Skip to Content.
Sympa Menu

opal - Re: [Opal] Dark Current Simulations - FE

opal AT

Subject: The OPAL Discussion Forum

List archive

Re: [Opal] Dark Current Simulations - FE

Chronological Thread 
  • From: "frederic.le.pimpec" <frederic.le.pimpec AT>
  • To: wangchuan02513 <wangchuan02513 AT>
  • Cc: "opal AT" <opal AT>
  • Subject: Re: [Opal] Dark Current Simulations - FE
  • Date: Fri, 02 Sep 2011 14:04:26 +0200
  • List-archive: <>
  • List-id: The OPAL Discussion Forum <>
  • Organization: PSI

Hi Chuan,

The simulation is queuing since 11 am this morning.
In the .data file
QBUNCH 0.1 # Bunch charge [pC]

before hand it was 200.
After QB will certainly deal with the number of QBUNCH .

So now I am expecting to see variation in the FE over 900 steps, 900 deg of RF S band phase.


wangchuan02513 wrote:

Hi, Frederic,
Yes you can try Npart = 50 but no less, which as far as I know it's OK. But
the unit of QB is C not pC, so QB=1e-13 will not scare people:-)
Best regards

At 2011-09-02 16:41:17,"frederic.le.pimpec" <frederic.le.pimpec AT> wrote:
Dear Chuan,

OK I understand about the emission time. I indeed looked at this 1e-14 s and i was wondering what it was since i specified 1e-12, which becomes correct after a while.

So Can I set Npart to 50 ?
This is small enough, and i don't care about the normal bunch. I am solely interested in how the dark current appears and get multiplied.
I can also set the bunch to 0.1pC in the data or tmpl file. I never remember.
Qinitial = Qi= 4e-14 [C] = QB(0.1)/Npart(54)
It should be OK with the simulation ? no cluster crashing ?


wangchuan02513 wrote:
Dear, Frederic, Sorry for the late reply. Basically, Qinitial = Qi=
0.002e-12 [C] = QB/Npart, where, QB [C] and Npart are defined in your .in
file. Currently, Npart can not be set to zero, there are two reasons
(at least these two are good enough:-??

One is OPAL need to know the charge, particle type, etc before run FE
model, so I adopted the old way to initialize the initial electron
particle bunch for tracking ( Although I've provided another way to
initialize particle bunch for cyclotron cavity multipacting
simulation which do not involve initial electron bunch). Please don't
set Npart to zero in FE simulation, this may crash the cluster.

Second is we can use the initial bunch as a time/position reference
during visualization for the dark current particles. In your case, I
think you can simply subtract the number of initial bunch
(Npart=5000), if the initial bunch will not hit the surface of the
cavity, then you can get the number of dark current particles in each
time step. Best regards Chuan

At 2011-08-31 17:41:33,"frederic.le.pimpec"
<frederic.le.pimpec AT> wrote:
Dear Chuan,

The png file which was sent yesterday shows : top plot, the total
number of particles at each time step and bottom plot the
difference between 2 systematic input from the PartStatistics.dat after extracting in the oxxx file for example step 666 OPAL {0}> *
NP= 31665 Qtot= 4.19804 [nC] Qi= 4e-14 [C] This matches the number
in PartStatistics.datat the same step.

The png file sent Partcl_Stat02.png, shows the number of emitted particle at every step extracted from the 0xxx file. This is very similar to what i sent yesterday from the PartStatistics.dat

So the Qinitial (Qi= 4e-14 [C]) is set by default in the .in file ?
this is in the data file or in the tmpl file ? I did not find it
in the .in file

OK for the Npart. In fact since i have no initial bunch then i
should not care Npart should be set to zero ? I only care about
electrons emitted above threshold !

About FN and the unsymmetrical surface mesh. I perfectly agree with
you "this unsymmetry should be small as the mesh is usually fine."

Bests Frederic

wangchuan02513 wrote:
Hi, Frederic, First, could you find the PartStatistics.dat file
in: your_simulation_directory/data/ ? Please have a quick look at
the last column of the PartStatistics.dat which contains the
number of physical particles in each time step, to see if the
number of physical particles is the same with the simulation
particle number data extracted from the .0xxx files.

Then please also send me the input file, i.e., the .in file you
use as well as the PartStatistics.dat file and .0xxx files
generated by the OPAL output. So I can better understand the
result you got.

And, the Npart you defined in the .in file is the particle number
of primary bunch, which only matters to the initial charge of
each simulation particles in FE based simulation. As the time
step goes by, there will be more more paritcles emitted by the FE

The Qreal I mentioned in the previous email can be different for every field emitted particles, because each field emitted
particle, namely particle i, which have charge Qreal_i, may
represent many more particles which have the same charge Qinitial
as the charge of particles in initial bunch (like the initial
bunch you have defined in beam1in your .in file).

As the Fowler-Nordheim formular only compute the current density,
so we have to multiply the triangle area as well as the time step
size to get the emitted charge per time step. So the unsymmetry
of the surface mesh (the triangle area) could cause the
unsymmetry of the emitted particle number. But this unsymmetry
should be small as the mesh is usually fine.

Best regards Chuan

At 2011-08-30 22:19:09,"frederic.le.pimpec" <frederic.le.pimpec AT> wrote:
Hi Chuan,

so if you see attach what gives (extracted) the .0xxx files it
is emitting always at the beginning the same number of
particles. My RF phase change every ps in S band so 1 deg
phase. The particle emission should look like what is going on
between 600 and 700. which is already after a full period.

when below threshold I should indeed have 0 emission particles.
This is correct.

Hence my question about the injection phase of the RF. It seems
that for the 1st (360 steps) period I do not emit properly.

Also in my file I have defined NPART 5000 # Number of simulation particles

at the end I endup with more than 50 000 !

in the .in file beam1: BEAM, PARTICLE=ELECTRON, pc=P0,

and what is Qreal = Qinitial*Nn/M

at t=0 Qinitial=0, at t=t0 is my Q initial the Q at =t0-1step
? bounded to 1 otherwise Q =0

I understand that M is the maximum number allowed per triangle.
How do you check that an iris in a structure when the field is
above threshold emitt the particles ? And all triangles mapping
the edge of the iris should emit homogeneously. They all should
emit the same way because I have a field cylindrical symmetric
around the Z axis of the machine !

So I am a bit lost ! like my particles.

bests and thanks Frederic

wangchuan02513 wrote:
Hi, Frederic, So the particle number you use in the .oxxx
file is the number of simulation particles, which may have
different charge for each particle and is not the number of
physical particles which you might be interested in. Please
try PartStatistics.dat in YOUR_SIMULATION_DIRECTORY/data/ or YOUR_SIMULATION_DIRECTORY/vtk/ (if you use older version of
OPAL) and the last column gives the correct number of
physical pariticles. Best regards Chuan

At 2011-08-30 19:43:42,"Andreas Adelmann" <andreas.adelmann AT> wrote:

On 30.08.2011, at 12:50, wangchuan02513 wrote:

Hi, Frederic First I'm happy someone other than Andreas and myself are interested in FE model in OPAL:-)

First I have a question about the .oxxx file, because as far
as I know, I didn't write any .oxxx file during the FE based simulation. Do you add this new feature in OPAL? If so,
please give me more information about the .oxxx file.

The .oxxx file is the std-out which appears in the output
file written by SGE.


W.r.t you question, I would first introduce what the FE model
in OPAL exactly do, which may help you to find out what the
problem is. In the FE model in OPAL, 1. we traverse all the
surface triangles in each time step, read and compute the
Efields normal to the triangles, compare the Efields with a
threshold defined by user in input file. 2. If
Ffield<threshold, skip following algorithms and see the next
triangle. 3. If Ffield>=threshold, compute the emission
current density J according to Fowler-Nordheim formular. Then
compute the nominal emitted particle number Nn. As the field
emission process is a quantum process which may generate so
many particles that beyond the memory limit of a cluster, so
we scale the real emission particle number Nr with a larger
charge Qreal = Qinitial*Nn/M, if Nn>M, where M is the allowed
maximum number of emitted particles per triangle per time
step. And in this case Nr=M. So probably, first you can try
to set the threshold=0, which will make OPAL go through the
full algorithms of FE model, then you should use tot charge/Qinitial instead of the number of simulation particle number in each time step. Hopefully these can help you, if
not, please let me know. Best regards Chuan

At 2011-08-29 23:59:36,"frederic.le.pimpec" <frederic.le.pimpec AT> wrote:
Dear Chuan,

I am wondering about the particle emission depending of the
RF phase of the accelerating structure, case of an RF gun,
2 and a 1/2 cell, S band, meaning 1ps=1deg of rf phase.
Pure Field emission (FE)

I have set in the ini file, 1000 steps. 1 ps = 1 step for
every ps.

The .oxxx file give the number of particle emitted per
step. let say for at least the first 25 steps, the emission
is constant and correspond to 27 particles. From the Field emission model and the RF phase , the number should either increase or decrease over 25 deg of RF phase.

Sometimes during the emission the number of particle
emitted is identical to 22 and then jump at 400, then it

So I am not really understanding the behaviour of the
emission !
Dr Frederic Le Pimpec
Paul Scherrer Institut
WBBA 315
5232 Villigen
Tel : +41 56 310 5130
FAX : +41 56 310 4488

Dr Frederic Le Pimpec
Paul Scherrer Institut
WBBA 315
5232 Villigen
Tel : +41 56 310 5130
FAX : +41 56 310 4488

Archive powered by MHonArc 2.6.19.

Top of Page