Document [original]

Ti:Tm:LiNbO3Waveguide Amplifiers And Lasers

Thesis

Submitted to the

Department of Physics,

Faculty of Science

University of Paderborn, Germany

for the degree

Doktor der Naturwissenschaften (Dr. rer. nat.)

mathew george

reviewers:

Prof. Dr. Wolfgang Sohler

Prof. Dr. Donat As

date of submission:06.09.2012

date of examination:02.10.2012

ABSTRACT

The fabrication of Ti:Tm:LiNbO3waveguides by diffusion doping is

briefly described. The waveguides thus fabricated have been charac-

terized by determining the propagation losses, near field intensity

profile, fluorescence spectrum, absorption spectrum and Tm depth

profile by secondary neutral mass spectroscopy. Emission cross sec-

tions of the gain medium Ti:Tm:LiNbO3have been determined by

McCumber theory from the experimentally determined absorption

cross sections. A sensitive experimental setup was developed to per-

form all spectral measurements. For the first time an in-band pumped

(λp=1650 nm)Ti:Tm:LiNbO3waveguide amplifier capable of broad-

band optical amplification in the spectral range 1750 nm < λs<1900

nm is demonstrated. Experimental results of small signal gain mea-

surements were found to be in good agreement with the modeling

results. Modeling results show (wavelength dependent) gain of up to

30 dB in a 10 cm long single pass pumped waveguide. A double pass

pumping scheme can improve the gain achievable significantly. Fabry-

Pérot type lasers (λs=1890 nm and 1850 nm) have been realized by

depositing specially designed mirrors at the waveguide amplifier end-

faces. Laser threshold (1890 nm) is at 4mW coupled pump power;

the slope efficiency is ∼13.3%. The slope efficiency (laser threshold)

is larger (smaller) by more than an order of magnitude than those

reported so far for Tm:LiNbO3waveguide lasers. Modeling results

show that the slope efficiency can be improved to 34% by redesign-

ing one mirror. The fabrication and characterization of Ti:Tm:LiNbO3

waveguide for quantum memory applications is included as an ap-

pendix.

ZUSAMMENFASSUNG

Die Herstellung von Ti:Tm:LiNbO3Wellenleitern durch Diffusions-

dotierung wird kurz beschrieben. Die so hergestellten Wellenleiter

wurden sehr genau mit verschiedenen Methoden experimentell unter-

sucht, um Ausbreitungsverluste, Nahfeldverteilungen, Fluoreszenz-

spektren, Absorptionsspektren und Tm-Tiefenprofile zu bestim-

men. Die Emissionswirkungsquerschnitte des Verstärkungsmediums

Ti:Tm:LiNbO3wurden mit Hilfe der McCumber Theorie aus den ex-

perimentell ermittelten Absorptionsquerschnitten berechnet.

Ein experimenteller Aufbau zur Durchführung äußerst empfind-

licher spektraler Messungen wurde entwickelt. Zum ersten Mal

gelang es, breitbandige Verstärkung mit ’in-band’-gepumpten (λp

=1650 nm) Ti:Tm:Li:NbO3Wellenleitern im Wellenlängenbereich

iii

1750 nm < λs<1900 nm zu erhalten. Die experimentell bestimmte

Kleinsignalverstärkung dieser Wellenleiterverstärker ist in guter

Übereinstimmung mit Modellierungsergebnissen. Weitere Rechen-

ergebnisse sagen eine (wellenlängenabhängige) Verstärkung von bis

zu 30 dB in einem 10 cm langen Wellenleiter voraus. Mit einer

’double-pass’ Pumpanordnung kann die optische Verstärkung weiter

deutlich verbessert werden.

Durch Aufdampfen speziell entwickelter Spiegelschichten auf

die Wellenleiterendflächen wurden Laser vom Fabry-Pérot Typ

hergestellt mit Emissionswellenlängen von 1890 nm (Resonator ho-

her Güte) und 1850 nm (Resonator geringerer Güte). Die Laser-

schwelle (λs=1890 nm) beträgt 4mW eingekoppelter Pumpleis-

tung; die ’slope efficiency’ ist 13.3%. Damit sind beide Werte

um mehr als eine Größenordnung größer (’slope efficiency’) bzw.

kleiner (Laserschwelle) als entsprechende Ergebnisse, die bisher zu

Tm:LiNbO3Wellenleiterlasern veröffentlicht wurden. Weitere Model-

lierungsergebnisse zeigen, dass die ’slope efficiency’ bis zu 34% er-

höht werden kann, wenn die Resonatorspiegel optimiert werden.

Die Herstellung und Charakterisierung von Ti:Tm:LiNbO3Wellen-

leiter für Anwendungen als Quantenspeicher werden im Anhang

diskutiert.

CONTENTS

1 introduction 1

1.1Background 1

1.2Motivation 2

1.3Organisation of Thesis 3

2 waveguide fabrication and characterization 5

2.1Introduction 5

2.2Fabrication of Ti:Tm:LiNbO3waveguide 5

2.2.1Tm Diffusion Doping 6

2.2.2Fabrication of Ti Waveguide 6

2.3Characterization of Ti:Tm:LiNbO3waveguide 7

2.3.1Waveguide Mode Intensity Distribution 7

2.3.2Scattering Losses 8

2.3.3Thulium Depth Profile 9

2.3.4Absorption Spectra 11

2.3.5Fluorescence Spectrum 12

2.3.6Transition Cross Sections 13

2.4Conclusions 14

3Ti:Tm:LiNbO3waveguide amplifier 17

3.1Introduction 17

3.2In-band Pumping Scheme 17

3.2.1General Considerations 17

3.2.2Choice of Pump Wavelength 19

3.3Modeling 20

3.3.1Gain Spectra 22

3.3.2Power Characteristics 28

3.3.3Double Pass Pumping 29

3.3.4Further Comments 31

3.4Experimental Investigations 31

3.4.1Experimental Setup 31

3.4.2Experimental Results 32

3.5Discussion 33

3.5.1Comparison of Experimental and Modeling Re-

sults 33

3.5.2Amplifiers for Laser Applications 37

3.5.3Conclusions 37

4Ti:Tm:LiNbO3waveguide laser 39

4.1Introduction 39

4.2Laser Cavity 39

4.3Experimental Setup 40

4.4Power Characteristics 40

4.4.1High-Q Operation at λs=1890 nm 40

4.4.2Low-Q Operation at λs=1850 nm 42

4.4.3Relaxation Oscillations 43

4.5Spectral Properties 44

4.5.1Emission at 1890 nm 44

4.5.2Radio Frequency Spectrum of Laser Output 45

4.5.3Emission at 1850 nm 45

4.6Optimization 48

4.7Conclusions 49

5 conclusions and outlook 51

5.1Conclusions 51

5.2Outlook 52

a thulium doped waveguides for quantum memory

applications 53

a.1Waveguide Fabrication 53

a.2Waveguide Characterization 55

a.3Quantum memory 55

b optical spectrum of Ti:Er:LiNbO3waveguide laser 57

6 acknowledgements 59

bibliography 61

INTRODUCTION

1.1 background

Lasers are sources of coherent, diffraction limited electromagnetic

beams. The acronym laser (light amplification by stimulated emis-

sion of radiation) is based on the fact that lasers rely on the phe-

nomenon called stimulated emission of radiation for amplification of

light. When reported for the first time by Maiman [1] in 1960, its

critics described lasers as a solution looking for a problem. Now, more

than five decades after that remarkable invention, lasers find applica-

tions in day to day life in devices like CD players, bar code scanners

etc. and in a variety of fields, like communications, metrology [2],

medicine [3], etc. to name a few.

Integrated optics refers to the integration of various optical devices

like lasers, modulators, beam splitters etc. into one substrate [4]. Its

essential feature is the fabrication of multifunctional chips, like in the

case of integrated circuits in the domain of electronics. Four decades

after the first proposal [5] of integrated optics, a lot of progress has

been made in fabricating various devices as well as in the integration

of several devices in one chip [6]. Integrated lasers have improved

properties in comparison to their bulk counter-parts [7] due to (i)

the reduction of cavity mode volume due to optical confinement, (ii)

higher optical gain, (iii) lower thresholds and (iv) the possibility for

integrating several devices (including the laser) in one chip.

Lithium niobate (LiNbO3) is a versatile substrate for integrated op-

tics because of its excellent electro-optic, acousto-optic, nonlinear op-

tic properties combined with the possibility to fabricate low loss opti-

cal waveguides [8]. Moreover, it can be easily doped with rare-earth

ions (to get a laser active material and) to take advantage of the favor-

able properties of both, the host material and the rare earth ions si-

multaneously. A prominent example is the demonstration of a whole

family of erbium doped waveguide lasers namely Fabry-Pèrot type

lasers, Distributed Bragg Reflector- (DBR-) lasers, acousto-optically

tunable lasers, electro-optically Q-switched lasers and harmonically

mode-locked lasers [9] and of neodymium doped waveguide lasers

[10,11]. Also attractive properties of rare earth-doped waveguide am-

plifiers in LiNbO3have been shown [11,12]. Another more recent

example is the demonstration of a thulium-doped waveguide quan-

tum memory in LiNbO3[13].

1.2 motivation

The research presented in this thesis started with the discovery of

self-pulsations from a Fabry-Pérot type Ti:Er:LiNbO3waveguide laser

(λp=1480 nm, λs=1611 nm) [14]. The laser was found to emit pulses,

as in the case of a passively mode locked laser, without the aid of

any intracavity elements to induce such pulsations. This observation

immediately raised several open questions.

Would it make any difference if the laser active ions are different? In or-

der to find the answer, Ti:Tm:LiNbO3waveguide amplifier and sub-

sequently lasers were developed. This is described in detail in this

thesis. Although the answer to the question posed above seems to be

no as hinted by the laser results, it is yet to be proved beyond doubt.

Apart from the academic interests which motivated this research

there are other equally interesting aspects from the point of view of

applications of the integrated lasers mentioned above. Lasers operat-

ing at wavelengths longer than 1380 nm are termed eye safe. This is

because of the strong absorption of these wavelengths by the front

part of the human eye (cornea and vitreous humor) resulting in some

sort of protection1to the retina [15]. There are several applications

where such eye safe lasers are needed because the radiation cannot be

guided. Laser radar, remote sensing, communications etc. are exam-

ples of such applications. In particular Tm-doped lasers, depending

on the host material, emit in the wavelength range 1650 nm to beyond

2000 nm. Two example applications of Tm lasers are mentioned in the

following. Several gas molecules have absorption bands falling in this

spectral region. Tm-doped lasers can be used for the spectroscopy

and detection of such gases. Tm lasers emitting at the appropriate

wavelength can be used for performing laser surgery. This is because

of the strong water absorption lines falling in the emission spectrum

of Tm.

There is a growing interest in Tm-doped LiNbO3waveguide lasers

utilizing the 3F4→3H6transition to get emission in the wavelength

range 1600 nm < λ < 1900 nm [16,17] and in Tm-doped waveg-

uide lasers in other hosts [18,19]. Up to now, Tm:LiNbO3lasers were

pumped at 795 nm wavelength exploiting the strong absorption by

the 3H6→3H4transition. This thesis discusses the development of

in-band pumped (λp=1650 nm) Ti:Tm:LiNbO3waveguide amplifiers

and Fabry-Pérot type Ti:Tm:LiNbO3waveguide lasers (λs=1890 nm

and 1850 nm).

1This applies only to stray radiations of low powers. With high power radiation ade-

quate saftey measures should be taken.

1.3 organisation of thesis

The core component of the waveguide amplifiers and laser cavities

mentioned in this thesis is a Ti:Tm:LiNbO3waveguide. The fabrica-

tion of Ti:Tm:LiNbO3waveguides and their characterization are men-

tioned in chapter 2. Chapter 3discusses how a Ti:Tm:LiNbO3waveg-

uide can be used as an amplifier. An in-band pumped Ti:Tm:LiNbO3

waveguide amplifier capable of achieving broadband optical gain in

the wavelength range 1750 nm < λ < 1900 nm is demonstrated. This

chapter includes a discussion of in-band pumping, extensive mod-

eling of amplifiers, the specially developed experimental setup for

small signal gain measurements, results of small signal gain measure-

ments which are in good agreement with modeling results and a dis-

cussion on amplifiers for laser applications. The waveguide lasers op-

erating near 1890 nm and 1850 nm developed using a Ti:Tm:LiNbO3

waveguide amplifier is presented in chapter 4. This chapter includes a

discussion of the formation of laser cavity, results of laser experiments

done with both lasers and optimization of the 1890 nm laser. The high-

light is the first demonstration of an in-band pumped Ti:Tm:LiNbO3

waveguide laser (i) emitting at the longest emission wavelength, (ii)

with the smallest laser threshold and the highest output power re-

ported from a Tm:LiNbO3waveguide laser, so far.

A Ti:Tm:LiNbO3waveguide can be used as a solid state quantum

memory [13]. Appendix Abriefly outlines the fabrication and charac-

terization of Ti:Tm:LiNbO3waveguides for quantum memory appli-

cations.

WAVEGUIDE FABRICATION AND

CHARACTERIZATION

2.1 introduction

By doping lithium niobate (LN) with rare earth (RE) ions, it is pos-

sible to fabricate integrated optical devices which make use of the

properties of the host material as well as the rare earth ion. RE doped

integrated lasers [20,10] in LN are examples of such devices where

the possibility to fabricate low loss waveguides in LN is combined

with the laser properties of the rare earth ion. The laser(s) described

in this thesis were fabricated by three processing steps: (i) indiffus-

ing the RE ion (Er or Tm) into congruent lithium niobate (CLN), (ii)

formation of Ti indiffused waveguides fabricated on the RE doped

surface and (ii) subsequently the laser cavity was formed by coating

suitable mirrors on waveguide endfaces.

Laser active ions can be incorporated into LN lattice by various

techniques. For example Er can be doped into LN lattice by indif-

fusion, ion implantation, pulsed laser deposition or by growing the

crystal from a doped melt [21]. Our group has performed Pr doping

[22] and Er doping [20,9] in the past as well as Tm doping [23,24]

of LN by indiffusion recently. Particularly, detailed investigations on

diffusion doping of Er were done and the results are described in [21].

Irrespective of the dopant used, we start by coating a planar layer of

the RE atoms on the surface of a CLN wafer. Due to the lithium defi-

ciency (Li/Nb=0.94) in CLN, the dopant atoms occupy regular lattice

sites and a dopant concentration of several mole % can be realized.

Afterwards, this layer is indiffused at a temperature lower than the

Curie temperature of LN (1140 0C), so that the ferro electric phase of

the lattice is not disturbed during the diffusion process.

In this chapter a brief overview on RE doping of LN by indiffu-

sion (with Tm-doping as an example), subsequent Ti waveguide fab-

rication and the characterization of the fabricated waveguides thus

formed are given.

2.2 fabrication of ti:tm:linbo3waveguide

The RE doped waveguides (and laser cavities) used for the research

mentioned in this thesis were fabricated by Raimund Ricken and Vik-

tor Quiring from the technology wing of our lab.

2.2.1Tm Diffusion Doping

Fig. 2.1depicts the different steps of waveguide fabrication which

were discussed above. Commercially available 0.5mm thick Z-cut

wafer of undoped optical grade CLN were Tm-doped near the +Z-

surface before waveguide fabrication. A vacuum deposited Tm layer

of 32 nm thickness was in-diffused at 1130 0Cduring 150 hours in an

argon atmosphere followed by a post treatment in oxygen (1hour).

2.2.2Fabrication of Ti Waveguide

On the Tm-doped surface 104 nm thick Ti layer was deposited first.

Subsequently stripes with widths ranging from 4.5µm to 8.5µm

in steps of 0.5µm were defined by a photolithography step and in-

diffused at 1060 0Cfor 9.6hours to form 60 mm long optical strip

waveguides. Subfigure 6of Fig. 2.1shows the co-ordinate system of

the crystal (capitals) with respect to the laboratory frame (small let-

ters).

LiNbO3

1LiNbO3

Tm:LiNbO3

LiNbO3

Photoresist

LiNbO3

Ti stripe

LiNbO3

Ti:Tm:LiNbO3

-Z,y

X,x

Y,z

Figure 2.1: Ti:Tm:LiNbO3waveguide fabrication steps. 1&2- Tm deposi-

tion and indiffusion, 3- Ti deposition, 4&5Ti stripe defenition

and 6- Ti indiffusion to realize Ti:Tm:LiNbO3waveguide.

The thickness of the RE layer, the diffusion temperature and the

indiffusion time are adjusted in such a way that the (i) resultant RE

depth profile has a good overlap with the fundamental waveguide

mode of both pump and signal emissions as well as (ii) the reservoir

(RE layer) is completely depleted resulting in a smooth surface which

is essential for the realization of low loss optical waveguides. The

thickness of the Ti layer, the diffusion temperature and the indiffusion

time are adjusted in such a way that the waveguides are mono mode

for wavelengths larger than 1500 nm. Indiffusion of RE ions is done

prior to the fabrication of Ti waveguides. This is due to the fact that

the Ti indiffusion occurs at a much faster rate in comparison to the

RE indiffusion.

Afterwards, the sample was cut to 5mm, 10 mm, 15 mm and 30

mm long pieces (breadth = 12 mm, in all cases) and the end faces

of each piece were polished perpendicular to the waveguides. The 15

mm long sample was used for investigations to determine waveguide

properties, except for the fluorescence measurement which was done

with the 5mm long sample. Subsequently, the 15 mm long sample

was used for amplifier experiments and the 30 mm long sample was

used for laser (fabrication and) experiments. The width of the waveg-

uide which was used for both amplifier and laser experiments is 6.5

µm.

2.3 characterization of Ti:Tm:LiNbO3waveguide

The various properties of the waveguide namely, mode intensity dis-

tribution, scattering loss, Tm doping profile, absorption spectrum,

fuorescence spectrum and transition cross sections were character-

ized as mentioned below.

2.3.1Waveguide Mode Intensity Distribution

Whether the waveguide is monomode as designed can be tested by

measuring the near field intensity distributions. The near field inten-

sity distribution of the waveguide mode was measured by imaging

the magnified (100x) waveguide mode onto an infrared camera. Am-

plified spontaneous emission from a 1650 nm diode laser operated

below threshold was used as the light source for this measurement

so as to minimze the potential errors that could arise from interfer-

ence effects. From the captured image (see Fig. 2.2) the dimensions

of the intensity profile is calculated using a specially developed soft-

ware1. The measured full width at half maximum of the fundamental

TE mode at 1650 nm is 7.8µm ×6.2µm. On the other hand the calcu-

lated FWHM of the TE mode at 1650 nm, with fabrication parameters

of Pb304z input to Focus [25] are 7.0µm ×5.2µm. This is slightly

smaller in comparison with the measured values. The software takes

the bulk density of Ti for performing the calculation whereas the ac-

tual density is slightly lower. This is the reason for the discrepancy.

1by Dr.Ansgar Hellwig.

Width (µ m)

Depth (µ m)

−5 0 5 10

−10

−5

Figure 2.2: Measured (left) and calculated (right) mode intensity profiles of

a6.5µm waveguide. The blue line (in measurement data) cor-

responds to 1µm. Isolines are drawn at intensity levels corre-

sponding to 90%,70%,50% etc.

2.3.2Scattering Losses

The propagation losses of Ti:Tm:LN waveguides were measured by

the Fabry-Pérot method which was developed in our group [26]. A

waveguide with polished endfaces perpendicular to the waveguide

itself form a low finesse Fabry-Pérot resonator because of the resid-

ual endface reflectivities arising from the air:LN interface. During the

experiment the transmitted optical power behind such a waveguide

resonator is measured as a function of wavelength (or as a function of

the resonator length) using the setup shown in Fig.2.3. A narrow band

laser emission with a coherence length longer than twice the optical

path length of the waveguide is used as the light source in the ex-

periment so that the Fabry-Pérot resonances of the low finesse cavity

can be measured. From the measured curve (see actual measurement

data in Fig.2.4), we can calculate the waveguide propagation losses

with the following relation,

α=4.34

L(ln[R] + ln[2] − ln[K]) (2.1)

where the contrast, K=Imax−Imin

Imax+Imin , R is the waveguide endface

reflectivity and L is the length of the waveguide in cm. With the above

expression αis obtained in units of dB/cm.

Narrowband

laser Polarization

controller

Ti:RE:LN

waveguide

InGaAs

Figure 2.3: Experimental setup used to measure waveguide propagation

loss.

Narrow band emission centered at 1505 nm (from a tunable ex-

tended cavity diode laser) was used for the measurement because

of (i) the weak Tm absorption at this wavelength and (ii) the fact

that only the fundamental mode will be supported by the waveg-

uide. Apart from the 6.5µm waveguide, the adjacent 7µm waveg-

uide was also used for the measurement. The result from the 7µm

waveguide for TE polarised light is shown in Fig.2.4. For this result,

1505 1505.05 1505.1

0.02

0.04

0.06

0.08

0.1

Wavelength (nm)

Photocurrent (a.u.)

K∼25.3%K∼25.3%

Figure 2.4: Measured Fabry-Pérot resonances from a Tm-doped 7µm wide

waveguide (l=15 mm).

the calculated scattering loss value is 0.3dB/cm. In the case of a 6.5

µm waveguide, the measured scattering loss value is slightly higher

(0.4dB/cm). On the other hand for an undoped (i.e., without Tm)

7µm wide waveguide the scattering loss value is only 0.2dB/cm.

The difference of the scattering losses could be due to surface rough-

ness arising from the Tm doping. One of the parameters that needs

to be input for laser modeling is the propagation loss of the waveg-

uide. It was observed that the calculated power characteristics (see

section 4.4) agrees reasonably well with the measured result, when

the scattering loss value is only 0.1dB/cm. Hence, the measured re-

sult could be slightly higher than the actual scattering losses of the

6.5µm waveguide.

2.3.3Thulium Depth Profile

The doping profile of Tm ions was measured by secondary neutral

mass spectroscopy (SNMS). The measurement was performed (by

Dr.Huberus Paulus from Fachhochschule Südwestfalen, Söest.) using

700 eV Argon-ions for ion milling. Ions and electrons were extracted

from the plasma source with a duty cycle of 4:1at a rate of 320 kHz

to avoid charging of the insulating CLN-substrate. SNMS was chosen

instead of secondary ion mass spectroscopy (SIMS) to significantly re-

duce matrix effects (see e.g. [27]). In Fig. 2.5, the concentration profile

versus depth has been recorded for thulium (Tm). It is expected that

0 0.4 0.8 1.2 1.6

−10

−9

−8

−7

−6

−5

−4

−3

−2

−1

Tm Conc. (1020/cm3)

Depth (µ m)

Figure 2.5: Measured Tm depth profile.

Tm occupies regular Li-sites similar to Er-ions when incorporated in

CLN by diffusion [28]. The maximum Tm concentration (at surface)

of the sample was about 1.25×1020cm−3corresponds to a concentra-

tion 0.68 mole % and a penetration depth (d1/e) of 9microns.

Thulium (69Tm) is the 13th element in the lanthanide series with

the electronic configuration - [Xe]4f136s2. In pure form it is a soft

metal. Upon incorporation into a dielectric host, it loses three elec-

trons to become [Xe]4f12 (or Tm3+). The unfilled 4f electrons interact

by spin-spin and spin-orbit interaction and this leads to different en-

ergy levels among the 4f electrons. As an example, the 3H6,3H4and

3F4energy levels of a Tm ion in LN lattice along with different tran-

sitions resulting from absorption or emission of photons are shown

in Fig. 2.6(left). When incorporated into a host, the crystal field lifts

the degeneracy of the different energy levels leading to a spread in

energies. The corresponding splitting of 3H6and 3F4energy levels

for a Tm ion incorporated into a LN lattice is shown on the right side

of Fig. 2.6. These energy levels of Tm3+ion are exploited from the

point of view of amplifier and laser applications which are discussed

in this thesis. Tm energy levels can be exploited for quantum memory

applications, as well.

3H4τ= 230µs

3H6

τ= 2.4ms

3F4

795 nm

1375 - 1530 nm

1625 - 1950 nm

3F4

3H6

500

5500

6000

Energy (cm−1)

1660 nm 1916 nm 1762 nm

Figure 2.6: Tm:LiNbO3energy levels (from [29]) and the various transitions

between them. The splitting of 3H6and 3F4energy levels is

shown in the right side.

2.3.4Absorption Spectra

As the next step both TE and TM polarized absorption spectra in

the wavelength range 1600 nm <λ<1900 nm. were measured. This

correspond to weak excitation of Tm ions from the 3H6level to the

3F4level (see Fig. 2.6).

Halogen

lamp

SM fiber Ti:Tm:LN

waveguide Polarizer

Monochromator ex-InGaAs

Chopper

Lock-in amplifier

Figure 2.7: Schematic of the experimental setup developed to measure the

absorption spectrum of the Ti:Tm:LiNbO3waveguide.

The schematic of the experimental setup used is shown in Fig. 2.7.

A halogen lamp was used as the light source. The filament was im-

aged onto the end face of a standard single mode fiber (SMF 28) and

the other fiber end was kept butt coupled to one endface of the waveg-

uide. In order to avoid any Fabry-Pérot etalon effects that might arise,

index matching fluid was applied at the interface. The collimated

beam from behind the waveguide was analysed using a monochro-

mator. A polarizer kept in front of the monochromator blocked the

orthogonal polarisation. As the detected optical power levels were

quite low, a lock-in amplifier together with a cooled extended InGaAs

photodiode were used in the detection circuit. The reference signal

was derived from a chopper kept in the beam behind the sample. The

measurements were controlled by a computer which also recorded

the data. To evaluate Tm absorption, two measurements were per-

formed, one with a doped and the second with an undoped waveg-

uide. The measurements were done with a resolution bandwidth of

2nm. From published data [30] it was known that at wavelengths be-

yond 1950 nm Tm ions in lithium niobate have negligible absorption.

Ideally both measured curves should overlap well in the spectral re-

gions without Tm absorption (assuming identical scattering losses in

both waveguides). But this was found to be a bit hard to be achieved

in practice (this could be due to the fact that coupling efficiency into

both doped and undoped waveguides might not be the same for the

entire spectral range). In order to take care of the level differences in

the measurement data (from doped and undoped waveguides) at the

long wavelength side, one of the data set was multiplied by a suitable

constant such that both curves overlap well at wavelengths beyond

1950 nm. Afterwards the absorption by the Tm ions was calculated

by dividing the measured curve from doped waveguide and that ob-

tained from an undoped waveguide. The results are shown in Fig. 2.8.

There are two absorption bands one centered at 1660 nm and the sec-

1600 1700 1800 1900

−10

−8

−6

−4

−2

Wavelength (nm)

Absorption (dB)

L = 15 mm

Figure 2.8: TE and TM polarised absorption spectra of a Ti:Tm:LiNbO3

waveguide corresponding to 3H6→3F4transition in the wave-

length range 1600 nm <λ<1900 nm. The measurement was

done with a resolution bandwidth of 2nm.

ond one at 1758 nm. The fine structure (which one might expect from

the energy levels shown in Fig. 2.6) is not visible in this measurement

as this measurement was done at room temperature.

2.3.5Fluorescence Spectrum

TE polarised fluorescence measurement was done by pumping the

waveguide with TE polarised laser light at 795 nm from a Ti:Sa (tita-

nium sapphire) laser (see Fig.2.6for the corresponding optical tran-

sitions). The Tm ions are excited to the 3F4level from where they

subsequently relax to the 3H6level by emitting a photon (radiative

transition. There could be nonradiative transitions also). For this mea-

surement, the setup was slightly modified from what is shown in Fig.

2.7. The pump beam was coupled into the waveguide (l = 5mm) using

free space optics. The signal emitted by the sample was coupled into

a PM (polarization maintaining) fiber kept butt coupled and aligned

with its slow axis in the vertical direction. The other end of the fiber

was aligned with its slow axis in vertical direction at the input side

of the monochromator. A lens kept behind the fiber end face colli-

mated the beam directed into the monochromator optics. A pump

blocking filter kept in the free space beam path attenuated the resid-

ual pump emission from behind the waveguide. As in the absorption

measurement, the measurement was done with a lock-in amplifier.

The chopper was placed in the collimated beam path. The measured

result is shown in Fig. 2.9. A similar measurement done by placing

the chopper in the pump beam path also yielded the same result.

The three prominent fluorescence peaks recorded by Cantelar [30] at

1762,1800 and 1850 nm from bulk Tm:LiNbO3are observed in this

measurement as well. These peaks also correspond to the integrated

laser wavelengths reported from Tm:LiNbO3so far and these are in-

1600 1700 1800 1900

0.2

0.4

0.6

0.8

Wavelength (nm)

Photocurrent (a. u.)

L = 5 mm

This work

Madrid

Southampton

Figure 2.9: TE polarised fluorescence spectrum of a Ti:Tm:LiNbO3waveg-

uide corresponding to 3F4→3H6transition in the wavelength

range 1600 nm <λ<1950 nm. The measurement was done with

a resolution bandwidth of 2nm. Ti:Tm:LiNbO3waveguide laser

lines reported by Madrid [17] and Southampton [16] are also

shown.

dicated in the figure. As the measurement was done by collecting

light from behind the waveguide, the actual shape of the fluorecence

spectrum may not be reflected in the measured curve. This is because

of potential distortions due to (i) stimulated emission (might result

in amplification of the peaks) and (ii) reabsorption (might attennuate

some part of the spectrum).

2.3.6Transition Cross Sections

Transition (more specifically, absorption and emission) cross sections

quantify the ability of a mterial to absorb/ emit light [31]. In order

to perform modeling of waveguide amplifiers (and lasers), we need

to know absorption and emission cross sections in addition to Tm

doping profile, pump and signal mode distributions and energy level

life times. Absorption cross sections of the Tm doped waveguide were

calculated from the absorption spectrum as follows:

σa(λ) = α(λ)

Nd(λ)(2.2)

where σa(λ)is the absorption cross section at the wavelength λ,α

is the corresponding absorption coefficient and Nd(λ)is the effective

Tm density inside the fundamental mode at the wavelength λ. The

absorption coefficients were deduced from the absorption spectrum

and the effective Tm density was calculated numerically from the

waveguide mode intensity profile and the Tm depth profile. Hence

absorption cross sections were calculated using equation 2.2. These

are shown in Fig.2.10.

The McCumber relation between emission cross section σeand ab-

sorption cross section σais

σe(ν) = σa(ν)e−hν

kT (2.3)

where, σe(ν)is the emission cross section at the optical frequency

ν(= c

λ),σa(ν)is the absorption cross section, kis Boltzmann’s con-

stant, his Planck’s constant, is the net free energy to excite one Tm

ion from the 3H6level to the 3F4level at temperature T. The net free

energy was determined as described by Cantelar [30] assuming that

the Tm energy level distribution in both cases are the same. This is

justified by the fact that the absorption cross section spectrum was

found to have a good agreement with that reported by Cantelar [30].

1600 1700 1800 1900

0.5

1.5

Wavelength (nm)

Cross section 10−20(cm2)

Absorption

Emiss. (Mc Cumber)

Figure 2.10: TE polarized emission cross sections calculated from absorption

cross sections of Ti:Tm:LiNbO3doped waveguide.

Using equation 2.3the emission cross sections from the correspond-

ing absorption cross sections can be calculated. The cross sections

thus obtained (at room temperature) are given in Fig. 2.10. One of

the assumptions of the McCumber theory is that the Stark splitting

of the higher excited state (3H4level, in this case) is of the order of

kBT (∼200 cm−1). In the case of Tm:LN, according to the energy level

structure given by Cantelar [30], the splitting of the 3H4level is 410.4

cm−1. In cases where the Stark splitting is larger than kBT, the Mc-

Cumber theory over estimates the lower energy part of the spectrum

[32]. Such effects are already reported in RE doped glasses [33].

2.4 conclusions

In the first part of this chapter the fabrication of Ti:Tm:Li:NbO3waveg-

uides by diffusion doping method is briefly described. The character-

ization of the waveguides fabricated by means of scattering loss and

near field measurements are subsequently discussed. The studies on

properties of the waveguide due to Tm doping (absorption and fluo-

rescence) are also presented. The transition cross sections of the Tm-

doped waveguide which are important from the point of view of am-

plifier/laser fabrication using the waveguide are also determined for

the first time via the McCumber theory. Emission cross sections can

be determined by measuring the fluorescence spectrum of emission

perpendicular (so that effects of re-absorption and amplification can

be minimized) to the waveguide. Such a measurement would allow a

cross check of the emission cross sections determined by McCumber

theory.

Ti:Tm:LiNbO3WAVEGUIDE AMPLIFIER

3.1 introduction

From the transition cross section spectra (Fig. 2.10) determined from

absorption measurements, it is clear that optical amplification is possi-

ble with a pumped Ti:Tm:LiNbO3waveguide for signal wavelengths

beyond 1750 nm. This is because in this spectral region the emis-

sion cross section is larger in comparison to the absorption cross sec-

tion for a given wavelength. Small signal gain measurements were

done with the 15 mm long Ti:Tm:LiNbO3waveguide. The results,

which are presented in this chapter, demonstrate that broad-band op-

tical gain in the wavelength range 1750 nm < λs<1900 nm can be

achieved using a Ti:Tm:LiNbO3waveguide as an amplifier.

3.2 in-band pumping scheme

3.2.1General Considerations

One of the preconditions to achieve optical amplification by making

use of stimulated emission is the creation of population inversion

(i.e., for the case of Ti:Tm:LiNbO3, the number of Tm atoms in ex-

cited state should be larger than in the ground state) in the gain

medium [34]. One possibility to create a population inversion in the

Ti:Tm:LiNbO3waveguide is by making use of an absorption band of

the gain medium (optical pumping). The most prominent absorption

band of Tm is centered at 795 nm. The corresponding atomic tran-

sitions are shown in Fig. 2.6(left). By using 795 nm laser light Tm

atoms can be excited to the 3F4level, creating a population inver-

sion between the the 3H6(ground) and the 3F4(laser) levels. Optical

pumping by laser light at 795 nm is the widely used approach in

realizing Tm lasers. Some examples can be found in [18,19,16,29].

The choice of the 795 nm pump wavelength is due to the strong ab-

sorption1at this wavelength. One of the interesting feature of this

pumping scheme is the Tm-Tm cross relaxation process, which re-

sults in the excitation of two Tm atoms at the expense of a pump

photon. Relatively large Tm concentration is needed to take advan-

tage of this process in the case of Tm:LiNbO3[29]. Indiffusion of Tm

to achieve the required Tm concentration would result in increased

1predominantly TE polarised in the case of Ti:Tm:LiNbO3and ∼3times larger in

comparison to that at 1660 nm [30].; see Fig. A.3for the corresponding absorption

spectrum of a Ti:Tm:LiNbO3waveguide fabricated at Paderborn.

surface roughness (as noted by our study in this direction; see Fig.

3.1) and consequently larger waveguide scattering losses.

Figure 3.1: Micrographs of waveguides (view from top) from sample Pb304z

(left; 32 nm thick Tm layer indiffused for 150 hours) and Pb227z

(right; 50 nm thick Tm layer indiffused for 150 hours). The in-

creased surface roughness due to thicker Tm layer is visible.

An alternative pumping approach is the so called in-band pump-

ing scheme, where the Tm atoms in the ground state manifold are

pumped directly to the manifold which includes the laser level (in-

stead of having an intermediate level as in the case of pumping Tm

atoms using 795 nm light). A recent example of an in-band pumped

Tm laser can be found in [35]. The measured absorption spectra (Fig.

2.8) show two strong absorption bands centered around 1660 nm and

1758 nm for TE polarised light. The corresponding absorption is not

so strong for TM polarised light. For in-band pumping, the strong

Tm absorption bands around 1660 nm or 1758 nm can be exploited.

The corresponding transitions are shown in Fig.2.6(right). The best

choice of pump wavelength would be 1660 nm, as will be shown

later. The Ti:Tm:LiNbO3waveguide amplifer and laser(s) mentioned

in this thesis are in-band pumped at 1650 nm. The choice of the pump

wavelength was mainly due to the availability of a commercial laser

diode operating at that wavelength. Pumping at 1660 nm would ex-

cite Tm atoms from the lower lying multiplets of 3H6level to the

higher lying multiplets of 3F4level and subsequently they relax very

fast (relaxation time ∼1ps) to the lower lying levels of 3F4multiplet

due to thermalisation. As the overall population of both multiplets

follows the Boltzmann distribution, a population inversion between

the lower levels of 3F4multiplet and higher lying levels of 3H6mul-

tiplet can be achieved by relatively low pump powers. In this case

the gain medium (Tm:LiNbO3) acts as a four level system. On the

other hand, short wavelength transitions (1762 nm, as an example)

correspond to a three level system. In that case population inversion

can be achieved only with a relatively higher pump rate. As the in-

band pumping scheme does not rely on cross relaxation processes (as

in the case of pumping at 795 nm) to achieve population inversion,

optical amplification can be achieved with moderate Tm concentra-

tions. In-band pumping scheme is already demonstrated in the case

of Ti:Er:LiNbO3waveguide amplifiers [36] by exploiting the strong

absorption of Er:LN at 1480 nm.

The attempts to make use of the 3H6→3H4transition (795 nm)

for pumping did not yield any promising results and this outcome

was a strong motivation to consider other pumping schemes. The

reasons for the failure are thought to be due to (i) photorefraction

occuring in lithium niobate at short wavelengths, (ii) poor overlap

between the fundamental waveguide mode at 795 nm and laser (sig-

nal) mode and (iii) poor overlap between the higher order waveguide

mode(s) at 795 nm which also would be excited (usually) and laser

(signal) mode. The effect of photorefraction is strongly reduced at

longer wavelengths and hence the in-band pumping scheme does not

suffer from detrimental effects due to photorefraction. A third reason

which favors the in-band pumping scheme is the fact that the overlap

of pump mode with the signal mode is much better compared with

the fundamental mode at 795 nm. This reduces the attenuation of sig-

nal mode due to absorption at unpumped regions of the waveguide

which improves the pumping efficiency.

3.2.2Choice of Pump Wavelength

Theoretical modeling of Ti:Tm:LiNbO3waveguide amplifiers and laser

presented in this thesis was done by the software R P Fiber Power

(RPFP) [37]. The theoretical model implemented by RPFP is described

briefly in section 3.3. Here modeling results concerning the choice of

pump wavelength for the case of an in-band pumped 15 mm long

Ti:Tm:LiNbO3waveguide amplifier are presented.

Three wavelengths from the 1660 nm absorption band (1642 nm,

1660 nm and 1674 nm) and 1758 nm were chosen as the pump wave-

lengths. A coupled pump power of 13.5mW was assumed. The out-

come of this set of calculations is shown in Fig. 3.2. Net gain cor-

responds to the calculated gain including the waveguide scattering

losses. From the calculated results, it is clear that 1660 nm is the best

pump wavelength for in-band pumping of a 15 mm long Ti:Tm:LiNbO3

waveguide amplifier, as at that pump wavelength, maximum gain can

be achieved with the same coupled pump power level. On the other

hand pumping at 1758 nm is not as efficient (Fig. 3.3). This is due

to the the fact that at 1758 nm the difference between absorption and

emission cross sections is the smallest among all the pump wave-

lengths considered. Consequently some of the excited Tm atoms con-

1600 1700 1800 1900

−2

Wavelength (nm)

Net Absorption/Gain (dB)

1674 nm (TE)

1650 nm (TE)

1660 nm (TE)

Figure 3.2: Calculated net absorption/gain spectra of a 15 mm long

Ti:Tm:LiNbO3waveguide amplifier, single pass pumped at 1642

nm, 1660 nm and 1674 nm (signal and pump are TE polarized)

with a coupled pump power of 13.5mW.

tribute back to the amplification of pump and this makes this pump-

ing scheme less efficient. The attempts to procure a 1660 nm diode

laser were not fruitful and hence it was decided to use a Fabry-Pérot

diode laser operating at 1650 nm for pumping the amplifier (and laser

cavities) mentioned in this thesis, as such a diode laser could readily

be obtained.

1600 1700 1800 1900

−4

−2

Wavelength (nm)

Net Absorption/Gain (dB)

1758 nm (TE)

1660 nm (TE)

Figure 3.3: Calculated net absorption/gain spectra of a 15 mm long

Ti:Tm:LiNbO3waveguide amplifier, single pass pumped at 1650

nm and 1758 nm (signal and pump are TE polarised) with a cou-

pled pump power of 13.5mW.

3.3 modeling

Theoretical modeling of waveguide amplifier and laser presented in

this thesis were done with the commercial software RP Fiber Power

[37]. The physical model implemented by the software assumes the

following [38]: (i.) The doping profile is constant in longitudinal direc-

tion, (ii.) the gain medium is homogeneously broadened, (iii.) spatial

hole burning is neglected (iv.) the intensity distribution of an optical

channel is preserved at all times (v.) there is no coherence between dif-

ferent optical channels (vi.) there are no nonlinear effects (stimulated

Raman and Brillouin scattering) arising in the laser or amplifier and

(vii.) the population distributions within each Stark level manifold of

laser active ions are in thermal equillibrium.

Consider the two atomic energy levels E1and E2of a laser active

medium shown in Fig. 3.4. There can be three processes shown in the

figure between these two energy levels which involves a photon. In

the case of absorption, an incoming photon (having an energy equiv-

alent to the difference of energies of the two levels, hν =E2−E1)

gets absorbed by the atom resulting in the atom getting excited to

the higher level. When an incoming photon of energy hν interacts

with an excited atom, the atom can relax to the lower energy level

emitting a second photon. This process is called stimulated emission.

The third process, called as spontaneous emission occurs when the

excited atom relaxes to the lower energy level by emitting a photon

of energy hν. In the case of stimulated emission the resulting pho-

Absorption Stimulated

Emission

Spontaneous

Emission

Figure 3.4: Depiction of absorption, stimulated emission and spontaneous

emission processes in a laser active medium.

ton has the same quantum mechanical properties as the stimulating

photon. An amplifier (and a laser) relies on stimulated emission for

its operation. On the other hand spontaneous emission (and the cor-

responding stimulated emision arising from spontaneous emission

which results in amplified spontaneous emission or ASE) is a source

of noise in an optical amplifier (and laser) which relies on stimulated

emission for its operation.

For the case of an optical waveguide doped with laser active ions

having an energy level scheme as shown in Fig. 3.4at position z(lon-

gitudinal direction), the local gain coefficient g is given by,

g(z,λ) = NRE[n2(z)σ21(λ) − n1(z)σ12(λ)] (3.1)

where NRE is the density of the laser active atoms, n2(z)is the frac-

tion of laser active atoms in level 2,n1(z)is the fraction of laser active

atoms in level 1,σ21(λ)is the emission cross section at wavelength λ

and σ12(λ)is the absorption cross section, zis the longitudinal posi-

tion in the waveguide. Equation 3.1is valid for an infinitesimal length

of the waveguide, where n2(z)(and consequently n1(z)) remains a

constant. For a waveguide length L, the total amplification is given

G(λ) = exp 



g(z,λ)dz

(3.2)

Taking transverse dimensions into account, the gain has contribu-

tions from different positions and is given by

g(z,λ) = ∞

−∞

NRE(y)[n2(x,y,z)σ21(λ)−n1(x,y,z)σ12(λ)]ψ(x,y,z)dxdy

(3.3)

where ψ(x,y,z)is the intensity distribution of the waveguide sig-

nal (or pump) mode being considered. In the numerical model im-

plemented by R P Fiber Power, the waveguide is sliced in x and y

directions (lateral dimensions). The contribution to gain (or absorp-

tion) at different z locations from each cube of volume dxdydz is nu-

merically added and in this way the amplification by a given length

of waveguide for a given absorbed pump power can be calculated, if

the wavelength dependent transition cross sections, RE doping pro-

file and waveguide mode intensity profiles for pump and signal are

known. All theoretical calculations of Tm:Ti:LiNbO3waveguide am-

plifier and laser presented in this dissertation have been done with

emission cross sections determined from McCumber theory.

3.3.1Gain Spectra

In this section the calculated gain spectra of two waveguide amplifiers

(l = 90 mm (Fig. 3.5) and 15 mm (Fig. 3.7) ) at three different coupled

pump powers (25 mW, 50 mW and 100 mW) are presented. The pump

wavelength is 1650 nm (TE polarized). Calculations were performed

with input signal power kept at 1µW and αs=0.1dB/cm.

The potential for broadband optical gain for signal wavelengths

from 1750 nm < λs<1900 nm is evident from the calculated gain

curves for a coupled pump power of 100 mW. As the waveguide

length is increased the ground state absorption also increases as ex-

pected. This is evident from the ’gain’ (or absorption) spectra cor-

responding to the unpumped case of the two amplifier lengths. In

the case of a longitudinally (single pass2) pumped waveguide, the

2residual pump beam from the waveguide end is not reflected back to the waveguide.

1600 1700 1800 1900

−60

−40

−20

Wavelength (nm)

Net Absorption / Gain (dB)

L = 90 mm

λp = 1650 nm (TE)

unpumped

Ppc = 25 mW

Ppc = 50 mW

Ppc = 100 mW

Figure 3.5: Calculated net absorption/gain spectra of a 90 mm long waveg-

uide amplifier at three different coupled pump powers (signal

and pump are TE polarised).

typical distribution of pump power is shown in Fig. 3.6. From the

pumped end of the waveguide, the pump power gradually becomes

smaller and smaller depending on the strength of pump absorption.

The pump power required to keep a length of waveguide pumped

(i.e., Tm atoms of gain media are maintained in the excited state) be-

comes larger and larger as the waveguide length is increased. The

distribution of pump power for the case of a 90 mm long waveguide

for three different coupled pump powers are shown in Fig. 3.6. With

0 20 40 60 80

100

Length (mm)

Pump power (mW)

λp = 1650 nm (TE)

Ppc = 25 mW

Ppc = 50 mW

Ppc = 100 mW

Figure 3.6: Distribution of pump power across a 90 mm long waveguide

amplifier at three different coupled pump powers.

25 mW of coupled pump power, the length of the waveguide which is

pumped is 8cm. In other words, the coupled pump power (25 mW in

this case) is absorbed completely by a 8cm long waveguide section.

Beyond 8cm, the waveguide is not pumped and consequently that

section of the waveguide absorbs signal radiation (λs>1750 nm).

With 50 mW of coupled pump power, the situation is improved. A

90 mm long waveguide is pumped throughout its length in this case

(14% pump transmission). With 100 mW coupled pump power, the

situation is improved even further. This is because of the fact that

the rate of pumping of Tm atoms (in an infinitesimal section of the

waveguide) is directly proportional to the incident pump power.

With pumping, Tm atoms in the ground state are excited and con-

sequently the absorption coefficient (at pump wavelength and signal

wavelengths) becomes smaller and smaller as the pump power is in-

creased. This is because of the fact that the population in the ground

state becomes smaller and smaller as pump power is increased grad-

ually. Absorption at any given wavelength will be bleached (i.e., ab-

soption coefficient becomes zero) when the atomic population of the

two sublevels (of the two manifolds) involved in contributing to the

emission (or absorption) of a given wavelength become equal. When

there is inversion between the two sublevels contributing to the emis-

sion of a given wavelength (i.e., when the pump power is larger than

what is required to bleach the absorption ), it is possible to amplify

the signal power corresponding to that wavelength. Whether a given

signal wavelength will be amplified or not, depends on the overall

gain (and absorption) experienced by the signal beam as it passes

through the amplifier. This becomes clear from the calculated gain

curve for 25 mW of coupled pump power in the case of a 90 mm

long waveguide. As said above, there is a 1cm long section of the

waveguide which is unpumped and consequently signal radiation (at

all wavelengths) will be absorbed by this section. However from the

gain spectra it is evident that a positive net gain is possible at signal

wavelengths larger than 1775 nm. This is because of the fact that the

gain experienced by these signal wavelengths while passing through

the pumped section is larger in comparison to the loss from the un-

pumped section. On the other hand the gain experienced by shorter

wavelengths while traversing through the pumped section is not suffi-

cient to overcome the absorption. Consequently they are attennuated

instead of being amplified. But at higher coupled pump powers (50

mW, for example) the gain experienced by the whole of the signal

spectrum (1750 nm < λs<1900 nm ) becomes positive.

As the absorption at longer wavelengths is relatively weaker (in

the unpumped case) bleaching of absorption at these wavelengths oc-

curs with relatively low coupled pump powers because the required

population of the higher lying sublevel can be realized at low pump

powers. The underlying reason for this is the Boltzmann distribution

of energy levels in the ground and excited state multiplets. By the

same reasoning it is clear that positive net gain is achieved first at

long wavelengths as the pump power is gradually increased. The

pump power required to achieve (bleaching and subsequently) net

gain at shorter wavelengths (below 1770 nm) is relatively larger in

comparison to that at longer wavelengths (for a given length of the

waveguide). This is a consequence of the stronger absorption at short

wavelengths in comparison to the longer wavelengths. However as

the number of atoms in the energy levels contributing to the emis-

sion at shorter wavelengths is relatively larger in comparison to those

contributing to the emission at longer wavelengths when population

inversion is achieved throughout the length of the waveguide, the net

gain achievable from a given waveguide length is always larger at

short wavelengths. This can be seen by comparing the gain spectra

for the two different amplifier lengths. As an example for 100 mW

1600 1700 1800 1900

−10

−8

−6

−4

−2

Wavelength (nm)

Net Absorption / Gain (dB)

L = 15 mm

λp = 1650 nm (TE)

unpumped

Ppc = 25 mW

Ppc = 50 mW

Ppc = 100 mW

Figure 3.7: Calculated net absorption/gain spectra of a 15 mm long waveg-

uide amplifier at three different coupled pump powers (signal

and pump are TE polarised).

coupled pump power the net gain at 1758 nm achievable from the 90

mm long amplifier is 25.7dB and that from the 15 mm long amplifier

is 4.7dB. For the same coupled pump power level, the gain at longer

wavelengths are smaller than these values.

With the diffusion technology developed in our lab, so far we have

fabricated waveguide samples up to 9cm in length. One of the best

waveguides developed so far (αs=0.03 dB/cm, waveguide length =

9cm) was used in the fabrication of a Ti:Er:LiNbO3waveguide laser

similar to that described in [14]. Due to its low scattering loss value,

the laser wavelength was 1614 nm, the longest emission wavelength

which was observed so far from a Fabry-Pérot type Ti:Er:LiNbO3

waveguide laser (see Fig. B.1). In the same way it should be possi-

ble to make Tm doped waveguides of good quality with lengths up

to 10 cm 3. In what follows the calculated results corresponding to the

net gain achieveable from different amplifier (Ti:Tm:LiNbO3waveg-

uide) lengths up to 10 cm at different coupled pump powers at four

different signal wavelengths are given. The four signal wavelengths

chosen were 1758 nm, 1800 nm , 1858 nm (three emission peaks; see

fluorescence spectrum Fig. 2.9) and 1890 nm (laser wavelength; see

section 4.5.1). Calculations were performed with input signal power

kept at 1µW and αs=0.1dB/cm.

3LN wafers with a diameter of 4inches are currently commercially available.

Fig. 3.8shows the calculated single pass gain at 1758 nm as a func-

tion of waveguide lengths for four different coupled pump powers

1 2 3 4 5 6 7 8 9 10

−32

−24

−16

−8

Length (cm)

Net absorption / gain (dB)

λp = 1650 nm (TE)

λs= 1758 nm (TE)

100 mW

50 mW

30 mW

10 mW

Figure 3.8: Calculated gain at 1758 nm (TE polarization) as a function of

waveguide length for four different coupled pump powers.

10 mW, 30 mW, 50 mW and 100 mW at 1650 nm. The coupled pump

power needed to just bleach the absorption (and achieve amplifica-

tion) becomes larger and larger as the waveguide length is increased.

At this wavelength with 10 mW of coupled pump power a waveguide

length of 2.3cm can be maintained at bleaching level (i.e., absorption

coefficient at 1758 nm is zero). With 30 mW of coupled pump power

the waveguide length which can be maintained at bleaching level be-

comes 96 mm. This is a consequence of the three level nature of the

gain medium at this wavelength. When the coupled pump power is

increased further, the net gain starts to become larger. The calculation

result show that with 100 mW of coupled pump power it is possible

to achieve a net single pass gain of 30 dB using a 10 cm long waveg-

uide.

Fig. 3.9shows the calculated gain at 1800 nm as a function of

waveguide lengths for different coupled pump powers. The length

1 2 3 4 5 6 7 8 9 10

−8

Length (cm)

Net absorption / gain (dB)

λp = 1650 nm (TE)

λs = 1800 nm (TE)

100 mW

50 mW

30 mW

10 mW

Figure 3.9: Calculated gain at 1800 nm (TE polarization) as a function of

waveguide length for four different coupled pump powers.

of waveguide which can be kept at bleaching level with 10 mW of

coupled pump power is 5.2cm at this wavelength. With a coupled

pump power of 30 mW, the absorption is not only bleached but am-

plification (i.e., positive net gain of 5.8dB) is possible even with a 10

cm long waveguide. These results point to the less prominent three

level nature of the gain medium, at this wavelength. With larger cou-

pled pump powers larger gain is achievable from long waveguide

lengths. In this case with a coupled pump power of 100 mW a 10 cm

long waveguide can yield a net gain of 18.3dB.

The calculated gain at 1848 nm as a function of different waveg-

uide lengths for different coupled pump powers at are shown in Fig.

1 2 3 4 5 6 7 8 9 10

−2

Length (cm)

Net absorption / gain (dB)

λp = 1650 nm (TE)

λs = 1848 nm (TE)

100 mW

50 mW

30 mW

10 mW

Figure 3.10: Calculated gain at 1848 nm (TE polarization) as a function of

waveguide length for four different coupled pump powers.

3.10. The length of the waveguide which can be kept just above the

bleaching level at this wavelength is 8.1cm with a coupled pump

power of 10 mW. This is because of the fact that ground state absorp-

tion at this wavelength is even smaller than at shorter wavelengths

and consequently the inversion necessary to achieve a net gain can

be realized at smaller pump powers (than what is required at shorter

wavelengths). With 30 mW of coupled pump power a net gain of

5.3dB can be achieved from a 100 mm long waveguide. At 100 mW

coupled pump power with a 10 cm long waveguide, it is possible to

achieve a net gain of 11.7dB at this wavelength.

Fig. 3.11 shows the calculated gain at 1890 nm as a function of

waveguide lengths for different coupled pump powers. Modeling re-

sults show that the maximum length of waveguide which can be kept

above bleacing level with 10 mW of coupled pump power is 7.3cm.

A coupled pump power of 30 mW would yield a gain of 1.5dB with

an amplifier length of 10 cm. With a 10 cm long waveguide and a cou-

pled pump power of 100 mW, a net gain of 3.5dB can be achieved.

The gain values are smaller than the corresponding values calculated

for 1848 nm signal wavelength. At this wavelength the gain medium

is more closer to a four level gain medium. Hence one would expect

1 2 3 4 5 6 7 8 9 10

−1

Length (cm)

Net absorption / gain (dB)

λp = 1650 nm (TE)

λs = 1890 nm (TE)

100 mW

50 mW

30 mW

10 mW

Figure 3.11: Calculated gain at 1890 nm (TE polarization) as a function of

waveguide length for four different coupled pump powers.

that an even longer waveguide length than that at 1848 nm signal

wavelength could be kept above bleaching level with 10 mW of cou-

pled pump power. But the modeling result does not favor this con-

clusion. The discrepancy is thought to arise from the inapplicability

of McCumber theory in the estimation of emission cross sections at

these long wavelengths and/or the inacuracies in the estimation of

waveguide propagation losses. Experimental results of small signal

gain measurements done with 1.5cm long amplifier also points this

conclusion.

3.3.2Power Characteristics

If the amplifier is pumped harder and harder (i.e., when pump power

is increased gradually) the gain of the amplifier at any given wave-

length also increases. However this process cannot go on infinitely.

This is because, at some point the rate of depletion of the excited

atoms will be faster than the rate of pumping (or replenishment of

the number of excited atoms). When this happens, the gain of the

amplifier becomes saturated. This is shown in Figs. 3.12 and 3.13. Fig.

3.12 shows the variation of gain of two amplifier lengths 30 mm and

90 mm at λs=1758 nm as a function of the coupled pump power. At

low pump powers the gain increases exponentially. But as the pump

power is increased further the gain of the amplifiers starts to get satu-

rated. The saturated gain of the 30 mm amplifier at 30 dBm coupled

pump power is 12 dB where as that for the 90 mm long amplifier is 36

dB. In this case, the saturated gain scales proportional to the length.

The situation is the same at a different signal wavelength λs=1850

nm. Fig. 3.12 shows the variation of gain of two amplifier lengths 30

mm and 90 mm at λs=1758 nm. At 30 dBm coupled pump power,

the 30 mm long amplifier yields a gain of 4dB whereas the 90 mm

long waveguide yields 12 dB.

−30 −20 −10 0 10 20 30

−60

−40

−20

Coupled pump power (dBm)

Net absorption / gain (dB)

L=90 mm

L=30 mm

λs= 1758 nm (TE)

Figure 3.12: Calculated net absorption/gain of 30 mm and 90 mm long

Ti:Tm:LiNbO3waveguide amplifiers as a function of coupled

pump power for signal wavelength λs=1758 nm (signal and

pump are TE polarised).

−30 −20 −10 0 10 20 30

−6

−3

Coupled pump power (dBm)

Net absorption / gain (dB)

L=90 mm

L=30 mm

λs= 1850 nm (TE)

Figure 3.13: Calculated net absorption/gain of 30 mm and 90 mm long

Ti:Tm:LiNbO3waveguide amplifiers as a function of coupled

pump power for signal wavelength λs=1850 nm (signal and

pump are TE polarised).

3.3.3Double Pass Pumping

The modeling results presented so far considered single pass pumped

amplifiers. The calculated internal distribution of pump power as a

function of position inside waveguide for a 15 mm long waveguide

amplifier is shown in Fig. 3.14. The residual pump power at the end

of a 1.5cm long amplifier is 8.4mW. It is possible to make better

use of this residual pump power by reflecting it back into the cav-

ity by placing a pump reflector mirror at the end of the waveguide.

The calculated result for a 1.5cm long amplifier with an ideal pump

reflector (pump reflection = 100% ) is also shown in Fig. 3.14. The

pump power distribution is at a higher level than with the case of

single pass pumping. This would mean that the pumping rate along

the waveguide is better than with the single pass pumping scheme

0 3 6 9 12 15

Length (mm)

Pump power (mW)

λp = 1650 nm (TE)

Figure 3.14: Calculated internal distribution of pump power as a function of

position inside waveguide for a 15 mm long waveguide for 13.5

mW coupled pump power for single pass (red) and double pass

(black) pumping schemes.

and consequently we can achieve an even higher gain where the sig-

nal is still single pass. Fig. 3.15 shows the comparison between both

single and double pass pumped amplifiers. For simulations, the ideal

back side mirror (pump reflectivity = 100% and signal reflectivity =

0% ) is considered. The double pass pumping scheme results in an

1650 1700 1750 1800 1850 1900

−6

−4

−2

Wavelength (nm)

Net absorption / gain (dB)

λp = 1650 nm (TE)

30 mm double pass

15 mm double pass

15 mm single pass

Figure 3.15: Calculated net absorption/gain spectra of a 15 mm long and

30 mm long Ti:Tm:LiNbO3waveguide amplifier for single and

double pass pumping at 1650 nm (signal and pump TE polar-

ized) for 13.5mW coupled pump power.

improved gain spectrum which is shown in black (in comparison to

the single pass pumped case shown in red) in the figure. An increase

in the waveguide length by a factor of 2together with double pass

pumping improves the situation even further as shown by the green

curve in Fig. 3.15.

Additional gain spectra for a 30 mm long waveguide with double

pass pumping scheme were calculated and the results are shown in

Fig.3.16. With just few milliwatts of coupled pump power (specifically

with 4mW), it is possible to achieve a net gain in the long wave-

length side. This shows the potential for a Fabry-Pérot type laser

fabricated with such a waveguide amplifier to operate in the long

wavelength side of the emission spectrum of Ti:Tm:LiNbO3. Such a

1600 1700 1800 1900

−20

−15

−10

−5

Wavelength (nm)

Net Absorption / Gain (dB)

L = 30 mm

λp = 1650 nm (TE)

unpumped

Ppc = 7 mW

Ppc = 13.5 mW

Ppc = 50 mW

Figure 3.16: Calculated net absorption/gain spectra for a double pass

pumped 30 mm long Tm doped waveguide. Both pump and

signal are TE polarised.

laser is demonstrated in chapter 4.

3.3.4Further Comments

The modeling results presented above did not take ASE into account.

Although, amplification of spontaneous emission is a stimulated pro-

cess, the energy of the excited atoms is not used in the amplification

of the input signal (or laser signal in the case of a laser), but is in a way

’lost’ and consequently it is a parasitic process. As a result when ASE

is also included in the calculation, the corresponding net gain values

will be lower than the case when ASE is not taken into account. ASE

becomes significant in the case where the gain of the amplifier is large

(<20 dB) [39] because then it can deplete the inversion thus reducing

the gain of the amplifier.

3.4 experimental investigations

3.4.1Experimental Setup

The schematic of the experimental setup which was developed to in-

vestigate the Ti:Tm:LiNbO3waveguide amplifier is shown in Fig. 3.17.

A fibre coupled Fabry-Pérot type diode laser with a peak wavelength

of 1650 nm was used as pump laser. Its output fibre was spliced to a fi-

bre optical polarization controller and a fibre optical isolator in series.

To measure optical absorption and gain spectra, unpolarized chopped

Pump laser

1650 nm Polarization

controller Isolator

3 dB coupler

Halogen

lamp

SM fiber Ti:Tm:LN

waveguide Polarizer

Pump blocking filter

Monochromator ex-InGaAs

Chopper

Lock-in amplifier

Figure 3.17: Schematic of the experimental setup which was used to investi-

gate the Ti:Tm:LiNbO3waveguide amplifier.

light (signal; infront of 3dB coupler power when measured with 2nm

resolution bandwidth at 1600 nm, the power density was ∼400 pW)

from a halogen lamp was superimposed to the pump radiation by

a fibre optical 3dB coupler and then butt coupled into the 15 mm

long Tm-doped waveguide. The output was imaged to the input of

a monochromator (resolution bandwidth set to 2nm; independently

measured with 1600 nm emission from an extended cavity laser). A

polarizer (specified extinction ratio of 60 dB) in the free space beam

path determined the polarization of the signal to be measured; a spe-

cial wavelength filter blocked to a large extent the transmitted pump

radiation (attenuation ∼18 dB). A thermo-electrically cooled ("wave-

length extended") InGaAs photodiode was used to measure with a

lock-in technique the (spectrally resolved) transmitted signal. The

measurements were controlled by a computer which also recorded

the data. To determine a gain spectrum at a given pump power level,

two measurements were performed, one with pumping (TE polar-

ized in all cases) and the second without pumping. By dividing the

measured data from the pumped case with those obtained without

pumping, the corresponding gain spectrum was obtained. The gain

spectrum thus determined do not include scattering losses.

3.4.2Experimental Results

Fig. 3.18 shows (as open circles) the measured gain spectra for three

different pump powers. At shorter wavelengths close to the pump

wavelength, gain measurements could not be made as the noise lev-

els were high enough to saturate the detection circuit. At those wave-

lengths, the pump blocking filter was not sufficient to attenuate the

transmitted pump power completely.

1700 1750 1800 1850 1900

−9

−6

−3

Wavelength (nm)

Absorption/Gain (dB)

Pinc= 27 mW

Pinc= 19.4 mW

Pinc= 9.4 mW

unpumped

L = 15 mm

Figure 3.18: Measured gain spectra of a 15 mm long Tm:Ti:LiNbO3waveg-

uide amplifier, pumped at 1650 nm (signal and pump TE polar-

ized). Parameter is the incident pump power Pinc (λp=1650

nm).

3.5 discussion

3.5.1Comparison of Experimental and Modeling Results

As the next step, the agreement between measured and calculated

gain spectra was checked. Along with the cross section data (ab-

sorption cross sections and McCumber emission cross sections), cal-

culated TE mode intensity profiles, upper state life time of 2.4ms

[29] and waveguide propagation loss of 0.1dB/cm for pump and sig-

nal beams were given as the inputs for performing calculations. The

signal input power used for the calculations is 1µW (small signal

amplification). Waveguide mode intensity profiles were calculated by

Focus, a software developed by Strake [25].

The outcome of modelling is shown in Fig. 3.19 (as lines). As this set

of measurements were done one after the other the pump coupling

efficiency was assumed to be the same in all cases. During the experi-

ments, the transmitted signal power level behind the waveguide was

checked before (and after) all measurements and in this way the con-

sistency of input coupling and over all stability of the experimental

setup was confirmed. The modeling was done by varying the cou-

pled pump power in order to find out the best choice which fits with

the measurement result (judgement by eye). From Fig. 3.19, we see

that there is a reasonably good agreement between measured and cal-

culated data sets for a pump coupling efficiency of 50%. In general,

the agreement of modeling and experimental results is very good,

demonstrating the reliability of the theoretical model used; this good

agreement is also a consequence of the experimentally determined in-

put parameters for the calculations. Remarkable is, that optical gain

surpasses the zero dB level in the whole wavelength range 1750 nm

<λ<1900 nm with a coupled pump power of 13.5mW only, indicat-

1700 1750 1800 1850 1900

−9

−6

−3

Wavelength (nm)

Absorption/Gain (dB)

Ppc= 13.5 mW

Ppc= 9.7 mW

Ppc= 4.7 mW

unpumped

L = 15 mm

Figure 3.19: Calculated net absorption/gain spectra (lines) of a 15 mm long

Tm:Ti:LiNbO3waveguide amplifier, pumped at 1650 nm (signal

and pump TE polarized), compared with experimental results

(open circles). Parameter is the coupled pump power Ppc (λp=

1650 nm).

ing the potential for tunable waveguide lasers in that range. The main

gain peaks, already known from the fluorescence data (Fig. 2.9), arise

at somewhat higher pump power levels. However, at the long wave-

length side (λ > ∼1880 nm), the agreement of theory and experiment

becomes poor. This could be attributed to the limited applicability of

the McCumber theory in this wavelength range.

Results of small signal gain measurements presented above, demon-

strate that broadband amplification in the spectral range 1750 nm <

λ < 1900 nm using a 15 mm long Ti:Tm:LiNbO3waveguide amplifier

is possible. In the remaining part of this section, additional model-

ing results, together with experimental results of the measurements

performed with the 15 mm long waveguide amplifier are presented.

The four signal wavelengths chosen were 1758 nm, 1800 nm , 1858

nm (three emission peaks; see fluorescence spectrum ) and 1890 nm

(laser wavelength; see section 4.5.1).

Fig. 3.20 shows the calculated net gain curve and measured net gain

data (i.e. gain including scattering losses of the waveguide) curves at

1758 nm signal wavelength for a waveguide length of 15 mm and

single pass pumping. From the measured gain data 0.3dB was sub-

tracted as scattering loss to obtain the net gain values. At this sig-

nal wavelength absorption is bleached with 9mW of coupled pump

power. This is because of the three level nature (and the consequent

reabsorption of signal wavelength) of the gain medium at this wave-

length range of the fluorescence spectrum of Ti:Tm:LiNbO3. A higher

coupled pump power (in comparison with longer signal wavelengths,

as will be shown later) is needed to achieve inversion. From Fig. 3.20,

it is clear that there is a reasonable agreement between calculated

and measured values. With 30 mW coupled pump power, more than

0 5 10 15 20 25 30

−10

−8

−6

−4

−2

Coupled pump power (mW)

Net absorption / gain (dB)

L=15mm

λs= 1758 nm (TE)

Measured

Calculated

Figure 3.20: Measured (red) and calculated single pass net gain at 1758 nm

(TE) for a waveguide length of 15 mm as function of coupled

pump power (TE).

3dB gain at λs=1758 nm with single pass pumping scheme could be

achieved.

The calculated net gain curve and measured net gain data at 1800

nm signal wavelength are shown in Fig. 3.21. At this signal wave-

0 5 10 15 20 25 30

−4

−2

Coupled pump power (mW)

Net absorption / gain (dB)

L=15mm

λs= 1800 nm (TE)

Measured

Calculated

Figure 3.21: Measured (red) and calculated single pass net gain at 1800 nm

(TE) for a waveguide length of 15 mm as a function of coupled

pump power (TE).

length absorption is bleached with only 5mW of coupled pump

power. The three level nature of the gain medium is less prominent

at 1800 nm and consequently signal re-absorption is much lower at

this wavelength in comparison to that at 1758 nm. From Fig. 3.21, we

see that there is a good agreement between calculated and measured

values. In this case with 30 mW coupled pump power, net gain more

than 2dB at λs=1800 nm could be achieved.

Fig. 3.22 shows the calculated net gain curve and measured net gain

data at 1850 nm signal wavelength. The three level nature of the gain

medium is even less prominent (absorption in the unpumped case

is only 0.7dB) than at 1800 nm at this wavelength and consequently

signal re-absorption is much lower. The absorption is bleached with

0 5 10 15 20 25 30

−1

Coupled pump power (mW)

Net absorption / gain (dB)

L=15mm

λs= 1850 nm (TE)

Measured

Calculated

Figure 3.22: Measured (red) and calculated single pass net gain at 1850 nm

(TE) for a waveguide length of 15 mm as a function of coupled

pump power (TE).

3mW of coupled pump power in this case. In this case also, there is a

reasonable agreement between calculated and measured values. With

30 mW coupled pump power, more than 1.5dB gain at λs=1850 nm

could be achieved.

The calculated net gain curve and measured net gain data at 1890

nm signal wavelength are shown in Fig. 3.23. In this case the gain

0 5 10 15 20 25 30

−0.5

0.5

Coupled pump power (mW)

Net absorption / gain (dB)

L=15mm

λs= 1890 nm (TE)

Measured

Calculated

Figure 3.23: Measured (red) and calculated single pass net gain at 1890 nm

(TE) for a waveguide length of 15 mm as a function of coupled

pump power (TE).

medium is close to an ideal four level system (absorption in the un-

pumped case is only 0.27 dB) than at 1850 nm and consequently sig-

nal re-absorption is practically zero. According to the modeling result,

the absorption is bleached with 4mW of coupled pump power. But

from the experimental result bleaching of absorption occurs with just

1.8mW of coupled pump. From Fig. 3.23, it is clear that the agree-

ment between calculated and measured values becomes poor in com-

parison to the corresponding outcomes at shorter wavelengths. This

surprising result is believed to be mainly due to the inapplicability of

McCumber theory in the estimation of (emission) cross sections (see

section 2.3.6) at these long wavelengths and/or the inacuracies in the

estimation of waveguide propagation losses (see section 4.4.1).

3.5.2Amplifiers for Laser Applications

The amplifiers discussed so far operated in the single pass pump

regime. It was also shown that the usage of the available pump power

can be improved by adopting double pass pumping scheme, where

the residual single pass pump is reflected back into the amplifier by

a pump reflector at the unpumped end of the amplifier. By adopting

the same strategy for signal as well (i.e., signal beam also is reflected

back) the achievable net gain from a given amplifier length and pump

power can be doubled, provided the amplifier is operated in the small

signal gain regime. By having a high reflector mirror for signal at both

endfaces of an amplifier, a laser can be realized. In that case the signal

is amplified as it traverses the length of the amplifier multiple times.

As the net gain crosses the zero level, lasing action sets in there after

as the pump power is increased the energy from the pump beam is

transfered to the laser beam. In practice there is no signal input (as in

the case of an amplifier) into a laser cavity. As the pump power is in-

creased, laser action starts from amplified stimulated emission. With

the Ti:Tm:LiNbO3amplifiers mentioned in this chapter, lasers oper-

ating at any wavelengths in the gain bandwidth can be realized. In

chapter 4, the realization of a laser with a 30 mm long Ti:Tm:LiNbO3

waveguide amplifier is discussed.

3.5.3Conclusions

The results which are presented in this chapter, demonstrate that

broadband optical gain in the wavelength range 1750 nm < λs<

1900 nm can be achieved using a Ti:Tm:LiNbO3waveguide as an am-

plifier. Results of small signal gain measurements are found to be in

good agreement with the corresponding modeling results for the case

of single pass pumping scheme. Modeling results show the possibil-

ity to achieve (wavelength dependent) gain of up to 30 dB from a 10

cm long single pass pumped waveguide. The gain achieveable with

a given pump power level can be improved by implementing double

pass pumping along with a careful choice of waveguide length. Such

double pass pumped amplifiers can be used to realize waveguide

lasers operating at any of the wavelengths in the broadband signal

spectrum where optical gain can be achieved.

Ti:Tm:LiNbO3WAVEGUIDE LASER

4.1 introduction

Modeling results presented in chapter 3show that it is possible to

achieve a positive net gain near 1890 nm from a double pass pumped

30 mm long Ti:Tm:LiNbO3waveguide. The realization and properties

of a waveguide laser operating at 1890 nm and 1850 nm with such an

amplifier are discussed in this chapter. The highlight of this chapter is

the first demonstration of an in-band pumped Ti:Tm:LiNbO3waveg-

uide laser (i) emitting at the longest emission wavelength, (ii) with

the smallest laser threshold and the highest output power reported

from a Tm:LiNbO3waveguide laser so far.

4.2 laser cavity

The laser cavity was formed by two dielectric multilayer mirrors com-

prising of alternating layers of TiO2and SiO2, vacuum deposited on

the end faces of the 30 mm long Ti:Tm:LiNbO3waveguide, enabling

double pass pumping. The pump coupler mirror has a high reflectiv-

ity (HR >90 %) at wavelengths >1800 nm, but a high transmission

(T >90 %) at the 1650 nm (pump wavelength). The output coupler

mirror has a broadband characteristic of high reflectivity (R>95%) in

the spectral range 1640 nm <λ<1900 nm . The measured mirror

reflectivity curves of both mirrors are given in Fig. 4.1. The calculated

finesse of the cavity thus formed is 32 (scattering losses are assumed

to be 0.1dB/cm, λs=1890 nm).

1600 1700 1800 1900

100

Wavelength (nm)

Reflectivity(%)

Output coupler

Pump coupler

Figure 4.1: Reflectivity of the input mirror (pump coupler;red) and of the

output mirror (output coupler;blue) in the wavelength range

1600 nm <λ<1900 nm.

4.3 experimental setup

The experimental set up to investigate the laser is shown in Fig. 4.2.

A fiber coupled diode laser with a centre wavelength of 1650 nm

was used as pump laser. At the output of the laser a fibre optic

isolator (isolation ⩾35 dB) and a fibre optic polarization controller

were spliced one behind the other to minimize losses. The output fi-

bre of the isolator was kept butt coupled to the end face of the Tm

doped waveguide with the pump coupler mirror. Light behind the

Tm doped waveguide was collimated by a lens and was coupled into

the input port of a 10/90 coupler. The 90% output was sent to a fiber

coupled monochromator and the 10% output was used to enable con-

trolled pumping. A thermo-electrically cooled wavelength extended

InGaAs photodiode kept behind the monochromator was used to de-

tect the emission of the waveguide laser. Alternatively the 90% beam

was send to a photodiode kept connected with an electronic spec-

trum analyser (or a fast oscilloscope), for performing frequency (or

time) domain measurements. The free space beam allowed the mea-

surement of pump and signal polarizations. A pump blocking filter

(∼18 dB) kept in the free space beam suppressed any residual pump

emission.

Pump laser

1650 nm Polarization

controller Isolator

Pump

coupler

Output

coupler

Ti:Tm:LN

waveguide

Polarizer

Pump blocking filter

10/90

Feedback

control

circuitry

90%

Monochromator

ex-InGaAs

Figure 4.2: Experimental setup used to investigate the 30 mm long

Ti:Tm:LiNbO3waveguide laser.

4.4 power characteristics

4.4.1High-Q Operation at λs=1890 nm

Laser emission was found near 1890 nm in TE-polarization only ir-

respective of the pump polarization. Fig. 4.3presents the measured

power characteristics for TE and TM polarized pump in cw operation

(?). The laser output consisted of undamped relaxation oscillations

was smoothed by a filter with a bandwidth of 22 Hz, for this mea-

surement. The experimental setup described above was modified in

such a way that the total laser output power (i.e., without spectral

dispersion behind the waveguide) was measured. Lasing sets in at 6

mW of incident pump power corresponding to about 4mW of cou-

pled pump power1only. With increasing pump power the spectrum

of the pump laser shifts to somewhat longer emission wavelengths

resulting in an even better pump absorption (see also Fig. 2.7). The

result is a slightly increasing slope efficiency up to 13.3% at 38 mW

coupled pump power. The output power of up to 4.5mW (limited by

the available pump power) was stable over time (Fig. 4.4.3).Both, the

output power of the laser and the slope efficiency surpass the results

reported so far for Tm:LiNbO3waveguide lasers [16,17] by an order

of magnitude.

0 10 20 30 40

Coupled pump power (mW)

Output power (mW)

TE pump (measured)

TM pump (measured)

TE pump (calculated)

Figure 4.3: Measured power characteristics of the Ti:Tm:LiNbO3waveguide

laser as output power (TE-polarization) versus coupled pump

power in TE- and TM-polarization, respectively. In addition, a

calculated characteristics is presented as output power versus

coupled pump power for the laser emission wavelength of 1890

nm.

0 50 100 150 200

Time (ms)

Output power (mW)

Ppc = 35.7 mW

Figure 4.4: The laser output power, measured with ∼1kHz bandwidth, ver-

sus time.

A low threshold for the long wavelength emission can also be ex-

pected from qualitative arguments: due to the quasi four level system

exploited, a low pump power is sufficient to get optical gain as al-

ready discussed in section 3.2.1(see also section 3.3.3). Therefore, in

a high-Q cavity lasing will start at that wavelength where the gain

1 66% coupling efficiency is assumed.

surpasses the overall losses first, when the coupled pump power is

increased.

The power characteristics have also been modeled. As input param-

eters all the measured and derived data were used, which have been

discussed above (waveguide propagation losses, mode distributions,

absorption and emission cross sections, lifetimes of energy levels). In

the case of laser we have endface mirrors and consequently R P Fiber

Power can implement the relevant boundary conditions and calculate

the optical gain arising from multiple round trips by the optical beam

through the cavity, so that laser action can be simulated. Surprisingly,

the modeling results show a somewhat larger threshold and a smaller

slope efficiency than the experiments. We attribute this discrepancy

to the calculated emission cross sections, which seem to be too small

at the long wavelength side. The green curve shown in Fig. 4.3is

calculated with waveguide propagation loss coefficient of 0.1dB/cm.

The agreement becomes better (without a change in threshold pump

power) when the calculation is done with waveguide propagation loss

coefficient kept at 0.07 dB/cm.

4.4.2Low-Q Operation at λs=1850 nm

The laser wavelength could be shifted to 1850 nm by inducing addi-

tional losses into the cavity (described in detail in section 4.5.3). The

corresponding power characteritics are given in Fig. 4.5. By the in-

creased cavity losses the (coupled) threshold power was increased to

∼9mW and the slope efficiency of the laser characteristic was re-

duced to ∼2.6%.

0 10 20 30 40

0.2

0.4

0.6

0.8

Coupled pump power (mW)

Output power (mW)

TE pump (measured)

Figure 4.5: Measured power characteristics of the Ti:Tm:LiNbO3waveguide

laser (λs=1850 nm) as output power (TE-polarization) versus

coupled pump power in TE polarization.

4.4.3Relaxation Oscillations

As mentioned above, the laser output at both pump polarizations

consisted of regular undamped relaxation spikings. Typical laser out-

put at 50 mW incident pump power is shown in Fig. 4.6(measure-

ment bandwidth ∼10 MHz). The left subfigure shows a regular pulse

0 5 10 15 20

0.25

0.5

0.75

Time (µs)

Output power (a. u.)

0 1 2 3 4 5 6 7

−100

−80

−60

−40

−20

Frequency (MHz)

Electrical power (dBm)

rbw 1 kHz

Figure 4.6: Output power of Ti:Tm:LiNbO3waveguide laser (λs=1890 nm)

for an incident power of 50 mW in time (left) and frequency

(right) domains (measurement bandwidth >10 MHz).

train with a repetiton frequency of 227 kHz and ∼6% dutycycle.

The right side figure shows the corresponding measurement in fre-

quency domain, where several harmonics of the fundamental fre-

quency are clearly visible. It has already been demonstrated that by

opto-electronic feedback controlled pumping, such relaxation oscilla-

tions could be suppressed [40,41]. A special example is the feed back

controlled pumping of actively mode locked Ti:Er:LiNbO3waveg-

uide laser[42]. In order to suppress the relaxation spikings of the

Ti:Tm:LiNbO3laser, the same controlled pumping scheme was used.

With this scheme, the peak corresponding to the fundamental relax-

ation oscillation frequency could be suppressed by more than 40 dB.

This result is shown in Fig.4.7(left). In this case, the incident pump

power was only 14 mW. Fig. 4.7(right) shows the integrated laser out-

put power (without wavelength dispersion) recorded as a function of

time. As evident from the figure, the laser out power was found to

be stable over time, as it was shown before (see Fig. ) with additional

smoothing. On the other hand when the emission was wavelength

dispersed with the monochromator, the relaxation spikings were ob-

served to be irregular (see Fig. 4.8). This observation would mean that

the gain experienced by different laser cavity modes are changing as a

function of time eventhough the total laser output power remains sta-

ble. This observation fits well with the constant fluctuations recorded

in the optical spectra (see section 4.5).

0 0.5 1 1.5

−110

−90

−70

−50

−30

Frequency (MHz)

Electrical power (dBm)

with controlled pumping

without controlled pumping

rbw 1 kHz

Figure 4.7: Laser output power in the frequency domain for 14 mW inci-

dent pump power with and without feedback controlled pump-

ing. The frequency spectrum was measured with a resolution

bandwidth (rbw) of 1kHz.

0 10 20 30 40 50

0.02

0.04

0.06

0.08

0.1

Time (µs)

Output power (a.u.)

Integrated power

Behind monochromator

(rbw 2 nm, 1904 nm)

Figure 4.8: Output power of Ti:Tm:LiNbO3waveguide laser (λs=1890 nm)

in time domain: comparison between integrated output and spec-

trally resolved output centered at 1904 nm (resolution band-

width 2nm).

4.5 spectral properties

4.5.1Emission at 1890 nm

Laser emission spectra were measured with a monochromator of 250

pm resolution (measured independendly with a line from extended

cavity laser). The output from the photodiode (kept behind the mono-

chromator) was recorded with a storage oscilloscope (measurement

bandwidth ∼53 kHz) while the monochromator was scanned in steps

of 140 pm. The acquisition time at each data point was 2s. In or-

der to minimize the influence of water vapour absorption lines, the

monochromator was purged with nitrogen. Irrespective of the pump

polarization, the laser emission was found to be near 1890 nm and TE

polarized. The fine structure of the laser spectra changed from scan

to scan and was also found to be pump power dependent. Fig. 4.9

(left) shows as example the optical spectrum for TE polarized pump

at 32 mW incident pump power. The envelopes of the optical spectra

were found to be (i) few nm wide, (ii) a function of pump power and

(iii) changing from scan to scan. At higher pump power levels, the

envelope of the spectrum became wider. An example is shown in Fig.

4.9(right).

1885 1890 1895 1900

0.2

0.4

0.6

0.8

Wavelength (nm)

Optical power (a.u.)

TE pump

TM pump

rbw 250 pm

1885 1890 1895 1900

0.2

0.4

0.6

0.8

Wavelength (nm)

Optical power (a.u.)

TE pump

TM pump

rbw 250 pm

Figure 4.9: Measured optical spectra of Ti:Tm:LiNbO3waveguide laser (TE

polarization; taken with a resolution bandwidth (rbw) of 250

pm) with emission near 1890 nm (high-Q cavity) for an incident

power of 32 mW (left) and 51 mW (right).

Lasing at the longer wavelength side of the Tm emission spectrum

can be understood by considering a four level system, i.e., pumping

of atoms at the lower levels of the ground state multiplet to the higher

lying levels of the excited state mutiplet and emission from the lower

levels of the excited state multiplet to the high lying levels of the

ground state multiplet. Due to thermalisation, the higher levels of

both multiplets are rapidly depopulated to their low levels. As there

are no wavelength selecting components inside the cavity and as the

propagation losses of the waveguide are small, the gain achieved by

longer wavelengths in the Tm emission spectrum, although low, is al-

ready sufficient to overcome the losses of the cavity and lasing action

sets in.

4.5.2Radio Frequency Spectrum of Laser Output

The radio frequency spectrum of the laser (λs=1890 nm) was recorded

by a fast detector. The measurement result is shown in Fig. 4.10. Beat

frequencies due to mode beating of longitudinal modes up to 8.8

GHz (free spectral range = 2.2GHz) is clearly visible. Due to the

broad optical spectrum, we can expect a large number of such beat

frequency lines in the spectrum.

4.5.3Emission at 1850 nm

By the integration of a tuneable wavelength filter into the cavity, a

tuneable laser could be developed exploiting the large bandwidth of

2 4 6 8 10

−75

−70

−65

Frequency (GHz)

Electrical power (dBm)

2.1 GHz 4.3 GHz

6.6 GHz

8.8 GHz

rbw 3 MHz

Figure 4.10: Measured electronic power spectrum of Ti:Tm:LiNbO3laser.

the gain (see Fig. 4.12). This approach has already been realized for

Er-doped waveguide lasers using an acousto-optical filter [43]. For

our Tm-doped laser, however, it would require the fabrication of an-

other, more complicated device, which is an attractive challenge for

the future. On the other hand, "tuneability" of the present device was

demonstrated in quite a different way. By inducing additional losses

in the cavity, the laser emission wavelength jumped to the neighbour-

ing gain peak at ∼1850 nm wavelength. This has been done by de-

positing a layer of silver paste on the waveguide surface. In this way

additional propagation losses are induced by (partial) absorption of

the evanescent tail of the laser mode in the metallic silver layer. Conse-

quently, a higher gain is required to compensate the round trip prop-

agation losses. However, by increasing the pump power, the required

threshold gain cannot be achieved at ∼1890 nm as in the high-Q

cavity, but only at ∼1850 nm, where laser emission sets in. A corre-

sponding emission spectrum is shown in Fig. 4.12. By the increased

cavity losses the (coupled) threshold power was increased to ∼9mW

and the slope efficiency of the laser characteristic was reduced to ∼

2.6%.

Emission at the long wavelength side of the Tm 3F4→3H6transi-

tions qualitatively explains the low threshold of the laser in case of

a high Q cavity. However, it does not explain the fine structure of

the observed emission spectrum, which should ideally consist of a

single line (single longitudinal mode emission), if a homogeneously

broadened laser material is used. Therefore, our conclusion is that

a significant inhomogeneous broadening has been induced by the

laser fabrication processes yielding concentration profiles of both, Tm

and Ti ions. Moreover, spatial hole burning and small photorefrac-

tive effects might contribute to mode coupling and in this way to a

time-varying broadening of the emission spectra. Nevertheless, we

are confident that single mode emission can be achieved by appro-

Figure 4.11: Experimental setup used to study the Ti:Tm:LiNbO3laser

(λs=1850 nm). The waveguide laser ’covered’ with silver paste

is shown in inset also. The different components in the setup

are 1: pump input fiber, 2: waveguide laser, 3: collimating lens

behind the output coupler mirror of the laser, 4: pump blocking

filter, 8: input side of the monochromator, 7: cooled ex-InGaAs

photodiode at the output side of the monochromator and 9: cur-

rent source for thermo-electric cooler of ex-InGaAs photodiode.

In addition, the white light source (5) chopper (6) and lock-in

amplifier (10) used in gain measurements are also visible in the

photograph (see section 3.4).

1846 1848 1850 1852

0.25

0.5

0.75

Wavelength (nm)

Optical power (a. u.)

rbw 250 pm

Figure 4.12: Measured optical spectrum (TE polarization; taken with a reso-

lution bandwidth (rbw) of 250 pm) of the Ti:Tm:LiNbO3waveg-

uide laser with emission near 1850 nm (low-Q cavity).

priate narrow-band spectral filtering as demonstrated for Er doped

waveguide lasers [9].

Southampton [16] Madrid [17] Paderborn

Gain Medium Ti:Tm:LiNbO3Zn:Tm:LiNbO3Ti:Tm:LiNbO3

λemission 1800 nm, 1850 nm 1762 nm 1890 nm, 1850 nm

λpump 795 nm 795 nm 1650 nm

Threshold 67 mW (launched) 34 mW (launched) 6mW (incident)

Slope Efficiency 0.8%1%10%

Table 4.1: Comparison of different laser parameters of integrated

Tm:LiNbO3lasers reported so far with those of the lasers

reported in this thesis.

Table 4.1gives a comparison of different laser parameters of in-

tegrated Tm:LiNbO3lasers reported so far with those of the lasers

reported in this thesis.

4.6 optimization

In order to find the best combination of mirrors and waveguide length

yielding maximum output power (and slope efficiency) for our dou-

ble pass pumping scheme and emission at 1890 nm wavelength, addi-

tional simulations were performed. The results are presented in Fig.

4.13 as calculated output power (λ=1890 nm) of the 3cm long waveg-

uide laser versus coupled pump power and in Fig. 4.14 versus device

length (Ppc =50 mW) with signal reflectivity Rso of the rear mirror

(output coupler) as parameter. It is evident that the highest reflectivity

results in the lowest threshold, but also in the lowest slope efficiency,

ηslope. Reducing Rso increases the threshold, but increases the slope

efficiency as well. With 50 mW coupled pump power the maximum

laser output power of 14.5mW is achieved with a mirror reflectivity

of Rso =80% (assuming that the resulting higher cavity loss will not

yet induce a jump of the emission to a shorter wavelength; otherwise,

a wavelength filter has to be incorporated). A further decrease of Rso

reduces the output power again; only somewhat longer structures are

a bit more efficient (not shown in Fig. 4.14 for reasons of clarity). The

results presented in Fig. 4.14 confirm, that the chosen length of our

waveguide laser of 3cm is very close to the optimum choice. On the

other hand, the mirror reflectivity of Rso =98% of the fabricated de-

vice yields a low threshold, but not the maximum output power nor

slope efficiency (green characteristic in Fig. 4.13). A significant opti-

mization could be achieved with Rso =80% instead of Rso =98%: the

slope efficiency would grow by a factor of four and nearly the same

can be expected by the maximum output power (black characteristic

in Fig. 4.13). And a reduction of the waveguide scattering losses (0.1

dB/cm have been assumed) and in this way of the cavity round trip

losses would contribute to a further improvement of the laser.

0 10 20 30 40 50

Coupled pump power (mW)

Output power(mW)

ηslope= 26%

ηslope= 34%

ηslope= 8%

80%

90%

98%

Rso =

Figure 4.13: Calculated output power (λ=1890 nm) of the Ti:Tm:LiNbO3

waveguide laser (l = 3cm) versus coupled pump power. The

signal (pump) reflectivity of the input mirror is Rsi =90% (Rpi

=6%); the pump reflectivity of the output coupler is Rpo =96%.

1 3 5 7 9

Length (cm)

Output power(mW)

80%

90%

98%

Rso=

current laser

Figure 4.14: Calculated output power (λ=1890 nm) of the Ti:Tm:LiNbO3

waveguide laser (l = 3cm) versus device length (Ppc =50 mW)

with signal reflectivity Rso of the rear mirror (output coupler)

as parameter. The signal (pump) reflectivity of the input mirror

is Rsi =90% (Rpi =6%); the pump reflectivity of the output

coupler is Rpo =96%.

An optimization for another emission wavelength and/or pump

power level would result in a modified mirror reflectivity and device

length. Moreover, it would require a wavelength selective component

inside the cavity.

4.7 conclusions

The 30 mm long free running laser (without wavelength selective com-

ponents in the cavity) emits at 1890 nm, the longest emission wave-

length of a Tm:LiNbO3laser reported so far; also 1850 nm emission

could be demonstrated by increasing cavity losses. Laser threshold

(1890 nm) of the double pass pumped device is ∼4mW coupled pump

power only; the slope efficiency is ∼11% resulting in a stable output

power of ∼4mW, if pumped with ∼40 mW (coupled power). Exten-

sive modeling also demonstrates the potential of the laser: both, slope

efficiency and maximum output power should grow by about a factor

of four by adjusting the reflectivity of the output mirror to ∼80%.

CONCLUSIONS AND OUTLOOK

5.1 conclusions

The fabrication of Ti:Tm:Li:NbO3waveguide by diffusion doping is

briefly described. The characterization of the waveguide by means

of scattering loss and near field measurements are subsequently dis-

cussed. The experimental studies of the properties of the waveguide

due to Tm doping (absorption and fluorescence spectra) are also

presented. The transition cross sections of the Tm-doped waveguide

which are important from the point of view of amplifier/laser fabri-

cation using the waveguide were determined for the first time using

the McCumber theory.

In-band pumping using 1660 nm laser radiation was found to be

the optimum for pumping the experimentally studied waveguide am-

plifier and laser discussed in this thesis. Modeling results demon-

strate that broad-band optical gain in the wavelength range 1750 nm

< λs<1900 nm can be achieved using the Ti:Tm:LiNbO3waveg-

uide as an amplifier, when pumped at 1650 nm. Results of small sig-

nal gain measurements are found to be in good agreement with the

corresponding modeling results for the case of single pass pumping

scheme. Modeling results point to the possibility to achieve (wave-

length dependent) gain of upto 30 dB from a 10 cm long single pass

pumped waveguide. The gain achieveable with a given pump power

level can be improved by implementing double pass pumping along

with a careful choice of waveguide length. Such double pass pumped

amplifiers can be used to realize waveguide lasers operating at any

wavelength in the broadband signal spectrum where optical gain can

be achieved.

The 30 mm long free running laser (without wavelength selective

components in the cavity) realized using a Ti:Tm:Li:NbO3waveguide

with dielectric mirrors deposited at the waveguide endfaces emits

at 1890 nm, the longest emission wavelength of a Tm:LiNbO3laser

reported so far; also 1850 nm emission could be demonstrated by in-

creasing cavity losses. Laser threshold (1890 nm) of the double pass

pumped device is ∼4mW coupled pump power only; the slope ef-

ficiency is ∼11% resulting in a stable output power of ∼4mW, if

pumped with ∼40 mW (coupled power). Extensive modeling also

demonstrates the potential of the laser: both, slope efficiency and

maximum output power should grow by about a factor of four by

adjusting the reflectivity of the output mirror to ∼80%.

5.2 outlook

It should be possible to realize a whole family of lasers using the

Ti:Tm:LiNbO3waveguide as the family of Ti:Er:LiNbO3waveguide

lasers demonstrated in the past[9]. An exciting possibility which has

not yet been realized in the case of Ti:Er:LiNbO3lasers is the demon-

stration of a passively mode-locked laser. This can be done by coating

a layer of carbon nanotube (CNT) on top of the waveguide, so that the

CNT layer interacts with the evanescent tail of the laser field as a sat-

urable absorber, leading to mode-locking. The 1850 nm laser demon-

strated in this thesis shows the potential of this approach. This has

already been realized in the case of fiber lasers. Another approach is

to use a suitably designed semiconductor saturable absorber mirror

butted at one of the endfaces to form the laser cavity.

THULIUM DOPED WAVEGUIDES FOR QUANTUM

MEMORY APPLICATIONS

a.1 waveguide fabrication

Commercially available 0.5mm thick Z-cut wafers of undoped op-

tical grade congruent lithium niobate (CLN) were used as starting

material. Samples of 12 mm x 30 mm size were cut from these wafers

and doped by thulium near the +Z-surface before waveguide fabrica-

tion. The doping was achieved by in-diffusing a vacuum deposited

(electron-beam evaporated) Tm layer of 19.6nm thickness. The diffu-

sion was performed at 1130◦C during 150 h in an argon-atmosphere

followed by a post treatment in oxygen (1h) to get a full re-oxidization

of the crystal.

To determine the diffusion coefficient of Tm into Z-cut CLN, sec-

ondary neutral mass spectroscopy (SNMS) was performed using 700

eV Argon-ions for ion milling. Ions and electrons were extracted from

the plasma source with a duty cycle of 4:1at a rate of 320 kHz to avoid

charging of the insulating CLN-substrate. SNMS was chosen instead

of secondary ion mass spectroscopy (SIMS) to significantly reduce

matrix effects (see e.g. [27]). In Fig. A.1, the concentration profiles

versus depth have been recorded for thulium (Tm), lithium (Li), nio-

bium (Nb) and oxygen (O). Interestingly, the Li-concentration slightly

increases towards the surface, although it is expected that Tm occu-

pies regular Li-sites similar to Er-ions when incorporated in CLN by

diffusion [28].

In Fig.A.2, the Tm concentration is plotted on a linear scale versus

the depth. The slight dip close to the surface is unexpected and needs

further investigations. Fitting a Gaussian profile to the concentration

curve leads to a 1/e-penetration depth d1/eof about 6.5µm.The max-

0 1 2 3 4 5 6 7 8

-2

-1

Depth (µm)

T m

C onc entrat ions (at % )

Figure A.1: Measured concentrations of Tm, Li, Nb, and O versus depth,

using SNMS with 700 eV Ar-ions.

Figure A.2: Scheme of the waveguide geometry with the measured Tm con-

centration profile on the left and the calculated intensity distri-

bution of the fundamental TM-mode superimposed on the pro-

file of the extraordinary index of refraction induced by the Ti-

doping. The latter data are for 795 nm wavelength. Iso-intensity

lines are plotted for both, the index and the mode profile, cor-

responding to 100%, 87.5%, 75% etc. of the maximum index in-

crease (∆nmax =4.0×10−3) and mode intensity, respectively.

imum Tm concentration of about 1.35·1020 cm−3corresponds to a

concentration 0.74 mole %, which - according to Ref. [28] - is consid-

erably below the solid solubility of Tm in CLN.

On the Tm-diffusion doped surface of the substrate, a 40 nm thick

titanium (Ti) layer was deposited using electron-beam evaporation.

From this layer, 3.0µm wide Ti stripes were photolithographically

defined and subsequently in-diffused at 1060◦C for 5h to form 30

mm long optical strip waveguides. In the wavelength range around

775 nm, the waveguides are single mode for TE- and TM-polarization

(see Fig. A.2).

The total waveguide propagation loss at room temperature, includ-

ing absorption and scattering loss, were measured by the Fabry-Pérot

method [26]. A stabilized, single frequency Ti:Sapphire laser was used

to measure the transmission of the low-finesse waveguide resonator

as function of a small temperature change at a number of fixed wave-

lengths in the range between 750 nm and 807 nm. From the contrast

of the measured Fabry-Perot response, the propagation loss was de-

duced for all wavelengths. The results are presented in Fig. A.3for

TM-polarization. The waveguide propagation loss was determined at

room temperature and 729 nm wavelength, where negligible absorp-

tion by the Tm-ions can be expected. Therefore, the measured loss

coefficients reflect the scattering loss alone; it is 0.2dB/cm for TE-

as well as for TM-polarization. As the scattering loss is only weakly

dependent on the wavelength, it can be regarded as a background for

the Tm-induced absorption loss.

a.2 waveguide characterization

The transmission through a 30 mm long waveguide for TE and TM-

polarized light is shown in Fig. A.3. It has been normalized to the in-

cident spectral power density of the broadband tungsten lamp used

in this experiment. We observe broad absorption, reflecting different

transitions between Stark levels in the 3H6and 3H4multiplets su-

perimposed with inhomogeneous broadening, and the thermal distri-

bution of the population in the electronic ground state. In addition, a

strong polarization dependence of absorption is observed, confirming

previous studies performed on bulk crystals [44,30].

760 780 800 820 840

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Relative Transmission

Wavelength (nm)

750 760 770 780 790 800 810

0.0

0.5

1.0

1.5

2.0

(cm

-1

)

(dB/cm)

Wavelength (nm)

0.0

0.2

0.4

0.6

0.8

1.0

Figure A.3: Relative transmission through the Ti:Tm:LiNbO3waveguide for

TM- and TE-polarization, respectively, as a function of wave-

length (left figure). The resolution bandwidth of the optical spec-

trum analyzer used in this experiment was 2nm due to the

low spectral power density of the thermal radiator.Measured

loss coefficient (α) for TM-polarization as a function of wave-

length (right side). Data points are connected by spline fitting as

a guide for the eye.

a.3 quantum memory

Our collaborators from the University of Calgary (Group of Prof.Tittel)

made spectroscopic measurements of this sample at 4K[24] and sub-

sequently succeeded in demonstrating the first waveguide quantum

memory[13] with this sample.

OPTICAL SPECTRUM OF Ti:Er:LiNbO3WAVEGUIDE

LASER

The optical spectrum of waveguide laser sample Pb149z (with end-

face mirrors on both endfaces) is presented below. The laser cavity

was formed by mirrors similar to those described in [14]. The scat-

tering losses the waveguide were estimated to be αs=0.03 dB/cm

Figure B.1: Optical spectrum of self-pulsing Ti:Er:LiNbO3waveguide laser

for TM polarized pumping at 1480 nm. The laser was lasing si-

multaneously at 1602 nm and 1614 nm.

1Measurement done by Dr.Selim Reza.

ACKNOWLEDGEMENTS

First of all I would like thank Almighty God for all the never ending

gracious blessings showered on my life.

I would like to thank the members of my Ph.D. commission Prof.

Wolf Gero Schmidt (chairperson), Prof. Wolfgang Sohler (first reviewer),

Prof. Donat As (second reviewer) and Dr. Simone Sanna (representa-

tive from scientific colleagues) for offering their service.

Prof. Sohler, my supervisor, gave me the invaluable opportunity

to work in his group. His guidance and constant encouragement al-

lowed me to reach higher levels of excellence which often I did not

envisage when I started with something, an experiment for example.

During the initial years of my stay in Paderborn, Selim Reza and

Hubertus Suche helped me in getting familiar with the lab and the

physics of various experiments in great detail. They helped me in

fine tuning my skills in such a way that I can ’stand on my own’.

Ms. Irmgard Zimmermann was always there to extend all the help

and guidance with all official matters and even with advice on liv-

ing at Paderborn, whenever it was sought. Harald Hermann was

always generous in spending his time whenever an opinion or ad-

vice was sought about my doubts on various topics of my research.

He extended his advice (in programming and) in getting the HR640

monochromator, which was kept idle for about 10 years, up and

running practically from scratch. Prof. As was kind enough to pro-

vide the user manual of HR640 upon my request. Raimund Ricken

helped me in getting all the samples prepared. Viktor Quiring pre-

pared all the special mirrors and special coatings on sample end

faces (and top surfaces) for me. Both Raimund and Viktor were al-

ways willing to extend a helping hand in getting things done by

the mechanical workshop (often saving valuable time). My friends

and former colleagues, Abu Thomas, Miguel Garcia Granda, Ansgar

Hellwig, Hui Hu, Li Gui, Rahman Nouroozi, Sergey Orlov, Daniel

Büchter, Dirk Mantei, Shantanu Pal, Kai-Huong Luo, Suhas Bhandare,

Bhaskar Bhandarapu, Vitali Mirvoda, Ali Al-Bermani, Omar Jan, Ben-

jamin Koch, Kidsanapong Puntsri and Xu Yang were always happy

to lend a helping hand whenever asked for. Prof. Silberhorn and her

group members extended all the support and encouragement they

possibly could to enrich my stay at Paderborn. Prof. Noé and David

Sandel always allowed me to use several of their laboratory equip-

ments for my research. My discussions with Dr. Rüdiger Paschotta

were often illuminating. I would like to extend my sincere gratitude

to all of the above mentioned persons. Perhaps I have missed to in-

clude some of the people who have enriched my stay at Paderborn.

Thanks to all of those as well.

I would like to to thank my parents and my wife for encouraging

me although I was far away from them.

The financial support extended to me by the people of NRW (via a

Ph.D. position), the people of the European Union (via QuRep project)

and Deutsche Forchungsgemeinschaft (via GRK-1464) is gratefully ac-

knowledged.

BIBLIOGRAPHY

[1] T. H. Maiman. Stimulated optical radiation in ruby. Nature,

187:493–515,1960.

[2] Th. Udem, R. Holzwarth, and T. W. Hänsch. Optical frequency

metrology. Nature,416:233–237,2002.

[3] Qian Peng, Asta Juzeniene, Jiyao Chen, Lars O Svaasand,

Trond Warloe, Karl-Erik Giercksky, and Johan Moan. Lasers in

medicine. Reports on Progress in Physics,71(5):056701,2008.

[4] Ginés Lifante. Integrated Photonics. Wiley VCH, 2003.

[5] S. E. Miller. Integrated optcs: An introduction. The Bell System

Technical Journal,48:2049–2069,1969.

[6] C. Gunn. CMOS photonics for high-speed interconnects. IEEE

Micro,26(2):58 –66, march-april 2006.

[7] Western Bolanos, Joan J. Carvajal, Xavier Mateos, Ganapa-

thy Senthil Murugan, Ananth Z. Subramanian, James S. Wilkin-

son, Eugenio Cantelar, Daniel Jaque, Ginés Lifante, Magdalena

Aguiló, and Francisco Díaz. Mirrorless buried waveguide laser

in monoclinic double tungstates fabricated by a novel combi-

nation of ion milling and liquid phase epitaxy. Optics Express,

18:26937–26945,2010.

[8] Wolfgang Sohler, Hui Hu, Raimund Ricken, Viktor Quiring,

Christoph Vannahme, Harald Herrmann, Daniel Büchter, Selim

Reza, Werner Grundkötter, Sergey Orlov, Hubertus Suche, Rah-

man Nouroozi, and Yoohong Min. Integrated optical devices in

lithium niobate. Optics and Photonics News,19:24–31,2008.

[9] W. Sohler, B. Das, D. Dey, S. Reza, H. Suche, and R. Ricken.

Erbium-doped lithium niobate waveguide lasers. IEICE Trans-

actions Electron. E,88-C:990–997,2005.

[10] Martin Hempstead, , James S. Wilkinson, and Laurence Reekie.

Waveguide lasers operating at 1084 nm in neodymium-diffused

lithium niobate. IEEE Photonics Technology Letters,4(8):852–855,

1992.

[11] E. Lallier, J.P. Pocholle, M. Papuchon, M.P. DeMicheli, M.J.

Li, Q. He, D.B. Ostrowsky, C. Grezes-Besset, and E. Pelletier.

Nd:MgO:LiNbO3channel waveguide laser devices. IEEE Jour-

nal of Quantum Electronics,27(3):618–625,1991.

[12] R. Brinkmann, I. Baumann, M. Dinand, W. Sohler, and H. Suche.

Erbium-doped single- and double-pass Ti:LiNbO3waveguide

amplifiers. IEEE Journal of Quantum Electronics,30(10):2356 –2360,

1994.

[13] Erhan Saglamyurek, Neil Sinclair, Jeongwan Jin, Joshua A. Slater,

Daniel Oblak, Félix Bussiéres, Mathew George, Raimund Ricken,

Wolfgang Sohler, and Wolfgang Tittel. Broadband waveguide

quantum memory for entangled photons. Nature,469:512–515,

2011.

[14] Mathew George, Selim Reza, Hubertus Suche, Raimund Ricken,

and Viktor Quiring Wolfgang Sohler. Self-pulsing Ti:Er:LiNbO3

waveguide laser. In European Conference on Lasers and Electro-

Optics 2009 and the European Quantum Electronics Conference.

CLEO Europe - EQEC 2009.,2009.

[15] Alphan Sennaroglu, editor. Solid-State Lasers and Applications.

CRC Press, 2006.

[16] de Sandro J.P., Jones J.K., Shepherd D.P., Hempstead M., J. Wang,

and Tropper A.C. Non-photorefractive CW Tm-indiffused

Ti:LiNbO3waveguide laser operating at room temperature. IEEE

Photonics Technology Letters,8(2):209–211,1996.

[17] E. Cantelar, J. A. Sanz-Garcia, G. Lifante, F. Cussó, and P. L. Per-

nas. Single polarized Tm3+laser in Zn-diffused LiNbO3channel

waveguides. Applied Physics Letters,86(16):161119,2005.

[18] D. G. Lancaster, S. Gross, H. Ebendorff-Heidepriem, A. Fuer-

bach, M. J. Withford, and T. M. Monro. 2.1µm waveguide

laser fabricated by femtosecond laser direct-writing in Ho3+,

Tm3+:ZBLAN glass. Optics Letters,37(6):996–998,2012.

[19] S. Rivier, X. Mateos, V. Petrov, U. Griebner, Y. E. Romanyuk, C. N.

Borca, F. Gardillou, and M. Pollnau. Tm:KY(WO4)2waveguide

laser. Optics Express,9:5885–5892.

[20] I. Baumann, S. Bosso, R. Brinkmann, R. Corsini, M. Dinand,

A. Greiner, K. Schäfer, J. Söchtig, W. Sohler, H. Suche, and R. Wes-

sel. Er-doped integrated optical devices in LiNbO3.IEEE Journal

of Selected Topics in Quantum Electronics,2:355–366,1997.

[21] I. Baumann, R. Brinkmann, M. Dinand, W. Sohler, L. Beckers,

Ch. Buchal, M. Fleuster, H. Holzbrecher, H. Paulus, K.-H. Müller,

Th. Gog, G. Materlik, O. Witte, H. Stolz, and W. von der Osten.

Erbium incorporation in LiNbO3by diffusion-doping. Applied

Physics A,64:33–44,1997.

[22] I. Baumann, F. Cussó, B. Herreros, H. Holzbrecher, H. Paulus,

K. Schäfer, and W. Sohler. Praseodymium-doped Ti:LiNbO3

waveguides. Applied Physics A,68:321–324,199.

[23] Mathew George, Raimund Ricken, Wolfgang Sohler,

E. Saglamyurek, N. Sinclair, C. La Mela, and W. Tittel.

Ti:Tm:LiNbO3waveguide for quantum memory applications.

In European Conference on Integrated Optics,2010.

[24] N. Sinclair, E. Saglamyurek, M. George, R. Ricken, C. La Mela,

W. Sohler, and W. Tittel. Spectroscopic investigations of a

Ti:Tm:LiNbO3waveguide for photon-echo quantum memory.

Journal of Luminescence,130:1586 –1593,2010.

[25] Strake E., Bava G.P., and Montrosset I. Guided modes of

Ti:LiNbO3channel waveguides: a novel quasi-analytical tech-

nique in comparison with the scalar finite-element method. IEEE

Journal of Lightwave Technology,6(6):1126 –1135,1988.

[26] R. Regener and W. Sohler. Loss in low-finesse Ti:LiNbO3optical

waveguide resonators. Applied Physics B,36:143–147,1985.

[27] H. Bubert and H. Jenett (eds.). Surface and Thin Film Analysis.

Wiley, 2002.

[28] M. Quintanilla, E. Cantelar, J.A. Sanz-García, and F. Cussó.

Growth and optical characterization of Tm3+-doped LiNbO3.

Optical Materials,30:1098 –1102,2008.

[29] E. Cantelar, G. Lifante, and F. Cussó. Modelling of Tm3+

doped LiNbO3waveguide lasers. Optical and Quantum Electron-

ics,38:111–122,2006.

[30] E. Cantelar, M. Quintanilla, P.L. Pernas, G.A. Torchia, G. Lifante,

and Cussó. Polarized emission and absorption cross-section cal-

culation in LiNbO3:Tm3+.Journal of Luminescence,128:988 –991,

2008.

[31] P. C. Becker, N. A. Olsson, and J R Simpson. Erbium Doped Fiber

Amplifiers, Fundamentals and Technology. Academic Press, 1999.

[32] Rodica. M. Martin. Reciprocity between emission and absorption for

rare earth ions in glass - Ph.D. thesis. Worcester Polytechnic Insti-

tute, 2006.

[33] M.J.F. Digonnet, E. Murphy-Chutorian, and D.G. Falquier. Fun-

damental limitations of the McCumber relation applied to Er-

doped silica and other amorphous-host lasers. IEEE Journal of

Quantum Electronics,38(12):1629 –1637,2002.

[34] Joseph T. Verdeyen. Laser Electronics. Prentice Hall, 2edition,

1989.

[35] Yong Wang, Deyuan Shen, Hao Chen, Jian Zhang, Xianpeng Qin,

Dingyuan Tang, Xiaofang Yang, and Ting Zhao. Highly efficient

Tm:YAG ceramic laser resonantly pumped at 1617 nm. Optics

Letters,36:4485–4487,2011.

[36] R. Brinkmann, I. Baumann, M. Dinand, W. Sohler, and H. Suche.

Erbium-doped single- and double-pass Ti:LiNbO3waveguide

amplifiers. IEEE Journal of Quantum Electronics,30:2356–2360,

1994.

[37] R P Photonics Consulting GmbH. www.rp-photonics.com.

[38] R P Fiber Power V 3.0- User Manual. pages 14 –15.

[39] Marc Eichhorn. High-gain Tm-doped fluoride fiber amplifier.

Optics Letters,30:456–458,2005.

[40] S. Valling, B. Ståhlberg, and Å.M. Lindberg. Tunable feed-

back loop for suppression of relaxation oscillations in a diode-

pumped Nd:YVO4laser. Optics and Laser Technology,39(1):82 –

85,2007.

[41] S. Taccheo, P. Laporta, O. Svelto, and G. De Geronimo. Intensity

noise reduction in a single-frequency ytterbium-codoped erbium

laser. Optics Letters,21:1747–1749,1996.

[42] Rudolf Wessel. Actively Mode Locked Waveguide Lasers in Erbium

Doped Lithium Niobate, Ph.D. thesis.2003.

[43] K. K. Schäfer, I. Baumann, W. Sohler, H. Suche, and S. Westen-

höfer. Diode-pumped and packaged acoustooptically tunable

Ti:Er:LiNbO3waveguide laser of wide tuning range. IEEE Jour-

nal of Quantum Electronics,33(10):1636 –1641,1997.

[44] L. Núñez, J. O. Tocho, J. A. Sanz-García, E. Rodríguez, F. Cussó,

D.C. Hanna, A.C. Tropper, and A.C. Large. Optical absorption

and luminescence of Tm3+-doped LiNbO3and LiNbO3(MgO)

crystals. Journal of Luminescence,55:253 –263,1993.