The P 2 O 5 /F combination allows a reasonably high concentration of P2O5 to be used which, besides giving a low deposition temperature, helps to reduce considerably diffusion of OH from the substrate tube. In the absence of fluorine, the P 2 O 5 would give an increase in refractive index of 000085 relative to that of silica. References 1 AINSUE, B. J., DAY, C. R., FRANCE, P. W., BEALES, K. J., a n d NEWNS, G. R.: 'Preparation of long lengths of ultra-low-loss single-mode fibre', Electron. Lett., 1979, 15, pp. 411-413 2 EDAHIRO, T., HORIGUCHI, M., CHIDA, K., and OHMORI, Y.: 'Spectral loss characteristics of GeO 2 -P 2 O 5 -doped silica graded index fibres in long wavelength band', ibid., 1979, 15, pp. 274-275 3 mode cutoff region 4 5 SENTSUI, S., FURUI, Y., YOSH1DA, K., KUROHA, T., a n d HOSAKA, T . : 'Low loss monomode fibres with P 2 O s -SiO 2 cladding in the wavelength region 1-2—1-6 fim'. Fifth European conference on optical communication, Amsterdam, Sept. 1979 GLOGE, D.: 'Weakly guiding fibers', Appl. Opt., 1971, 10, pp. 2252-2258 ARONSON, B. s., POWERS, D. R., and SOMMER, R. G.: 'Chlorine drying of a doped deposited silica preform simultaneous with consolidation'. Topical meeting on optical fibre communication (Optical Society of America), Washington, 1979, p. 54 03 6 1000 1200 1400 wavelength, nm 1600 1800 Fig. 2 Loss spectra for three lengths of single-mode fibre, measured using tungsten-lamp/monochromator source and germanium detector 171 km 125 km 9-8 km In Table 1 we have summarised results for three lengths of fibre pulled from three consecutive sleeved preforms. The quantity of material deposited was adjusted in each case to allow for the sleeving tubes to be used, in order to achieve the desired mode cutoff position. The 'start' ends of the preforms were fused into the sleeving tubes and the composite structure was collapsed and drawn into fibre in one operation. The loss spectra for the three fibres are shown in Fig. 2, and excellent reproducibility of loss was achieved at 1-3 /zm and at 1-6 fim. The estimated OH content of the fibres lies in the range 30 to 100 parts in 109, inferred from the peaks at 1-25 fim in long fibre lengths and from measurements at 1-39 /itn made on 1 km lengths of fibre. This low OH level was achieved by using chlorine as a drying agent during the collapse of the preform tube. The use of chlorine or thionyl chloride as drying agents is well known in the o.v.p.o.5 and v.a.d.6 processes. Pearson (private communication) has observed a similar effect in the m.c.v.d. process by continuing the flow of GeCl4 during the collapse of the substrate tube. In the m.c.v.d. process there is always chlorine present during the deposition owing to the decomposition of SiCl4 and other halides. At the collapse stage, however, the halide flow is normally stopped and if no Cl2 is introduced a high OH concentration can appear at the centre of the fibre due to the OH present in the carrier gas or by back-diffusion. This is particularly important in monomode fibre where the power is concentrated at the centre of the fibre. The intensity of the OH overtone at 1-39 /jm has been reduced to < 3 dB/km, and further reduction should be possible. The monochromatic pulse broadening of the 171 km length of fibre was measured using a Nd:y.a.g. laser at 1-32 /un. No broadening due to birefringence effects was detectable. Conclusion: An optimised structure has been evolved to permit long lengths of single-mode fibre to be produced with losses of 0-6 dB/km at 1-3 pm and 0-34 dB/km at 1-6 nm. The OH impurity level has been reduced to 30-100 parts in 109 by using chlorine as a drying agent in the m.c.v.d. process. Acknowledgments: The authors wish to thank W. A. Merlo and D. A. Colthorpe for technical assistance, J. V. Wright for help in computing the single-mode power profiles, M. C. Brierley and K. I. White for refractive-index profile measurements, B. P. Nelson for dispersion measurements, and the Director of British Telecom Laboratories for permission to publish this work. B. J. AINSLIE C. R. DAY J. RUSH K. J. BEALES British Telecom Laboratories Martlesham Heath, Ipswich, Suffolk, England 10th July 1980 ELECTRONICS LETTERS 28th August 1980 SUDO, S., KAWACHI, M., EDAHIRO, T., IZAWA, T , SHIODA, T., a n d GOTOH, H.: 'Low OH content optical fibre fabricated by vapourphase axial-deposition method', Electron. Lett., 1978, 14, pp. 534-535 0013-5194/80/180692-O2$U0/0 EFFECT OF p-DOPING ON CARRIER LIFETIME AND THRESHOLD CURRENT DENSITY OF 1-3 jim GalnAsP/lnP LASERS BY LIQUID-PHASE EPITAXY Indexing terms: Doping, Transport processes, Lasers A correlation was found between the variation of the threshold current density and carrier lifetime with acceptor concentration in the active layer. An injected electron concentration of 2-5-3 x 10 18 /cm 3 , independent of the acceptor concentration in the active layer, was found at the lasing threshold. GalnAsP-InP double heterostructure (d.h.) lasers are considered as the most promising light source for optical communication between 10 and 1-55 nm. The dependence of their threshold current density on active layer thickness has been reported previously.1-2 Recently, Itaya et a/.3 reported that, for a given active layer thickness i-the threshold current density Jth of the lasers was a sensitive function of the p (zinc) doping in the active layer. However, the reason for the dependence was not given. In this letter, we report the variation of the spontaneous lifetime of injected carriers at the lasing threshold as a function of the acceptor concentration in the active layer. The increase of carrier lifetime rs with decrease in p-doping will be shown to be the major cause of the reduction in Jth, with effects such as free carrier absorption playing a minor role. The d.h. lasers were grown by liquid-phase epitaxy at a temperature of ~ 65O°C using the single phase technique. They consist of a heavily Sn-doped (> 1018/cm3) buffer layer, an unintentionally doped GalnAsP active layer (015 to 0-3 /im), a zinc-doped p-InP confinement layer, and a p+-quaternary cap layer for good ohmic contact. The quaternary layers are lattice-matched to the n+-InP substrate (Aa/a ~ 12 x 10 ~ 4 ). To prevent cross contamination of the melts by zinc,4 all the wells of the boat were covered by graphite caps during the growth. The wafers were subsequently processed into oxidedefined stripe lasers with stripe widths between 10 ^m and 100 nm. Using electron-beam-induced current techniques, we found that, for the doping range of 1017/cm3 to 2 x 1018/cm3, the p-n junction was located at the interface of the quaternary active layer and the n + -InP buffer layer. The active layer was thus unintentionally doped by the diffusion of zinc from the p-InP confining layer during the growth. This is in good agreement with Reference 5, in which the p-doping was determined to be uniform in the active and p-InP confining layer. Vol. 16 No. 18 693 Fig. 1 shows the variation of the threshold current density Jth as a function of p-doping. The doping concentration was determined from the mole fraction of zinc in the In melt.6 The value of Jth was evaluated from the threshold current of lasers with cavity lengths of ~ 300 /im and stripe widths of more than 30 /im. As shown, Jlh was reduced by approximately a factor of two as the p-doping was varied from 1018/cm3 to 1017/cm3. The lowest threshold current density achieved was ~ 1-5 kA/cm2 for devices with an active layer thickness of 015 /im and an acceptor concentration of ~l-5 x 1017 cm" 3 . To understand the above observation, the spontaneous carrier lifetime of the d.h. lasers at threshold was determined by measuring their 'turn-on' delay time. The diodes were Insoldered p-side down on copper heat sinks which were then mounted at the end of a stripline. The rise time of the package and driving pulse were estimated to be ~ 0-5 ns. The optical output was detected with a GaAlAsSb/GaSb avalanche photodiode with a rise time of 100 ps or less. The turn-on delay time td was plotted against In {//(/ — /,,,)} in Fig. 2 for three devices, each with a different p-doping in the active layer. To avoid complications caused by the appearance of filaments,7 only lasers with stripe widths of 10 /im were used in the above experiment. The spontaneous carrier lifetime T, of the lasers at threshold was estimated by the well known relation t s = tjln {1/(1 - /,„)}• ! 3 a, 2 li 0 KX) 200 300 current ,mA 13100 13116 V3132 13148 V316A wavelength, jjm wavelength, jjm 13172 rzoeTJI Fig. 3 Emission spectrum and power/current characteristics of typical device during c.w. operation = 03jjm I = HI/,,,; T = 23°C; stripe width = 22 /im; /,„ (pulsed) = 185 mA;/ r t (c.w.) = 200 mA £3 O d= 0-15 pm 10 2 3 A 5 6 7 8 9 P-doping 10 18 fco^HI Fig. 1 Variation of threshold cWrrent density with acceptor concentration in active layer (thickness d) A d = 0-3 fim • d = 0-2 /im O rf = 0-15/im H122 r s = 2-6 ns 7 //V H153 / // ^r s = 4ns rs r 4-1 ns 5 10 delay time t d , ns Fig. 2 Plot of In {1/(1 — Ith)} against 'turn-on' delay time of lasers O O A 694 H122; p = 1-2 x 10 18 , s = 12 /im, d = 0-3 /im, T, = 2-6 ns H177;p = 2 x 10 17 , s= 12 /im, d = 0 1 5 /im, TS = 4 1 ns H153; p = 3 x 10 17 , s = 13 /im, d = 0-3 /im, t , = 4 ns For devices with p-doping of 1-2 x 1018 cm" 3 , the spontaneous carrier lifetime at threshold (Jth ~ 10 kA/cm2, s = 10 /im, d = 0-3 /im) was estimated to be 2-6-3 ns. This is to be compared with a value of ~ 4 ns obtained for devices with a p-doping of ~ 3 x 1017 cm" 3 (Jth = 5-55 kA/cm2, s = 10 /im, d = 0-3 /im). The number of electrons injected into the active layer to achieve lasing can be easily estimated by using the relation nlh = Jthxs/qd. We found that, independent of the original p-doping, nth is ~ 2-5-3 x 1018 cm" 3 , after taking into account that the threshold current densities given above were approximately twice that of broad-area devices because of current spreading. Similar conclusions were reached by Hwang and Dyment8 for GaAs-GaAiAs lasers. Since TS =* l/(Beffpth), where pth is the total number of holes in the active layer at threshold, and Beff is the effective recombination constant,8 the ratio of the carrier lifetimes for devices with different p-doping NA in the active layer can be reestimated once nth is known. Taking pth = nth + NA for charge neutrality, a ratio of ~ 4/3 is found between lasers with NA ~ 1018/cm3 and those with NA =* 1017/cm3. This is in good agreement with our experimental results given above. A value of Bejf — 8-3 x 1 0 " ' ! cm3/s is also obtained. While the reduction in p-doping of the confining layer reduces the pulsed threshold current density at room temperature, we found that To, where J,h(T) oc exp (T/To\ is also affected significantly by the change in doping. The effect of this variation on their c.w. operation range will be reported in a separate publication. A p-doping of ~ 4-5 x 1017/cm3 is found to be optimum for c.w. operation when both the threshold current density and its temperature dependence is taken into account. Fig. 3 shows the emission spectrum of a d.h. laser with a stripe width of 22 /tm. A single longitudinal mode was observed between 106 to 117 times the threshold current. During c.w. operation at room temperature, a typical differential quantum efficiency of 20-24% per facet was obtained. C.W. operation has been achieved for devices of this stripe width at heat sink temperatures of ~ 63°C. In conclusion, a good correlation was obtained between the variation of the spontaneous carrier lifetime and threshold current density with the acceptor concentration in the active layer. At threshold, the injected electron concentration was ~ 2-5-3 x 1018 cm" 3 . D.H. lasers with lower p-doping were found to have longer spontaneous carrier lifetimes. They were therefore able to generate more gain for the same injection current. This, in turn, leads to a decrease of the threshold current density at room temperature. W. NG 17th July 1980 Y. Z. LIU Rockwell International/Electronics Research Center 1049 Camino Dos Rios, Thousand Oaks, Ca. 91360, USA ELECTRONICS LETTERS 28th August 1980 Vol. 16 No. 18 References 1 NAHORY, R. E., and POLLACK, M. A. : Threshold dependence on active layer thickness in InGaAsP/InP d.h. lasers', Electron. Lett., 1978, 14, pp. 727-729 2 YANO, M., NISHI, H-, and TAKUSAGAWA, M.: Theoretical and exper- imental study of threshold characteristics in InGaAsP/InP DH lasers', IEEE J. Quantum Electron., 1979, 15, pp. 571-579 3 ITAYA, Y., SUEMATSU, Y., KATAYAMA, S., KISHINO, K., a n d ARAI, S.: 'Low threshold current density (100) GalnAsP/InP DH lasers for wavelength 1-3 nm\ Japan. J. Appl. Phys., 1979, 18, pp. 1795-1805 4 COLEMAN, J.}., and NASH, F. R.: 'Zinc contamination and misplaced p-n junctions in InP-GalnAsP d.h. lasers', Electron. Lett., 1978, 14, pp. 558-559 5 ROSENTHAL, A., ITAYA, Y., and SUEMATSU, Y.: 'Measurement of Zn doping level in InGaAsP/InP DH lasers', Japan. J. Appl. Phys., 1979, 17, pp. 1655-1656 6 L P n mode cutoff. The cutoff wavelength determined by this technique represents the local (2-3 mm offibrelength) physical guide parameters and unlike the techniques using transmitted power it is not affected by the fibre length and the associated uncertainties. Experiment: The experimental set-up, shown in Fig. 2, includes a tungsten lamp/monochromator combination for variable wavelength source, the output of which is focused to a 5-6 /im spot on the fibre end. The normalised refracted power as a function of wavelength is determined by comparing the refracted power under two conditions: (i) focused spot on core, and (ii) focused spot on cladding. For (i), good results are obtained when the focused spot is less than or equal to the core ABRAMS, E. B., SUMSKI, S., B O N N E R , W . A., a n d COLEMAN, J. J.: ' B e doping of liquid-phase-epitaxial InP\ J. Appl. Phys., 1979, 50, pp. 4469-4470 7 DYMENT, J. c , RIPPER, J. E, and LEE, T. p.: 'Measurement and inter- pretation of long spontaneous lifetimes in DH lasers', ibid., 1972, 43, pp. 452-457 8 HWANG, c. J., and DYMENT, J. C : 'Dependence of threshold and electron lifetimes on acceptor concentration in GaAs-Ga, _ , Al^ As laser', ibid., 1973, 44, pp. 3240-3244 2 0013/5194/80/180693-03$!.50/0 Si u z \ <b Ifl \ > c REFRACTED POWER TECHNIQUE FOR CUTOFF WAVELENGTH MEASUREMENT IN SINGLE-MODE WAVEGUIDES Indexing terms: Optical fibres, Waveguides A technique to measure the local (2-3 mm) values of cutoff wavelength Xc in single-mode waveguides is described. The technique, insensitive to the length of waveguide used, involves the spectral measurement of refracted power. The feasibility of using the cutoff wavelength determined by this technique to predict the zero dispersion wavelength is also studied. Introduction: The determination of the cutoff wavelength for the LPn mode is important for the design and verification of the different models developed for single-mode fibres. Various techniques1 3 have been developed to determifle this important parameter. Most of the presently available techniques (e.g. far-field,1 near-field, and bending methods2)involve the use of transmitted power and the results have been shown to be sensitive to the fibre length used. For fibres longer than 5 cm bending effects become significant, and for shorter lengths leaky mode effects are important.3 In this letter, a technique is described that is insensitive to these effects and which gives an accurate value of the L P n mode cutoff wavelength Xc that is representative of the local equivalent step-index guide parameters. The technique involves measurement of the normalised refracted power as a function of wavelength and from such data the LPn mode cutoff is determined. Principle: The radiation incident on the waveguide input face excites different bound, leaky and radiation modes. As the wavelength of the source is increased, some of the bound modes become radiation modes and some transform to leaky modes. For example, in a step-index waveguide the LPn mode has 100% of the power propagating in the cladding at a Vparameter value of 2-405.4 By immersing the waveguide in a fluid with a higher refractive index than the waveguide cladding, the radiation modes in the cladding can be refracted out within a few millimetres of the waveguide input face and the refracted power measured as a function of wavelength. As indicated in Fig. la, the fraction of power in the cladding, and hence the normalised refracted power, alters sharply at the ELECTRONICS LETTERS 28th August 1980 • R O 1 700 1 C i I 800 900 1000 wavelength ,nm 1100 k32/i| Fig. 1 a Variation of ratio [Pctad/P],tne fraction of total power in cladding as a function of K-parameter b Normalised refracted power log [P(cladding focus ii)/F(core focus i)] as a function of wavelength for fibre A listed in Table 1 prism 500W monochromator tungsten I lamp I chopper \ e capillary tube U32/2J microscope lock-in amplifer Fig. 2 Schematic diagram of experimental set-up for refracted power technique diameter. The fibre is enclosed in an opaque capillary tube, with only 1-2 mm of fibre exposed, to prevent the leaky mode contribution to the measured refracted power. This arrangement is essential to obtain a sharp change in refracted power at cutoff. This assembly is pfaced in a cell containing a fluid with slightly higher index than the cladding. The refracted power is measured as a function of wavelength by using a silicon detector. The refracted power measurement with focused spot on cladding [condition (ii)] provides the reference power, which can compensate for source coupling fluctuations and absorption in the index fluid. Results: Fig. 16 shows the normalised refracted power as a function of wavelength, with a sharp change in the curve at a wavelength X? ~ 860 nm. The change at cutoff is not instantaneous as indicated in Fig. la. A major reason for this is the instrument resolution, and the rest may be due to the depar- Vol. 16 No. 18 695
1/--страниц