��� ���� � �

������ ������� �� ����� ������� ����� ������ ��� �������

�� � �� �������� ����

���� ����� ����� ������ ��� ����������

���� ��������

��������������� �������� ������

���������� �� �����������
������������ ���� ���
������ �������
 Winter College on Fibre Optics, Fibre Lasers and Sensors
International Centre for Theoretical Physics
12 – 23 February 2007

High Power Fibre Lasers and Amplifiers

Andy Clarkson

Optoelectronics Research Centre

University of Southampton
Southampton, SO17 1BJ
United Kingdom
1
 Contents

1. Introduction:

Background 7
Brief history of rare-earth doped fibers 8
Operating wavelengths for rare-earth doped fiber lasers 11
Principles of rare-earth doped fiber lasers 12
Simple model for four-level lasers 14
References 18

2. Cladding-pumped fibers:

The ‘cladding-pumping’ concept 20

Inner-cladding geometry 21
Main laser transitions for high-power fiber devices 23
Theory for amplifier and laser performance 27
External feedback cavity design 37
References 39

2
3. Pump sources and pump coupling:

Introduction 41
Brightness and the M2 parameter 42
Calculating pump beam sizes 43
Focussing pump light into double-clad fibers 44
Diode laser pump sources 46
Pump launching schemes 50
References 61

4. Thermal effects in fibers:

Heat generation 63
Thermal effects 65
Summary 79
References 80

5. Nonlinear processes:

Introduction 82
Stimulated Raman scattering 83
Raman Lasers 87
Stimulated Brillouin scattering 89
3
References 93
6. Scaling mode area:

Novel core designs 97

Mode-selection in multimode cores 98
Mode selection by bending 102
References 105

7. Wavelength selection and tuning:

Introduction 107
Wavelength selection and tuning schemes 109
Single-frequency fiber sources 112
References 114

8. High power CW and pulsed fiber sources:

High power CW Yb-doped fiber lasers 116

High power Q-switched Yb-doped fiber lasers 118
Fiber Facet damage 119
MOPA configurations 120
References 122

4
10. Power and brightness scaling via beam combination:

Introduction 124
Incoherent beam combining 124
Coherent beam combining 129
References 133

5
1. Introduction

6
 Background
Scaling the output power from lasers has been an activity that has preoccupied many within
the laser community ever since the laser’s invention. This has been driven not simply by
curiosity, but also to fulfil the needs of huge range of applications, and by the prospect of even
more applications:

Application areas include:

• Scientific applications
• Medical applications
• Remote monitoring and sensing
• Free-space communications
• Materials processing
• Defence applications

In addition to high power, many of these applications place other demands on the laser
source, for example, in terms of beam quality, efficiency, linewidth, mode of operation, etc,
which may be quite difficult to achieve.

For a long time, the high-power laser area has been dominated by gas lasers and conventional
‘bulk’ solid-state lasers. Fiber lasers and amplifiers are relatively new arrivals to this area!

7
 Brief history of rare-earth doped fibers

1964 First rare earth doped fiber active fiber device1

Coiled Nd-doped fiber side-pumped by a flashlamp
C. Koester and E. Snitzer, American Optical Co

1974 First fiber laser pumped by a laser diode2

CW operation of a Nd-doped silica fiber laser pumped by a GaAs laser diode
J. Stone and C. Burrus, Bell Labs

1985 Fabrication of low-loss RE-doped silica fibers by MCVD and solution doping
resulting in the first low-loss Nd-doped fiber laser pumped by a laser diode3
R. J. Mears, L. Reekie, S. B. Poole and D. N. Payne, University of Southampton

1987 First erbium-doped fiber amplifier4

R. Mears, L. Reekie, I. Jauncey and D. N. Payne, University of Southampton

1988 First cladding-pumped fiber laser5

Nd-doped fiber laser with offset core
E. Snitzer, H. Po, F. Hakimi, R. Tumminelli, B. C. McCollum, Polaroid

1999 First >100W cladding-pumped fiber laser

V. Dominic, S. MacCormack, R. Waarts, S. Sanders, S.Bicknese, R. Dohle, E.
Wolak, P. S. Yeh and E. Zucker, SDL

8
 2002 First fiber laser with > 1 kW output power7
K.-I. Ueda, H. Sekiguchi, H. Kan, University of Electrocommunications, Hoya, and
Hamamatsu, Japan

2005 Ytterbium-doped silica fiber laser with >2kW output8

V. P. Gapontsev, D. V. Gapontsev, N. S. Platonov, O. Shkurihin, V. Fomin, A.
Mashkin, M. Abramov and S. Ferin, IPG Photonics

2500

2000
The dramatic rise in output
power from cladding-pumped

Power (W)
1500
Yb-doped silica fiber lasers

1000

500

0
1996 1998 2000 2002 2004 2006
Year
9
Energy levels is rare-earth ions

From ref. 9

10
 Operating wavelengths for rare-earth doped fiber
lasers in silica and non-silica glasses

3+ 3+
Tm3+ Pr3+ Tm3+ Nd3+ Yb3+ Nd3+ Tm Er Tm3+ Ho3+ Tm3+ Er3+
3+
Ho

500 1000 1500 2000 2500 3000

Wavelength (nm)

11
 Principles of RE-doped fiber lasers

RE-doped core Silica cladding

Pump Output

Outer coating Output coupler

High reflector
Transmission = T

Core: Radius = ra 2πra 2πra NA

Refractive index = na V= n a2 − n 2b =
λ λ
Cladding: Radius = rb For single-mode propagation in the core, V<2.405
Refractive index = nb ra = 4µm for NA = 0.1 and λ = 1.06µm

Gaussian transverse intensity profile approximation for fundamental mode10

Mode radius = wL ra/(logeV)0.5 = 4.3µm

Typical value for rb is 62.5µm 12

 Core composition
The core is fabricated from SiO2 with various dopants including:

• Rare earth ions (e.g. Nd3+, Yb3+, Er3+, Tm3+ ) in the form of RE2O3 with typical
concentrations in the range of ~100ppm to >10,000ppm
• Dopants such as Al and P are generally added to modify the environment to increase
the RE solubility in SiO2
• Dopants (e.g. Al, Ge) are added to increase the refractive index of the core relative to
surrounding pure silica cladding

The addition of these extra dopants varies the host composition and hence can have an
effect on the spectroscopy
3e-24

3e-24

Aluminosilicate
Cross-section [m2]

Phosphosilicate
For various Yb-doped 2e-24 Boro-aluminosilicate

silica glass compositions11: 2e-24

1e-24

5e-25

0
850 900 950 1000 1050 1100 1150 13
Wavelength [nm]
 Simple model for four-level lasers
Overview of the advantages and disadvantages of RE-doped fiber lasers versus
conventional ‘bulk’ solid-state lasers using simplified theory for laser performance
Pumped Laser
region x
medium
Output
z
y

High reflectivity l Mirror with

mirror transmission, T

Stark splitting of energy levels due crystal field.

Fast non-radiative decay
Resulting levels are further split / broadened by
other mechanisms11.

RE-doped crystals generally have relatively discrete

sub-levels and relatively narrow homogeneously-
broadened emission and absorption spectra.

RE-doped glasses generally have very broad

(overlapping) sub-levels with very broad
Fast non-radiative decay continuous emission and absorption spectra.
Broadening involves both homogeneous and
inhomogeneous mechanisms, but homogeneous
broadening usually dominates. 14
Rate equations:

Population densities for the lower and upper laser levels are: N1 = 0 (from four-level approximation)
N2 = N = Inversion density

Under steady-state conditions:

dN ( r, z ) N ( r, z )
= R p ( r, z ) − − c n σe N ( r, z )s( r, z ) = 0
dt τf

dS S
= c n σe N ( r , z )s( r , z )dV − = 0
dt cavity τc

where cn is the speed of light in the laser medium, τf is the fluorescence lifetime of the upper level, τc
is the cavity photon lifetime, Rp(r,z) is the pump rate density, s(r,z) is the photon density, S is the total
number of photons in the laser mode inside the resonator and σe is the emission cross-section for the
transition. σe is usually much smaller for a glass host than for a crystal host.

Material Nd:YAG Nd:YVO4 Nd:YLF Nd:silica Nd:phosphate

glass glass
L(µm) 1.064 1.064 1.047 ( ) 1.060 1.060
1.053 ( )

e (10-23m2) 3.4 25 ( ) 1.9 ( ) ~0.14 ~0.4

1.2 ( )
15
f (µs) 230 90 520 500 350
Simplifying assumptions:

• Gaussian transverse profiles for fundamental laser mode, pump and inversion distribution
• Neglect diffraction spreading of pump and laser mode for bulk laser
• Negligible ground-state depletion
• Low resonator loss: T = transmission of output coupler, L = round-trip loss (excluding output
coupler) (L + T) << 1

Can solve rate equations to obtain expressions for threshold and slope efficiency as follows12,13:

• Threshold pump power

Low emission cross-section for fiber is
πh ν p ( L + T )( w + w )
2 2
more than offset by the small core/beam
Ppth ≈
p L
size compared to bulk laser configurations
4 σ e τ f η p η abs Much lower threshold for fiber laser
Threshold powers can be < 1mW
• Slope efficiency
T νL In practice, slope efficiencies for fiber lasers
ηs ≈ ηp ηabs ηPL
L+T νp tend to be higher than for bulk lasers - Why?

where wp is the pump beam size or fiber core radius, wL is the laser mode radius, ηp is the pumping
quantum efficiency, ηPL is the pump – laser mode overlap factor and ηabs = [1-exp(-αpl)] is the
of pump light absorbed (αp is the absorption coefficient for the pump). 16
 Overlap factor:
w 2L (2 w 2P + w 2L )
ηPL ≈ <1 at low powers (I/Isat <<1)
( w 2P + w 2L ) 2
and
ηPL ≈ 1 at high powers (I/Isat >>1)
hν L
where Isat is the saturation intensity given by: Isat =
σe τf
Laser mode Inversion
Fibre Bulk

< 10µm > 200µm

N(r, z) Rate of stimulated emission I
R p (r, z) = + c n σ e N(r, z)s(r, z) = c n s ( r , z )σ e τ f =
τf Rate of spon tan eous emission I sat
Spontaneous emission Stimulated emission

Stimulated emission will dominate when I/Isat >>1 over the entire inverted region. This is much easier
to achieve in a fiber because of the small core size (in spite of the large value for Isat). Hence fiber lasers
tend to have higher slope efficiencies than comparable bulk solid-state lasers. At high power levels, bulk 17
lasers also suffer from detrimental thermal effects (notably thermal lensing), which also reduce the efficiency.
 References

1. C. J. Koester and E. Snitzer, “Amplification in a fiber laser,” Appl. Opt., vol.3, p.1182-1186 (1964).
2. J. Stone and C. A. Burrus, “Neodymium-doped fiber lasers: room temperature CW operation with an
injection laser pump,”Appl. Opt., vol.13, p.1256-1258 (1974).
3. R. J. Mears, L. Reekie, S. B. Poole and D. N. Payne, “Neodymium-doped silica single-mode fibre
lasers,” Electron Lett., vol.21, p.737-738 (1985).
4. R. J. Mears, L. Reekie, I. M. Jauncey and D.N.Payne, “Low-noise erbium-doped fibre amplifier
operating at 1.54µm,” Electron. Lett., vol.23, p.1026-1028 (1987).
5. E. Snitzer, H. Po, F. Hakimi, R. Tumminelli and B. C. McCollum, “Double-clad, offset-core Nd
fiber laser,” Proc. Conf. Optical Fiber Sensors, Postdeadline paper PD5, (1988).
6. V. Dominic, S. MacCormack, R. Waarts, S. Sanders, S.Bicknese, R. Dohle, E. Wolak, P. S. Yeh and
E. Zucker, “110W fiber laser,” Electron. Lett., 35, (1999), p1158-1159.
7. K. Ueda, H. Sekiguchi, and H. Kan, Conference on Lasers and Electro-Optics (Optical Society of
America, Washington, D.C.2003), Postdeadline paper CPDC4.
8. V. P. Gapontsev, D. V. Gapontsev, N. S. Platonov, O. Shkurihin, V. Fomin, A. Mashkin, M.
Abramov and S. Ferin, “2kW CW ytterbium fiber laser with record diffraction-limited brightness,”
CLEO-Europe 2005, paper CJ1-1-THU.
9. P. C. Becker, N. A. Olsson, and J. R. Simpson, “Erbium-doped fiber amplifiers: fundamentals and
technology” (Academic Press 1999).
10. A. W. Snyder and J. D. Love, “Optical Waveguide Theory,” (Kluwer Academic Publishers 2000).
11. W. J. Miniscalco, “Optical and electronic properties of rare earth ions in glasses,” in Rare-Earth-
Doped Fiber Lasers and Amplifiers, edited by M. J. F. Digonnet, 2nd edition, (Marcel Dekker, Inc.
New York, 2001).
12. K. Kubodera and K. Otsuka, J. Appl. Phys., vol.50, p.653 (1979).
13. W. A. Clarkson and D. C. Hanna, “Effect of transverse-mode profile on slope efficiency and
relaxation oscillations in a longitudinally-pumped laser,” J. Mod. Opt., Vol.36, p.483-498.
18
2. Cladding pumped fibers

19
 The ‘cladding-pumping’ concept
(a) Core pumping
Core Cladding
Output

Pump

Outer-coating Fiber laser

(Pump beam size × Far-field beam divergence) < (Core size × arcsin (Core NA))
If the core is single-mode then the pump source must be nearly single-mode

(b) Cladding pumping1

Core Inner-cladding
Output

Pump

Outer-cladding Fiber laser Double-clad fiber

(Pump beam size × Far-field beam divergence) < (Inner-cladding size × arcsin (Inner-cladding NA))
Can pump with high power + poor beam quality (i.e. low brightness) diode pump sources20
 Inner-cladding geometry:
Circular inner-cladding + off-set core
• Core must be off-set by a large amount
for efficient pump absorption
• More difficult to fabricate
Circular inner-cladding + centred core • Difficult to splice
• Easy to fabricate + splice
• Rays with trajectories that do not
pass through core Poor pump absorption
• Bending fiber helps but is not very effective
with large fibers + high NA’s

Polygon-shaped inner-cladding + centred core

• More difficult to fabricate
• Very effective way to increase pump
absorption efficiency
• Axially-symmetric
• Quite easy to splice

D-shaped inner-cladding + centred core Rectangular inner-cladding + centred core

• Quite easy to fabricate • More difficult to fabricate
• Very effective way to increase pump • Very effective way to increase pump
absorption efficiency absorption efficiency
• Difficult to splice • Helps with asymmetric pump beams
• Difficult to splice 21
 Silica versus other glasses

Silica is generally the material of choice for high-power fiber lasers and amplifiers
for the following reasons:

• Very high melting temperature and mechanical strength

• Relatively simple and well-established techniques for fiber fabrication
• Very low loss
• Good handling properties (e.g. splicing and cleaving)
• Compatibility with existing silica-based active and passive low power (e.g. telecom)
components
• RE-doping Useful range of operating wavelengths

Other glasses (e.g. fluoride glass) offer an extended range of operating wavelengths (e.g.
in visible and mid-infrared regimes), but fibers have a relatively poor power handling
capability and are not compatible with silica-based devices.

22
 Main laser transitions for high-power silica fiber devices

Nd3+ doping

λp ~800nm
λL ~1.36µm
λL 1050 – 1090nm

λL 905 – 940nm

Nd3+

Yb3+ doping
3 .0

2 .5

Cross-section (pm )
2
2 .0

λp ~900 - 980nm λL ~ 980nm, 1030 – 1120nm 1 .5

Absorption
1 .0 Emission
0 .5

0 .0
850 900 950 1000 1050 1100 1150
W a v e le n g th ( n m )

Yb3+ 23
Er3+ doping

In cladding-pumped fiber configurations

Yb3+ is used as a sensitizer2. Pump
light is absorbed by Yb3+, and Er3+ is
excited by energy transfer from Yb3+

λp ~910 - 980nm
λL ~ 1530 – 1620nm

Yb3+ Er3+
6

Cross-sections [x10-25 m2]

5 Emission
Absorption
Absorption (a.u.)

1
0
850 900 950 1000 1050
0
1400 1450 1500 1550 1600
Wavelength (nm)
Wavelength [nm] 24
 Tm3+ doping

Cross-relaxation

λp ~780 – 810nm
λp ~1200nm λL ~ 1730 – 2100nm
λp ~1550 - 1750nm
Tm3+ Tm3+

Emission cross-section (arb. units)

10

0
1500 1600 1700 1800 1900 2000 2100 2200 2300
W avelength (nm)

Pumping at ~ 780 – 800nm can be very efficient with careful optimisation of the core composition3,4:

• High Tm3+ concentration Efficient ‘two-for-one’ cross-relaxation

Higher absorption coefficient for pump
Short device length + reduced loss due to OH- impurity
25
• High Al3+ concentration Reduces clustering
Tm3+ doping

Can also use Yb3+ as a sensitizer5. Pump light is absorbed by Yb3+, and
Tm3+ is excited by energy transfer from Yb3+

λp ~900 – 980nm
λL ~ ~1900 – 2100nm

Yb3+ Tm3+

26
 Theory for amplifier and laser performance
Yb-doped silica fibers:
N1 and N2 are the population densities for the lower and upper
manifolds and N is the total RE ion concentration. λp and λL are
pump and lasing/signal wavelengths respectively.

N1 and N2 are determined by the rate equations for upper and

lower levels under steady-state conditions6:

= (R 12 + W12 )N1 − (R 21 + W21 + A 21 )N 2 = 0

dN 2
(1)
dt

= −(R 12 + W12 )N1 + (R 21 + W21 + A 21 )N 2 = 0

dN1
Yb3+ (2)
dt

where the transition rates are: R 12 = σa (λ p )I p / hν p , R 21 = σe (λ p )I p / hν p and A 21 = 1 / τ f

W12 = σ a (λ L ) I L / hν L and W21 = σ e (λ L ) I L / hν L

σa and σe are the absorption and emission cross-sections for lower and upper levels respectively and
τf is the lifetime of the upper level
N (R 12 + W12 )
N1 + N 2 = N N2 = (3)
R 12 + R 21 + W12 + W21 + A 21
27
 The pump loss and growth of the laser/signal light along the fiber are given by:

dPp
dz
( )
= ηp σ e (λ p ) N 2 − σ a (λ P ) N1 Pp (4)

dPL+
= ηL (σ e (λ L ) N 2 − σa (λ L ) N1 ) PL+ (5)
dz
dPL−
= − ηL (σ e (λ L ) N 2 − σ a (λ L ) N1 ) PL− (6)
dz

where ηp and ηL are overlap factors for the pump and signal/laser light with the doped core.
For a double-clad fiber ηp Aco/Acl, where Aco is the doped area of the core and Acl is the
inner-cladding area and ηL 1. Equations (4), (5), and (6) can be solved numerically to yield
amplifier gain6 and laser output power.

If we make some simplifying assumptions, then we can obtain approximate expressions for amplifier
small signal gain and laser threshold as follows7:

In the absence of laser or signal power, and assuming that ASE is negligible, then:

dPp (z) A co hν p N 2 (z)

=− ( 7)
dz ηq τ f
28
From (4) and (7):

Pp (z) Pp ( z) − Pp (0)
log e + = − ηp σ a (λ p ) N z (8)
Pp (0) Ppsat
where
hν p A co
Ppsat = is the pump saturation power
[σa (λ p ) + σe (λ p )] τ f ηq ηp

In general, equation (8) must be solved numerically to calculate the pump power absorbed in a
length of fiber. If Pp(0) << Ps (i.e. there is negligible ground-state depletion) then equation (8) can
be simplified to
(
Pp ( z) = Pp (0) exp − ηp σ a (λ p ) N z )
The single-pass gain G for the fiber is given by:

l
G = exp g( z)dz − α Ll where g ( z ) = N 2 ( z )σ e (λ L ) − N 1 ( z )σ a (λ L )
0

where αL is the attenuation coefficient for laser/signal light

29
Using (7), we obtain the following simplified expression for small signal gain:

(σe (λ L ) + σa (λ L )) τf ηq Ppabs
G = exp − σ a (λ L ) N l - α L l
A co hυp

Re-absorption loss Core propagation loss

where Ppabs is the absorbed pump power

G can also be expressed in dB as

(σ e (λ L ) + σa (λ L ) ) τf ηq Ppabs
10 log10 G = 4.34 − σ a (λ L ) Nl - α L l
A co hυp
103

Small signal gain (dB)

Example: Cladding-pumped Yb-doped fiber amplifier
102
Parameters: ra = 10µm, ηq = 1, τf = 1ms, λL = 1100nm
σe(λL) 2 × 10-23m2
101

Assume: Negligible propagation loss and σa(λL)Nl 0

(i.e. four-level approximation) 100
0 2 4 6 8 10

In practice, the small signal gain will be limited by the onset Absorbed pump power (W)

of parasitic lasing or amplified spontaneous emission to < ~30dB 30

Small-signal gain regime: G+ = G-
Saturated gain regime: G+ < G- Use counter-propagating pump and signal beams in power amplifier

Co-propagating pump and signal beams

Counter-propagating pump and signal beams

hν L A co
Signal saturation power: PLsat ≈ ≈ 0.28W for previous example
[σ a (λ L ) + σ e (λ L )] τf

Need input signal to exceed saturation power for amplifier to achieve efficient power extraction
Limits gain for practical power amplifiers to ~ 10dB – 30dB 31
 Power scaling with a master-oscillator power-amplifier (MOPA)

Need a chain of amplifiers to scale the output power from a low power master-oscillator

10 – 20dB 10 – 20dB 10 – 20dB

Output

Isolator Fiber amplifier

Master-oscillator

Increasing core area + Increasing pump power

32
Threshold pump power:

T = transmission of output coupler

L = Cavity loss due to imperfect feedback at opposite end of cavity

Round-trip gain at threshold:

(σe (λ L ) + σa (λ L )) τf ηq Ppabs
2

G = exp − σ a ( λ L ) Nl - α L l (1 − T ) (1 − L ) = 1
A co hυ p

A co hυp
Ppabs , th = [− log e (1 − T) − log e (1 − L) + 2σ a (λ L ) Nl + 2α Ll ]
2τ f ηq (σe (λ L ) + σ a (λ L ) )

Previous example: Cladding-pumped Yb-doped fiber laser

Parameters: ra = 10µm, ηq = 1, τf = 1ms, λL = 1100nm

Ppabs ,th ≈ 0.46 W
σe(λL) 2 × 10-23m2, L = 0, T = 96%

Assume: Negligible propagation loss and σa(λL)Nl 0 33

(i.e. four-level approximation)
Slope efficiency:

At power levels well above threshold (i.e. PL >> PLsat) and assuming that the core
propagation loss is negligible, then the slope efficiency is given by the following
approximate expression:

dPLout T 1− L λp
ηs = ≈ ⋅ ⋅ ηq ηabs
dPp T 1− L + L 1− T λL
When T + L <<1, this simplifies to

dPLout T λp
ηs = ≈ ⋅ ⋅ ηq ηabs
dPp T + L λL

In a typical cladding-pumped fiber laser, the transmission of the output coupler is very
high and the fiber length is selected so that the absorption efficiency (ηabs) is ~1, so
ηs ηqλp/λL providing there is sufficient feedback at the opposite (non-output) end of
the cavity.

34
 Effect of feedback efficiency on laser performance

P1 = Output power P2 = total power loss in feedback

optical arrangement

Active fiber of length l Feedback mirror or FBG

Output coupler
or external cavity

Rigrod analysis8: Power in fiber

P1
1 − R1
- P1R 1G + P2
+ =
1 − R1 1− R 2

P1R 1R 2 G +
P1R 1
1 − R1
1 − R1
z=0 z=l
+
P1+ P1R 1R 2 G + G −
= → R 1R 2 G + G − = 1
1 − R1 1 − R1
35
G+ = G- = Effective single-pass gain (taking into account core propagation loss)
Simple result:
P1 1 − R 1 R2
=
P2 1 − R 2 R1

For a high lasing efficiency we require P1>>P2

Example: Cladding-pumped fiber laser with feedback for lasing provided by a

perpendicularly-cleaved fiber end facet and an external feedback cavity.

Question: What feedback efficiency for the external cavity is required for efficient operation?

• Relatively modest feedback efficiency for 50

external cavity is needed
• Constraints on feedback cavity design can 40
further relaxed by reducing R1 (e.g. by anti-
30
reflection coating the fiber end facet)

P1/P2
R1 = 1%
20
Very high gain Enormous flexibility in cavity design
without compromising efficiency 10
R1 = 3.6%
Warning: Using a very low reflectivity output coupler 0
0 20 40 60 80 100
makes the fiber laser very susceptible to
R2 (%) 36
optical feedback from external components
 External feedback cavity design
Generalised feedback arrangement:
Optical element for extending functionality
(e.g. Q-switch, wavelength tuning element)

Main design issues: f1 f2 f3

• Broadband feedback from end-facet adjacent HR
to external cavity should be suppressed so that it
does not compete with external cavity’s function
Angled end-facet
• Low loss + AR coated intracavity components required
• Degradation in beam quality (M2) due to lens aberration9
can dramatically reduce the feedback efficiency or
8πC 4 w i4 8πC 4 f θ4
M = ( M ) + (M )
2 2 2 2 2
where M = 32
≈
f λL 2 λL 2
f i q q
End-cap

where wi is the beam radius on the lens, θ is the far-field beam

divergence and C4 is the quartic phase aberration coefficient Most of the reflected light is guided in the
( = 0.29 for a plano-convex lens for infinite conjugate ratio) inner-cladding where the propagation loss
is generally much higher and the effective
gain is much lower Dramatic reduction in
Need to keep beam divergence <0.05 rad with singlet lenses feedback from end-facet
to avoid degradation in beam quality or use aspheric or 37
multi-element lenses
Simple feedback cavity:
P2 lost in feedback arrangement

• Power in feedback cavity << Output power ( P1)

• Power handling capability of feedback cavity components may not be an issue

Simple + effective method for scaling to very high (>1kW) power levels10,11

Alternative resonator feedback schemes:

(a) Butted mirror

• Simple, but limited scope for

power scaling due to mirror
damage
HR
(b) Fiber Bragg grating • All fiber
• Very high feedback efficiency
FBG • Power handling??
38
 References
1. E. Snitzer, H. Po, F. Hakimi, R. Tumminelli and B. C. McCollum, “Double-clad, offset-core Nd fiber
laser,” Proc. Conf. Optical Fiber Sensors, Postdeadline paper PD5, (1988).
2. J. D. Minelly, W. L. Barnes, R. I. Laming, P. R. Morkel, J. E. Townsend, S. G. Grubb and D. N. Payne,
“Diode-array pumping of Er/Yb co-doped fiber lasers and amplifiers,” IEEE Photon Tech. Lett., vol.5,
p.301-303 (1993).
3. S. D. Jackson, “Cross relaxation and energy transfer upconversion processes relevant to the
functioning of 2µm Tm-doped silica fibre lasers,” Opt. Comm., vol.230, p.197-203 (2004).
4. G. Frith, D. G. Lancaster and S. D. Jackson, “85W Tm3+-doped silica fibre laser,” Electron. Lett.,
vol.41, p.687-688 (2005).
5. Y. Jeong, P. Dupriez, J. K. Sahu, J. Nilsson, D. Y. Shen, W. A. Clarkson and S. D. Jackson, “ Power
scaling of 2µm ytterbium-sensitised thulium-doped silica fibre laser diode-pumped at 975nm,”
Electron Lett., vol.41, p.173-174 (2005).
6. R. Paschotta, J. Nilsson, A. C. Tropper and D. C. Hanna, “Ytterbium-doped fiber amplifiers,” IEEE J.
Quantum Electron., vol.33, p.1049-1056 (1997).
7. H. Pask, R. J. Carman, D. C. Hanna, A. C. Tropper, C. J. Mackechnie, P. R. Barber and J. M. Dawes,
“Ytterbium-doped silica fibers: Versatile sources for the 1-1.2 µm region,” IEEE J. Selected Topics in
Quantum Electronics, vol.1, p.2-12 (1995).
8. W. W. Rigrod, “Saturation effects in high-gain lasers,” J. Appl. Phys. vol.36, p.2487-2490 (1965).
9. A. E. Siegman, “Analysis of laser beam quality degradation caused by quartic phase aberrations,”
Appl. Opt., vol.32, p.5893-5901 (1993).
10. Y. Jeong, J. K. Sahu, D. N. Payne and J. Nilsson, “Ytterbium-doped large-core fiber laser with
1.36kW continuous-wave output power,” Opt. Express, vol.12, p. 6088-6092 (2004).
11. A. Tunnermann, T. Schreiber, F. Roser, A. liem, S. Hofer, H. Zellmer, S. Nolte and J. Limpert, ”The
renaissance and bright future of fiber lasers,” J. Phys. B: At. Mol. Opt. Phys., vol.38, p.681-693
(2005).
39
3. Pump sources and pump coupling

40
 Introduction

• Efficient coupling of pump light from one or more high power diode laser pumps into
double-clad fibers is essential for efficient power and brightness scaling of cladding-
pumped fiber sources.

• This requires careful selection of the pump diode(s), and very careful design and precise
alignment of the pump light collection, delivery and coupling optical arrangement.

• The choice of pump source and incoupling scheme is a major element of the overall fiber
system design and can have a huge impact on the overall system performance and on
flexibility in mode of operation.

• Pump brightness is usually the key factor. The higher the pump brightness, the more
flexibility there is in the choice of fiber design and mode of operation, and the easier it is
to scale to high output powers.

• Advances in high-power diode pump sources and pump coupling techniques over the last
decade have been dramatic, and this has probably been the single biggest factor in taking
fiber-based sources to the power levels reported to date

41
 Brightness and the M2 parameter

Power
Brightness (or radiance) ≡ ( Units : Wm − 2sr −1 )
Area × Solid angle

Power
Fundamental transverse mode (diffraction-limited): B∝
λ2
Power
Non-diffraction-limited beam (Mx,y2 >1): B∝
M 2x M 2y λ2

where Mx,y2 is the beam propagation factor1

• B determines the maximum focussed intensity

• B is invariant as beam propagates through a (perfect) lens system

B determines the maximum pump power that can be coupled into a double-clad fiber
42
 Calculating pump beam sizes and beam divergences

Mx2, My2 Mx2, My2

Optical component

As Bs
Ray Transfer Matrix for optical system : M s = = M n M ( n −1) ..........M 3 M 2 M1
Cs Ds
A s q in x , y + Bs
Using ABCD law2: q out x , y =
Cs q in x , y + D s

1 1 jM 2 λ
where q is the complex beam parameter given by : = −
q R πw 2
w = beam radius and R = radius of curvature of wavefronts

Simple procedure for calculating output beam parameters from an optical system
given knowledge of the input beam parameters and Ms.

Important assumption: Mx2 and My2 are preserved 43

 Focusing pump light into double-clad fibers
u

w2 f
2rb =
w1 z o21 + (u − f ) 2

πw12
z o1 = 2
Beam waist
Perfect
Beam waist Mλ
radius = w1 radius = w2
lens

For efficient launching of pump light into the inner-cladding:

λM 2p
rb ≥
πθna γ uf

where na = arcsin(NA) and uf is a factor which takes into account the need to
underfill the inner-cladding and inner-cladding’s NA to avoid pump-induced
damage to the outer-coating. The value for uf depends on the situation and, in
particular, the pump power.
Rough guide: For efficient pump coupling, the beam propagation factor
for pump source must satisfy:

πθ na rb γ uf
M 2p ≤
λp

Typical situation:
2rb (µm) na (rad) Mp2
Double-clad fiber with
125 0.41 66
• NA = 0.4
• p = 980nm 200 0.41 105
• uf = 0.8
400 0.41 210

600 0.41 315

45
 Diode Laser Pump Sources

Wavelength options:
GaN 380-nm – 480nm Pr3+
GaInP, AlGaInP 640nm – 680nm Cr:LiCAF, Cr:LiSAF
AlGaAs, GaAs 780nm – 860nm Nd3+, Tm3+
InGaAs 900nm – 980nm Yb3+, Er3+
InGaAsP/InP 1.47µm – 1.6µm Er3+
InGaAsP 1.8µm – 1.96µm Ho3+

Diode laser types:

(a) Single-stripe (single-mode) diode lasers

Emitter size ~1µm × few µm
Beam divergence (FWHM): y 25°- 30º (perpendicular to junction) and
x 7º (parallel to junction)
My2 = Mx2 = 1
Max. cw output power ~0.5 – 0.8W (limited by catastrophic failure)

46
(b) Broad area diode lasers
Emitter size ~1µm × ~100µm
Beam divergence (FWHM): y 25°- 30º and x 8º
My2 = 1 and Mx2 ~15-20
Max. cw output power ~ 7 - 8W

(c) Diode-bars
Emitting region ~1µm × 10mm
Beam divergence (FWHM): y 25°- 30º
and x 6°- 9º
My2 = 1 and Mx2 ~ 1300-1800
Max. cw output power ~ 40-120W

(d) Diode-stacks
Emitter region ~ N × bar spacing × 10mm
Beam divergence (FWHM): y 25°- 30º and x 9º
My2 [(N-1) × bar spacing/emitter height] + 1
and Mx2 ~ 1300-1800
Max. cw output power ~ 40-120W × N
47
 Main requirements for efficient pump coupling:

1. Selection of appropriate diode pump laser(s)

2. Pump light collection and aperture filling

3. Re-formatting of the beam using a ‘Beam - Shaper’ to roughly equalise the

M2 parameters in orthogonal planes preferably without decreasing
the brightness

Beam shaping

Mxi2 >> Myi2

Mxf2 Myf2 (Mxi2Myi2)0.5

4. Scheme for launching into fiber

5. Management of stray pump light

48
 How much pump light can be launched into a double-clad fiber?

The maximum launched pump power depends on:

• Brightness of a constituent emitter of the pump sources
• Pump collection, delivery and coupling scheme
• Fiber’s inner-cladding size and NA

Theoretical upper limit (i.e. for a ‘perfect’ pump launching scheme) is given by:
2
Ps πrb θna γ uf
Pp max ≈
M 2x M 2y λp 105
NA=0.6

104
where Ps is the power of a single (constituent) NA=0.4

Ppmax (W)
emitter and Mx2 and My2 are its beam
propagation factors in orthogonal planes 103 NA=0.2

Example: Ps = 8W, Mx2=20, My2=1

102

Very high launched pump powers, but not

possible to achieve in practice due to lens 101
imperfections and tight alignment tolerances 0 100 200 300 400 500 600
Inner-cladding diameter (µm) 49
 Pump launching schemes:

1. End-pumping:
- Higher brightness pumping
- Short device length Higher threshold for unwanted nonlinear loss processes
Lower propagation loss
- Flexibility in fiber design + better fiber laser/amplifier performance
- Free-space beams optical components needed
- Alignment issues
- High pump deposition density at fiber end

2. Side-pumping:
- Multi-point pump injection and distributed pump injection configurations
- Distributed heat loading
- Easy access to fiber ends for splicing to other components
- All fiber architecture Robust + fewer optical components + no alignment issues
- Lower brightness pumping
- Longer device length Higher propagation loss
Lower threshold for nonlinear loss processes

50
 Examples of pump launching schemes

V-groove scheme:

• Simple side pumping scheme that requires few optical components3,4

• Beam divergence, arcsin(NA) Limits use to broad-area diode pumps
• Relatively large degradation in pump brightness on launching
• Power scaling via the use of multiple pump injection points

51
Multimode fiber coupler5:

Fiber-coupled broad-area diode

Multimode
coupler

Tapered Fiber bundle

bundle
Core Splice
Fiber-coupled
diode
Buffer

Inner-cladding (rθna )final

2
≥ (rθna )input
2

n
Outer-cladding
n = number of input fibers
or
7 pump fibers into 6 pump fibers into
1 pump fiber 1 active fiber

52
Can be used for end-pumping and/or multi-injection-point side pumping
 Distributed pump coupling scheme7:

High NA multimode Active fiber + pump fibers

pump Fibers in optical contact Pump port (× 4)

Laser
Output

∆n(r)

• Pump and active fibers are in optical contact Breaks symmetry Efficient pump absorption
• More uniform pump deposition than for end-pumping or multi-point injection
• Access to fiber ends for splicing to other fibers 53
 Side-pumped coiled fiber8:

Laser
output Coiled fiber

Diode
bars

• Coiled fiber with rectangular inner-cladding embedded in a disk that acts as a waveguide for
pump light
• Can be pumped by diode bars (or stacks)
• Disk geometry Simple thermal management
• More suitable for four-level transitions

54
 Pumping schemes for diode arrays
• Side pumping or end pumping schemes may be used
• In both cases, the output from the diode array must be re-formatted to allow
efficient coupling into the active fiber or pump delivery fiber
• Fiber end termination must be designed to eliminate the risk of damage due to
stray (uncoupled) pump light

Example of fiber end termination:

Main features:
• Section of uncoated fiber protruding from heat-sink so that uncoupled pump light
spreads out by diffraction and is not incident on the outer-coating
• Heat-sink over first section of coated fiber minimises the risk of damage due to
high divergence pump light leaking into the outer-coating 55
 Fiber-bundle-coupled diode-bars: (see refs. 9 and 10)

(a)

• Simple and robust way to equalise

M2 parameters for diode-bars
• Fibers are under-filled
Low brightness
• Output from bundle can be coupled
into a single delivery fiber using
lenses or a multimode coupler
(tapered fiber bundle)

(b)
• Combination of free-space optics
and fiber bundle approach
• Can be used to equalise M2 parameters
for diode-bars or stacks
• Higher brightness than method (a),
but more complicated and expensive
• Output from bundle can be coupled
into a single delivery fiber as for
method (a) 56
 Aperture filling
Diode bars and stacks have a large area of ‘dead space’ between the actual emitting regions. For
applications requiring high brightness beams it is usually necessary to use cylindrical microlenses and
cylindrical lens arrays to collimate (reduce the beam divergence) from individual bars and individual
emitters to increase the brightness. This is often referred to ‘Aperture filling’ and is crucial for most
fiber laser/amplifier pumping applications.

Fast-axis collimation: wxf,yf = beam width after collimation in the x and y directions
wxi,yi = beam width at emitter in the x and y directions
xf,yf = far-field beam divergence after collimation in the x and y directions
xf,yf = far-field beam divergence without collimation in the x and y directions

M 2yf θyf w yi
= ≈
M yi θyi w yf
2

Slow-axis collimation:

M 2xf θxf w xi
= ≈
M xi θxi w xf
2

57
 Free-space beam shaping techniques

(a) Two-mirror beam shaper11

(1)-(5)
Plan view
Output
HR beams

x z

HR
y

(1) A
(2) B
(3)
Incident (4) • Two mirrors used to slice beam in poor beam quality
beam (5)
direction and stack resulting N beams in the orthogonal
direction

Side view
y M 2xf ≈ M 2yf ≈ 1.3M 2xi M 2yi

x
Output
by choosing N ≈ M 2xi / 1.3M 2yi
HR beams
z
(1)-(5)
HR • Simple + low loss
B
A • Can be used for bars or stacks
• Output can be focussed into the active fiber or delivery
fiber 58
(b) Stacked-plate beam-shaper12

• Beam is sliced and beam components

are then re-directed and stacked
(by refraction) using two orthogonal
stacked multi-glass-plates

• Mx2 is reduced by factor N, and

My2 is increased by factor N, where
N is the number of plates in each stack

• Can be used for bars or stacks

59
 Summary

• There are many diode pump source architectures and many different coupling schemes
that may be employed to launch pump light into double-clad fibers

• Emitter brightness is ultimately the limiting factor on how much pump power
can be launched into a given fiber

• Still a great deal of scope for improvement in pump launching schemes

• Alignment tolerances are difficult to satisfy Compromise between cost/simplicity

and performance

60
 References

1. International standard: ISO 11146 (1999).

2. H. Kogelnik and T. Li, “Laser beams and resonators,” Appl. Opt., vol.5, p.1550-1567 (1966).
3. R. J. Ripin and L. Goldberg, “High efficiency side-coupling of light into optical fibres using
imbedded v-grooves,” Electron. Lett., vol.31, (1995), p2204-2205.
4. L. Goldberg, J. P. Koplow and D. A. V. Kliner, “Highly efficient 4W Yb-doped fiber amplifier
pumped by a broad-stripe laser diode,” Opt. Lett., vol.24, (1999), p673-675.
5. http://www.itfoptical.com
6. http://www.sifamfo.com
7. www.spioptics.com
8. H. Sekiguchi, G. G. Vienne, A. Tanaka, Y. Senda, K. Ito, Y. Matsuoka, H. Toratani, S. Takahashi, H.
Miyajima, H. Kan, and K. Ueda, “New Concept: fiber embedded disk and tube lasers”, International
Forum on Advanced high-power lasers and Applications (AHPLA'99) Osaka, Japan, Nov. 1-5. 1999.
9. T. Baer and D. F. Head, “Apparatus for coupling a multiple emitter laser diode to a multimode optical
fibre,” US patent No.5579422.
10. D. C. Shannon and R. Wallace, “Multiple diode laser stack for pumping a solid-state laser,” US
patent No. 5,299,222 (1994).
11. W. A. Clarkson and D. C. Hanna,”Two-mirror beam-shaping technique for high-power diode bars,”
Opt. Lett., vol.21, (1996), p375-377.
12. C. Ullmann and V. Krause, “Diode optics and diode laser,” US patent No. 5986794, (1999).

61
4. Thermal effects in fibers

62
 Heat generation
Various sources: N2

hνp hνL
• Quantum defect heating:
N1

E=h p –h L HEAT RE3+

• Excited-state absorption (ESA)

At pump and/or signal wavelength

RE3+

• Energy-transfer-upconversion (ETU)

Depends on RE3+ concentration and excitation density

High conc. + high excitation density can lead to a
significant decrease in efficiency + extra heating due
to ETU. RE3+ RE3+ 63
Various sources (cont’d):
• Impurities
• Non-radiative sites
• Absorption of fluorescence and/or stray pump
light in outer-cladding and/or mount

Quantum defect heating is often considered to be the main source of heat, but
this is not always correct. If it is then:

Fraction of absorbed pump converted to heat: h =1– L/ p

Rare earth ion Pump wavelength Lasing wavelength γh

(nm) (nm)
Nd3+ ~800 ~ 920 - 940 ~ 0.13 – 0.15
~ 1050 - 1080 ~0.24 – 0.26
~1360 ~0.41
Yb3+ ~915 - 980 ~ 980 – 1140 ~0.05 – 0.20

Er3+,Yb3+ ~915 - 980 ~ 1530 – 1620 ~0.36 – 0.44

Tm3+ ~790, ~1550 ~ 1720 – 2100 ~0.10 – 0.62

64
 Thermal effects

Pump Heating Laser

power output

Heat
Heat removal Heat removal

Temperature increase Temperature gradient

• Damage due to melting • Thermal lensing/guiding

and/or burning or anti-guiding
• Change in spectroscopy beam distortion + loss
• Change in thermo-optical • Thermally-induced stress
and thermo-mechanical Birefringence + lensing
properties Fracture

Thermal effects become more pronounced at high pump powers

Degradation in beam quality + reduced efficiency + eventually damage 65
Impact of thermal loading on power scalability depends on laser medium geometry

Heat
Laser Heat
Laser

ROD SLAB

Laser
Heat
Pump Pump
Laser

Pump

Heat
THIN DISC CLADDING PUMPED FIBRE 66
Attractions of fibers:

• Generated heat can be spread over a long device length

(typically a few metres to a few tens-of-metres)
• Large surface area/core volume facilitates heat removal
• Waveguiding properties of core (i.e. refractive index profile)
usually dominate over thermally-induced changes in
refractive index.

Thermal management is easier than for other laser medium geometries,

but fibers are not completely immune to the effects of thermal loading1,2,3,4,5.

As power levels from cladding-pumped fiber lasers and amplifiers rise,

thermal loading is starting to become a serious issue (especially in rare
earth-doped fibers with a relatively large value for h).

Also, thermal effects in external components (e.g. Faraday isolators) may

impact on the overall performance of the system
67
Typical fiber heat-sinking configurations:

• For fiber end-section in an end-pumped configuration

• For a side-pumped fiber configuration

68
 HEAT
HEAT Outer-cladding
Core radius = ra

r
Pump

2rb z
2rc

Inner-cladding
HEAT
HEAT

Region 1 Region 2

Under steady state conditions6: ∇ .h( r , z ) = Q( r , z )

h( r , z ) = − K c ∇ T ( r , z )

where h(r,z) is the heat flux, Q(r,z) is the heat deposition density, T(r,z) is the
temperature and Kc is the thermal conductivity
69
r h(r,z) Assuming heat flow is purely radial, then
total heat radial flow from volume bounded
z
by z and z + z is equal to the net heat
generated in this region

z z+ z
z + ∆z r
2π r ∆z h( r, z ) = 2π r ' Q( r ' , z )dr ' dz '
z 0

Pp (z) α p γ h r
h (r , z ) = ( r ≤ ra )
2π r a
2

Pp (z) α p γ h
h (r , z ) = (r > ra )
2π r

where Pp(z) = Pp(0)exp( pz) is the pump power at position z and p is the absorption
coefficient for pump light in the active ion doped core. 70
Boundary conditions: h(r,z) and T(r,z) are continuous at boundaries between layers

Thermal conductivity: Kc=Kic for r rb (i.e. core and inner-cladding have the same
thermal conductivity)
Kc=Koc for r > rb

Resulting temperature distribution can be obtained from:

r
1
∆T ( r , z ) = T ( r , z ) − T(0, z ) = − h ( r ' , z )dr '
Kc 0
Region 1:
Ph (z) 1 2 rb 2
T(0, z) − Ts = + log e +
4π K ic K ic ra rb H1

Region 2:
Ph ( z) 1 2 rb 2 rc 2
T(0, z) − Ts = + log e + log e +
4π K ic K ic ra K oc rb rc H 2

where Ph(z) = Pp(z) p h is the heat generated per unit length, Ts is the ambient
temperature of the surroundings, T(0,z) is the temperature at the core’s centre and
H1,2 is the heat transfer coefficient for the inner/outer cladding in regions 1,2 71
Maximum heat deposition per unit length before onset of softening or melting:

Region 1:
−1
1 2 r 2
Ph max = 4π(Tm − Ts ) + log e b +
K ic K ic ra rb H1

For convective cooling, this term dominates

and hence a larger diameter fiber with a larger
surface area facilitates heat removal

Region 2:
−1
1 2 r 2 r 2
Ph max = 4π(Tm − Ts ) + log e b + log e c +
K ic K ic ra K oc rb rc H 2

For double-clad fibers with polymer outer-claddings

Koc<<Kic, so the outer-cladding should be as thin as
possible to avoid a larger temperature rise.

where Tm is the softening/melting temperature of the core 72

 Thermal loading limit for a typical double-clad silica fiber
Fibre design: Core diameter = 25µm
Inner-cladding diameter = 200, 300, 500µm
Outer-cladding diameter = 300, 400, 600µm For silica: Tm 2000K and
Pure silica inner-cladding: Kc 1W/mK assuming Ts=293K
Polymer outer-cladding: Kc 0.1W/mK

Region 1 Region 2
10000 10000

1000 rb=500µm
1000 rb=500µm

Ph (Wm-1)
Ph (Wm-1)

rb=300µm rb=300µm

rb=200µm rb=200µm
100 100

Convection cooling Conduction cooling

10 10
101 102 103 104 105 101 102 103 104 105

Heat Transfer Coefficient (Wm-2K-1) Heat Transfer Coefficient (Wm-2K-1)

For convection cooling1: H ~ 10Wm-2K-1

Upper-limit on power that can be extracted from fiber: Pmax Ph(1 - h) / h

e.g. For an Yb fiber laser at 1080nm pumped at 980nm: Pmax 10Ph
For an Er,Yb fiber laser at 1550nm pumped at 980nm: Pmax 1.7Ph 73
Suggests that typical double-clad fibers can handle heat deposition densities in the range
~20 – 50W/m to >1kW/m depending on the cooling scheme before the onset of damage
due to melting.

However, this is not the whole story since in Region 2 the threshold for damage to the
outer-coating is reached first. For a typical polymer outer-coating the maximum
temperature, Td, that can be tolerated before the coating begins to degrade is ~150°C.
This imposes the following upper-limit on the heat deposition density:
−1
2 r 2
Ph max = 4π(Td − Ts ) loge c +
K oc rb rc H 2

1000

Aggressive cooling needed for 100 rb=500µm

Ph (Wm-1)
scaling to high power levels rb=300µm

rb=200µm
Thinner coating + larger inner 10

-cladding facilitates thermal

management 1
101 102 103 104 105
Heat Transfer Coefficient (Wm-2K-1) 74
 Temperature profile across fiber

120

Temperature rise (K)

100
Ph = 100W/m
Core diameter = 25 m 80
Inner-cladding diameter = 400µm
Outer-cladding diameter = 600µm 60

40

20

0
-300 -200 -100 0 100 200 300
Radial position (µm)

• Large temperature drop over outer-coating

Use thinner coating
• Temperature drop between outer-coating and heat-sink / surroundings depends
on the heat-transfer coefficient and can be very large

75
 Jacketed-Air-Clad Fibers

• Becoming increasingly popular for high power operation

• All glass structure with high NA for inner-cladding
• Pump light does not interact directly with outer-coating
• Outer-coating can be selected for thermal properties

Ph ( z ) 2 r Ph ( z ) t
Temp rise across air cladding ∆T = log e c ≈
4π K eff rb 2πK eff rb

where Keff is the effective thermal conductivity of the ‘air’ layer and t is the
thickness of the air cladding. If the ‘struts’ that suspend the inner-cladding
are very thin, then Keff Kair 0.025-0.025Wm-1K-1.

For a small temp rise we require: t < Keff rb

76
t ~ few m
 Thermal guiding
Two possibilities:
1. dn/dT > 0
Step-index profile
for core
• Tighter mode confinement
• Degradation in beam quality
• Reduced efficiency

2. dn/dT < 0
• Less mode confinement
• Mode distortion
• Increased propagation loss
• Reduced efficiency

For silica glass: dn/dT > 0 and for phosphate glass dn/dT < 0 77
 Case 1:

Step-index waveguide has a fundamental mode radius wo given by7 w o ≈ ra / loge V

where V 4 nora n/
Ph dn / dT
n( r ) = n o − n th r 2 where n th ≈
4πra2 K c

1/ 4
2 K cλ2 ra2
Thermal guiding with mode radius: w th ≈
πn o Ph dn / dT

Hence, thermal guiding starts to have a significant impact when wth wo

i.e. when:
1000
2 K cλ2 (loge V) 2
Ph ≥
πra2 n odn / dT 100
V=2.4

Ph (W/m)
10
Thermal guiding will have significant effect
on guiding and hence performance in
large-core fibers 1
0 10 20 30 40 50 60 78
Core radius (µm)
 Summary

• Double-clad fibers have excellent thermal properties compared to

conventional solid-state lasers, but are not immune from the effects of heat
generation.

• Degradation and/or damage to the fiber outer-coating due to the high

temperatures which result from heat generated in the core and poor thermal
management is one of the main failure mechanisms at high power levels.
Improved heat-sinking can remedy this allowing scaling to much higher power
levels

• Thermal guiding is not a serious problem in conventional (small) core fiber

designs, but will start to impact on performance in very large core / short length
devices

• Thermal lensing in ‘free-space’ components (e.g. Faraday isolators) is an issue

at high power levels and requires the use of very low absorption materials and
careful design to eliminate and/or compensate for beam distortion

79
 References

1. D. C. Brown and H. J. Hoffman, “Thermal, stress and thermo-optic effects in high average power
double-clad silica fiber lasers,” IEEE J. Quantum Electron., vol.37, p.207-217 (2001).
2. L. Li, H. Li, T. Qiu, V. L. Temyanko, M. M. Morrell, A. Schulzgen, A. Mafi, J. V. Moloney and N.
Peyghambarian, “3-Dimensional thermal analysis and active cooling of short-length high-power fiber
lasers,” Opt. Express, vol.13, p.3420-3428 (2005).
3. N. A. Brilliant and K. Lagonik, “Thermal effects in a dual-clad ytterbium fiber laser,” Opt. Lett.,
vol.26, p.1669-1671 (2001).
4. D. C. Hanna, M. J. McCarthy, and P. J. Suni, “Thermal considerations in longitudinally pumped fibre
and miniature bulk lasers,” in Fiber laser sources and amplifiers, M. J. F. Digonnet, eds., Proc. SPIE
1171, 160-166 (1989).
5. Y. Wang, “Thermal effects in kilowatt fiber lasers,” IEEE Photon. Tech. Lett. vol.16, p.63-65 (2004).
6. M. E. Innocenzi, H. T. Yura, C. L. Fincher and R. A. Fields, “Thermal modelling of continuous-wave
end-pumped solid-state lasers,” Appl. Phys Lett., vol.56, p.1831-1833 (1990).
7. A. W. Snyder and J. D. Love, “Optical Waveguide Theory,” (Kluwer Academic Publishers, 2000).

80
5. Nonlinear processes

81
 Introduction
The response of a dielectric medium to light becomes increasingly nonlinear as the intensity grows.
In optical fibers, the combination of tight optical confinement of light in the core, a long interaction
length and low loss leads to conditions where nonlinear processes can have a very dramatic impact
on the fiber device performance1. Nonlinear effects in fibers can be beneficial or detrimental to
performance depending on the particular nonlinear process and the desired mode of operation.

Light incident on a medium Polarisation P induced by the electric field E according to the relation:

P = εo χ(1) E + εo χ( 2 )EE + εo χ( 3)EEE + ..............

Linear response Nonlinear response

where εo is the permittivity of free space

χ(1) the linear susceptibility Main contribution to P
χ(2) is the second-order susceptibility and is zero for materials (e.g. silica) with inversion
symmetry
χ(3) is the third-order susceptibility third harmonic generation, four-wave mixing and
nonlinear refraction
Nonlinear refraction is generally the most important χ(3) process in fibers n (ν, I) = n o (ν ) + n 2 I
n2 is the nonlinear refractive index coefficient and is related to χ(3). In silica, n2 2.2 – 3.4×10-20m2/W

Self-phase modulation (SPM) and cross-phase modulation (XPM) 82

Spectral broadening of ultrashort pulses
 Stimulated Inelastic Scattering

In addition, there are also nonlinear processes that result from stimulated inelastic scattering.
Two important processes in optical fibers are stimulated Raman scattering and stimulated
Brillouin scattering.

(a) Stimulated Raman scattering (SRS)

Scattering of an incident (pump) photon (νp)

to a lower frequency photon (νs) as the molecule
makes a transition between two vibrational states

Energy difference ∆E = hνR = hνp – hνs is absorbed in

the medium as heat.
Optical
phonon
Frequency down-shifted wave (νs) is the Stokes wave
and the frequency shift (νR ) is known as the Raman
Spontaneous Raman scattering
shift

Phase-matching is not required for this process

83
 Raman amplification

dIs
= g R I p Is − α s Is
dz
dI p g R I p Is ν p
=− − α pIp
dz νs
Stimulated Raman scattering αp and αs are the loss coefficients for the pump and
Stokes wave, and gR is the Raman gain coefficient

• The values for gR and νR, and the frequency range over which Raman gain extends
depend on the material.

• Amorphous materials have a very broad Raman gain spectrum. For silica, Raman
gain can be achieved over a frequency range of up to ~ 40THz

• For silica the max value for gR ~ 1×10-13m/W and corresponds to a Raman frequency
shift of ~13THz

• gR can vary significantly with the core composition (i.e. adding dopants can have a
significant effect on the value for gR and the value for the Raman frequency shift at
84
which gR is a maximum.
Stimulated Raman scattering has a number of important applications and there are a
number of fiber-based devices which make use of SRS:

• Fiber Raman amplifiers2 Distributed + broadband amplification

• Fiber Raman lasers3 Extension to wavelength regimes that are not covered
by RE-doped fiber lasers
• Broadband wavelength generation

However, SRS can also be detrimental to the performance of RE-doped fiber lasers and
amplifiers since it act as a loss.

Threshold for SRS1:

16A eff
SRS
Ppth ≈
g R leff
where leff =
1
αp
[
1 − exp(− α pl ) ] is the effective length and Aeff is the effective core area.

This assumes that the Raman gain spectrum has a Lorentzian lineshape and the
polarisation of the pump and Stokes waves are maintained. Once the threshold for SRS is
reached, the Stokes wave can grow very efficiently at the expense of pump power. The
Stokes wave may then act as the pump for a second-order Stokes wave and so on.
This is one of the limiting factors on the maximum power that can be obtained from a
cladding pumped fiber laser or amplifier 85
Example: Typical cladding-pumped Yb-doped silica fiber laser with

Core diameter = 15µm Aeff 1.77×10-10m2

leff = 20m
SRS
Ppth 1.4kW

This is only a very rough guide, but it shows that SRS is potentially a problem in very
high power cw fiber systems and quite modest peak power pulsed fiber sources.

Possible solutions include:

• Increasing the core area – Limited scope without degrading beam quality

• Shorten device length – More demanding on diode pump beam quality and
thermal management

• Distributed loss filter to suppress Stokes generation

86
 Raman Lasers
Attractions: - Access to wavelength regimes not available in RE-doped lasers
- Low thermal loading if wavelength shift is small

Cascaded Raman Laser:

e.g. Cladding-pumped
Yb-doped fiber laser

Example: Yb fiber pump laser at 1.06µm

Phosphosilicate Raman fiber: λs1=1.24µm and λs2=1.48µm
Requires very long lengths of fiber (~1km) for a low threshold

>1W power levels at 1.48µm have been generated4,5,6 with efficiencies up to 48% wrt
Yb fiber laser power. 87
 Cladding-pumped Raman fiber laser

The cladding-pumping concept can also be applied to Raman fiber lasers as a means
for enhancing brightness as well as extending wavelength coverage7,8. However, direct
diode pumping with currently available high power diode sources is not practical
due to their low brightness, so an intermediate ‘brightness-enhancement’ stage is needed.

Basic idea:

• Brightness enhancement in two stages

• Design constraints on RE-doped fiber
laser are more relaxed
• Heat loading is distributed over two
stages and can be very low in the Raman
fiber laser
• Small cladding-to-core area ratio in Raman
fiber laser
• Long fiber length needed for Raman laser,
so core propagation loss must be very low 88
 (b) Stimulated Brillouin scattering1 (SBS)
Pump wave generates acoustic Periodic variation in
waves via electrostriction refractive index

Pump wave
vA = acoustic velocity
of grating
Stokes wave

Pump light is scattered via Bragg

diffraction and down-shifted in frequency
due to the Doppler shift Backward
propagating Stokes wave

2nv A
→ Frequency shift ν B = ν p − νs =
λp

• For silica, vA 6km/s νB 17.5GHz at λp = 1µm. Hence frequency shifts for SBS are much
smaller than for SRS.
• Scattering involves low energy acoustic phonon for SBS and higher energy optical phonons for SRS
• Stokes wave is backward propagating for SBS 89
Brillouin gain:
Due to the very small Brillouin frequency shift:

dIs
= −g B I p I s + α I s • Assuming νp νs and hence αp αs α
dz
• Note the ‘minus’ sign has been included
dI p to account for the propagation direction of the
= −g B I p I s − α I p Stokes wave (c.f. SRS)
dz

If the propagation loss (α) is negligible then dIs/dz = dIp/dz

gB is the Brillouin gain coefficient. If the acoustic waves decay as exp(-t/τB), where τB is
the phonon lifetime, then

∆ν 2B g B (ν B ) 2πn 7 p122
g B (ν ) = where g B (ν B ) = 2
4 (ν − ν B ) 2 + ∆ν 2B cλ p ρ v A ∆ν B

∆νB = 1/πτB =Brillouin gain bandwidth (FWHM)

p12 = longitudinal elasto-optic coefficient
ρ = density
For bulk silica, ∆νB ~ 10MHz and gB ~ 1×10-10m/W. For silica-based fibers, the Brillouin gain
may be reduced due to the presence of dopants (e.g. RE ions) and inhomogeneities, and hence the
Brillouin linewidth is generally much broader in silica fibers 90
Brillouin gain vs pump bandwidth:
If the pump bandwidth ∆νp is comparable or larger than ∆νB, then the maximum value
for Brillouin gain coefficient will decrease significantly

∆ν B
g B max = g B (ν B )
∆ν B + ∆ν p

SBS is mainly an issue in narrow-linewdth sources

Brillouin threshold:
The power in the Stokes wave increases exponentially in the backward (-z) direction according to:

I s (0) = I s (L) exp(g B I pleff − αl)

where leff =
1
α
[
1 − exp(− α l ) ]

21A eff
Threshold pump power for SBS: PSBS
pth ≈
g Bleff
91
Example: Typical cladding-pumped Yb-doped silica fiber amplifier with

Core diameter = 15µm Aeff 1.77×10-10m2

leff = 20m
Narrow-linewdth input signal (∆νp < 1MHz) at 1.06µm
SBS
Ppth < 2W

This is only a very rough guide. In practice the SBS threshold would be somewhat higher
due inhomogeneities in the fiber core. Nevertheless, this does show that SBS is a significant
problem to overcome when scaling the output power from narrow-linewidth single-
frequency cw fiber sources.

Possible solutions include:

• Increasing the core area – Limited scope without degrading beam quality

• Shorten device length – More demanding on diode pump beam quality and
thermal management

• Deliberately modify the fiber properties along its length (e.g. using longitudinal
variation in temperature9,10) 92
 References
1. G. P. Agrawal, “Nonlinear fiber optics,” Second Edition, (Academic Press 1995).
2. J. Stone, “CW Raman fiber amplifier,” Appl. Phys. Lett., vol. 26, p.163-165 (1975).
3. S. A. Skubchenko, A. Y. Vyatkin and D. V. Gapontsev, “High-power CW linearly polarized all-fiber
Raman laser IEEE Photonics Tech. Lett., vol. 16 p.1014-1016 (2004).
4. D. Inniss, D. J. DiGiovanni, T. A. Strasser, A. Hale, C. Headley, A. J. Stenz, T. Pedrazzani, D. Tipton,
S. G. Kosinaki, D. L. Brownlow, K. W. Quoi, K. S. Kranz, R. G. Huff, R. Espindola, J. D. Legrange
and G. Jacobovitz-Veselka, “Ultra-high-power single-mode fiber lasers from 1.065-1.472µm using
Yb-doped cladding-pumped and cascaded Raman Lasers,” CLEO 1997, Postdeadline paper CPD-31
(1997).
5. V. I. Karpov, E. M. Dianov, V. M. Paramonov, O. I. Medvedkov, M. M. Bubnov, S. L. Semyonov, S.
A. Vasiliev, V. N. Protopopov, O. N. Egorova, V. F. Hopin, A. N. Guryanov, M. P. Bachynski and W.
R. L. Clements, “Laser-diode-pumped phosphosilicate-fiber Raman laser with an output power of 1W
at 1.48µm,” Opt. Lett., vol.24, p.887-889 (1999).
6. N. S. Kim, M. Prabhu, C. Li, J. Song and K. Ueda, “1239/1484nm cascaded phosphosilicate Raman
fiber laser with CW output power of 1.36W at 1484nm pumped by a CW Yb-doped double-clad fiber
laser at 1064nm and spectral continuum generation,” Opt. Comm., vol176, p.219-222 (2000).
7. J. Nilsson, J. K. Sahu, J. N. Jang, R. Selvas, D. C. Hanna and A. B. Grudinin, “Cladding-pumped
Raman fiber amplifier,” Optical Amplifiers and their Applications (OAA 2002), post-deadline paper
PD-2 (2002).
8. J. N. Jang, Y. Jeong, J. K. Sahu, M. Ibsen, C. A. Codemard, R. Selvas, D. C. Hanna and J. Nilsson,
“Cladding-pumped continuous-wave Raman fiber laser,” OFC 2006, paper OThJ2.
9. V. I. Kovalev and R. G. Harrison, “Suppression of stimulated Brillouin scattering in high-power
single-frequency fiber amplifiers,” Opt. Lett., vol.31, p161-163 (2006)
10. Y. Jeong, J. Nilsson, J. K. Sahu, D. B. S. Soh, C. Alegria, P. Dupriez, C. A. Codemard, D. N. Payne, R.
Horley, L. M. B. Hickey, L. Wanzyck, C. E. Chryssou, J. A. Alverez-Chavez and P. W. Turner,
“Single-frequency, single-mode, plane-polarised ytterbium-doped fiber master-oscillator power
amplifier source with 264W of output power,” Opt. Lett., vol.30, p.459-461 (2005). 93
6. Scaling mode area

94
Scaling the core area and fundamental mode size whilst preserving single-spatial-mode
operation and hence good beam quality is an important requirement for further power
scaling, and hence is a topic which is currently attracting much interest. The main reasons
for this are the following:

• Reduce the cladding-to-core area ratio to reduce fiber length

• Increase the threshold for unwanted nonlinear loss processes (e.g. SRS, SBS)

A eff
SBS,SRS
Ppth ∝
leff

• Raise the power/energy damage limit (especially for pulsed operation)

Surface damage limit for bulk silica glass is ~2GW/cm2 and is lower (~1GW/cm2) for
doped silica glass

• Increase the ‘stored’ energy for pulsed operation

• Decrease fiber propagation loss
• Flexibility in operating wavelength
• Reduce output beam divergence Easier collimation of output with lower spec. lenses

95
Example: Double-clad Yb-doped fiber with a conventional step-index single-mode
core design1

2π ra 2π ra Core
V= n −n
2 2
= NA < 2.405
λ λ
a b
Inner-cladding

For λ = 1.06µm and NA = 0.15, then ra = 2.6µm

If rb = 200µm and [Yb3+] = 7000ppm (i.e. for a typical double-clad fiber design) then:

Inner-cladding-to-core area ratio is 5900:1

Fiber length for efficient pump absorption (>10dB) at 975nm is ~ 120m
Threshold for SRS is ~ 28W
CW damage threshold is ~ 200W 1 0000

Conventional SRS
8000 single-mode
Power limit (W)
Very rough guide to power limits 6000
(excluding thermal limits) Damage
4000
Difficult to fabricate conventional fibers
with core NA’s <0.04 - 0.05 2000
Limited scope for power scaling SBS
0
For higher power levels a different approach 0 5 10 15 20 25 30
96
is needed M ode radius (µm )
Novel core designs:
3.0

Refr. index difference x 10-3

• Large-mode-area core 2.5

- Complex refractive index profiles 2.0

with ring structure to expand the 1.5

1.0
fundamental mode area2,3
0.5
- More difficult to fabricate
0.0
-200-150-100 -50 0 50 100 150 200
• Microstructured (holey) fibers Position [micron]

Solid core
Holey-cladding

= hole-to-hole spacing
relative hole size
Aeff ~ 490µm at λ = 1µm
NA ~ 0.03 and can be reduced
further by decreasing hole size.
Light is guided by the effective refractive
index difference between the core and 97
cladding regions Bend loss is the main limiting factor4,5
 Mode selection in multimode cores

An alternative strategy for scaling mode area is to employ a multimode core design and
restrict the number of modes, preferably to just the fundamental mode.

This is generally quite straightforward for cores that are only slightly multimode, but
becomes increasingly difficult as the core size increases.

If we make the simplifying approximation that the fundamental LP01 mode has a Gaussian
intensity profile: ILP01(r) = Aexp(-2r2/wo2), then for relatively small V values (V < 8) the
mode radius wo is given by6
1.4

1.2
1.619 2.879
w o = ra 0.65 + 3 / 2 + 1.0
V V6
0.8

wo/ra
0.6

0.4
LP01 mode area increases with as core
radius increases, but it occupies a smaller 0.2

fraction of the core area as V increases 0.0

2 4 6 8 10 12
V parameter 98
Main problems:

(a) Mode-coupling
In a ‘perfect’ multimode fiber there is no energy conversion from one guided mode to
other guided modes7. However, in practice, multimode fibers have perturbations in
refractive index due imperfections in the fiber, curved waveguide trajectory or bending
which leads to coupling between modes. Thus, if a single-mode laser beam is launched
into a multimode, then energy is coupled into higher order modes and the beam quality
deteriorates as the light propagates along the fiber.

As a rough guide8: 2 2
16 r Dz
M 2 ( z) ≈ 1 + a

λ2

where D is the mode-coupling coefficient and D ∝ ra8 /( rb6λ4 ) . D also depends on other
details of the fiber (e.g. imperfections), which may be influenced by the fabrication
procedure.

Multimode fibers with a core diameter of 45µm (NA = 0.13) and a low enough values
for D for single-mode propagation over 20m at 1.55µm have been reported5. However,
scaling to even larger core sizes and extending to shorter wavelengths around 1µm is
difficult. 99
(b) Mode overlap

Single-mode core (V = 2.4) Multimode core

RE-doped RE-doped
core Cladding core Cladding

V increasing

LP01 mode LP01 mode

As the V parameter increases and the core becomes more multimode and the fundamental
(LP01) mode occupies a smaller fraction of the core area
Unused population inversion
RE-doped
Gain for higher order modes region
Can promote multimode lasing in oscillators
and is also a problem in amplifiers

One way to alleviate this problem is to use a core design with

the RE dopant confined to the central region to improve the
spatial overlap. However, this has the disadvantage of
increasing the device length by virtue of increasing the
100
effective absorption coefficient for cladding pumping.
 Mode selection using a taper

Inner-cladding

Multimode core
Single-mode core

Basic idea: Taper fiber so that the multimode core is reduced in diameter so that only
the fundamental mode can propagate in the core with low loss9

• Low loss for fundamental mode + high loss for higher order modes
• Requires low mode-coupling otherwise there is a significant decrease in efficiency
• Pump light in the cladding will experience a high loss in the tapered section, so pump
light should be launched into the opposite end of the fiber
• Core damage may still be an issue due to the small core size in the tapered region

Distributed mode-filtering along the entire length of the fiber is generally a more effective
technique (especially when mode-coupling is strong).
101
 Mode selection by bending
Basic idea:
Bending a fiber results in increased propagation loss for all guided modes. However, the
propagation loss for the fundamental mode (LP01) is lower than for higher order modes,
so bend-induced loss can be used as a very effective means for suppressing higher order
modes in a multimode fiber10.

The loss coefficient, , for the fundamental mode and higher-order modes in a
bent fiber with a ‘step-index’ core can be estimated from11:
1/ 2
1 πρ U2 2 βRbW 3
γ = ⋅ ⋅ exp −
ρ Rb 2 3/ 2
eν V W Kν −1 (W ) Kν +1 (W ) 3(knclρ ) 3
where eν= 2 if ν = 0 (LP01, fundamental mode) and eν=1 for higher order modes, U, V, and W are the
waveguide modal parameters1, K(W) is the modified Hankel function of W, β is the propagation
constant of the mode in the fiber, k is the free space propagation constant, ρ is the fiber core radius
and Rb is the bend radius of the fiber.

The fiber can be wound on a circular heat-sink Distributed mode filter

Bend loss increases with:
• Decreasing bend radius
• Lower NA
• Longer wavelength
• Increasing core radius 102
Example: Bend loss versus bend radius for LP01 and LP11 modes at 1.06µm for a core
with NA = 0.05
ra = 10µm ra = 20µm
1000 1000

Propagation loss (dB/m)

Propagation loss (dB/m)

100 100
LP11
10 10
LP11
1 1
LP01 LP01
0.1 0.1

0.01 0.01
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Bend radius (mm) Bend radius (mm)

ra = 30µm ra = 40µm
1000
1000
Propagation loss (dB/m)

Propagation loss (dB/m)

100
100

10
10 LP11
1 LP01 LP11 LP01
1

0.1
0.1

0.01
0.01
0 20 40 60 80 100 120 140 160
0 20 40 60 80 100 120 140 160
Bend radius (mm)
Bend radius (mm)

• Loss versus bend radius curves get steeper as the core radius is increased
103
Tighter tolerance on bend radius
Important points:
• Calculating the bend radius required for suppression of high-order modes is
quite complicated as many factors need to be considered.
• Easier to estimate bend radius from bend loss calculation and then optimise design by
experiment
• As the V value increases Number of guide modes ~V2/2. If all the modes are
excited then M2 V/2, but the fundamental mode still occupies a reasonable fraction
of the core area Spatial overlap between LP01
mode and higher order modes increases
Increasingly difficult to suppress higher order
modes as the core size is increased
The lower limit on core NA in conventional
fibers is ~ 0.04. This limit is currently determined
ra = 40µm
by the fabrication procedure. Microstructured 1000
fibers offer the possibility of even lower NA’s and

Propagation loss (dB/m)

100
hence larger core areas. However, bend loss for the NA = 0.02
10
LP01 mode becomes a serious problem as the NA is NA = 0.04
decreased further. 1

The use of bend loss as a means for suppressing 0.1 NA = 0.06

higher order modes in large-core fibers has facilitated 0.01
scaling of fiber cw powers to the ~kW regime12 and 0 500 1000 1500 2000 2500
Bend radius (mm) 104
pulse energies to the ~ mJ regime13
 References
1. A. W. Snyder and J. D. Love, “Optical Waveguide Theory,” (Kluwer Academic Publishers, 2000).
2. N. G. R. Broderick, H. L. Offerhaus, D. J. Richardson, R. A. Sammut, J. Caplen and L. Dong, “Large
mode area fibers for high power applications,” Opt. Fiber Tech., vol. 5, p.185-196 (1999).
3. H. L. Offerhaus, J. Alvarez-Chavez, J. Nilsson, W. A. Clarkson, P. W. Turner, D. J. Richardson and A.
B. Grudinin, “Characteristics of Q-switched cladding-pumped ytterbium-doped fibre laser with different
high energy fibre designs,” J. Quantum Electron., vol.37, p.199-206 (2001).
4. J. M. Fini, “Bend-resistant design of conventional and microstructure fibers with very large mode
area,” Optics Express, vol. 14, p.69-81 (2006).
5. http://www.crystal-fibre.com
6. D. Marcuse, “Loss analysis of single-mode fiber splices,” Bell System Technical Journal, vol. 56. p.703-
718 (1977).
7. D. Marcuse, “Theory of Dielectric Optical Waveguides,” (Academic Press, San Diego, 1974).
8. M. E. Fermann, “Single-mode excitation of multimode fibers with ultrashort pulses,” Opt. Lett., vol.23,
p.52-54 (1998).
9. J. Alvarez-Chavez, A. B. Grudinin, J. Nilsson, P. W. Turner and W. A. Clarkson, “Mode selection in high
power cladding pumped fibre lasers with tapered section,” Conference on Lasers and Electro-Optics 1999,
Optical Society of America, Washington D.C., paper CWE7, p.247-248 (1999).
10. J. P. Koplow, D. A. V. Kliner, L. Goldberg “Single-mode operation of a coiled multimode fiber
amplifier”,Opt. Lett., vol. 25, p. 442-444 (2000).
11. D. Marcuse, “Curvature loss formula for optical fibers”, J. Opt. Soc. Am., vol. 66, p.216-220 (1976).
12. A. Tunnermann, T. Schreiber, F. Roser, A. liem, S. Hofer, H. Zellmer, S. Nolte and J. Limpert, ”The
renaissance and bright future of fiber lasers,” J. Phys. B: At. Mol. Opt. Phys., vol.38, p.681-693 (2005).
13. M. Y. Cheng, Y. C. Chang, A. Galvanauskas, P. Mamidipudi, R. Changkakoti and P. Gatchell, “High-energy
and high-peal-power nanosecond pulse generation with beam quality control in 200-µm core highly
multimode Yb-doped fiber amplifiers,” Opt. Lett., vol.30, p.358-360 (2005).

105
7. Wavelength selection and tuning

106
One of the attractions of fiber-based sources is the flexibility in operating wavelength owing
to the broad transition linewidths for a glass host. The range of operating wavelengths
depends on a number of factors, including:
• Spectroscopy of the laser transition (i.e. emission and absorption cross-sections,
upper-state lifetime, linewidth, loss processes)
• Wavelength selection scheme
• Fiber configuration (RE ion doping level, pumping scheme, pump wavelength, device
length)

Yb-doped silica:

N1 and N2 are the total population densities of the lower and

upper manifolds respectively and N = N1 + N2 is the doping
concentration.

For transparency we require: N1σ a (λ L ) = N 2 σe (λ L )

Nσ a (λ L )
N2 =
σ a (λ L ) + σ e (λ L )
where σe(λL) and σa(λL) are the emission and absorption cross-sections at lasing wavelength
λL respectively. As σa(λL)/σe(λL) increases, the transition has a stronger three-level character and
hence the threshold increases Short wavelength limit
107
 3 .0

2 .5 Pump power required for transparency is given by1

Cross-section (pm )
2

2 .0
A c hν p
1 .5 Ptrans =
Absorption σ (λ ) σ (λ )
Emission ηp τf e L a p − σe (λ p )
1 .0
σa (λ L )
0 .5

0 .0
850 900 950 1000 1050 1100 1150 where Ac is the core area, τf is the lifetime of the
W a v e le n g th ( n m ) upper level and ηp is the overlap factor for pump light
with the doped core
Cross-section
σ e (λ L ) σ e ( λ p )
>
σ a (λ L ) σ a (λ p )

Threshold increases dramatically as the lasing wavelength

approaches the emission peak at ~980nm
Very short wavelength operation requires pumping at shorter
wavelengths (e.g. ~915nm) and a high value for ηp
λp λL
Cladding pumping Long device lengths and high reabsorption loss at short wavelengths
Moves the effective gain peak towards the longer wavelength end
of the emission spectrum
Flexibility in operating wavelength is reduced compared to core pumping 108
 Wavelength selection and tuning schemes

(a) External cavity with diffraction grating2

Angle-polished end
to suppress broadband Diffraction grating
feedback
in Littrow configuration
Output

Line spacing = Λ
Fiber
θ

• Lasing wavelength can be selected by adjusted grating angle: λ L = 2Λ sin θ

• Wavelength tuning range depends on many factors (see previous slides)

• Bandwidth of emission depends on the grating, the feedback cavity design and the fiber

λ L M 2 Λ cos θ
Spectral selectivity of feedback cavity: ∆λ L ∝
f sin −1 ( NA)
Linewidths < 0.1nm are easy to achieve

Note: Damage due to self-pulsing can occur if the grating angle is adjusted to select a
109
wavelength beyond the tuning range.
 Example:

Typical tuning curves (see refs. 2,3,4):

Yb-doped alumino-silicate fiber Er,Yb-doped phospho-silicate fiber Tm-doped alumino-silicate fiber
60
120
10
Output power (W)

50
100

Output Power (W)

Output power (W)

8
40 80
30 6
60
20 40 4

10 20 2

0 0 0
1050 1060 1070 1080 1090 1100 1110 1530 1540 1550 1560 1570 1850 1900 1950 2000 2050
110
Wavelength (nm) Wavelength (nm) Wavelength (nm)
(b) In-fiber Bragg grating5,6,

• Require photosensitive core or photosensitive ring

around the core

• FBG may be written directly in active fiber or may be

written in another fiber and spliced on to the end of
the active fiber Spacing = λ B / 2n eff

• Reflectivity at central wavelength5: R = tanh 2 π δn lG g ηG

λ
where δn is the refractive index change, lG is the length of the grating, g is the average
value of the envelope weighting function and ηG is the modal overlap factor
FBG reflectivity >99% can be achieved
Wavelength tuning over a few % of λB can be achieved by compressing/stretching the grating6
Narrow linewidth (<0.01nm) achievable, but is generally much broader in multimode devices
111
Power handling issues at high power levels may be avoided using a MOPA configuration
 Single-frequency fiber sources

There are a number of different approaches for achieving single-frequency operation7.

One of the most popular is to use a distributed feedback (DFB) architecture8

DFB fiber laser:

lG
Typical operating characteristics:

• Low temperature sensitivity (<0.01nm/K or ~1GHz/K)

• Ultra high Signal-to-Noise-Ratio, SNR (>70dB)
• Narrow line-width (~1-100kHz)
• Tuneable over >20nm
• High polarisation purity (>30dB)
• Typical output power (10 – 100mW)
• Simple + low cost +monolithic construction

Power scaling of DFB fiber lasers is difficult since a high power single-mode pump
source is needed and the pump light must be absorbed within a length of ~10cm or
less. 112
Use a MOPA to scale to higher power levels
 High-power single-frequency fiber sources

Typical arrangement:
Large-core double-clad
Yb-doped fiber

Isolator
Isolator
λ/2 plate
Output

Single-frequency
master-oscillator Intermediate Yb
(e.g. DFB fiber fiber amplifier stages Unabsorbed High-power
laser or bulk pump light diode pump
oscillator) @ 975nm

• Number of intermediate amplifiers depends on master-oscillator power

• Large core fiber needed for final power amplifier stage to increase the threshold for SBS
• Polarisation maintaining fiber is required for a single-polarisation output

Using this generic approach single-frequency powers of over 100W have been demonstrated9,10
113
 References
1. R. Paschotta, J. Nilsson, A. C. Tropper and D. C. Hanna, “Ytterbium-doped fiber amplifiers,” IEEE J.
Quantum Electron., vol.33, p.1049-1056 (1997).
2. J. Nilsson, W. A. Clarkson, R. Selvas, J. K. Sahu, P. W. Turner, S-U. Alam and A. B. Grudinin,
“High-power wavelength-tunable cladding-pumped rare-earth-doped silica fiber lasers,” Optical
Fiber Technology, vol.10, p.5-30 (2004).
3. W. A. Clarkson, N. P. Barnes, P. W. Turner, J. Nilsson and D. C. Hanna, “High-power cladding-
pumped Tm-doped silica fiber laser with wavelength tuning from 1860 to 2090nm,” Opt. Lett., vol.27,
p.1989-1991 (2002).
4. D. Y. Shen, J. K. Sahu and W. A. Clarkson, “Highly efficient Er,Yb-doped fiber laser with 188W
free-running and >100W tunable output power,” Opt. Express, vol.13, p.4916-4921 (2005).
5. K. O. Hill and G. Meltz, “Fiber Bragg grating technology fundamentals and overview,” IEEE J.
Lightwave Technology, vol.15, p.1263-1275 (1997).
6. Y. Jeong, C. Alegria, J. K. Sahu, L. Fu, M. Ibsen, C. A. Codemard, M. R. Mokhtar and J. Nilsson, “A
43-W C-band tunable narrow-linewidth erbium-ytterbium co-doped large-core fiber laser,” IEEE
Photon. Tech. Letters, vol.16, p.756-758 (2004).
7. N. Langford, “Narrow-linewidth fiber sources,” in Rare-Earth-Doped Fiber Lasers and Amplifiers,
edited by M. J. F. Digonnet, 2nd Edition, (Marcel Dekker, Inc., NewYork, 2001).
8. W. H. Loh and R. I. Laming, “1.55µm phase-shifted distributed feedback fibre laser,” Electron. Lett.,
vol.31, p.1440-1442 (1995).
9. A. Liem, J. Limpert, H. Zellmer and A. Tunnermann, “100W single-frequency master-oscillator fiber
power amplifier,” Opt. Lett., vol.28, p.1537-1539 (2003).
10. Y. Jeong, J. Nilsson, J. K. Sahu, D.B. S.Soh, C. Alegria, P. Dupriez, C. A. Codemard, D. N. Payne, R.
Horley, L. M. B. Hickey, L. Wanzcyk, C. E. Chryssou, J. A. Alvarez-Chavez and P. W. Turner,
“Single-frequency, single-mode, plane-polarised ytterbium-doped fiber master-oscillator power
amplifier source with 264W of output power,” Opt. Lett., vol.30, p.459-461 (2005).
114
8. High-power CW and pulsed
fiber sources

115
 Example: Cladding-pumped Yb-doped fiber laser with 1.4kW output power1
Signal output
HT @975 nm @~1.1 µm HT @975 nm
HR @~1.1 µm HR @~1.1 µm

Diode stack Diode stack

@975 nm, 0.6 kW @975 nm, 1.2 kW

HT @975 nm
HR @~1.1 µm Double-clad
Yb-doped fibre

1.4
M2 = 1.4, λ = 1100 Pure silica D-shaped inner-
Measured
nm Linear fit cladding (Diameter = 650µm,
1.2
Pump-power limited NA = 0.4-0.5)
1.0 Slope efficiency: 83%
Signal power [kW]

0.8
Core diameter = 40µm, NA <0.5

0.6
Using polarisation-maintaining fiber
0.4
(inner-cladding diameter = 380µm and
0.2 core diameter = 25µm) and a modified
external cavity design with a polariser +
0.0 half-wave plate 633W of linearly-
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
polarised output for ~1kW of launched
Launched pump powwer [kW]
pump power2

116
For further examples of high-power cw fiber lasers see refs. 3, 4 and 5
Owing to the:
• Wide range of operating wavelengths
• Broad emission linewidth
• Long upper-state lifetime
• High-power handling capability
• Good beam quality

Cladding-pumped fiber sources are also very attractive for generating high average
output power and high peak power in pulsed mode

Two main approaches:

• Laser oscillator (Q-switched or mode-locked)

- Single gain element + external cavity required Simple + low cost
- Limited flexibility
- Limited peak power

• Master-oscillator power-amplifier (MOPA)

- Low power oscillator + multiple gain stages
- Flexibility in pulse duration, pulse shape, repetition rate, etc
- Power scalable, but can be quite complicated
117
 High-power Q-switched Yb-doped fiber lasers
Extractable energy6:
Nlσ a (λ L )
l
E ext = hν L A co N 2 (z)dz −
0
σ e (λ L ) + σ a (λ L )

As a very rough guide, the extractable energy stored in a fiber is limited by ASE to around ten times
the saturation energy E sat ≈ hν L A co /[σa (λ L ) + σe (λ L )] ηL . In other words, the upper-limit on pulse energy is
determined mainly by the core area. Eext is ~2 - 5mJ for typical Yb-doped fiber with a 20µm diameter
core. In practice, the maximum pulse energy also depends on other factors as well (e.g. cavity design,
fiber length, nonlinear loss processes (SRS, SBS) and fiber damage).
Example7:

8.4 mJ @ 0.5 kHz, pulse duration = 460ns

0.6 mJ @ 200 kHz (120 W)
Beam quality M2 ~ 4 118
Shorter pulse durations can be achieved using a shorter fiber8
 Fiber Facet damage

Damage example
• Damage threshold of pure silica:
~2 GW/cm2 (20 W/µm2)
~50J/cm2 (~ 10ns pulse)
• Dependent on surface quality, dopants,
impurities, inhomogeneities, especially
at facets → Large core + end caps needed
• Suppression of self-pulsing

Double-clad fiber
Splice

Matched core-less
Lens silica end-cap
119
 MOPA configurations
Master-oscillator (diode, fiber or bulk)
- low power cw oscillator Multi-stage fiber amplifier
- Pulsed (Q-switched)
- Short pulse

Output

Modulator Isolator

Increasing mode area + increasing pump power

May also require band-pass filters between amplifier stages to suppress ASE

For long pulse operation: High gain + low saturation energy Strong pulse re-shaping
2

Input pulse 0

2 0 0 0
Output pulse shape can be controlled by
1 5 0 0 modifying the input pulse shape
Output pulse
1 0 0 0

5 0 0

- 5 0 0
120
- 1 0 1 2 - 1 0 1 2
 Example: High energy Yb fiber MOPA9
Bandpass
filter
Amplified diode laser
@ 1064nm
AOM
Output

Isolator
Single-mode Multi-mode
Yb fiber amplifier Yb fiber amplifier Multi-mode
ra = 25µm Yb fiber amplifier
rb = 175µm ra = 100µm
Pp = 40W rb = 300µm
Pp = 200W

Pulse energies ~ 82mJ for 500ns pulses were reported, and coiling the fiber in the
final amplifier with a radius of ~12cm produced strong enough mode-filtering to reduce
the M2 parameter to 6.5 (see ref. 9)

Cladding-pumped fiber amplifier schemes employing chirped pulse amplification have

also been used amplify ultrashort pulse oscillators to average power levels >100W
(see ref. 10).
121
 References
1. Y. Jeong, J. K. Sahu, D. N. Payne and J. Nilsson, “Ytterbium-doped large-core fiber laser with
1.36kW continuous-wave output power,” Opt. Express, vol.12, p. 6088-6092 (2004).
2. Y. Jeong, J. Nilsson, J. K. Sahu, D. B. S. Soh, P. Dupriez, C. A. Codemard, S. Baek, D. N. Payne, R.
Horley, J. A. Alvarez-Chavez and P. W. Turner, “Single-mode plane polarised ytterbium-doped
large-core fiber laser with 633-W continuous-wave output power,” Opt. Lett., vol.30, p.955-957
(2005).
3. A. Tunnermann, T. Schreiber, F. Roser, A. liem, S. Hofer, H. Zellmer, S. Nolte and J. Limpert, ”The
renaissance and bright future of fiber lasers,” J. Phys. B: At. Mol. Opt. Phys., vol.38, p.681-693
(2005).
4. C-H. Liu, B. Ehlers, F. Doerfel, S. Heinemann, A. Carter, K. Tankala, J. Farroni, and A.
Galvanauskas, ”810 W continuous-wave and single-transverse-mode fibre laser using 20 /spl mu/m
core Yb-doped double-clad fibre,” Electron. Lett., vol.40, p.1471-1472 (2004).
5. http://www.ipgphotonics.com/ and http://www.spioptics.com/
6. C. C. Renaud, H. L. Offerhaus, J. A. Alvarez-Chavez, J. Nilsson, W. A. Clarkson, P. W. Turner, D. J.
Richardson, A. B. Grudinin, “Characteristics of Q-switched cladding-pumped ytterbium-doped fiber
lasers with different high-energy fiber designs,” IEEE Journal of Quantum Electronics, vol.37(2)
p.199-206 (2001).
7. Y. Jeong, J. K. Sahu, M. Laroche, W. A. Clarkson, K. Furusawa, D. J. Richardson and J. Nilsson,
“120W Q-switched cladding-pumped Yb-doped fibre laser”, CLEO/Europe-EQEC 2003, paper CL5-
4 (2003).
8. J. Limpert, N. Deguil-Robin, S. Petit, I. Manek-Hönninger, F. Salin, P. Rigail, C. Hönninger and
E. Mottay, “High power Q-switched Yb-doped photonic crystal fiber laser producing sub-10 ns
pulses,” Appl. Phys. B., vol.81, p.19-21 (2005).
9. M. Y. Cheng, Y. C. Chang, A. Galvanauskas, P. Mamidipudi, R. Changkakoti and P. Gatchell,
“High-energy and high-peak-power nanosecond pulse generation with beam quality control in 200µm
core highly multimode Yb-doped fiber amplifiers,” Opt. Lett., vol.30 (4), p.358-360 (2005)
10. F. Röser, J. Rothhard, B. Ortac, A. Liem, O. Schmidt, T. Schreiber, J. Limpert, and A. Tünnermann, 122
“131W 220 fs fiber laser system,” Opt. Lett., vol. 30, p.2754-2756 (2005).
9. Power and brightness scaling
via beam combination

123
 Introduction
Scaling the output power and radiance beyond the upper limit for a single fiber core
can be achieved via the use of multiple fiber sources or a multi-core fiber source
and beam combining1. Beam combining schemes fall into one of two categories:

(a) Incoherent beam combining:

Spatial beam combining

Far-field beam divergence of array = Α
Far-field beam divergence of single element = S

No phase relationship between elements A= S

πwθS πNdθS
M S2 ≈ and M 2A ≈
d 2λ 2λ
w PA = N PS

BA w
≈ = ηfill
BS d
N sources
PS = power of a single element, PA = power of array, BS = brightness of
124
a single element, BA = brightness of array and fill is the fill-factor
 Simple and robust way to scale output power, but brightness cannot exceed the brightness of
a single element

Examples:

(a) Multi-fiber sources

Multimode
delivery fiber

Multi-kilowatt power levels

are achievable2

(b) Multi-core fiber sources

HT@ 976nm F1=20mm F2=20mm
HR@1040nm
Pump in at
976nm

F3=25mm

Laser out at 1040nm

Yb-doped multi-core
ribbon fiber3 125
Output power > 320W demonstrated (limited by pump power)3
Polarization beam combining

B A = 2ηc BS
where c is the combining efficiency, which takes
into account loss and misalignment

Wavelength beam combining

or

B A = Nη c B S 126
 Example:

Common output coupler ensures

that beams at different wavelengths
N fiber sources sharing are all incident on the grating at the
a common external cavity same angle

1+ 2+ ..... N

Spacing between 1st and Nth fiber cores = d, grating period = , spacing of adjacent cores = s

Wavelength for ith fiber laser: λi =Λ(sin θi + sin φ) and dispersion of grating: d /d = 1/(Λcos )

Wavelength spread for fiber laser array: ∆λA = Λd[(cos θ)/f] ∆λL
Wavelength spacing between adjacent fibers is ∆λs = Λs[(cos θ)/f]

where ∆λL is the gain bandwidth for the transition and can be quite large (>100nm) in a glass host
127
 Power scaling limit Ps L/ s

where Ps is the power of a single element

This approach has been used to combine diode4 and fiber laser arrays5,6. For the latter, power levels
of >100W have been realised for an array of three cladding-pumped Yb-doped fibers using a fused
silica transmission grating as the dispersive element6.

Main challenges for power and brightness scaling7:

• Accurate positioning of fiber cores in a linear array to avoid degradation in beam quality

• Lens design – Minimising degradation in beam quality due to aberrations

• Dispersive element design – Good wavelength discrimination, high efficiency, thermal handling

• Tight alignment tolerances

128
(b) Coherent beam combining:

Filled-aperture beam combining All elements have the same wavelength and are in phase
or there is a defined phase relationship between adjacent
elements

Io
Strehl ratio = S =
Io '

where Io is the actual on-axis intensity and Io’ is

Beam combiner
the on-axis intensity for a perfect ‘top-hat’ beam
with the same overall aperture size

Tiled-aperture beam combining Sphased array = NSincoherent array

Can be viewed as trying to synthesize a plane wave

w Fill-factor must be as large as possible to minimise

d proportion of light in side-lobes

Phase error must be <<2 to avoid a reduction in on-axis

intensity
129
Example: 1D array of single-frequency fiber lasers with the same wavelength and in phase
(N = 5, d = 500 m)
2
η sin β sin Nα πw πd
2

I(θ) ∝ fill2 where β = sin θ and α = sin θ

N β sin α λ λ

1.0 1.0
fill = 1 fill = 0.75

Relative intensity
Relative intensity

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
-10 -5 0 5 10 -10 -5 0 5 10
Angle (mrad) Angle (mrad)

1.0 1.0
fill = 0.5 fill = 0.25

Relative intensity
Relative intensity

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
-10 -5 0 5 10 -10 -5 0 5 10 130
Angle (mrad) Angle (mrad)
 There are a number of possible implementation schemes. Most employ a MOPA architecture with
one or more amplification stages (e.g. see refs. 8 and 9)

Example:

Optical path differences between array elements are detected and actively controlled using acousto-optic
frequency shifters (or some alternative means) to compensate variations in OPD due to differences in
fiber length, temperature variations, etc 131
Coherent beam combining offers a route to very high power and very high brightness
from a fiber-based source. Power levels up to 470W have been demonstrated for a
4-element array9.

Some practical issues and challenges:

• Many narrow-linewidth, polarized fiber sources (each comprising a multi-stage

fiber amplifier) are required.
• Maximum power for a single-frequency polarised fiber source is currently ~400W
(limited by pump power)10.
• SBS is likely to be an issue for further power scaling.
• As the number of array elements increases, the alignment tolerance of output beams
from individual elements in the near-field and far-field (i.e. to avoid pointing errors)
will be much more difficult to meet.
• Very high fill-factor + ‘top-hat’ beam profile are needed to ensure that there is very
little power in the side-lobes.
• Power per element needs to be carefully controlled to avoid non-uniformity across
the array.
• Accurate monitoring and robust control of the relative phase between elements
is needed.
132
 References
1. T. Y. Fan, “Laser beam combining for high-power high-radiance sources,” IEEE J. Selected
Topics in Quantum Electronics, vol.11, p.567-577 (2005).
2. http://www.ipgphotonics.com
3. L. J. Cooper, P. Wang , R. B. Williams, J. K. Sahu, W. A. Clarkson, A. M. Scott and D. Jones,
Opt. Lett., vol.30, p2906-2908 (2005).
4. C. Hamilton, S. Tidwell, D. Meekhof, J. Seamans, N. Gitking and D. Lowenthal,” High power
laser source with spectrally beam combined diode laser bars,” Proc. SPIE, vol.5336 (2004).
5. S. J. Augst, A. K. Goyal, R. L. Aggarwal, T. Y. Fan and A. Sanchez, “Wavelength beam
combining of ytterbium fiber lasers,” Opt. Lett., vol.28, p331-333 (2003).
6. F. Röser, S. Klingebiel, A. Liem, T. Schreiber, S. Höfer, J. Limpert, T. Peschel, R. Eberhardt,
A. Tünnermann, “Spectral combining of fiber lasers”, Proc. SPIE Vol. 6102, p. 196-201, Fiber
Lasers III: Technology, Systems, and Applications; Andrew J. Brown, Johan Nilsson, Donald
J. Harter, Andreas Tünnermann; Eds. (2006).
7. E. J. Bochove, “Theory of spectral beam combining of fiber lasers,” IEEE J. Quantum
Electron., vol.38, p.432-445.
8. S. J. Augst, T. Y. Fan, A. Sanchez, “Coherent beam combining and phase noise measurements
of ytterbium fiber amplifiers,” Opt. Lett., vol.29, p.474-476.
9. J. Anderegg, S. Brosnan, E. Cheung, P. Epp, D. Hammons, H. Komine, M. Weber, M.
Wickham, “Coherently coupled high-power fiber arrays,” Proc. SPIE Vol. 6102, p. 202-206,
Fiber Lasers III: Technology, Systems, and Applications; Andrew J. Brown, Johan Nilsson,
Donald J. Harter, Andreas Tünnermann; Eds. (2006).
10. D. N. Payne, Y. Jeong, J. Nilsson, J. K. Sahu, D. B. S. Soh, C. Alegria, P. Dupriez, C. A.
Codemard, V. N. Philippov, V. Hernandez, R. Horley, L. M. B. Hickey, L. Wanzcyk, C. E.
Chryssou, J. A. Alvarez-Chavez, and P. W. Turner, “Kilowatt-class single-frequency fiber
sources”, in Fiber lasers II: technology, systems, and applications, L. N. Durvasula, A. J. W.
Brown, and J. Nilsson, Eds., Proc. SPIE vol. 5709, p. 133 -141 (2005).
133