Cavity Length Dependence of Wavelength Conversion ... · 236 FUJITSU Sci. Tech. J.,34, 2,(December...

235FUJITSU Sci. Tech. J.,34,2,pp.235-244(December 1998)

UDC 621.373.8: 621.375.826: 621.391.6

Cavity Length Dependence of WavelengthConversion Efficiency of Four-wave Mixingin λ /4-shifted DFB Laser

VTakasi Simoyama VHaruhiko Kuwatsuka VHiroshi Ishikawa(Manuscript received June 17,1998)

Non-degenerate four-wave mixing in a semiconductor DFB laser using its lasing waveas pump waves is a promising method of attaining wide-range and high-bit-rate wave-length conversion in a single device. The structure dependence of the wavelengthconversion efficiency in a lasing λ/4-shifted DFB laser has been analyzed for the firsttime. Systematic calculation of the optical field profile in DFB lasers has shown that astructure with a small grating coupling coefficient κ and a large cavity length L is ahopeful candidate for obtaining a high conversion efficiency. Experiments showedgood agreements with the analytical results. For a structure of κ = 11 cm -1 and L =1,300 µm, a very high conversion efficiency of -5 dB and extremely low noise charac-teristics at a large detuning of 1.6 THz have been attained.

1. IntroductionA wide-range wavelength converter will be

required for the wavelength division multiplexed(WDM) optical communication systems of the nearfuture. Wavelength conversion using non-degen-erate four-wave-mixing (NDFWM) has been stud-ied extensively because of its potential for realiz-ing wide-range and high-bit-rate wavelengthconversion. The NDFWM process can convert fre-quency modulated signals as well as amplitudemodulated signals. NDFWM can also be used asthe fiber dispersion compensater in long-distancetransmission systems1) because this process pro-vides a conjugate, i.e., time reversal output of theinput.

NDFWM in carrier injected semiconductormaterials is an attractive candidate for attaininga high conversion efficiency with a small device(~1 mm) because of its large nonlinear suscepti-bility (over 1 × 10 -16m2/V2 )2) and linear gain, whichamplifies not only converted waves but also inputwaves.

There are two possible devices that can take

full advantage of these materials: semiconductoroptical amplifiers (SOAs) and semiconductor la-sers. From the viewpoint of the conversion effi-ciency, NDFWM in SOAs has already been stud-ied extensively and a conversion efficiency of morethan 0 dB has been accomplished,3)-5) while the la-ser structure is not yet optimized for the NDFWMprocesses. NDFWM in a semiconductor laser, com-pared to NDFWM in an SOA, has the great ad-vantage of practicality. In the case of the laser,the pump wave is provided by the laser itself, sothere is no need to prepare an external pump wavesource. Thus, a wavelength converter consistingof a single device is readily obtained. To makethese devices really applicable, however, they mustachieve a much higher conversion efficiency. Thispaper presents guidelines for optimizing the la-ser structure for NDFWM processes. The guide-lines are based on theoretical analysis and exper-imental results and are designed to optimize theconversion efficiency to noise ratio rather than theefficiency itself.

This paper is organized as follows. In Chap-

236 FUJITSU Sci. Tech. J.,34, 2,(December 1998)

T.Simoyama et al.: Cavity Length Dependence of Wavelength Conversion Efficiency of Four-wave Mixing in λ/4-shifted DFB Laser

ter 2, the features of the NDFWM in a λ/4 -shiftedDFB laser and the kind of characteristics requiredfor this device are explained. In Chapter 3, ana-lytic results are shown. The calculation methodused to estimate the conversion efficiency and thenoise level is explained, and the laser structuredependencies of the conversion efficiency to noiseratio are discussed. Chapter 4 describes experi-ments that were done to check the validity of theanalysis. The experimental results of the conver-sion efficiency versus cavity length are shown, andthe results are compared to the analysis. Chap-ter 5 compares these results with a typical reportof NDFWM in an SOA.

2. Four-wave mixing in λ/4 -shifted DFBlaserA λ/4-phase-shifted distributed feedback

(DFB) laser with anti-reflection (AR) coating onboth facets (Figure 1) is the most appropriatechoice for single device operation of the NDFWMprocess.6),7) Stable lasing waves working as pumpwaves could be generated in the middle of the stopband, as opposed to the uniform corrugation DFBlasers, which require another master laser for sta-bilizing the pump.8) The DFB laser also works as

an SOA for input signal waves or output conju-gate waves whose wavelengths lie outside the stopband. As a result, conversion efficiency is not af-fected by sharp Fabry-Perot resonances, and a widerange of wavelength signals can be converted.

Using this device as a single device operatedwavelength converter, some fundamental experi-ments have already been performed.7) In theseexperiments, a device with a grating coupling co-efficient of 12.2 cm-1 and a cavity length of 900 µmwas used. A relatively high conversion efficiencyof -6 dB at 1.75 nm (220 GHz) detuning betweenthe pump and the signal was obtained. The de-tuning dependence of the conversion efficiency wasmeasured, and wavelength conversion over arange of more than 60 nm was reported (Figure2). The third order optical nonlinear susceptibil-ity χ(3) was estimated, and it was clear that thevalue of χ(3) was determined by the sum of non-linear effects such as the carrier modulation ef-fect, carrier heating effect, and spectral hole burn-ing effect. Also, the saturation characteristicswere investigated and a maximum conjugate pow-er of -7.3 dBm was obtained at an input signalpower of 2.2 dBm. In terms of the noise charac-teristics, a conjugate output to noise ratio of 37dB was demonstrated at an input of 0 dBm and adetuning of 2.5 nm. These results are, for funda-mental research, fairly good and promising. How-

Figure 2Signal wavelength dependence of conversion efficiency.

Figure 1λ/4-phase-shifted DFB laser with AR coated facets.

AR coating

Current

λ/4 phase shift

n-InP substrate

MQW active layer

AR coating

Signal

Signal. Pump. Conjugate

550

to 1

,300

µm

-40

-30

-20

-10

0

1,500 1,520 1,540 1,560 1,580 1,600

Con

vers

ion

effic

ienc

y (d

B)

Signal wavelength (nm)

Lasing wavelength

237FUJITSU Sci. Tech. J.,34,2,(December 1998)


ever, for a practical wavelength converter, they arenot sufficient, especially in terms of the conver-sion efficiency.

Also, a preliminary experiment demonstrat-ing fiber dispersion compensation was performedusing this device.1) The outline of the experimentis as follows. A 10 Gb/s NRZ signal was intro-duced into a single mode optical fiber (SMF). Af-ter 50 km of transmission, the signal was convert-ed into the conjugate by FWM in the λ/4-shiftedDFB laser and then amplified by an erbium dopedfiber amplifier (EDFA). The converted signal wasretransmitted for another 50 km in SMF; thus, atotal of 100 km of optical transmission was com-pleted. The waveform was observed before andafter the 100 km transmission. The eye patternsof these transmissions are shown in Figure 3.After 50 km of transmission, the waveform is dis-torted due to finite chromatic dispersion. But af-

ter an additional 50 km (total 100 km), the conju-gate wave is inversely affected by the chromaticdispersion and the eye pattern is restored[Figure 3(b)]. The result confirms that NDFWMin the DFB laser can be used for dispersion com-pensation.

However, the eye pattern after 100 km issomewhat noisy. Figure 4 shows the bit errorrate (BER) characteristics before and after trans-mission. The power penalty of 2 dB is observed ata 10-11 BER level. This is due to the noise fromthe amplified spontaneous emission (ASE) in thelaser. In carrier injected materials, the genera-tion of ASE noise is an intrinsic problem. Thus, alaser structure which has both a high conversionefficiency and a low ASE level is required.

The conversion efficiency of NDFWM inSOAs or lasers is decided by (1) the absolute val-ue of the third order non-linear susceptibility χ(3),(2) the intensity of pump waves, (3) the intensityof input signal waves, (4) the linear gain which

-35 -30

Minimum receivable power (dBm)

10-7

10-6

10-8

10-9

10-10

10-11

10-12

Bit

erro

r ra

te

After 100 km trans.

Before trans.

Figure 3Eye patterns of 10 Gbit/s NRZ signal: (a) Before trans-mission (b) After transmission.

Figure 4Bit error rate of 10 Gbit/s NRZ signal.

(a)

(b)

80 160 (ps)

160 (ps)

0

800



amplifies input signal waves or output conjugatewaves, and (5) the length for non-linear interac-tion between pump waves and input signal waves.Among those factors, (4) and (5) are obtained dif-ferently in SOAs than in lasers. In SOAs, a larg-er interaction length and a higher linear gain canbe obtained simultaneously by lengthening thedevice, which also improves the conversion effi-ciency. But, this is not the case with NDFWM inlasers. The linear gain in the laser is clamped atthe threshold gain, and the threshold gain is in-versely proportional to the cavity length. Conse-quently, in the case of lasers, it is not obvious whichstructure is optimal for a good conversion efficiency.

Although a large linear gain increases theconversion efficiency, it also increases the ASElevel. In the case of an SOA, it is well known thathigh-power saturated operation with a long de-vice can suppress the ASE. It is also necessary,therefore, to consider the ASE in the case of lasers.

3. Theoretical analysis of conversionefficiency

3.1 Assumption for analysisThe authors made the following assumptions

in the analysis to abstract the essence of theNDFWM process:1) The phase-matching condition between sig-

nal and pump waves propagating in the samedirection is satisfied inside the entire cavity.The coherent length for phase matching is1.8 mm, even in the case of the large wave-length detuning of 40 nm when the pumpwavelength is near 1.55 µm. In this paper,laser structures whose cavity length is lessthan 1.5 mm are considered, thus, this as-sumption is valid.

2) The signal and the conjugate waves outsidethe stop band are not affected by the gratingin the DFB laser waveguide, i.e., the laserworks as though it is a traveling wave SOAfor these waves.

3) The pump and the signal optical field profileare not affected by the pump-signal nonlin-

ear interaction. Under the experimental con-ditions described later, the amplified signaloutput is -15 dB smaller than the pump out-put. The interaction is three orders of mag-nitude smaller than the linear gain and couldbe negligible.

4) The third-order optical nonlinear suscepti-bility χ(3) was assumed to be constant alongthe propagation direction in the cavity. It isthe carrier density profile that makes thelargest contribution to the profile of χ(3). Weconsidered only the case of uniform currentinjection in the entire cavity, so the carrierdensity, which is decided by the quasi-fermi-level, is approximately uniform.

3.2 Calculation of optical field profilein DFB lasersFrom the above assumptions, we can obtain

the profiles of the pump wave and signal waveindependently. Then, the conjugate wave can becalculated in a straight forward fashion.

The electric field profile of the pump wave inthe longitudinal direction evolves according toMaxwell’s wave-equation. In the grating wave-guide, the whole electric field is described as thesum of a pair of waves traveling in opposite direc-tions9) :

Ep(z)=R(z)exp(–iß0z)+S(z)exp(iß0z), (1)

where the z axis is the optical axis in the laserwith z = 0 at one end and z = L (L : cavity length)at the other end, Ep is the electric field of the pumpwaves without the time evolution terms, R (S) isthe envelope function of the optical wave propa-gating in the positive (negative) z-direction, andß0 is the wavenumber of the corrugation. Max-well’s equation and Equation (1) reduce to the cou-pled mode wave equations for R and S as follows:

(2)– dR + [gp –j(ß–ß0)]R = jκe-jΩS dz 2



(3)

where gp is the linear gain, ß is the wavenumberof the lasing wavelength, κ is the coupling coeffi-cient of the corrugation, and Ω is the phase shiftof the corrugation; which is equal to π for the λ/4-shifted DFB laser. Equations (2) and (3) can besolved under the conditions that R (S) is equal tozero at the z = 0 ( z = L ) position and R and S arecontinuous at the phase shift position. The thresh-old gain gth is also calculated at the same time.

The signal wave evolves as in an SOA:

dEs = Γg(ωs) Es, dz 2

(4)

where Γ is the optical confinement factor for theactive layer in the waveguide.

From the assumptions given in the previoussubsection, the development of the conjugate waveis given as:

dEi = Γ[g(ωi) Ei + i 3µ0ε0ωi

2χ(3)(ωi;ωp,ωp,-ωs)R2Es], dz 2 　　　 8ki

(5)

where χ(3) is the third-order optical nonlinear sus-ceptibility.

Figure 5 shows the development of thepump, the signal, and the conjugate wave obtainedfrom the above equations. Three cases are shown:case (a): κL = 0.65, case (b): κL = 1.43, and case(c): κL = 3.90. These figures tell us useful infor-mation about the NDFWM processes in the laser.As the parameter κL becomes larger, the pumpwave’s profile becomes more concentrated at thecenter of the cavity, and, because of the smallerthreshold gain, the signal wave’s growth becomesslower. As a result, for a large κL cavity the con-jugate wave’s growth position is shifted to the cen-ter of the waveguide.

We need to know the ASE level for the signalto noise ratio (SNR) at the conjugate wave’s fre-quency. We can calculate the ASE intensity IASE

from the following equation10) :

dIASE = ΓgIASE + Crhω , dz 　　 4π

(6)

where C is the fraction of spontaneous emissioncoupled into the guide mode, g is the net gain, andr is the radiative recombination rate.

3.3 Dependence of conversionefficiency on the laser structureFrom the method described above, some char-

acteristics, i.e., the threshold gain, wavelength con-version efficiency, and noise level, can be calcu-lated. Systematic calculation of these quantitiesreveals the structural dependence of these char-acteristics.

Figure 5Calculated optical intensity profiles in DFB laser:(a) κL=0.65, (b) κL=1.43, (c) κL=3.90.

dS + [gp –j(ß–ß0)]S = jκejΩR, dz 2

(a)

(c)

(b)

z axis (µm)

z axis (µm)

z axis (µm)

1.0

0.8

0.6

0.4

0.2

0.00

0 400 800 1,200

100 200

× 1,000,000

× 4

× 40

× 100

× 400,000

× 20

300 400 500 600

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

1.2

1.0

0.8

0.6

0.4

0.2

0.0

Inte

nsity

(a.

u.)

Inte

nsity

(a.

u.)

Inte

nsity

(a.

u.)

0 400 800 1,200

Pump(traveling to right)Pump(total)SignalConjugate



κ=11 cm-1,L=550 µm

κ=11 cm-1,L=1,300 µm

κ=30 cm-1,L=1,300 µm



Figure 6 shows the cavity length dependenceof several characteristics when the grating cou-pling coefficient κ is taken to be the relativelysmall value of 11 cm-1. The threshold gain, ASElevel, normalized conversion efficiency, and nor-malized conversion efficiency to noise ratio areplotted. Normalization was done based on thesquare of the maximum pump power along the zaxis to eliminate the effect of differences in pumppower.

From the figure, we can see that the normal-ized conversion efficiency increases as the cavityis lengthened, while the threshold gain becomessmaller. This means that, for structures with asmall κ coefficient, when the interaction length is

made large, the beneficial effect on the conversionefficiency outweighs the negative effects that aredue to the smaller linear gain. In addition, in con-trast to the case in SOAs, the ASE level is lowerfor a longer cavity length. As a result, the conver-sion efficiency to noise ratio, which is the mostimportant characteristic of a wavelength convert-er, becomes very high in a longer cavity device;for example, increasing the length from 550 µmto 1,300 µm improves the ratio by 13 dB.

Next, Figure 7 shows the κ coefficient de-pendence of the characteristics when the cavity istaken to the relatively long length of 1,300 µm.This figure shows that a small κ structure is ad-vantageous in terms of these characteristics. Thisis because the concentration of the pump waveprofile at the cavity center results in the suppres-sion of total nonlinear interaction between thepump and the signal and also because the lineargain is reduced.

Thrershold gainASE levelNorm. conv. effNorm. conv. eff. to noise

30 10-510-17

10-18

10-19

10-6

10-7

10-8

1011

1010

109

108

25

20

15

10

200 400 600 800 1,000 1,200 1,400 1,6005

AS

E le

vel (

W/H

z)

Thr

esho

ld g

ain

(dB

)

Nor

m. c

onve

rsio

n ef

ficie

ncy

(a.u

.)

Nor

m. c

onve

rsio

n ef

ficie

ncy

to n

oise

(a.

u.)

Cavity length (µm)

Figure 6Cavity length dependence of conversion efficiency.

12

10

8

6

4

2

010 15 20 25 30 35 40

AS

E le

vel (

W/H

z)

Thr

esho

ld g

ain

(dB

)

Nor

m. c

onve

rsio

n ef

ficie

ncy(

a.u.

)

Nor

m. c

onve

rsio

n ef

ficie

ncy

to n

oise

(a.

u.)

10-610-18

10-19

1012

1011

1010

10-7

10-8

10-9

Threshold gainASE levelNorm. conv. effNorm. conv. eff. to noise

Grating coupling coefficient (cm-1)

Figure 7Grating coupling coefficient dependence of conversionefficiency.

κ = 11 cm-1

p-clad : InP 50 nm

p-SCH : InGaAsP 100 nm

λg = 1.15 µm

n-SCH : InGaAsP 30 nm

λg = 1.15 µm

n-guide : InGaAsP 70 nm

λg = 1.23 µm

Grating

Sub. : InP

Well : InGaAsP 5.1 nm

0.8% compressed

Barrier : InGaAsP 10 nm

λg = 1.3 µm

10

Figure 8Layer structure of laser.



4. Experimental results4.1 Experimental setup

Based on the findings described in the previ-ous chapter, we fabricated λ/4-phase-shifted DFBlasers with a small κ and various cavity lengths.The wavelength conversion efficiencies and noiseratios were evaluated for these devices.

Figure 8 shows the layer structure of thedevices. The active layer is a multiple quantumwell (MQW) having 10 InGaAsP 0.8% compres-sive quantum well layers and a photoluminescencespectrum with a center wavelength of 1.565 µm.The corrugation was made on an InP substrate,and a separate optical guide layer was made di-rectly on the substrate. Figure 8 also shows var-ious other information, for example, the width andband gap wavelength of each layer. The opticalconfinement to the active layers is estimated tobe 10%, and the coupling coefficient to the grat-ing is estimated to be 11 cm-1. We fabricated thesemi-insulated planar buried hetero (SI-PBH)structure shown in Figure 1. The cleaved facetsare anti-reflection-coated by the TiON/MgF bilay-er. Samples with cavity lengths of 550, 800,1,050, 1,300 µm were made.

The measurement was performed as follows.A uniform current of 100 mA /300 µm was inject-ed into all of the samples except the 1,300 µm sam-ples. The 1,300 µm samples were operated at 70mA/300 µm because they exhibited multimode las-ing at a large injection current. The temperaturewas held at 15˚C. The lasing wavelengths were1.48 to 1.51 µm , and the threshold currents wereabout 10 mA for the 550 µm samples and about20 mA for the 1,300 µm samples. Input signalwaves from a variable wavelength light sourcewere polarization-controlled to maximize the con-version efficiency and coupled into one side facetof the laser. The input wave power was typically40 to 100 µW. The dependence of the conversionefficiency on the input power was negligible be-cause saturation did not occur at this range ofpower. The input wave’s wavelength was adjust-ed to 5, 10, or 20 Trad/s (0.8, 1.6, or 3.2 THz) de-

tunings from the lasing pump wave. Output wavesconsisting of the pump, signal, and conjugatewaves from the other side of the facet were ob-served with an optical spectrum analyzer, and theconversion efficiencies were estimated for eachdetuning.

4.2 Dependence of conversionefficiency on the cavity lengthFigure 9 shows typical measured spectra for

550 µm and 1,300 µm samples. The figures arefor 10 Trad/s detuning between the input signaland the pump. Comparing these two figures, theratio of conjugate wave intensity to ASE level isobviously higher in the longer cavity sample. Aperiodic resonant structure was also seen in thesespectra for wavelengths longer than the pumpwavelength. The periods are equal to the free spec-tral range (FSR) of the laser cavities, so these sam-ples have a Fabry-Perot resonant structure dueto the residual reflection from the AR coated fac-ets. With a small κ, these devices are easily influ-enced by even a small amount of reflectance. Toeliminate the effect of this resonant structure,

Figure 9Spectra of NDFWM in DFB lasers: (a) L=550 µm, (b)L=1,300 µm

0-10

-20

-30

-40

-50

L=550 µm

Conjugate

Pump Signal

∆ω=10 Trad/s

-60

-701,520

Inte

nsity

(dB

)

1,530 1,540 1,550(a)

1,560 1,570 1,580

L=1,300 µm

Conjugate

Pump Signal

-70-60-50-40-30-20-10

0

1,520

Inte

nsity

(dB

)

1,530 1,540 1,550

(b)

Wavelength (nm)

1,560 1,570 1,580



measurements were performed for several signalwaves of slightly different wavelengths, then theefficiency was obtained by averaging the results

of each measurement.Figure 10 shows the cavity length depen-

dence of the wavelength conversion efficiency fordetunings between the pump and signal of 5, 10,and 20 Trad/s (0.8, 1.6, and 3.2 THz). Qualitativetendencies agreed with the analysis described inChapter 3, i.e., the longer the cavity, the higherthe efficiency. A very high conversion efficiencywas observed for the 1,300 µm sample; that is, -5dB at a 5 Trad/s detuning and -9 dB at a 10 Trad/s detuning. Compared to earlier results obtainedfor the κ = 12.2 cm-1 and L = 900 µm device de-scribed in Reference 7), the efficiency is 8 to 9 dBbetter.

The cavity length dependence of the ASE lev-el is shown in Figure 11. Again, the results agreequalitatively with the analysis in Chapter 3, i.e.,the longer the cavity, the lower the ASE level. TheASE of the 550 µm device is five times larger thanthat of the 1,300 µm device. Only a 3 dB changewas expected from the analysis. The differencebetween the analysis and the experiment is dueto the shift of the gain peak wavelength due tothe different threshold current densities of thedifferent cavity lengths.

From the above results, the conversion effi-ciency to noise (ASE) ratio was obtained. Figure12 shows the cavity length dependence of the nor-malized conversion efficiency to noise ratio. Thefigure shows that the ratio is exponentially im-proved as the cavity length is increased. From550 µm to 1,300 µm, there is a 10 dB improve-ment. This value agrees with the value of 13 dBthat was expected from the analysis and showsthat the rather simplified model we used was val-id. The figure also shows that the ratio is slightlyhigher for a 10 Trad/s detuning than for a 5 Trad/s detuning in long-cavity (over 1,000 µm) devices.This is due to the correlation between the peakgain and the signal wave’s wavelength. That is,because a long-cavity device’s peak gain is closerto the signal wave’s wavelength, the linear gain ishigher and consequently signal waves grow morerapidly.

Con

vers

ion

effic

ienc

y (d

B) 0

-5

-10

-15

-20

-25

-30400 600 800 1,000 1,200 1,400

Cavity length (µm)

∆ω=5 Trad/s∆ω=10 Trad/s∆ω=20 Trad/s

Figure 10Experimental results of conversion efficiency for differentcavity lengths.

AS

E le

vel (

W/H

z)

10-16

10-17

10-18

10-19

400 600 800 1,000 1,200 1,400Cavity length (µm)


Figure 11Experimental results of ASE level for different cavitylengths.

Nor

m. C

onve

rsio

n ef

ficie

ncy

to n

oise

(a.

u.)

10-4

10-15

10-16

400 600 800 1,000 1,200 1,400Cavity length (µm)


Figure 12Experimental results of conversion efficiency to noise ra-tio for different cavity lengths.



5. Comparison with SOAsWe will now compare these devices with

SOAs in respect to the conversion efficiency andnoise ratio. A conversion efficiency of +6 dB hasbeen reported for an SOA at a detuning of 1 THz.5)

Although the experimental result of -5 dB at adetuning of 0.8 THz for the 1,300 µm DFB laser isstill inferior to the SOA, there is still some spacefor improvement. In the experiments, the laserwas used with only a 21 mW pump wave. Thereason for this limit was that at higher levels ofcarrier injection, single-mode operation brokedown and Fabry-Perot resonant peaks began tolase. We could achieve a larger pump power byfurther optimizing the facet AR coating or by us-ing the slanted facet technique commonly usedfor SOA devices. If the pump power is improvedby 3 dB, the conversion efficiency may be improvedby 6 dB and over 0 dB of conversion would be re-alized.

In terms of the conversion efficiency to noiseratio, our results are superior to those reported inReference 5) and are some of the best results forNDFWM in a carrier injected semiconductor de-vice. Reference 5) reports a conversion efficiencyof about 0 dB and a conjugate/ASE ratio of 21.1dB at a detuning of 1 THz and a bandwidth of 5.6GHz for a pump power of -2.2 dBm and an inputsignal power of -18 dBm. This ASE level corre-sponds to 4.2 × 10-17 W/Hz, while our results forthe 1,300 µm DFB laser are 2.1 × 10-18 W/Hz at adetuning of 0.8 THz and 1.0 × 10-18 W/Hz at a de-tuning of 1.6 THz. The conversion efficiency toASE ratio in the SOA corresponds to 2.4 × 1016

Hz/W, while the results for the 1,300 µm DFB la-ser are 1.5 × 1017 Hz/W at a detuning of 0.8 THzand 1.6 × 1017 Hz/W at a detuning of 1.6 THz. Thus,the noise level of the DFB laser is 13 to 16 dBlower and the ratio is about 8 dB greater as com-pared to the SOA. These differences between theDFB laser and the SOA are due to the extremelylow ASE noise level of the laser.

6. SummaryThis paper is the first to report on the struc-

tural dependence of wavelength conversion effi-ciency when the NDFWM process is used in a DFBlaser. Some practical assumptions were intro-duced to calculate the optical field profile in λ/4-shifted DFB lasers. Theoretical analysis basedon a simplified model reveals that a structure witha small grating coupling coefficient κ and a longcavity length L can achieve a very high conver-sion efficiency. To examine the validity of thisanalytical result, DFB lasers with a small κ of 11cm-1 and lengths of 550, 800, 1,050, and 1,300 µmwere fabricated. The dependences of the conver-sion efficiency to noise ratio on the cavity lengthobtained from experiments showed good agree-ment with the analysis. For the samples with thelongest cavity length (1,300 µm), a very high con-version efficiency of -5 dB at a wide detuning of1.6 THz and an extremely low noise level of 10-19

to 10-18 W/Hz were observed.These results show that a λ/4-shifted DFB

laser is suitable as a single-device wavelength con-verter. The high conversion efficiency to noiseratio obtained from a laser with a long cavity anda small grating coupling coefficient meets the pre-condition for a high-bit-rate signal converter.

References1) S. Watanabe, H. Kuwatsuka, S. Takeda, and

H. Ishikawa: Polarisation-insensitive wave-length conversion and phase conjugation us-ing bi-directional forward four-wave mixingin a lasing DFB-LD. Electron. Lett., 33,pp.316-317 (1997).

2) M. Yamada: Theoretical analysis of nonlin-ear optical phenomena taking into accountthe beating vibration of the electron densityin semiconductors. J. Appl. Phys., 66, pp.81-89 (1989).

3) A. D’Ottavi, E. Iannone, A. Meccozi, S. Scotti,and P. Spano: 4.3 terahertz four-wave mix-ing spectroscopy of InGaAsP semiconductor



amplifiers. Appl. Phys. Lett., 65, pp.2633-2635(1994).

4) J. Zhou, N. Park, K. J. Vahara, M. A. Newkrik,and B. I. Miller: Four-wave mixing wave-length conversion efficiency in semiconduc-tor traveling-wave amplifiers measured to 65nm of wavelength shift. IEEE. Photon. Tech-nol. Lett., 6, pp.984-987 (1994).

5) A. D’Ottavi, F. Martelli, P. Spano, A. Mecozzi,S. Scotti, R. Dall’Ara, J. Eckner, and G. Guekos:Very high efficiency four-wave mixing in asingle semiconductor traveling-wave ampli-fier. Appl. Phys. Lett., 68, pp.2186-2188 (1996).

6) H. Kuwatsuka, H. Shoji, M. Matsuda, and H.Ishikawa: THz frequency conversion usingnondegenerate four-wave mixing in an In-GaAsP multi quantum well laser. Electron.Lett., 31, 24, pp.2108-2110 (1995).

7) H. Kuwatsuka, H. Shoji, M. Matsuda, and H.Ishikawa: Nondegenerate four-wave mixingin a long-cavity λ/4 -shifted DFB laser usingits lasing beam as pump beams. IEEE J.Quantum Electron., 33, 11, pp.2002-2010(1997).

8) E. Cerboneschi, D.Hennequin, and E.Arimondo: Frequency conversion in externalcavity semiconductor lasers exposed to opti-cal injection. IEEE J. Quantum. Electon, 29,pp.1477-1487 (1996).

9) S. Akiba, M. Usami, and K. Utaka: 1.5- µmλ/4-shifted InGaAsP/InP DFB lasers. J. Light-wave Technol., LT-5, pp.1564-1573 (1987).

10) M. J. Adams, J. V. Collins, and I. D. Henning:Analysis of semiconductor laser optical am-plifiers. in Inst. Elec. Eng., 132, pp.58-63(1985).

Takasi Simoyama received the B.S.and M.S. degrees in Applied Physicsfrom the University of Tokyo, Japan in1993 and 1995, respectively. He joinedFujitsu Laboratories Ltd., Atsugi in 1995,where he has been engaged in the re-search of optical semiconductor devic-es for optical communication systems.He is a member of the Japan Society ofApplied Physics (JSPA).

Hiroshi Ishikawa received the B.S. andM.E. degrees in Electronics from theTokyo Institute of Technology, Tokyo,Japan in 1970 and 1972, respectively.He joined Fujitsu Laboratories Ltd.,Kawasaki in 1972, where he has beenengaged in the research and develop-ment of semiconductor lasers for opti-cal communications. He received theDr. degree from the Tokyo Institute ofTechnology in 1984. He is a member

of the Institute of Electronics, Information and CommunicationEngineers (IEICE) of Japan, the Japan Society of Applied Phys-ics, and the Optical Society of America. Also, he is a seniormember of the Institute of Electrical and Electronics Engineers(IEEE). He received the Young Engineers Award from the IE-ICE in 1976 and the Invention Prize for Encouragement fromthe Japan Institute of Invention and Innovation in 1990.

Haruhiko Kuwatsuka received the B.S.and Dr. degrees in Applied Physics fromthe University of Tokyo, Japan in 1985and 1993, respectively. He joined FujitsuLaboratories Ltd., Atsugi in 1985, wherehe has been engaged in the researchof optical semiconductor devices foroptical communication systems. He isa member of the Japan Society of Ap-plied Physics (JSPA) and the Instituteof Electronics, Information and Commu-

nication Engineers (IEICE) of Japan.

Cavity Length Dependence of Wavelength Conversion ... · 236 FUJITSU Sci. Tech. J.,34, 2,(December...

Documents

Transcript of Cavity Length Dependence of Wavelength Conversion ... · 236 FUJITSU Sci. Tech. J.,34, 2,(December...