123

DISTRIBUTED FEEDBACK LASERS WITH PHOTON-PHOTON-RESONANCE-ENHANCED MODULATION BANDWIDTH

M. Dumitrescu*, A. Laakso*, J. Viheriala*, T. Uusitalo*, M. Kamp**, P. Uusimaa***

*Optoelectronics Research Centre, Tampere University of Technology, Tampere, Finland E-mail: Mihail.Dumitrescu@tut.fi, Antti.I.Laakso@tut.fi, Jukka.Viheriala@tut.fi, Topi.Uusitalo@tut.fi

**Technische Physik, Wurzburg University, Wurzburg, Germany E-mail: Martin.Kamp@physik.uni-wuerzburg.de

***Modulight Inc., Tampere, Finland E-mail: Petteri.Uusimaa@modulight.com

Abstract–Multi-section distributed-feedback lasers with surface gratings have been fabricated without re-growth by employing ultraviolet nanoimprint lithography. High-frequency photon-photon resonance was exploited to extend the direct modulation bandwidth beyond the conventional limits set by the carrier-photon resonance. Keywords: distributed feedback lasers, surface gratings, nanoimprint lithography, photon-photon resonance, extended modulation bandwidth.

1. INTRODUCTION

The conventional buried-grating distributed feedback (DFB) lasers require two or more epitaxial growth steps, complicating the fabrication, affecting the device performance, yield and reliability (especially when Al-containing materials are used) and, ultimately, increasing the device cost. To avoid the problematic overgrowth we have employed laterally-corrugated ridge-waveguide (LC-RWG) surface gratings, illustrated in Fig. 1, which are applicable to different materials, including Al-containing ones, and can be easily integrated in complex device structures and photonic circuits.

 

Fig. 1. a) Sketch of the laterally-corrugated ridge-waveguide grating structure; b) SEM image of the

facet of a laser with LC-RWG gratings

A supplementary advantage of the LC-RWG gratings derives from the fact that there is only a limited interaction between the defect-prone processed grating interfaces and the carriers, which leads to more stable devices with better

performances and increased reliability. Despite substantial efforts that have been

undertaken to increase the direct modulation bandwidth, no significant breakthrough has been made when the direct modulation bandwidth has been linked to the carrier-photon resonance (CPR), largely because the CPR has inherent physical limitations. However, since the direct modulation bandwidth can be extended considerably by introducing a supplementary high-frequency resonance of the laser, we have managed to consistently and systematically induce a high-frequency photon-photon resonance (PPR) in order to increase the direct modulation bandwidth of our lasers.

2. PHOTON-PHOTON RESONANCE

MODELING AND SIMULATION

A modified rate-equation model has been developed to include the PPR by treating the longitudinal confinement factor as a dynamic variable [1]. The differential rate equations including the extra term resulted from taking the quantum well confinement factor as a dynamic variable are the following:

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡

Γ++⎥

⎦

⎤⎢⎣

⎡⎥⎦

⎤⎢⎣

⎡−−−

=⎥⎦

⎤⎢⎣

⎡

dRgvN

dIqV

dNdN

dNdN

dtd

spgp

i

pPPPN

NPNN

p )( '

η

γγγγ (1)

where γNN, γNP, γPN and γPP are rate coefficients, as defined in [2]. By following the same analysis of the small-signal response to modulation as in [2], the small-signal photon density, including the influence of the extra term, results as:

( )tj

NNspgp

PNip e

djRgvNqV

IN ω

ωγγη Γ⋅

Δ+

⋅++Δ

⋅= )( '11

(2)

where Δ = (γNN + jω)·(γPP + jω) + γNPγPN. The modulation transfer function including

the influence of the extra term resulted from the

a b

978-1-4673-0738-3/12/$31.00 © 2012 IEEE

124

(space and) time variation of the confinement factor, can be written as:

∫∫∫

∫ Γ⋅

⋅Δ

+⋅+⋅

⋅+

Δ=

T

tjspgpNN

TT

TPN

i dtdtd

eRgvNj

dtIdt

dt

qVH

0

'

01

0

0 )()(1)( ω

ωγγ

ηω

(3)

where T is the time interval for which the phase difference between the dominant longitudinal modes is maintained. The first term in (3) resembles the traditional modulation transfer function, with γPN and Δ taken as time-dependent, while the second term is resulted from considering the (space and) time dependence of the confinement factor. This second term introduces the supplementary PPR peak placed at a frequency equal with the frequency difference between the two dominant longitudinal modes. The model indicates that a primary condition for achieving the PPR is to have the dominant modes phase-locked for long enough (quasi-phase-locked). Therefore, besides the large frequency difference between modes usually encountered in multi-mode lasers, the main reason for not achieving a significant PPR peak is that conventional multimode lasers do not provide a mechanism to maintain the phase difference between modes for long enough.

The difficulty in extending the direct modulation bandwidth of the lasers is not related to placing the PPR at high frequencies but to achieving a flat modulation bandwidth between the CPR and PPR. Fig. 2 illustrates this difficulty by showing the modulation bandwidth for a multi-section DFB laser with LC-RWG gratings having PPR at 31.5 GHz (targeting a -3 dB modulation bandwidth of 35 GHz and a transmission rate capability of 43 Gb/s).

0 10 20 30 40 50-30

-20

-10

0

10

20

30

Mod

ulat

ion

resp

onse

[dB

]

Modulation frequency [GHz]

0 dB 9 dB 10 dB 20 dB 30 dB

-3 dB

Fig. 2. Simulated small-signal modulation response of a multi-section DFB laser with LC-RWG gratings having variable mode-suppression ratios between

the two grating modes.

The simulated modulation response in Fig. 2 is above -3 dB at 35 GHz when the mode-suppression ratio between the grating modes is 10 dB. However, a slightly smaller (8 dB) mode-suppression ratio is needed in order to fill the gap between CPR and PPR peaks. The figure illustrates that, besides having the grating modes phase-locked for long-enough, a reasonably good balance of power between the grating dominant modes is needed in order to extend the modulation bandwidth.

Fig. 3 shows simulated and measured small-signal modulation responses for two DFB lasers with LC-RWG surface gratings fabricated from the same epiwafer but having different structures. The first one is a single-longitudinal-section DFB laser that does not exhibit PPR, while the second one is a much longer multiple-longitudinal-section DFB laser with PPR at 14 GHz. Besides the direct modulation bandwidth increase, Fig. 3 illustrates that a major difficulty for extending the modulation bandwidth is to reduce the dip between the CPR and the PPR.

5 10 15 20-20

-15

-10

-5

0

5

10

15

20

increase inmodulation bandwidth

Measured with PPR Simulated with PPR Measured without PPR Simulated without PPR

photon-photonresonance peak

Mod

ulat

ion

resp

onse

[dB

]

Modulation frequency [GHz]

carrier-photonresonance peak

-3 dB

Fig. 3. Measured and simulated small-signal modulation

responses from DFB lasers with and without photon-photon resonance.

3. LASER FABRICATION

Multi-section 1.3 and 1.55 µm DFB lasers with LC-RWG gratings have been fabricated from InP-substrate legacy epiwafers (with epilayer structures designed for Fabry-Perot lasers). PPR frequencies around 20 GHz have been targeted due to the difficulty in achieving a relatively flat modulation response with PPR at higher frequencies.

An important requirement for achieving a stable PPR was found to be a high-enough grating coupling coefficient. Several methods have been investigated for increasing the LC-RWG grating coupling coefficient, κ: adjusting the epilayer structure, reducing the width of the

125

un-etched central section of the ridge W (Fig. 1a), increasing the lateral extension of the grating D (Fig. 1a), using lower order gratings and increasing the grating filling factor. Reducing the p-side cladding thickness to enable a higher κ with less deep etching was not applied since the p-side cladding was (in most cases) already close to the thickness that would induce optical field coupling into the contact. Altering the epilayer structure for extending the optical field penetration into the p-side cladding in order to increase the grating confinement factor was also ruled out because, besides optical field coupling with the contact, it leads to a decrease in the quantum well confinement factor, which affects adversely the laser characteristics. Also reducing too much the width of the un-etched central section of the ridge, W, affects adversely the electrical characteristics of the laser (increased resistance and operating voltage) and increases the difficulty of aligning the contact opening. Regarding the lateral extension of the gratings, D, the simulations have pointed out that the increase in κ saturates with increasing D but have also shown that the experimentally-determined LC-RWG grating coupling coefficients for the first device batches were significantly smaller than predicted.

The analysis of the discrepancies between the simulated and experimentally-determined coupling coefficients showed that technologically-induced imperfect grating profiles play a significant role in reducing the grating coupling coefficient. Fig. 4a shows a scanning electron microscopy (SEM) image of an imperfect LC-RWG grating profile caused by aspect-ratio-dependent-etching (ARDE).

Fig. 4. SEM images of various LC-RWG grating geometries.

The imperfect grating profiles shown in Fig. 4a exhibit un-etched pockets at the bottom of the grating trenches, towards the ridge, exactly in the areas where the optical field should couple with the gratings. Figure 4b shows the improved etching profiles obtained by adjusting the process parameters and by using a more uniform distribution of the areas to be etched across the surface. Current isolation trenches through the gratings and parallel with the ridge were used to limit the lateral current leakage through the grating wings (Fig. 4c-4d).

3rd-order LC-RWG gratings with improved profiles and ~0.5 filling factors, defined by UV nanoimprint lithography [3], have been used in our latest experiments, leading to experimentally evaluated coupling coefficients of ~20 cm-1.

4. EXPERIMENTAL RESULTS

The experiments have confirmed that the PPR frequency is largely determined by the longitudinal structure of the multi-section DFB lasers. Supplementary, we have studied the possibility to adjust the PPR position (and the flatness of the modulation response between the CPR and PPR) by adjusting the bias of the laser sections. Fig. 5 shows the tuning of the PPR frequency by adjusting one of the bias currents for a 1.6 mm long multi-section DFB laser emitting at 1.55 µm. It can be observed how the CPR-PPR gap is filled when the PPR is brought closer to the CPR.

5 10 15 20

-20

-10

0

10

20photon-photonresonance peak

I1=I2=35mAMod

ulat

ion

resp

onse

(dB

)

Frequency (GHz)

I3=55mA I3=65mA I3=70mA I3=75mA

carrier-photonresonance peak

Fig. 5. Measured small-signal modulation response with PPR frequency adjusted by the bias applied to one of the

laser sections in a 1.6 mm long multi-section laser with LC-RWG gratings emitting at 1.55 µm.

Fig. 6 shows the changes in the CPR and

PPR positions and the flattening of the CPR-PPR gap obtained by adjusting the bias in one of the

a

b

c

d

126

sections of a 1.5 mm long multi-section DFB laser with LC-RWG gratings emitting at 1.3 µm.

5 10 15 20-10

0

10

20

30photon-photonresonance peak

carrier-photonresonance peak

Mod

ulat

ion

resp

onse

[dB

]

Frequency [GHz]

I3= 60 mA I3= 70 mA I3= 80 mA I3= 90 mA I3=100 mA I3=110 mA

-3dB

Fig. 6. Measured small-signal modulation response with PPR frequency adjusted by the bias applied to one of the

laser sections in a 1.5 mm long multi-section laser with LC-RWG gratings emitting at 1.3 µm.

The intensity noise measurements for these

lasers confirm that, although sometimes the lasers exhibit a multimodal emission (with several longitudinal modes lasing due to the un-optimized ridge geometry and relatively small coupling coefficient), the laser structures induce only two quasi-phase-locked longitudinal modes. Fig. 7 shows a sample intensity noise (IN) spectrum (not normalized to the optical power of the laser) obtained from a 1.6 mm long laser emitting in several modes at 1.55 µm. A clear and sharp PPR peak in the IN spectra indicates that the grating-determined longitudinal modes, spaced at 20 GHz are indeed quasi-phase-locked. The difference between the CPR and PPR features, resulted from phe phase-locking of the modes inducing the PPR is clearly observable.

5 10 15 20-44

-42

-40

-38

-36

-34

-32 

1 54 6 1 548 1550 155 2 1 554 155 6 1 55 8 1 560 15 62

-7 0

-6 0

-5 0

-4 0

-3 0

-2 0

-1 0

Inte

nsity

(dB

)

W av ele ngth ( nm)

5

Photon-Photon Resonance

Inte

nsity

noi

se [d

B]

Frequency [GHz]

Laser noise System background

Carrier-Photon Resonance

Fig. 7. Intensity noise spectra and emission spectrum (inset) measured for an imperfect 1.6 mm long multi-section DFB

laser with LC-RWG gratings emitting at 1.55 µm.

Structures with quasi-phase-locked longitudinal modes spaced at frequencies, from

47 GHz to 1.3 THz have also been fabricated, Figure 8, but their modulation responses were hampered by the large and deep dip between CPR and PPR.

1299 1300 1301 1302-35

-30

-25

-20

-15

-10

-5

0 aMode spacing

47 GHz

Inte

nsity

[dB

]

Wavelength [nm]

measured simulated

1525 1530 1535 1540 1545-45-40-35-30-25-20-15-10-50

measured simulated

bMode spacing

1.0 THz

Inte

nsity

[dB

]

Wavelength [nm] Fig. 8. Simulated and experimental dual-mode emission spectra with quasi-phase-locked longitudinal modes having: a) 47 GHz and b) 1.0 THz frequency spacing.

3. CONCLUSIONS

Multi-section lasers employing LC-RWG surface gratings at 1.3 and 1.55 µm have been designed, fabricated and characterized. Photon-photon resonances, largely determined by the longitudinal structure of the devices and adjustable by bias, were achieved consistently and systematically, according to the model and simulations. The results prove that the PPR can be exploited for increasing the direct modulation bandwidth beyond the limitations set by the carrier-photon resonance. Lasers with one or several photon-photon resonances at much higher frequencies than the CPR, can be exploited as high-frequency optical clocks or in advanced modulation schemes that do not require a relatively flat modulation response over the whole frequency range.

Acknoledgement–The results reported in the paper have been obtained within the European Commission FP7 ICT-224366 STREP “Development of low-cost technologies for the fabrication of high-performance telecommunica-tion lasers (DeLight)” (www.delightproject.eu).

References [1] Laakso, M. Dumitrescu, "Modified rate equation

model including the photon-photon resonance", Opt. Quantum Electron, 42(11-13), pp. 785–791 , 2011.

[2] L.A. Coldren and S.W. Corzine, “Diode Lasers and Photonic Integrated Circuits”, Wiley, New York, 1995.

[3] J. Viheriälä, M-R. Viljanen, J. Kontio, T. Leinonen, J. Tommila, M. Dumitrescu, T. Niemi, M. Pessa, ”Soft stamp ultraviolet-nanoimprint lithography for fabrication of laser diodes”, J. Micro/Nanolith., 8(3), pp. 033004/1–8, 2009.