REVIEW OF SCIENTIFIC INSTRUMENTS 79, 10F338 !

2008"

In situ window cleaning by laser blowoff through optical fibera

A. Alfier,1 S. Barison,3 T. Danieli,2 L. Giudicotti,2 C. Pagura,3 and R. Pasqualotto1
1

Consorzio RFX, Euratom-ENEA Association, C.so Stati Uniti 4, 35127 Padova, Italy
Department of Electrical Engineering, Padova University, Via Gradenigo 6/A, 35131 Padova, Italy
3
Istituto per lEnergetica e le Interfasi, CNR, C.so Stati Uniti 4, 35127 Padova, Italy
2

!Presented 14 May 2008; received 7 May 2008; accepted 9 June 2008;

published online 31 October 2008"
The feasibility of a window cleaning system based on the laser blowoff technique is investigated to
remove the impurity deposition on vacuum windows of the modified reversed field experiment
fusion device. The laser pulse is sent to the window through a fused silica fiber optic !! = 1 mm",
then focused on its internal surface, single shot ablating up to #5 mm2 of the impurity layer; the
focused pulse is scanned across the window to clean its entire surface. The composition of the
deposited layer is studied through the secondary ion mass spectrometry and profilometry techniques.
Effectiveness of cleaning is analyzed in terms of quality of the cleaned spot, its dimension,
repetition rate of the laser, and its wavelength. The energy damage threshold of the fiber optic is also
investigated. Three different lasers !microjoule Nd:YAG, Nd:YLF, and ruby" are first tested directly
on the window; then only the ruby laser beam is propagated through an optical fiber and tested.
2008 American Institute of Physics. $DOI: 10.1063/1.2955858%

INTRODUCTION

In the present fusion devices, the plasma interaction with

the inner wall of the vacuum vessel causes deposition of thin
layers of wall material on plasma viewing surfaces during
both standard and conditioning plasma discharges. This impurity deposition process affects also vacuum windows used
by optical diagnostics, such as Thomson scattering !TS" or
spectroscopy, gradually degrading their optical transmission
and causing a distortion of the spectral transmission curve.
The modification of the collection optics spectral response
represents a threat for measurement accuracy since it decreases signal amplitude and introduces systematic errors in
data analysis: for a TS diagnostic this corresponds respectively to a systematic underestimation of electron density ne
and to a wrong electron temperature Te !higher or lower
depending on the arrangement of spectral channels".1 In particular, the effect on Te is more pronounced at higher temperatures and for larger scattering angles, because of the
broader TS spectrum. In the specific case of the main TS
diagnostic installed2 on the modified reversed field experiment !RFX-mod",3 where core temperature ranges between
0.2 and 1.2 keV,4 this impurity deposition mainly affects the
signal level, gradually increasing the minimum density operational limit of the diagnostic. The effect is the same on all
spatial channels associated with each of the three collection
windows, because each window is next to the lens that images the scattering volume onto the fiber optics, i.e., in the
field position.2
To cope with this problem, various approaches have
a"

Contributed paper, published as part of the Proceedings of the 17th Topical

Conference on High-Temperature Plasma Diagnostics, Albuquerque, New
Mexico, May 2008.

0034-6748/2008/79!10"/10F338/4/$23.00

been proposed and implemented in other fusion experiments,

such as protecting windows with mechanical shutters when
diagnostics are not in use,5 and in situ window cleaning by
laser blowoff.6,7 However, these techniques can hardly be
adopted on RFX-mod, because of severe space constrains at
several vessel vacuum pipes, like those used for TS. Hence
on RFX-mod the solutions have been so far to periodically
monitor the transmission of TS collection windows by scanning an internal dc light source along the scattering volume,
and to replace the darkened windows with clean ones during
machine shut downs, about twice a year.
In this paper we propose a further improvement of conventional in situ window cleaning system by laser blowoff,
where the laser radiation is carried through an optical fiber: if
this method were adopted on RFX-mod, window replacement with vacuum break would be avoided. First the characteristics of the impurity layer typical of RFX-mod are investigated. Then a feasibility study of the improved blowoff
method is presented, using three different pulsed lasers available at RFX-mod, investigating also the energy damage
threshold of the fiber optic, and finally presenting a conceptual design of the system.
IMPURITY LAYER COMPOSITION

Collection windows of the main TS system on RFX-mod

are made of fused silica and have different shapes: that considered here is an oblong window of #150" 100 mm2,
10 mm thick. A dichroic filter is deposited on the outer side
!T # 95% at 530 nm$ % $ 1200 nm", and a broadband antireflection !AR" single layer MgF2 coating, tens of nanometer
thick, on the inner side. The spectral region considered for
TS measurements is between 850 and 1070 nm: the diagnostic is, in fact, based on a Nd:YLF !yttrium lithium fluoride"
laser !% = 1053 nm" and its detection system is optimized for

79, 10F338-1

2008 American Institute of Physics

Downloaded 29 Sep 2009 to 192.101.166.230. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

10F338-2

Rev. Sci. Instrum. 79, 10F338 !2008"

Alfier et al.
100
a)

40
Clean window

Transmission (%)

Transmission (%)

80

60

40
Dirty window
20

600

800
1000
1200
Wavelength (nm)

1400

b)

30
20
10

0
6
Lo 4
ca
tio 2
no 0
nw 2
ind
ow 4 6
(cm
)

1000
800
)
m
(n
ngth
vele

600
Wa

FIG. 1. !Color online" !a" Transmission of the window before and after plasma exposure !i.e., respectively, without and with impurity layer". Note the effect
of the dichroic coating cutting off the short wavelength range. !b" Transmission of the dirty window measured along the minor axis.

boron glow discharges. Other metals, mainly Zr+ or Ni+,

come from the interaction of plasma with in-vessel metallic
components. This composition may explain the effect of the
deposited layer on the transmission of the window: most of
the attenuation is determined by the compounds of light elements such as H, C, or B, as stated in Ref. 10, while heavy
elements may cause the chromatic variation. On the basis of
the crater depth measurements with SIMS, a relevant result is
that the impurity deposition process degrades the single layer
AR coating: the presence of Mg+ and F+ components on the
surface of impurities !600 nm from the AR coating" can be
explained either by sputtering of the AR layer by heavy components from the plasma, or by penetration of impurities in
the AR coating. Since the MgF2 layer is only 20 nm thick,
this latter is quite unlikely. In particular, the cause of similar
in-depth distribution profile of both light elements B and C
requires a specific investigation: whether B acts as an attractor for C, or vice versa, or MgF2 attracts both, is out of the
aim of this work.

MgF coating

10

B+

10
Intensity (a.u.)

measuring Te in the range of 20 1500 eV for ne # 1

" 1019 m3. A typical transmission curve of collection windows before being exposed to plasma discharges !clean
window" is shown in Fig. 1, frame a !blue line", superposed
to an example of transmission curve obtained as the impurity
layer is deposited on the dichroic filter !red line" after 1000
RFX-mod pulses at #600 800 kA, one wall conditioning
discharge in boron !boronization, B2H6" and 15 glow discharges in helium !10" and hydrogen !5": it is evident the
reduction in the average transmission, from #90% to #35%,
and the deformation of the spectral curve, with higher absorption toward the blue side of the spectrum. In other cases
the reduction has been found to be worse !#10% transmission" or, on the opposite, irrelevant, depending on the position of the window on the machine and on the experimental
campaigns during which it has been exposed. This degradation is not uniform on most of the window surface: it decreases at its edges where the mechanical structure of
vacuum pipe vignettes the impurity flux !see Fig. 1, frame b"
and its distribution mainly follows the plasma flow.
In order to investigate the composition at various depths
of the stack of layers deposited during plasma operation,
glow discharges, and boronization, dynamic secondary ion
mass spectrometry !SIMS" analyses were performed with a
customized instrument:8 a monochromatic !6 keV" O2+ primary ion beam was generated in a mass-filtered duoplasmatron ion gun !model DP50B, VG Fisons, UK" and an
EQS1000 !Hiden, UK" mass energy analyzer with a sector
field electrostatic energy analyzer and a quadrupole mass filter were used for negative- and positive-ion detection in
counting mode. SIMS crater depth was measured by a Tencor P-10 !Tencor, USA" mechanical profiler having a 2 nm
vertical resolution. Results are shown in Fig. 2. Except for
Mg+ and F+ which compose the AR coating, it is found that
carbon !C" is the main element of the coating, as demonstrated by the presence of several carbon-related signals
!e.g., Cx+ or CH+": it comes from graphite tiles which completely cover the inner wall of the RFX-mod vessel and
which are eroded by the plasma !plasma-wall interaction in
RFX-mod can be very intense9". The carbon content also
determines most of the transmission attenuation. Boron is the
other main element and it is used for wall conditioning with

C+

10

Mg+

Cr+

10

Zr+

10

Ni+

10

F+

100

200

300
400
Depth (nm)

500

600

700

FIG. 2. !Color online" In-depth investigation of the impurity layer obtained

with the combination of the SIMS and profilometry analyzer techniques.
The peak of Mg+ and F+ at #650 nm depth corresponds to the AR coating.
The hardness of the fused silica window determines the subsequent drop of
all signals.

Downloaded 29 Sep 2009 to 192.101.166.230. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

10F338-3

Rev. Sci. Instrum. 79, 10F338 !2008"

Window cleaning by laser blowoff

TABLE I. Specifications of the three pulsed lasers used.

% !nm"
E / pulse available
Pulse duration !FWHM"!ns"
Pulse repetition rate
Beam divergence !mrad"

Microjoule
Nd:YAG

Nd:YLF
oscillator

Ruby
oscillator

1064
8 &J
$1
8 kHz
30

1053
200 mJ
15
1 Hz
0.5

694
40 mJ
50
0.1 Hz
0.4

Blowoff tests
A. Direct on the window

Three different pulsed lasers are available at RFX-mod:

a passively Q-switched Nd3+-microchip Nd:YAG !yttrium
aluminum garnet" laser, the oscillator of the ruby laser used
for the edge TS and the blowoff impurity injection systems,11
and the oscillator of the Nd:YLF laser used for the main TS
system.2 Their main specifications are reported in Table I.
Their beams have been first focused directly on the window
at atmospheric pressure, in order to determine which available laser was most effective and suitable to be coupled to an
optical fiber. For each laser, the spot size has been varied to
determine the minimum diameter and energy density required to completely ablate a portion of the impurity layer.
A limited portion of the window has been cleaned with each
laser, scanning the window with an xy-micropositioning
stage, with a single laser shot per position. The transmission
function of the cleaned region has been measured and it has
been compared to that of the clean window !see Fig. 3". A
direct comparison with previous works in other experiments
is not possible,6,7,10 since the composition of the impurity
layer and experimental setup are probably different.
The microchip Nd:YAG laser is capable of ablating with
an energy density of #0.6 J / cm2 and with a high repetition
rate !#8 kHz". Its main disadvantage is the reduced spot
diameter, #50 &m, of the focused beam. The recovered
transmission is spectrally almost flat and similar to that of
the clean window, but only 67% on average !see Fig. 3": a
relevant amount of residual impurities left on the window
surface is due to the resolution !30 &m" of the cleaning scan

step with the adopted translation system !10 &m accuracy",

so that the adjacent ablation spots are probably not uniformly
cleaning the surface. A motorized scanning system based on
such a laser would produce a more thorough cleaning of the
surface with a more resolved scan step !e.g., 5 &m accuracy,
10 &m resolution".
The oscillator of the Nd:YLF laser produces trains of
pulses with energy up to 200 mJ at 1 Hz repetition rate. The
main limitation is that in a single shot an energy density of
about 0.2 J / cm2 on a 2 mm diameter spot produces a recovering of the average transmission only to #72%, and this is
also not spectrally uniform !see Fig. 3": higher transmission
is obtained in the spectral range closer to the laser wavelength !1000 1400 nm", while transmission is still too low
at shorter wavelength. This may be explained by the irregular
beam energy density, with several peaks of energy corresponding to better cleaned portions of the beam spot. Residuals of the deposited layer are not ablated and left on the
window, because locally the beam intensity is below the ablation threshold: the measured transmission across a partially
cleaned surface, retains the original feature, with lower transmission at lower wavelength values. However, this has not
been investigated in more details, since the main disadvantage of the Nd:YLF is its high damaging risk: high energy
absorption at energy peaks from transition surfaces both at
the fiber and at the window may result in the appearing of
localized damages.
The oscillator of the ruby laser represents a good compromise between repetition rate, beam energy, and quality of
the recovered transmission: applying an energy density of
0.2 J / cm2 on a 1.5 mm diameter spot, transmission has been
completely restored in a single pulse !see Fig. 3". For this
reason, despite the limitation represented by the low repetition rate of 0.1 Hz, the ruby laser has been chosen to be
coupled to an optical fiber.
B. Optical fiber coupling

Impurity layer ablation by propagating the laser beam

directly to the collection windows of the RFX-mod experiment is not easily achievable. Several beam paths should be

Blowoff effectivenees (%)

110
100

Ruby

90
Nd:YLF

80
70
60

Nd:YAG
50
600

700

800
900
Wavelength (nm)

1000

1100

FIG. 3. !Color online" !a" Recovered transmission measured on the window cleaned with each of the three lasers: Nd:YAG, Nd:YLF, and ruby. !b" Effect of
ablation of a single ruby oscillator pulse on a dirty window.

Downloaded 29 Sep 2009 to 192.101.166.230. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp

10F338-4

Rev. Sci. Instrum. 79, 10F338 !2008"

Alfier et al.

installed, with lenses to image the beam !free propagation is

not advisable if the beam is not single transverse mode" and
mirrors to steer it, one path for each window to be cleaned,
as they are in different locations around the torus. This would
also imply additional precautions for safety. Hence, the use
of an optical fiber to guide the laser emission would represent a more flexible and safer system. Main problems in
coupling lasers to fibers are !a" misalignment between the
laser beam and the optical fiber and !b" beam divergence
higher than the fiber acceptance numerical aperture !NA" or
!c" overfilling of the core input surface. In the first case, the
beam may be focused on the fiber clad, while in the other
two a relevant amount of energy may be dissipated in the
clad, in any case risking to damage it. In these cases, fiber
damages are localized in the first 5 25 mm length, i.e., at
the first internal reflection point.12 A further relevant constrain is the energy density, with damage thresholds depending on both the core and clad materials: even with the proper
alignment and input spot size, damages may be localized at
transition surfaces !input face or core/clad interface" where
part of the energy beam is absorbed.
In our case, the ruby laser oscillator has been directed
onto a ! = 1 mm, 0.37 NA, low OH quartz fiber. The laser is
coupled to the fiber without any lens, after passing through a
0.1 mm diameter spatial filter !pinhole" on which the ruby
oscillator radiation is focused, before entering the following
amplification stages. The resulting beam divergence was
used to expand the beam, enough to reduce damaging risks,
but still fitting the fiber core: two sets of fibers have been
considered, one entirely made of fused silica, and the other
with plastic clad and fused silica core. Tests over 300 consecutive pulses have shown that with the correct alignment
and imaging on the input face of the fiber !! = 0.8 mm", no
damage is caused with 35 mJ energy in both kinds of fibers.
On average, 26 mJ are measured at the output of the fiber,
#75% of the input energy. A complete ablation with a single
laser shot is obtained focusing the beam to an #2 mm diameter spot with a f = 25 mm lens. Similar results are obtained
connecting an additional fiber with a SMA to SMA connector, except for the lower output energy, #23 mJ !65%". In
both cases, the recovered transmission is almost identical to
that reported in Fig. 3 in the case of direct ablation with the
ruby laser.
CONCEPTUAL DESIGN

We have demonstrated the feasibility of a window cleaning system based on the laser blowoff technique, adopting
the available ruby laser oscillator employed for the edge TS
system of RFX-mod. The laser radiation can be easily transferred through optical fibers directly to the window surface,
with enough input energy and reasonable spot size. The main
limitation is represented by the low repetition rate of the
laser, 0.1 Hz: cleaning the entire window surface of about
8500 mm2 would require #15 h. Two parallel improvements
will be considered to overcome this issue. The first consists

in adopting a commercial high damage threshold fiber

!1 GW/ cm2", increasing the total beam energy, and hence
the size of the focused beam on the window, but preserving
the beam energy density previously determined: this will reduce the number of required pulses to scan the entire window. The second consists in adopting a higher repetition rate
laser, with an unvaried beam spot size at the window, thus
reducing the required time to complete the scan. The combination of the two options is commercially available, as a
laser beam directly coupled to a high power fiber. These
approaches could reduce the window cleaning process to less
than 1 2 h, being limited mostly by the duty cycle of the
scanning system.
The laser pulse duration may play a relevant role as it
represents the power absorbed by the impurity layer so that a
faster pulse could allow a larger spot; nevertheless, as stated
in Refs. 12 and 13, this could damage the fiber, and the
proper compromise between energy, power, and laser repetition rate should be determined accurately.
The focusing of the laser on the window is obtained
placing three groups of optics. A fast lens collimates the laser
radiation from the fiber output along one window axis to a
first mirror; this deflects the beam along the other window
axis to a second mirror coupled to a fast lens, which, respectively, redirects the beam perpendicularly to the window and
focuses it onto the internal window surface. Dimension of
optics !1 in. " and the focal length of the lens !$20 30 mm"
are imposed by space constrains around the installed window, mainly imposed by coils. A complete scan of the window can be accomplished through a two-dimensional translation stage remotely controlled: two commercial stepper
motors, with 0.1 mm accuracy, are used to move the last two
groups of optics.
ACKNOWLEDGMENTS

This work was supported by the European Communities

under the contract of Association between EURATOM/
ENEA. The views and opinions expressed herein do not
necessarily reflect those of the European Commission.
1

D. H. McNeill, Rev. Sci. Instrum. 61, 1263 !1990".

A. Alfier and R. Pasqualotto, Rev. Sci. Instrum. 78, 013505 !2007".
3
R. Paccagnella et al., Phys. Rev. Lett. 97, 075001 !2006".
4
A. Alfier, R. Pasqualotto, G. Spizzo, A. Canton, A. Fassina, and L. Frassinetti, Plasma Phys. Controlled Fusion 50, 035013 !2008".
5
D. Johnson, D. Dimock, B. Grek, D. Long, D. McNeill, R. Palladino, J.
Robinson, and E. Tolnas, Rev. Sci. Instrum. 56, 1015 !1985".
6
B. W. Brown, C. W. Gowers, P. Nielsen, and B. Schunke, Rev. Sci.
Instrum. 66, 3077 !1995".
7
K. Narihara and S. Hirokura, Rev. Sci. Instrum. 63, 3527 !1992".
8
A. Tolstogouzov, S. Daolio, and C. Pagura, Surf. Sci. 441, 213 !1999".
9
S. Ortolani, Plasma Phys. Controlled Fusion 48, B371 !2006".
10
H. Yoshida, O. Naito, T. Hatae, and A. Nagashima, Rev. Sci. Instrum. 68,
256 !1996".
11
A. Alfier and R. Pasqualotto, Rev. Sci. Instrum. 77, 10E501 !2006".
12
S. W. Allison, G. T. Gillies, D. W. Magnuson, and T. S. Pagano, Appl.
Opt. 24, 3140 !1995".
13
R. Wood, Laser Damage in Optical Materials !Hilger, Bristol, 1986".
2

Downloaded 29 Sep 2009 to 192.101.166.230. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp