Henini M. Handbook of Self Assembled Semiconductor Nanostructures for Novel devices in Photonics and Electronics

Подождите немного. Документ загружается.

Quantum Dot Superluminescent Diodes 599

59. K. Gündogdu , K.C. Hall , T.F. Boggess , D.G. Deppe , and O.B. Shchekin , Appl. Phys. Lett. 8 5 ( 2 0 ) ,

4570 ( 2004 ) .

60. M. Sugawara , K. Mukai , and H. Shoji , Appl. Phys. Lett. 71 ( 19 ) , 2791 ( 1997 ) .

61. S. Sanguinetti , M. Henini , M. Grassi Alessi , M. Capizzi , P. Frigeri , and S. Franchi , Phys. Rev.

60 ( 11 ) , 8276 ( 1999 ) .

62. A. Markus , J.X. Chen , O. Gauthier-Lafaye , J.G. Provost , C. Paranthoen , and A. Fiore , IEEE J. Sel.

Topics Quantum Electron. 9 ( 5 ) , 1308 ( 2003 ) .

63. A.A. Dikshit and J.M. Pikal , IEEE J. Quantum Electron. 40 ( 2 ) , 105 ( 2004 ) .

64. P. Dawson , O. Rubel , S.D. Baranovskii , K. Pierz , P. Thomas , and E.O. Göbel , Phys. Rev. 7 2 ( 2 3 ) , 1

( 2005 ) .

65. R. Ulbrich and M.H. Pilkhun , Appl. Phys. Lett. 16 ( 12 ) , 516 ( 1970 ) .

66. S. Hasuo and T. Ohmi , J. Appl. Phys. 13 ( 9 ) , 1429 ( 1974 ) .

67. E. Schöll , IEEE J. Quantum Electron. 24 ( 2 ) , 435 ( 1988 ) .

68. J.E. Bowers , P.A. Morton , A. Mar , and S.W. Corzine , IEEE J. Quantum Electron. 2 5 ( 6 ) , 1426

( 1989 ) .

69. N.S.K. Kwong , K.Y. Lau , and N. Bar-Chaim , IEEE J. Quantum Electron. 25 ( 4 ) , 696 ( 1989 ) .

70. I.E. Itskevich , M.S. Skolnick , D.J. Mowbray , I.A. Trojan , S.G. Lyapin , L.R. Wilson , M.J. Steer ,

M. Hopkinson , L. Eaves , and P.C. Main , Phys. Rev. 6 0 ( 4 ) , R2185 ( 1999 ) .

71. T. Müller , W. Parz , F.F. Schrey , G. Strasser , and K. Unterrainer , Semiconductor Science and Technology

19 ( 4 SPEC. ISS: ) , S287 ( 2004 ) .

72. J. Siegert , S. Marcinkevicius , and Q.X. Zhao , Phys. Rev. 72 ( 8 ) , 5316 ( 2005 ) .

73. Y. Zhao , W. Han , J. Song , X. Li , Y. Liu , D. Gao , G. Du , H. Cao , and R.P.H. Chang , J. Appl. Phys. 8 5 ,

3945 ( 1999 ) .

74. M. Bayer and A. Forchel , Phys. Rev. B 65 ( 4 ) , 413081 ( 2002 ) .

75. Y.-C. Xin , A. Martinez , T. Saiz , A.J. Moscho , Y. Li , A. Nilsen , A.L. Gray , and L.F. Lester , IEEE Photon.

Technol. Lett. 19 , 501 ( 2007 ) .

76. D.C. Heo , J.D. Song , W.J. Choi , J.I. Lee , J.C. Jung , and I.K. Han , Electron. Lett. 39 , 863 ( 2003 ) .

77. Z.Y. Zhang , Z.G. Wang , B. Xu , P. Jin , Z.Z. Sun , and F.Q. Liu , IEEE Photonics Techn. Lett. 1 6 , 2 7

( 2004 ) .

78. T. Akiyama , M. Ekawa , M. Sugawara , K. Kawaguchi , H. Sudo , A. Kuramata , H. Ebe , and

Y. A r a k a w a , IEEE Photonics Techn. Lett. 17 , 1614 ( 2005 ) .

79. S. Fafard , K. Hinzer , S. Raymond , M. Dion , J. McCaffrey , Y. Feng , and S. Charbonneau , Science 274 ,

1350 ( 1996 ) .

80. H.Y. Liu et al., Appl. Phys. Lett. 80 , 3769 ( 2002 ) .

81. Y.M. Manz , O.G. Schmidt , and K. Eberl , Appl. Phys. Lett. 76 , 3343 ( 2000 ) .

82. Y. Arakawa and S. Kako , Phys. Stat. Sol. a Applications and Materials Science 203 , 3512 ( 2006 ) .

83. T. Taliercio , P. Lefebvre , A. Morel , M. Gallart , J. Allegre , B. Gil , H. Mathieu , N. Grandjean , and

J. Massies , Materials Science and Engineering B – Solid State Materials for Advanced Technology 8 2 ,

151 ( 2001 ) .

CH019-I046325.indd 599CH019-I046325.indd 599 6/27/2008 5:21:42 PM6/27/2008 5:21:42 PM

CHAPTER 20

Quantum Dot-based Mode-locked Lasers and Applications

A. Martinez ,

C. Gosset

K. Merghem

G. Moreau

F. Lelarge ,

and A. Ramdane

CNRS Laboratory for Photonics and Nanostructures (CNRS LPN) Route de Nozay,

91460 Marcoussis, France;

ALCATEL THALES III–V Lab, Route de Nozay, 91460 Marcoussis, France

20.1 Introduction

Self-assembled semiconductor quantum dot (QD)-based lasers are attracting considerable inter-

est owing to their remarkable optoelectronic properties that result from the three-dimensional

carrier conﬁ nement. They are indeed expected to exhibit much improved performance over that

of quantum well devices [1] . Extremely low threshold currents [2] as well as high temperature

stability [3–4] have readily been demonstrated in the InAs/GaAs material system with tempera-

ture-independent transmission experiments already carried out at 10 Gbit/s [5] .

Progress in the growth of self-assembled nanostructures has allowed the stacking of several layers

of QDs into the active region, thus enabling the achievement of high modal gain. Fast carrier dynamics

such as sub-picosecond ground state (GS) gain recovery time of QD semiconductor optical ampliﬁ ers

has been demonstrated as well as sub-picosecond absorption recovery time for QD saturable absorbers.

Low relative intensity noise (RIN) is yet another interesting property for QD based lasers.

QD-based active layers have hence recently been used in the fabrication of mode-locked lasers.

Inhomogeneous broadening of the gain spectrum due QD size distribution is indeed an asset for

MLLs as this should result in pulse shortening. Indeed, the sub-picosecond barrier has recently

been crossed with the demonstration of 400 fs pulsewidth at a 20 GHz repetition rate based on

an InAs/GaAs QD material system [6] , and 500 fs pulsewidth at 53 GHz generated by InAs/InP

QDash lasers [7] . Ultra-fast carrier dynamics [8, 9] is also a major advantage in the short pulse

generation process. Interestingly, the low optical conﬁ nement factor, a high degree of population

inversion and low loss [10] should lead to low phase noise and results in low timing jitter.

Small jitter of the generated pulses is further expected owing to the theoretically predicted low

linewidth enhancement factor (Henry factor) of low-dimensional active layer-based lasers.

Section 20.2 discusses short pulse generation and low timing jitter of In(Ga)As quantum dot

mode-locked lasers emitting in the 1.1–1.3 μ m range. Section 20.3 describes the performances of

InAs/InP quantum dash mode–locked lasers emitting in the 1.55 μ m window. Section 20.4 high-

lights the potential of In(Ga)As/GaAs QD MLLs for optically interconnected systems. Finally, 40

Gb/s all optical clock recovery using InAs/InP QDash MLLs will be reported.

20.2 InAs/GaAs quantum dot mode-locked lasers

The ﬁ rst demonstration of monolithic passively mode-locked lasers using the QD-based active

region was reported by Huang et al. [11] . The authors showed stable mode locking at 1278 nm

CH020-I046325.indd 600CH020-I046325.indd 600 6/25/2008 10:03:34 AM6/25/2008 10:03:34 AM

Quantum Dot-based Mode-locked Lasers and Applications 601

at a repetition rate of 7.4 GHz in a two-section device (a gain section electrically isolated from

an absorber section) at room temperature in a two dot-in-a-well (DWELL) layer. This ﬁ rst paper

evidenced some of the fundamental trends observed in QD MLLs: shortest pulses are obtained

just above threshold while pulsewidth quickly broadens with the injection current. They demon-

strated 17 ps pulsewidth with a photocurrent radio frequency spectrum of 370 kHz width, which

is an indication of phase noise level. The authors pointed out that higher ground state gain (cor-

responding to  1 . 3 μ m laser emission) should lead to improvement in the MLL performances and

that QD MLLs are potentially attractive for ultra-short pulse generation.

In this chapter we will restrict our discussion to monolithic two-section conﬁ gurations made of

a gain and absorber part as illustrated in Fig. 20.1 . However, some of the properties have also been

reported for external cavity MLLs implemented from a two-section laser placed into an external

cavity. Our discussion will try to emphasize device properties arising from the use of QD material.

Planes of In

0.4

GaAs QD

QD size ~7–10 nm

33 nm GaAs

Forward bias

Reverse bias

gain

absorber

Substrate

Figure 20.1 Schematic view of a two-section quantum dot mode-locked laser. The gain section is forward biased

and AR coating is applied to the facet, the saturable absorber is reverse biased and the facet is HR coated [6] . (Reused

of Physics.)

While In(Ga)As/GaAs QD MLLs always rely on a two-section conﬁ guration, we will show that

mode locking in an InAs/InP QDash material system (section 20.3) can occur in a single-section

FP laser without resorting to a saturable absorber. Moreover, in our discussion, we are reminded

here that passive mode locking means pulses are generated by driving the gain and absorber sec-

tions in continuous wave mode; while hybrid mode locking corresponds to the application of an

external RF signal for loss modulation of the absorber.

20.2.1 Quantum dot active medium for ultra-short pulse generation

Some of the main advantages of QD-based active medium can easily be understood by analys-

ing pulse duration versus different parameters of MLLs. Using a model based on analytic theory,

it was demonstrated that for two-section conﬁ guration (gain and absorber) the full width half

maximum (FWHM) of the pulse width can be described by [12–15] :

αω

FWHM

 176

⎛

⎝

⎜

⎞

⎠

⎟

(20.1)

CH020-I046325.indd 601CH020-I046325.indd 601 6/25/2008 10:03:35 AM6/25/2008 10:03:35 AM

602 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

where g represents the gain of the laser,

is the saturation energy of the absorber, α

is a con-

stant loss of the absorber, Δ ω

the width of the gain spectrum and A

the envelope of the pulse at

the initial time and position.

The modal gain of QD lasers is lower than that of QW-based lasers due to a reduced optical

conﬁ nement factor Γ . It will be shown later that low Γ may prove to be an advantage as long as

GS gain saturation can be avoided through stacking of multiple layers of QDs up to 10 or 15. The

amount of gain also determines the minimum cavity length which allows GS emission and hence

sets a limit on the highest achievable repetition rate.

Relation (20.1) shows that an absorber with low saturation energy

should yield reduced

pulsewidth. This essential property will be discussed in section 20.2.3.2 through the carrier

dynamics in the absorber section.

Inhomogeneous broadening of the gain spectrum due QD size distribution (Gaussian-like dis-

tribution of the QD size) is indeed an asset for mode-mocked lasers as this should result in a wider

effective gain bandwidth and thus pulse shortening.

In conclusion, the section devoted to monolithic QD MLLs grown on GaAs substrates is organ-

ized as follows. The ﬁ rst paragraph (section 20.2.2) will present the properties related to the

modal gain of InAs/GaAs QDs and the carrier dynamics in the gain regime. We will consider

their consequences on the pulse duration as well as the role of the Henry factor in the pulse

shortening/broadening.

In a second part (section 20.2.3), the possibility of using QD active material as a saturable

absorber will be highlighted. It will be shown that increasing the reverse voltage in the absorber

section reduces by one order of magnitude the absorber recovery time, allowing sub-picosecond

pulse generation.

Moreover, section 20.2.4 will present the noise properties of the QD material system and dem-

onstration of sub-picosecond timing jitter will be reported.

Finally, the performances of QD MLLs will be summarized to compare the results reported so

far and discuss perspectives.

20.2.2 Gain properties of quantum dot mode-locked lasers

20.2.2.1 The role of inhomogeneous broadening

Owing to their wider inhomogeneous broadened gain spectrum, QD MLLs are expected to pro-

duce shorter pulses compared to their QW-based counterparts. The ﬁ rst demonstration of passive

QD MLLs showed pulsewidths of 17 ps [11] at a 7.4 GHz repetition rate and the average power did

not exceed 2 mW. Due to GS gain saturation in a two dot-in-a-well (DWELL) layer, the authors

could not reach a broad gain regime and pulses were not obtained for injection currents greater

than 45 mA.

While the 3 dB optical bandwidth (  1 nm) does not exceed that of QW MLL lasers, one solu-

tion to reduce the pulsewidth would be to exploit the potentially wider effective gain bandwidth,

achievable at high injection current in high modal gain structures. This could be obtained by

stacking a much larger number of QD layers, as pointed out by the authors.

Owing to the progress in growth and quality of this material system, the possibility of stacking

more QD layers into the active region opens the door to shorter cavity lasers emitting on the GS.

Kuntz et al. demonstrated improved performances of QD MLLs using higher GS modal gain by

stacking ﬁ ve QD layers [16] . Passive MLL at 20 GHz was evidenced by using a 2 mm long cavity

(1500 μ m long gain/500 μ m long absorber). Careful adjustment of the bias/voltage can provide

transform-limited pulses [16] . Indeed, assuming Gaussian pulse shape, Kuntz et al. measured an

FWHM of 12 ps which is in good agreement with the Fourier limit of 13 ps corresponding to a

spectral FWHM of 0.180 nm. Based on the same active layer [16] , the authors report GS emission

in a two-section conﬁ guration for cavities as short as 1130 μ m (980 μ m and 150 μ m, respec-

tively, for the gain and absorber lengths) leading to passive ML at a repetition rate of 35 GHz.

Transform-limited pulses were generated with a minimum pulse duration of 7 ps, which is half

of the ﬁ rst value reported in a two QD layer structure. These results show that QD MLLs at high

repetition rate can be obtained together with pulse duration of few ps and Fourier-limited pulses.

CH020-I046325.indd 602CH020-I046325.indd 602 6/25/2008 10:03:35 AM6/25/2008 10:03:35 AM

Quantum Dot-based Mode-locked Lasers and Applications 603

Kuntz et al. remarked that, in the case of 7 ps pulses, they were using only 2% of the intrinsic

spectral width of the QD gain medium. According to formula (20.1), the gain bandwidth can

further be increased to achieve sub-picosecond pulses [16] .

It is worth noticing that these devices exhibited mode-locking in regions where the gain cur-

rent did not exceed 70 mA (for a threshold current of 30 mA) thus limiting the width of the

optical spectrum. Finally, for the 7 ps transform-limited pulze, the peak power from one facet

amounts to 6 mW.This is rather low as some applications such as optically interconnected sys-

tems require peak powers up to a few watts [16] .

Optimized design of the MLL cavity includes careful adjustment of the gain/absorber length

to avoid GS gain saturation at high current and simultaneous enhanced absorber saturation

owing to reduced optical losses, using HR coating [6] . This allowed the demonstration of pico-

second as well as sub-picosecond pulse duration with an average power of 45 mW (  1.1 W peak

power) at 1260 nm at 21 GHz repetition rate. Through a careful adjustment of the gain current

and reverse voltage (185 mA,  7.6 V), an FWHM of 14 nm was achieved (Fig. 20.2) leading to

pulses as short as 390 fs, the lowest pulsewidth ever reported in the InAs/GaAs QD material sys-

tem [6] . However, the time bandwidth product of  1 indicated substantial residual chirping. The

optical spectrum is indeed two orders of magnitude larger than previous results based on a simi-

lar ﬁ ve QD layer structure [16] . The average power in this regime amounts to 25 mW, i.e.  3 W

peak power. These results highlight the potential of QD MLLs for ultra-short pulse generation by

exploiting the intrinsic wide gain spectrum of the active medium. This reﬂ ects the impact of the

gain spectrum on the shortening of the pulses and also shows that to beneﬁ t from the potentially

wide gain spectrum, one has to ensure high GS modal gain to be able to drive the section at high

currents ( Figs 20.2 and 20.3 ). It relies in part on a combination of three criteria in this case:

high modal gain (ﬁ ve QD layers), long gain section (1.8 mm), and HR coating which reduces the

mirror loss and provides lower threshold net gain into the cavity.

0.0

0.5

1.0

1.5

2.0

2.5

3.0

Normalized intensity (arb. units)

4 3 2 1012

Time delay (ps)

Wavelength (nm)

1240 1250 1260 1270 1280

Intensity (arb. units)

 ~ 14 nm

Experiment

Gaussian fit – 550 fs

t ~ 393 fs

Figure 20.2 Measured intensity autocorrelation and the corresponding optical spectrum of the QD MLL (inset).

Note the spectrum showing a large bandwidth [6] . (Reused with permission from E.U. Rafailov et al., Appl. Phys.

20.2.2.2 The role of Henry factor in the pulsewidth broadening

As previously mentioned, the pulsewidth is limited by the gain bandwidth of the active medium.

However, there may be deviation from this trend due to chirping of the pulses [17] , where the

chirp illustrates the wavelength shift of a longitudinal mode when the gain is changed. Indeed,

the role the linewidth enhancement factor (LEF) in QW-based mode-locked lasers was theoreti-

cally investigated [17] . Simulations showed that for QW MLLs the time–bandwidth product scales

linearly with the LEF, assuming that this parameter is the same for gain and absorber sections.

CH020-I046325.indd 603CH020-I046325.indd 603 6/25/2008 10:03:35 AM6/25/2008 10:03:35 AM

604 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

Several theoretical and experimental investigations showed a near-zero Henry factor of QD

lasers below threshold. But above threshold, excited state band ﬁ lling and carrier injection into

the wetting layer breaks the symmetry of the gain spectrum which results in the divergence of

Henry factor with current. For some speciﬁ c applications, Fourier-limited very short pulses with

high peak power are desirable. This implies working within the range of high injection current

to generate output peak power of typically a few W. However, it was shown that the Henry factor

of directly modulated QD lasers strongly depends on the injection rate. When GS gain saturation

occurs in QD lasers, the Henry factor starts diverging [18, 19] . Even giant linewidth enhance-

ment factors (LEF) of 50 for GS lasing were measured at high currents in low modal gain struc-

tures (three QD layers) [20] . This behaviour would induce chirping of the pulses in QD MLLs

with signiﬁ cant pulse broadening. A solution to overcome the divergence of the Henry factor α

relies on increasing the optical conﬁ nement factor Γ by stacking several QD layers, which in turn

increases the GS modal gain. It was indeed shown that stacking up to ten layers into the active

region prevents the Henry factor from diverging and yields α

similar to the best values obtained

in QW lasers [21] .

The ﬁ rst evidence of the impact of the Henry factor in the broadening of the pulses in QD MLLs

was reported by Choi et al. [22] . The authors showed that the LEF of pulses generated by external

cavity two-section QD MLL increases with bias current while the opposite trend is observed for the

absorber section when the bias voltage increases. It was pointed out that the total chirp can be con-

trolled through adjustment of the current and reverse voltage of the gain/absorber section.

More recently, measurement of the pulse duration in monolithic QD MLLs versus injection cur-

rent showed an increase in the pulse width for a constant reverse bias in the absorber. The pulses

broaden from 4 ps to  14 ps when the current increases from 50 mA to  270 mA [23, 24] . This

trend is explained by the increase of the Henry factor due to GS gain saturation as long as the

laser emits on the GS. The time bandwidth product of pulses generated in diode lasers is given

ΔΔτν⋅+ K(),

212

w h e r e K is a constant depending on the pulse shape [12] . While the

authors did not give values of the increase of the optical bandwidth Δ ν with current, these

results seem to be compatible with Henry factors measured in similar layer structures [19] .

Nevertheless, a systematic investigation of the Henry factor while the absorber is biased

would be required to evaluate the impact of this latter section on the chirp of pulses. The

impact of the Henry factor on the pulsewidth in self-pulsating lasers based on a single section

FP laser in the InAs/InP QDash material system has not yet been investigated (section 20.3).

QD active medium provides a means to achieve both very short pulses (few ps or less) and

high peak power at high repetition rate (  20 GHz). The exploitation of the inherent large gain

1245 1250 1255 1260 1265

Wavelength (nm)

absorber

 8V

Increasing current →

(42–240 mA)

Figure 20.3 Evolution of the output spectra of the QD MLL versus the injection current in the gain section for

a constant reverse voltage [6] . (Reused with permission from E.U. Rafailov et al., Appl. Phys. Lett. 87 , 081107

CH020-I046325.indd 604CH020-I046325.indd 604 6/25/2008 10:03:36 AM6/25/2008 10:03:36 AM

Quantum Dot-based Mode-locked Lasers and Applications 605

bandwidth ultimately relies on stacking several QD layers to avoid GS gain saturation, responsi-

ble for pulse broadening through an increase of the Henry factor.

However, applications such all optical clock recovery may require higher repetition rates in

excess of 40 GHz.

20.2.2.3 High modal gain for high repetition rate (40–80 GHz)

As the repetition rate is given by the inverse of the round trip cavity time, the cavity length needs

to be in the range 0.5–1 mm to match the 40–80 GHz range. Increasing the modal gain is man-

datory to obtain GS emission on such short cavities. The solution obviously relies on stacking

a high number of QD sheets compatible with single-mode operation and/or HR coating, which

decreases the threshold gain.

Fourier-limited pulses were, for instance, generated at high repetition rate (40–50 GHz) using

the same ﬁ ve QD layer structure as in [16] owing to an HR coating on the absorber facet [25] .

Passive mode locking was hence demonstrated at 40 GHz and 50 GHz with a typical pulse dura-

tion of 3 ps. Taking into account the spectral FWHM, the time bandwidth product showed

Fourier-limited pulses for both repetition rates.

Further improvement in the modal gain was achieved through stacking ten layers of QDs. This

resulted in the demonstration of passive 40 GHz mode locking with typical pulse duration of 2.5

ps and time bandwidth products ranging from 0.5 to 0.9 [26] . In this case, the device consisted

of a 940 μ m long gain and a 100 μ m long HR-coated absorber section.

Enhancement of the optical gain using 15 QD layers allowed a record repetition rate of 80 GHz

to be achieved associated with a 1.9 ps pulsewidth obtained from a 0.5 mm long passive mode-

locked laser based on InAs/GaAs [27] . The time bandwidth product amounts to 1.5, leaving

room for further pulsewidth shortening.

As a conclusion, most of the effects that take place in the gain region of QD MLLs and affect

the pulse duration and the time bandwidth product have been presented. Another advantage of

QD material system arises from the low saturation energy and fast recovery time (down to a few

hundred fs) of the absorber, which makes QD saturable absorbers ideal candidates for monolithic

MLLs [28] .

20.2.3 Quantum dot saturable absorber

20.2.3.1 Saturable absorber for short pulse generation

When a pulse enters the absorber, the leading and trailing edges of the pulse experience high

losses (unsaturated loss), while the peak pulse faces lower losses because the absorber exhibits a

sudden decrease of the absorption losses (saturation of the loss) [17] . The time during which the

absorber loss is saturated corresponds to a temporal window for which net gain is obtained. The

temporal width of this saturated loss is directly determined by the carrier lifetime in the absorber.

The shorter the carrier lifetime is, the narrower the window will be and hence the pulsewidth.

Two conditions are required to obtain mode locking with a two-section conﬁ guration. As pre-

viously discussed, shorter pulsewidths are obtained with an absorber that can be more easily sat-

urated [12] because the pulse duration scales with the squared root of saturation energy of the

absorber

Moreover, short pulses imply absorber recovery time faster than gain recovery time

[17] . In other words, the carrier lifetime in the absorber has to be shorter than  1 picosecond to

ensure the generation of sub-ps pulses. Broadband absorption spectrum related to inhomogene-

ous broadening of the QD ensemble is also a key parameter for short pulse generation [29] .

Lester and co-workers recently measured the saturation energy of a six-DWELL layer struc-

ture emitting pulses at  1210 nm at 5 GHz repetition rate [30] . The absorption saturation pow-

ers were calculated for different reverse bias voltages: they vary from  0.5 W to 2.2 W for a  5 V

applied voltage.

In the following, we will brieﬂ y mention the ﬁ rst works based on the use of the QD medium as

a saturable absorber for pulse generation and we will focus on the absorber recovery time. Then

the impact of the fast carrier dynamics of the absorber will be discussed regarding the perform-

ances of QD MLLs.

CH020-I046325.indd 605CH020-I046325.indd 605 6/25/2008 10:03:36 AM6/25/2008 10:03:36 AM

606 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

20.2.3.2 Ultra-fast QD saturable absorber dynamics

While the ﬁ rst use of a QD saturable absorber was reported by Lester and co-workers in a mon-

olithic conﬁ guration, the fast character of the quantum dot saturable absorber was further

evidenced in stable passive mode locking of a solid-state laser [29, ref. 5 therein]. The authors

demonstrated that pulses can be shortened by applying a reverse voltage to the absorber. Recent

work also demonstrated that the careful adjustment of the reverse voltage of a QD saturable

absorber allows the control of the pulsewidth in a monolithic two-section device [16, 25] . These

last papers evidenced standard passive mode locking, where the pulsewidth can be reduced from

18 to 8 ps by increasing the reverse bias from 0 V to  2 V for a constant gain current at a 20 GHz

repetition rate.

The electric ﬁ eld dependence of carrier dynamics of a QD saturable absorber waveguide has

been investigated by means of sub-picosecond pump–probe experiments at room temperature

[9] . Differential time spectroscopy aimed at measuring the absorber recovery in a degenerate

scheme at the 1.28 μ m GS transition. The investigated waveguides were fabricated from a ﬁ ve-QD

layer structure, comparable to the p-i-n diode used for MLL in [16] .

The authors demonstrated a decrease in the carrier recovery time from 62 ps to 700 fs, nearly

two orders of magnitude, with an increasing reverse voltage from 0 V to 10 V ( Fig. 20.4 ). The

absorption recovery time scales as a simple exponential function of the reverse voltage. The

experiments showed that at low bias, thermionic emission is the dominant carrier escape mecha-

nism while above  4 V, the tunnelling process becomes dominant.

024681012

Reverse bias (V)

0.1

100

Carrier recovery time (ps)

Figure 20.4 Absorption recovery time of QD saturable absorber as a function of the reverse voltage [9] . (Reused

Institute of Physics.)

In [9] , to saturate the QD absorber, pump pulse energy of 580 fJ was employed. Previous work

performed on InGaAs/InGaAlAs QW electro-absorption modulators demonstrated similar satu-

ration energy of 0.4 pJ and absorption recovery time of 5 ps at 4 V for an optimized band gap

engineered structure [31] .

These very interesting results highlight the potential of ultra-fast QD saturable absorbers for

sub-ps pulse generation. This also gives a general trend for the adjustment of the pulsewidth

because as the reverse voltage is increased, the average output power falls off approximately

exponentially. The next section is dedicated to the implementation of QD saturable absorber and

the adjustment of the reverse voltage to achieve ultra-short pulses.

20.2.3.3 QD saturable absorber for sub-picosecond pulse generation

The ﬁ rst demonstration of the impact of absorber carrier lifetime on the pulsewidth in mono-

lithic mode-locked lasers was reported by Kuntz et al. [16] . The authors performed a systematic

investigation of the pulse duration versus applied voltage in the absorber section for a range of

CH020-I046325.indd 606CH020-I046325.indd 606 6/25/2008 10:03:36 AM6/25/2008 10:03:36 AM

Quantum Dot-based Mode-locked Lasers and Applications 607

constant gain currents ( Fig. 20.5 ). Passive mode locking was investigated using a ﬁ ve QD layer

structure in a two-section conﬁ guration (1500 μ m long gain and 50 μ m long absorber), which

allowed mode locking at 20 GHz [16] . They demonstrated a reduction of the pulsewidth from

18 to 8 ps when the reverse voltage is decreased from 0 V to 3 V. These lasers exhibit a common

behaviour already observed in QW-based MLLs, implying that InAs/GaAs QD MLLs are governed

by similar physical mechanisms.

Moreover, Kuntz et al. showed that for a constant current, decreasing the reverse voltage from

0 V to  3 V forces the laser to operate in three different regimes: incomplete mode locking (low

V), mode locking (intermediate V) and no lasing (high V). These three different regimes were also

reported by Todaro et al. [32] . In this latter regime ( Fig. 20.5 , [16] ), increasing the reverse bias in

the range 2.5–3 V induces an increased absorption which prevents lasing when the gain current

is not large enough to sustain net gain into the cavity. Providing a higher modal gain through

stacking more QD layers is a solution to further reduce the pulse duration, as discussed in section

20.2.2.2 through the use of higher reverse voltages.

Incomplete

mode locking

lasing

0.0

30 40 50 60

0.5

1.0

1.5

2.0

2.5

FWHM pulsewidth (ps)

Reverse bias voltage (V)

Gain current (mA)

Figure 20.5 Plot of pulse width dependence on reverse bias voltage and gain current [16] . (Reused with permis-

To obtain much shorter pulsewidths by increasing the reverse voltage implies increasing the

available gain into the cavity, to compensate for the extra losses. This can be achieved by length-

ening the gain section and adding an HR coating on the absorber facet [6] as previously dis-

cussed in section 20.2.2.1.

Zhang et al. also reported a decrease in the pulsewidth from 8 ps to 5.5 ps with increasing the

reverse bias [33] from 6 V to 7.5 V, for a constant gain current of 110 mA in passive QD MLLs.

Further experiments on QD MLL lasers showed an exponential-like dependence of the pulsewidth

with the reverse bias [28, 34] ( Fig. 20.6 ). While the authors demonstrated nearly transform-lim-

ited 780 fs pulses at  7 V and 250 mA injection current [28] , deviation from this point shows an

increase of the pulsewidth when the reverse voltage is increased. The authors suggested an expo-

nential dependence of the absorber recovery time on the applied reverse bias, which was indeed

conﬁ rmed later in Malins et al. [9] . On the contrary, increasing the bias current in the gain section

induces a linear increase in pulsewidth, which could be attributed to self phase modulation and/or

increase [9, 17] .

Finally, reduction in the FWHM autocorrelation traces when increasing the reverse voltage of

the absorber was also reported in an external cavity mode locking of a quantum dot two-section

diode laser [35, 36] . The external cavity mode-locked laser is based on a two-section device

(3 mm long gain and 300 μ m long absorber) and a semiconductor optical ampliﬁ er (SOA) to

amplify the pulse energy. Both devices are fabricated from a ten QD layer structure. These inves-

tigations focused on a fundamental cavity frequency of 4.9 GHz. For a ﬁ xed gain current of

CH020-I046325.indd 607CH020-I046325.indd 607 6/25/2008 10:03:37 AM6/25/2008 10:03:37 AM

608 Handbook of Self Assembled Semiconductor Nanostructures for Novel Devices in Photonics and Electronics

115 mA ( Fig. 20.7 , [35] ), the FWHM of the autocorrelation trace decreases from 38 ps to 3.5 ps

when the reverse voltage increases from 2.5 V to 8 V. Nearly transform-limited pulses of 3.5 ps

with a 1.3 nm spectral bandwidth are reported.

12345678

Reverse bias (V)

Pulsewidth (ps)

(a) I  250 mA

Figure 20.6 Pulsewidth for constant gain current versus reverse voltage [28] . (Reused with permission from

Delay (ps)(a) (b) Reverse bias voltage (V)

60

2.5 V

5V

8V

40 200204060

100

2345678



Autocorrelation FWHM (ps)

Autocorrelation

Figure 20.7 Full width half maximum of autocorrelation signal of pulses generated from an external cavity QD

mode-locked laser versus reverse voltage [35] . (Reused with permission from Myoung-Taek Choi et al. , Appl. Phys.

The demonstration of sub-picosecond pulsewidth using high reverse bias is, however, similar

to simulated trends observed in QW-based material systems reported earlier [17] . The authors

further showed that in QW MLLs the pulse-width scales exponentially with the ratio of the non-

linear absorber saturation coefﬁ cient to the non-linear gain saturation coefﬁ cient (each coefﬁ -

cient is proportional to the inverse of the associated saturation energy), and that pulse duration

scales linearly with the absorber lifetime. This proves that the absorber saturation needs to be

stronger than the gain saturation to ensure a pulse shortening mechanism in QW MLLs [17] .

The experimental results identiﬁ ed from these QD MLLs [16, 26, 28, 33, 34] conﬁ rm that mon-

olithic two-section QD MLLs are governed by similar physical mechanisms as QW MLLs in terms

of carrier dynamics in the gain and absorber sections. But the achievable faster carrier dynamics

of QD saturable absorbers allows one order of magnitude improvement in the pulsewidth reduc-

tion compared to standard QW MLLs emitting at 1.5 μ m [17] .

A key parameter of MLLs is the small ﬂ uctuation in the arrival time of pulses, illustrated by the

timing jitter. The next section will present a recent demonstration of reduced timing jitter in QD

MLLs compared to standard QW MLLs.

CH020-I046325.indd 608CH020-I046325.indd 608 6/25/2008 10:03:37 AM6/25/2008 10:03:37 AM