 100

Conditioning curves of pure metals


Selection of new materials for RF structure fabrication was the original purpose of the experiment

Breakdown field of materials (after conditioning)


• In addition to other properties, also importance of crystal structure?• reminder : Cu < W < Mo same ranking as in RF tests (30 GHz)

hc

p

hc

p bc

c

bc

c

bc

c

bc

c

bc

c

bc

c

fcc

fcc

fcc : face-centered cubicbcc : body-centered cubichcp : hexagonal closest packing

9

Surface treatments of Cu


• Surface treatments on Cu only affects the very first breakdowns

rolled sheet / heat treatm. milling Subu electro-polishing

b before 1st spark ~ 15 - 20 ~ 20 ~ 25 - 30 ~ 15 - 20

1st brkd field [MV/m] ~ 200 - 400 ~ 300 - 500 ~ 150 - 200 ~ 300 - 400

• After a few sparks: ~ 170 MV/m, b ~ 70 for every samples

The first sparks destroy rapidly the benefit of a good surface preparation and result in deconditioning. This might be the intrinsinc property of copper surface

In RF, sparks are distributed over a much larger surface, and conditioning is seen. Might be due to extrinsic properties.

• More foreseen in the near future to assess the effect of etching, brazing, etc.

10

Oxidized copper


Cu2O is a p-type semiconductor, with a higher work function than Cu : 5.37 eV instead

of 4.65 eV

1. Cu oxidized at 125°C for 48h in oven (air): purple surface ↔ Cu2O layer ~15 nm

2. Cu oxidized at 200°C for 72h in oven (air):

BDR = 1 for standard Cu @ 300 MV/m

BDR = 10-3 – 10-4 for oxidized Cu @ 300 MV/m, but last only a few sparks

1 2

11

Breakdown rate experiments

• A target field value is selected and applied repeatedly for 2 seconds• BDR is as usual: #BD / total attempts• Breakdown do often appear in clusters (a simple statistical approach can

account for this)


Breakdown Rate : DC & RF (30 GHz)

g DC RF

Cu 10 - 15 30

Mo 30 - 35 20

BDR ~ Eg Same trend in DC and in RF,

difficult to compare ‘slopes’


Time delays before breakdown


delay• Voltage rising time : ~ 100 ns

• Delay before spark : variable

• Spark duration : ~ 2 ms

14

Time delays with different materials


Cu Ta Mo SS

R = fraction of delayed breakdowns (excluding conditioning phase, where

imediated breakdowns dominate)

R = 0.07 R = 0.29 R = 0.76 R = 0.83

R increases with average breakdown field

Eb = 170 MV/m Eb = 300 MV/m Eb = 430 MV/m Eb = 900 MV/m

(but why ?!?)

15

Correlation pre-current & delays


From: Kartsev et al. SovPhys-Doklady15(1970)475

With our simple thermal model (based on Williams & Williams J. Appl. Phys. D 5 (1972) 280) which lead to the establishment of the scaling quantity “Sc” Phys.Rev.ST-AB 12 (2009) 102001

Gas released during a breakdown


0.8 J / spark0.95 J / spark

• Same gases released, with similar ratios

Outgassing probably dominated by Electron Stimulated Desorption (ESD)

• Slight decrease due to preliminary heat treatment

• Data used for estimates of dynamic vacuum in CLIC strucures

(heat treatment: ex-situ, 815°C, 2h, UHV)

17

H2 outgassing in Breakdown Rate mode (Cu)


‘quiet’ period

consecutivebreakdowns

Outgassing peaks at breakdowns

Slight outgassing during ‘quiet’ periods ESD with FE e- at the anode

No visible increase in outgassing just before a breakdown

18

Local field: · Eb (Cu)


• Measurements of b after each sparks (Cu electrodes)

b ↔ next

E b

correlatio

n

b · Eb = const

b ↔ previous E

b

no correlation

b · Eb is the constant parameter

(cf. Alpert et al., J. Vac. Sci. Technol., 1, 35 (1964))19

Evolution of & Eb during conditioning experiments


(± 32%)

(± 36%)

Eb = 159 MV/m

b = 77

Local field = const = 10.8 GV/m for Cu

(± 16%)

b · Eb = 10.8 GV/m

conditioning ?

good surface state

20

Evolution of during BDR measurements (Cu)


Spark

• General pattern : clusters of consecutive breakdowns / quiet periods (BDR = 0.11 in this case)

• b slightly increases during a quiet period if E is sufficiently highThe surface is modified by the presence of the field (are « tips » pulled?)

No spark

21

Evolution of during BDR measurements (Cu)


• Breakdown as soon as b > 48 ( ↔ b · 225 MV/m > 10.8 GV/m)

• Consecutive breakdowns as long as b > bthreshold

length and occurence of breakdown clusters ↔ evolution of b

b·E = 10.8 GV/m

Spark

No spark

22

Effect of spark energy - Cu


• EBRD increases with lower energy (less deconditioning is possible)

• Local breakdown field remains constant

Effect of spark energy - Mo


• EBRD seems to increase with higher energy (better conditioning possible)

• Local breakdown field remains constant

• However, we have doubts on representative the measurement is in this case

• The diameter of the damaged area depends on the energy available– Area mostly determined by the conditioning phase– Decreases with decreasing energy; saturates below a given threshold


Cu Mo

25

Energy scaling of the spot size

The future• Ongoing work:

– Finalise the work on the effect of spark energy on BD field, and understand the beta measurements for Mo

– Trying to understand “worms”( “flowers”?)


The future• Effect of temperature on evolution and other

properties– To verify the hypothesis and the dynamic of dislocation

• Fast electronics– Defined pulse shape and duration, without and during

breakdowns– Repetition rate up to 1 kHz (107 pulses -> less than 3 hours)


Fast electronics


Mike BarnesRudi Henrique Cavaleiro Soares

The future• Effect of surface treatment and in general of the

fabrication process on BD

– To study the influence of etching and its link with machining (preferential etching at dislocations, field enhancement or suppression, smoothening etc.)

– To study the influence of H2 bonding (faceting, etc)– (In parallel, ESD studies on the same samples)


Depends on capability of

accurately measure th

e gap

without p

rior s

urface contact



1st oxidation30 sparks2nd oxidation

2 4 6 8 10Energy (keV)

0

50

100

150

cps

O

Cu

Cu

Cu Cu

O element%=1.3-1.7


0

50

100

150

cps

OCu

Cu

Cu Cu

O element% = 3.5-3.7


0

20

40

60

80

100

cps

OCu

Cu

Cu Cu

O element% = 0.9-2.1


0

50

100

150

cps

OCu

Cu

Cu Cu

O element% = 1.6

1st oxidation150 sparks2nd oxidation180 sparks


0

50

100

150

cps

OCu

Cu

Cu Cu

O element% = 3.8


0

50

100

150

cps

O

Cu

Cu

Cu Cu

O element% = 1.0


0

50

100

150

cps

O

Cu

Cu

Cu Cu

O element% = 0.3

31

• A sparked (damaged) surface was reoxidised by heating and was sparked then again

– Was not able to recover the initial high EBRD

– Oxidised, smooth surface high EBRD

– Oxidised, sparked surface no improvement– Connection to the oxidation process?

Reoxidation

Field profile on the cathode surface

• Tip – plane geometry

• dependence with the gap distance

x

E


b · Eb is the constant parameter

(cf. Alpert et al., J. Vac. Sci. Technol., 1, 35 (1964))

Gap dependence of Eb, b and b · Eb (Cu)


Surface damage (Cu)


Gap measurement damage


Helsinki Sergio Calatroni - 19.11.2010 36CLIC Breakdown Workshop Sergio Calatroni TS/MME 36

Heating of tips by field emission - I

• The tip has a height/radius (field enhancement factor)• For a given value of applied E the Fowler-Nordheim law gives a current density

J(E)=A*(E)2*exp(-B/E)• This current produces a power dissipation by Joule effect in each element dz of the

tip, equal to dP = (J r2)2 (z) dz / r2

• The total dissipated power results in a temperature increase of the tip (the base is assumed fixed at 300 K). The resistivity itself if temperature dependent

• Using the equations we can, for example, find out for a given what is the field that brings the “tip of the tip” up to the melting point, and in what time.

• Used for “Sc” derivation Phys.Rev.ST-AB 12 (2009) 102001

J(E)h=height

r=radius

dz

T = 300K

T = Tmelting


Heating of tips by field emission - II• If the resistivity is considered temperature-independent, a stable

temperature is always achieved [Chatterton Proc. Roy. Soc. 88 (1966) 231]• If the resistivity (other material parameters play a lesser role) is temperature

dependent, then its increase produces a larger power dissipation, resulting in a further temperature increase and so on [Williams & Williams J. Appl. Phys. D 5 (1972) 280]

• Below a certain current threshold, a stable regime is reached• Above the threshold, a runaway regime is demonstrated• The time dependence of the temperature can be calculated.

High Gradient Workshop 2008 38Sergio Calatroni - CERN

Simulation for Mo cone: diameter 20 nm, beta = 30E=378 MV/mE=374 MV/m


Time constant to reach the copper melting point (cylinders, =30)

10-4

10-3

10-2

10-1

100

101

102

10-12

10-10

10-8

10-6

10-4

0.01

1

101 102 103 104 105

Parameters to attain the melting point of the tipof a Cu cylinder of given radius and =30

I peak [A]

P peak [W]time constant [sec]

Energy [J]

radius [nm]

The tips which are of interest for us are extremely tiny, <100 nm (i.e. almost invisible even with an electron microscope)


Power density at the copper melting point (cylinders, =30)

102

103

108

109

1010

1011

1012

1013

1014

101 102 103 104 105

Parameters to attain the melting point of the tipof a Cu cylinder of given radius and =30

E peak [MV/m]Power density [W/m2]Current density [A/m2]

radius [nm]

Power density (power flow) during the pulse is a key issue.See talk by A. Grudiev for RF structures scaling based on Poyinting vector

Documents

DC breakdown measurements Sergio Calatroni Present team: Jan Kovermann, Chiara Pasquino, Rocio Santiago Kern, Helga Timko, Mauro Taborelli, Walter Wuensch