852
IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 7, JULY 2004
Strongly Sub-Poissonian Electrical Noise in 1.55-m DBR Tunable Laser Diodes Mikhaël Myara, Philippe Signoret, Member, IEEE, Jean-Philippe Tourrenc, Jean-Philippe Perez, Bernard Orsal, and Joël Jacquet
Abstract—We report terminal electrical noise measurements on 1.55- m DBR tunable laser diodes in the 1 Hz–1 MHz frequency range, performed using an electrical correlation method. These measurements are compared with a comprehensive electrical model based on rate equation formalism. Taking into account diffusion phenomenon and structural parameters, we obtain a complete agreement between the model and the measurements above threshold and a quite similar tendency below threshold. The influence of Bragg section bias is also discussed. Index Terms—DBR tunable lasers, sub-Poissonian noise, terminal electrical noise. Fig. 1. Three-section tunable laser.
I. INTRODUCTION
T
ERMINAL electrical noise (TEN) is a key diagnostic tool for analyzing the technological characteristics of semiconductor laser diodes. Moreover, such experimental investigation allows feedback-free measurements of optical parameters [1]. In this letter, we present 1 Hz–1 MHz TEN characterization applied to key components in optical communications: multisection integrated DBR tunable laser diodes emitting in the range of 1.55 m. All of the experimental results are compared to a comprehensive theoretical analysis, based on the Harder and Yariv [2] formalism. This model is here completed by a set of excess noise sources in order to explain the laser behavior below and far above threshold. Below threshold, the model takes into account the carrier diffusion noise [3], whereas far above threshold, structural parameters are to be considered. We also discuss the influence of Bragg section bias on TEN results, especially around threshold. II. TUNABLE DBR LASERS The most used tunable laser in communication systems is the two- or three-section DBR laser, which structure is shown in Fig. 1 [4]. In such a laser structure, each section has a specific influence on the emitted beam. We have to take into account two kinds of sections: on the one hand, the active section provides the optical gain thanks to quantum wells; on the other hand, the tunability of the laser is relevant to both “passive” sections. Manuscript received January 5, 2004; revised March 10, 2004. M. Myara, P. Signoret, J.-P. Tourrenc, J.-P. Perez, and B. Orsal are with CEM2, Universite Montpellier II, 34095 Montpellier Cedex 5, France (e-mail:
[email protected]). J. Jacquet is with Alcatel/Opto+ ”Research and Innovation,” 91460 Marcoussis, France. Digital Object Identifier 10.1109/JQE.2004.830203
Fig. 2.
Bragg section effects.
Both passive sections use the fact that carrier injection into a semiconductor produces an increase of the absorption and thus a decrease of optical index . As long as the prevalent phenomenon in the considered section is carrier injection (so that thermal effects can be ignored), both and changes follow a passive bias current square root rule [5] because of intervalence and band absorption (IVBA) effects: . The Bragg section filter—designed to “filter” the Fabry–Perot comb—will be simply shifted under the effect of in Fig. 2). carrier injection ( This principle allows us to select another Fabry–Perot wavelength. As a consequence, the tunability obtained with the Bragg section is discontinuous, with an overall tunability of about 15 nm around an average emission wavelength of 1.555 m. III. THEORY A. Laser Cavity Model In this section, we describe an electrical noise model of the laser cavity itself. 1) Rate Equation Model: We build a monomode corpuscular rate-equation model of the tunable laser diode. This rate equation system is given as follows: (1) is the active section bias, is the carrier density where in the active section quantum wells, is the photon density in
0018-9197/04$20.00 © 2004 IEEE
MYARA et al.: STRONGLY SUB-POISSONIAN ELECTRICAL NOISE IN 1.55- m DBR TUNABLE LASER DIODES
853
TABLE I SIMULATION PARAMETERS
Fig. 3. Equivalent electrical circuit.
(7) where is the fluctuating diode voltage, and is the stimulated emission branch current (Fig. 3). Introducing (6) and (7) into (1) written in pure dynamic mode, and neglecting the electrical pump fluctuations , we obtain
the active optical guide, is the classical carrier recombi), and and nation effect ( are the Langevin sources causing the noise inside the laser strongly depends on the passive cavity. The photon lifetime section carrier injection and can be described by the following equation: (2) where is the IVBA-phase-section absorption due to is the DBR reflectivity phase current injection, and for a stop-band-centered mode: it is numerically evaluated thanks to Orfanos’ methods [6], [7], taking into account the IVBA-originated absorption excess. The main DBR feature is . All other useful parameters are given in Table I, the values of which have been evaluated by fitting optical and electrical static-experimental measurements. 2) Langevin Sources: In the following, we write each fluctuating variable as a sum of a static and a dynamic part. For example, with the carrier density, we have
(8)
Then, following the Harder and Yariv treatment [2], we can deduce the equivalent electrical circuit given in Fig. 3 with the component values given by (9) (10) (11) (12) and noise sources
(3) The Langevin source spectral densities are evaluated thanks to a McCumber treatment [8], counting the appearance and disappearance of photons and carriers inside the laser cavity. Neglecting the cross-population terms, we can establish the expression of both Langevin source spectral densities as follows:
(13) (14) The overall voltage noise is obtained thanks to Millman’s theorem applied to the equivalent electrical circuit. The voltage noise spectral density is transformed into current noise spectral density thanks to the dynamic resistance as follows:
(4) (15) (5) 3) Harder and Yariv Electrical Model: In order to build an equivalent electrical model of the laser diode, we now use the following expressions [9], [2], [10]: (6)
The dyamic resistance electrical circuit.
is computed thanks to the equivalent
B. Excess Noise Sources In this section, we describe several models of the laser cavity excess noise sources.
854
IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 7, JULY 2004
Fig. 4. Diffusion inside a double heterostructure.
1) Subthreshold Diffusion Model: We use a theoretical relation based on Haug’s model [3] giving the current noise spectral density of the laser junction [11] as
Fig. 5.
Bulk, interface, and cavity noise effects.
(16) is the minority carrier diffusion length, and is the where active layer thickness. This noise source takes into account the shot-noise source and the diffusion phenomenon inside the laser cavity, due to the double heterostructure (Fig. 4). The diffusion length is much longer (several micrometers) nm), but the diffuthan the active layer thickness ( sion process is limited due to the barriers of the aforementioned double heterostructure. Therefore, the actual diffusion length is reduced to . Equation (16) can thus be simplified in the following way: (17) We only consider the electron diffusion effects because of the low mobility of the holes in the InGaAsP material which constitutes the active layer. Moreover, such a diffusion behavior cannot be observed above threshold: any pump-originating carrier entering in the cavity is involved in the stimulated light emission mechanism and so cannot be submitted to diffusion laws. 2) Series and SCLC Effects: Static measurements have brought to the fore a space charge limited current (SCLC) effect above threshold [12], [13]. This effect is classically explained by low barrier crossing phenomenon: a p -p junction for holes and p -p junction for electrons (Fig. 4), thanks to Mott–Gurney’s theory [14]. This effect induces a carrier transport limitation and thus implies an overall resistive behavior. This has been here empirically modeled by a decrease of the above-threshold dynamic resistance. In terms of induced fluctuations, the SCLC process could exhibit two different signatures: a current dependent one, described by Van der Ziel in [15], or a thermal one. The following noise measurements prove that, in our case, the only significant contribution is the thermal one. Therefore, we can model both series and SCLC noise sources as a global Johnson term as follows: (18) 3) Overall Equivalent Circuit: Taking into account all of the noise source contributions, we can establish a new electrical equivalent circuit as shown in Fig. 5. This equivalent circuit will be the model for all the following measurement-to-theory comparisons.
Fig. 6. Very low noise measurements setup.
IV. MEASUREMENT SETUP A. Scheme We give in Fig. 6 the electrical noise measurement scheme based on a correlation system using a set of two 1-Hz–1-MHz bandwidth amplifiers. Such a scheme allows us to remove the measurement limitation due to amplifier intrinsic noise. Indeed we consider the incoming signal in each fast Fourier transform (FFT) analyzer and , respectively, as the sum of the device channel and the background amunder test (DUT) noise term and , respectively. The cross-correlaplifier noise tion product can be written as follows: Corr
(19) assuming that Corr
,
, and
are uncorrelated: (20)
The FFT analyzer then computes the FFT of this cross-correlation product. Eventually, the displayed signal corresponds to the DUT power spectral density. B. Calibration We show in Fig. 7 the calibration of the correlation system. The given noise curves are averaged over 256 measurements. The standard deviation increase when thermal noise decreases is due to the deviation between the amplifier noise
MYARA et al.: STRONGLY SUB-POISSONIAN ELECTRICAL NOISE IN 1.55- m DBR TUNABLE LASER DIODES
Fig. 7. Cross-correlation measurement calibration.
Fig. 9. Electrical noise results for I
855
= 0 mA and I
= 0 mA.
Fig. 8. Several measured spectra.
and the DUT noise level: the lower the device noise is, the longer the integration time must be [16], [17]. With such a measurement system, we can measure noise levels down to the thermal noise produced by a 1.6- resistance. This nonzero value is relevant to two main effects: leakage currents between both amplifiers and FFT analyzer channel crosstalk. mA. Fig. 8 shows three representative spectra for mA), the noise spec• Below laser threshold ( noise level. The white noise trum shows a very strong level is then obtained by difference between the measured asymptotic slope, assuming that data at 1 MHz and a contrithe noise spectrum is only made of white and butions. mA), the white noise level • At threshold ( measurement is easily measured from 100 kHz to 1 MHz. mA), the white noise • Above threshold ( level is lower than the amplifier noise and therefore completely justifies the use of a correlation system. To be more accurate, the real white noise has been extrapolated from noise added to white noise. a fitting considering V. COMPARISON TO THEORY A. Null Passive Section Bias We plotted in Fig. 9 the white electrical current noise power spectral density as a function of the active section bias. • Below threshold, the model and the measurements exhibit a quite similar tendency: the noise level is proportional
Fig. 10.
Electrical noise results for I
= 50 mA and I
= 0 mA.
), but a bit lower than the to a shot noise level ( . expected • Around threshold, the model and the measurements are in good agreement. Such an agreement shows that the fitting parameters, particularly the -value, were well estimated during the preliminary static measurements. • Above threshold, the measurement and the model both follow the series resistor and SCLC diode thermal noise sources and exhibit very subshot noise levels (about a biases). This result also shows decade for strong that the SCLC diode follows here a thermal resistor behavior and not a barrier crossing statistic. B. Strong Bragg Section Bias We plot in Fig. 10 the electrical current noise power spectral density as a function of the active section bias when Bragg section biased. We observe that the reported noise levels behave exactly the same way for both subthreshold and above-threshold parts. We nevertheless notice a difference around threshold: the measured noise level does not follow the model predictions. This phenomenon is intrinsically linked to a mode-hopping phenomenon appearing around the laser threshold for such structures, because of the spontaneous emission in the Bragg section [18]. The reported noise level is in fact the flat part of a Lorentzian shaped noise spectrum: the white noise level is indeed not acces-
856
IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 40, NO. 7, JULY 2004
sible in this active section bias range because of the Lorentzian spectrum width, with a cutoff frequency higher than 10 MHz (so above the available analysis frequency range), as reported by Ohtsu et al. [19].
VI. CONCLUSION In summary, we proposed a theoretical model for the TEN in multisection integrated tunable DBR laser diodes. This rate-equation-based formalism also takes into account the below threshold diffusion and the above threshold structural induced processes. TEN measurements in the 1-Hz–1-MHz TEN range were carried out using a correlation system in order to get rid of the amplifier intrinsic noise. Below threshold, quite a good agreement between model and measurements shows the carrier diffusion noise predominance, level is not reached. even if the theoretical Around threshold, the Harder and Yariv’s model validity is completely confirmed in a null passive injection case. When a Bragg section is biased, we noticed a discrepancy between the model and the measurements. It is explained by a Lorentzian shaped mode-hopping noise with a cutoff frequency far above the analysis bandwidth. Above threshold, we turned into light a strong sub-Poissonian noise level in perfect concordance with theoretical simulations. It is also shown that the SCLC diode noise behavior is mainly driven by thermal effects instead of a common junction shot noise.
REFERENCES [1] P. A. Andrekson, P. Andersson, A. Alping, and S. T. Eng, “In situ characterization of laser diodes from wideband electrical noise measurements,” J. Lightwave Technol., vol. LT-4, pp. 804–812, July 1986. [2] C. Harder, J. Katz, S. Margalit, J. Shacham, and A. Yariv, “Noise equivalent circuit of a semiconductor laser diode,” IEEE J. Quantum Electron., vol. QE-18, pp. 333–337, Mar. 1982. [3] H. Haug, “Population and current noise in semiconductor laser junctions,” Z. Phys., vol. 173, no. 206, pp. 163–176, 1967. [4] L. Coldren, “Monolithic tunable diode lasers,” IEEE J. Select. Topics Quantum Electron., vol. 6, pp. 988–999, Nov./Dec. 2000. [5] P. Brosson, C. Labourie, L. L. Gouezigou, J. Lievin, J. Jacquet, F. Leblond, A. Olivier, and D. Leclerc, “Experimental determination of carrier-induced differential loss in 2-section GaInAsP/InP laser-waveguide,” Electron. Lett., vol. 25, no. 24, Nov. 1989. [6] I. Orfanos, T. Sphicopoulos, A. Tsigopoulos, and C. Caroubalos, “A tractable above-threshold model for the design of DFB and phase-shifted DFB lasers,” IEEE J. Quantum Electron., vol. 27, pp. 946–956, Apr. 1991. [7] A. Tsigopoulos, T. Sphicopoulos, I. Orfanos, and S. Pantelis, “Wavelength tuning analysis and spectral characteristics of three-section dbr lasers,” IEEE J. Quantum Electron., vol. 28, pp. 415–426, Feb. 1992. [8] D. E. McCumber, “Intensity fluctuations in the output of cw laser oscillators. I,” Phys. Rev., vol. 141, no. 1, pp. 306–321, Jan. 1966. [9] M. Morishita, T. Ohmi, and J. Nishizawa, “Impedance characteristics of double-heterostructure laser diodes,” Solid-State Electron., vol. 22, pp. 951–962, 1979. [10] J. Katz, S. Margalit, C. Harder, D. Wilt, and A. Yariv, “The intrinsic electrical equivalent circuit of a laser diode,” IEEE J. Quantum Electron., vol. QE-17, pp. 4–7, Jan. 1981. [11] B. Orsal, P. Signoret, J. Peransin, K. Daulasim, and R. Alabedra, “Correlation between electrical and optical photocurrent noises in semiconductor laser diodes,” IEEE Trans. Electron Devices, vol. 41, pp. 2151–2161, Nov. 1994.
[12] M. Wintrebert-Fouquet and B. Orsal, “Temperature investigation of dark current and its electrical noise in GaAs/AlGaAs multiquantum well photodiodes,” J. Appl. Phys., vol. 85, no. 2, pp. 1211–1215, Jan. 1999. [13] V. Arkhipov, H. von Seggern, and E. Emelianova, “Charge injection versus space-charge-limited current in organic light-emitting diodes,” Appl. Phys. Lett., vol. 83, no. 24, pp. 5074–5076, Dec. 2003. [14] M. Lambert and P. Mark, Current Injection in Solides. New York: Academic, 1970. [15] A. V. der Ziel, Noise in Solid State Devices and Circuits. New York: Wiley, 1986. [16] M. Sampietro, L. Fasoli, and G. Ferrari, “Spectrum analyzer with noise reduction by cross-correlation technique on two channels,” Rev. Sci. Instrum., vol. 70, no. 5, pp. 2520–2525, May 1999. [17] M. Sampietro, G. Accomando, L. Fasoli, G. Ferrari, and E. C. Gatti, “High sensivity noise measurement with a correlation spectrum analyzer,” IEEE Trans. Instrum. Meas., vol. 49, pp. 820–822, Aug. 2000. [18] M. Myara, P. Signoret, J.-P. Tourrenc, J. Jacquet, B. Orsal, and R. Alabedra, “Electrical noise power spectrum behavior around threshold in DBR 2 and 3 section tunable lasers emitting arount 1.55 m,” Proc. SPIE, vol. 5111, pp. 498–505, June 2003. [19] M. Ohtsu et al., “Precise measurements and computer simulations of mode-hopping phenomena in semiconductor laser,” Appl. Phys. Lett., vol. 46, no. 2, pp. 108–110, Jan. 1985.
Mikhaël Myara was born in Montreuil-sous-Bois, France, on November 2, 1975. He received the engineer degree in applied computer science and optoelectronics and the Ph.D. degree in optoelectronics from Montpellier University, Montpellier, France, in 1998 and 2003, respectively. He is currently with the Centre d’ Electronique et de MicroOptoElectronique de Montpellier, where he is mainly involved in research on the study, through the amplitude noise and the linewidth, of multielectrode DBR tunable lasers for long-haul optical-fiber transmission systems.
Philippe Signoret (M’03) was born in Marseille, France, on December 29, 1965. He received the engineer degree in radioelectricity and electronics from the Institut National Polytechnique de Grenoble, Grenoble, France, in 1989 and the Ph.D. degree in optoelectronics from Montpellier University, Montpellier, France, in 1994. He is currently with the Centre d’ Electronique et de MicroOptoElectronique de Montpellier, where he is engaged in research on the characterization by amplitude and frequency noise measurements of optotelectronics devices for optical-fiber transmission systems. He is especially involved in the study of DBR tunable lasers and vertical-cavity surface-emitting lasers.
Jean-Philippe Tourrenc was born in Mende, Lozère, France, on May 18, 1978. He received the engineer degree in radioelectricity and electronics from the Institut National Polytechnique de Grenoble, Grenoble, France, in 2001. He is currently working toward the Ph.D. degree in optoelectronics at the Centre d’ Electronique et de MicroOptoElectronique de Montpellier, Montpellier, France. His main research interest is the frequency noise study of vertical-cavity surface-emitting lasers.
Jean-Philippe Perez was born in Montpellier, France, on December 26, 1972. He received the M.S. degree in physics from Montpellier University, Montpellier, France, in 1995. He is currently working toward the Ph.D. degree in optoelectronics at the Centre d’Electronique et de MicroOptoélectronique de Montpellier Laboratory. His main research interest is the study of HgCdTe photodiodes.
MYARA et al.: STRONGLY SUB-POISSONIAN ELECTRICAL NOISE IN 1.55- m DBR TUNABLE LASER DIODES
Bernard Orsal was born in Espalion, France, on June 17, 1953. He received the engineer degree in nuclear electrical engineering from the Institut National Polytechnique de Grenoble, Grenoble, France, in 1980 and the third cycle Doctorate degree and the state Doctorate degree in physics from Montpellier University, Montpellier, France, in 1982 and 1986, respectively. In 1982, he joined the Centre d’Electronique de Montpellier. His research includes the characterization by noise measurements of optoelectronic devices for optical fiber transmission systems, lasers, photodiodes, and optical amplifiers.
857
Joël Jacquet was born in Courbebief, France, on January 12, 1962. He received the engineer diploma from Ecole Nationale Supérieure de Physique de Marseille, Marseille, France, and the Ph.D. degree from Ecole Nationale Superieure des Telecommunications, Paris, France, in 1986 and 1992, respectively. In 1986, he joined Laboratoires de Marcoussis, Marcoussis, France, where his field of interest was 1.5-m semiconductor tunable lasers for coherent application within the European Race 1027 project. From 1992 to 1994, he was involved in research on tunable wavelength converters for optical switching in the Photonic Component Division within the European Race ATMOS project. As a Technical Adviser, he participated to the industrial transfer of high-speed 1.55-m DFB lasers to the fabrication unit of Alcatel in 1994. He was leading the Vertical Cavity Surface Emitting Laser project in the Corporate Research Center at Laboratoires de Marcoussis and was the coordinator of the ACTS VERTICAL project. He has contributed to more than 80 papers in international journals or conferences and holds 10 patents. He is currently the Pump & WDM Lasers Group leader at OPTO+, Alcatel Research & Innovation, Marcoussis, France.
