measurement of the normal incidence, inter-band photoconductivity. For a device containinga single layer of dots the maximum modal gain of the ground state transition is found to be insu cient for lasing action. Asa consequence lasing occurs for excited state transitions, which have a larger oscillator strength, with the precise transitionbeing dependent upon the device cavity length. The second technique uses the Hakki{Paoli method to determine the spectraland current dependence of the gain. A quasi-periodic modulation of the below threshold gain is observed. This modulation isshown to be responsible for the form of the lasing spectra, which consist of groups of lasing modes separated by non-lasingspectral regions. Possible mechanisms for this behaviour are discussed.?2000 Elsevier Science B.V. All rights reserved.PACS:42.55.Px; 78.66.−w; 42.60.LhKeywords:Quantum dots; Semiconductor lasers; Electro-optic devices; Modal gain; III{V semiconductorsInjection lasers with self-organised quantum dot(QD) active regions are attracting considerable atten-tion due to their potential for low threshold currentdensity (Jth) and temperature insensitiveJthdevices[1,2]. In this paper we describe the use of two comple-mentary techniques to study the gain characteristicsof InAs{GaAs self-organised QD lasers. These tech-niques allow the magnitude of the ground state modalCorresponding author. Tel.: +44-114-222-4561; fax: +44-114-272-8079.E-mail address:d.mowbray@she eld.ac.uk (D.J. Mowbray)gain and the dependence of the gain spectra on injec-tion current to be determined.Self-organised InAs QDs were grown by molecularbeam epitaxy on a (001) GaAs substrate at a tem-perature of 500 C [3]. The QDs have a base lengthof 15 nm, height 3 nm and density51010cm−2.Two laser devices were studied containing eithera single QD layer con ned on either side by 1375A of GaAs or 10 QD layers separated by 250AofGaAs and con ned by 1000A GaAs layers. 16,000Athick Al0:6Ga0:4As cladding layers were used in bothdevices. Devices for photocurrent measurements1386-9477/00/$-see front matter?2000 Elsevier Science B.V. All rights reserved.PII: S1386-9477(99)00371-9
490D.J. Mowbray et al./Physica E 7 (2000) 489{493Fig. 1. Photocurrent (PC) and
electroluminescence (EL) spectra ofa mesa device. The inset shows the ground state (GS) photocurrenttransition after removal of the background. The solid points in theinset show the result of a Gaussian t to the experimental data.consisted of 400m diameter circular mesas withannular metal top contacts. Laser devices were ofthe form of SiN-coated ridges of width 20m [4].Photocurrent was excited using monochromated lightfrom a tungsten-halogen projector lamp (power den-sity3 mWcm−2) and was detected using standardlockin techniques. High-resolution emission spec-tra were recorded using a double grating 0.85mspectrometer and a liquid-nitrogen-cooled Ge p{i{nphotodiode. All measurements were performed for adevice temperature of 80K.Fig. 1 shows a photocurrent spectrum of a reversedbiased mesa device (Vbias−2:0 V) and an elec-troluminescence spectrum of a forward biased mesa(Vbias+1:5V;I=500 mA), both recorded for thesingle-layer QD structure. In both spectra a series offeatures are observed in the range 1.2{1.4 eV, whichareattributedtotransitionsbetweenQD-con nedholeand electron states. A broad, rising background is ob-served in the photocurrent spectrum, the origin ofwhich is unclear.For suitable temperatures and=or bias conditionsall the photoexcited carriers escape from the QDsbefore recombining. Under these conditions the ab-sorption strength (A) of the QDs can be determinedfrom the magnitude of the photocurrent (I) and therelationshipI=APe=hc, wherePis the total inci-dent optical power at wavelength. To determine theground state absorption the background photocurrentis rst removed by assuming a linear variation withenergy(seeinsettoFig.1).Thisallowsthemagnitudeof the photocurrent associated with the QD groundstate transition to be determined which, with the inci-dent optical power (measured with a calibrated powermeter) and the above equation, gives a peak absorp-tion ofA=(20:6)10−4.1ThevalueA=(20:6)10−4representsthefrac- tionalplanewaveabsorptionfornormalincidenceonasingle-dot plane. To determine the modal gain (g) forinplane propagation of light in a waveguide structure,the corresponding absorption coe cient ((−g)) iscalculated. This is related toAby=ln((1−A)1=D)ef;(1)whereDis the average dot spacing. The rst termin Eq. (1) (ln((1−A)1=D)A=DasA.1) representsthe plane wave absorption for an in nite stack of dotlayers. The second term,ef, is the geometric conver-sion factor from the plane wave-to-wave guide geom-etry, and is given byef=I(0)Pi=1i=−1I(iD);(2)whereI(x) is the transverse optical intensity pro le,determinednumericallyusingathree-layermodel.Fora measured dot density of 51010cm−2, a value=(73) cm−1is obtained for the ground state modalabsorption.BecausethemaximumgainofaQDtransi-tion has a numerical value equal to the absorption [5],the maximum ground state modal gain,gmaxmod, for the1This is the value appropriate to the bias conditions for las-ing action (Vbias=+1:5 V) and is larger than the value de-termined from the present photocurrent measurements whichcorresponds toVbias=−2 V. The value forVbias=+1:5 V is ob-tained from an extrapolation of values measured over the biasrange−1V6Vbias68 V. This variation of the absorption withbias arises from an electric- eld- induced separation of the elec-tron and hole wave functions which results in a reduction of thetransition oscillator strength for increasing reverse bias.
D.J. Mowbray et al./Physica E 7 (2000) 489{493491presentsingle-layerdeviceis(73)
cm−1.Althoughthe light propagation direction for the photocurrentmeasurements is normal to that for inplane waveguidepropagation, measurements of laser devices demon-strate that the ground state emission has an inplane(TE) polarisation. Hence the polarisation direction forthe two con gurations is identical.The value ofgmaxmod=(73) cm−1determined fora single-dot layer is considerably smaller than thatof comparable InGaAs quantum well lasers, wherevalues of50{100 cm−1are typical [6]. This rela-tivelysmallvalueofgmaxmodhassigni cantconsequencesforthepropertiesoflaserdevicesasitiscomparabletotheinternalcavityloss(i),wh ichtypicallyhasavaluein the range2{10 cm−1[7,8]. As a consequencelasing on the ground state transition may not be possi-ble and instead lasing may occur on an excited state,which generally exhibit a higher oscillator strength.The relative oscillator strengths of the lowest threetransitionsareshownintheupperinsetofFig.2wheretheir spontaneous emission intensities are plotted as afunction of current. With increasing current the tran-sition intensities saturate as a result of state lling [9].Higher transitions exhibit a higher saturated intensity,which may re ect a higher degeneracy of the under-lying states [9], consistent with a higher maximumgain.Thebehaviourofthesingle-layerQDlaserisshownin Fig. 2 where emission spectra fora2mmcavitydevice are displayed. The emission from the groundstate transition saturates at low currents, indicatingthat the gain of this transition is insu cient to over-comeiplus mirror losses ((1=L)ln(R)5:7cm−1).A similar behaviour is observed for the rst excitedstate transition and lasing eventually occurs via thesecond excited state transition. A 5 mm cavity deviceexhibitsasimilarbehaviour(uppermostspectrainFig.2) except that the reduced mirror loss ((1=L)ln(R)2:3cm−1) now permits lasing via the rst excitedstate. For none of the cavity lengths studied is lasingon the ground state transition possible for the singlelayer device [10]. The present results demonstrate thatthemaximumgroundstategainofaquantumdotlaseris relatively small and that unless care is taken withthe design, ground state lasing may not be possible.Lasing on the ground state transition is desirable ascarriers in the corresponding dot states are the moststrongly con ned, hence minimising thermal carrierFig. 2. Emission spectra of a single QD layer, 2 mm cavitylaser device showing lasing occurring on the second excited statetransition. The uppermost spectrum shows the lasing emission froma 5 mm cavity occurring on the rst excited state. The upper insetshows the evolution with current of the spontaneous emission fromthe QD transitions. The lower inset shows an emission spectrum(I=1:3Ith) of the 10 QD layer, 2 mm cavity device, recordedfrom a series of small windows formed in the top contact.loss from the dots and resulting in the optimum tem-perature performance [11].The second technique applied to study QD gaincharacteristics uses the Hakki{Paoli method [11].Here the net loss coe cient,, of the cavity is relatedto the peak-to-valley ratio,r, of the Fabry{Perot-likeoscillations, observed in the below threshold sponta-neous emission, by=1Lln(R)+1Llnr1=2+1r1=2−1; (3)whereRis the mirror re ectivity andLis the cavitylength. Hence by measuringras a function of wave-length the gain spectrum (=−) can be determined.Fig. 3 shows a series of sub-threshold emissionspectra for the 10 QD layer device and a 0.5 mmcavity length (Ith=21 mA). Close to threshold theemission intensity is modulated by the Fabry{Perotoscillations (see lower inset), the peak-to-valley ratio
492D.J. Mowbray et al./Physica E 7 (2000) 489{493Fig. 3. Below threshold emission spectra of
a 0.5 mm cavity deviceat 80 K. The lower inset shows the Fabry{Perot oscillations. Theupper inset shows a calculated gain spectrum forI= 20 mA anda lasing spectrum (I=40mA).of which increases with increasing current, consistentwith decreasing loss (increasing gain). The upper in-set to Fig. 3 shows a gain spectrum for a current of 20mA. This spectrum is calculated by rst extracting thespectraldependenceoftheFabry{Perotpeak-to-valleyratio (r) from the corresponding spectrum of Fig. 3,and then using Eq. (3) with a calculated value for themirror term of−23:5cm−1(corresponding to anRof0.31) to determine the gain (−). Close to thresh-old the gain maximum occurs above the ground stateenergy, the corresponding gain for which saturates toa negative value. This result demonstrates that for thepresent, short cavity device that the ground state gainis insu cient for lasing action to occur.The gain spectrum shown in the inset to Fig. 3exhibits a quasi- periodic modulation, which is alsovisible in the emission spectra of Fig. 3. This modula- tion,whichhasaperiod50Aanddepth10 cm−1,is closely linked to the form of the above thresholdlasing spectra, an example of which (I=40mA)isshown in the upper inset of Fig. 3. The spectrum con-sists of three groups of longitudinal cavity modes,separated by non-lasing spectral regions,2the posi-tions of which coincide with the maxima of the belowthreshold gain spectrum. A similar modulation of thebelow threshold gain has been observed in a quan-tum well (QW) laser by Arzhanov et al. [12]. Thisbehaviour was explained in terms of the penetrationof the cavity optical mode through the nite thicknesscladding layers and a resultant periodic modulation,which feeds back in to the cavity gain, due to inter-ference e ects in the substrate. A similar mechanismhas recently been proposed by O' Reilly et al. [13] toexplain the lasing spectra of QD lasers. In contrast tothe case of a QW laser the modulation of the gain isexpected to be observable in a QD laser above thresh-old due to strong inhomogeneous spectral broaden-ing which results from the presence of non- interactingcarriers localised in di erent dots [4]. Although theperiod of the gain modulation observed in the presentdevice agrees reasonably with the predictions of Refs.[12,13], the depth of the modulation appears to be toolarge. Using Eq. (2) of Ref. [13] with a calculatedcladding layer penetration depth of 9:3m−1and acladding layer thickness of 1:6m, a gain modula-tion of only10−9cm−1is calculated, very muchsmaller than the observed modulation (10 cm−1).Although the model of Refs. [12,13] is thereforeable to explain the periodicity of the observed gainmodulation,3when used with parameters relevant tothe present device it predicts a very small leakage oflight into the substrate and hence an extremely smallgain modulation. However there exists experimen-tal evidence for signi cant light leakage through thecladding layers, as demonstrated in the upper insetof Fig. 2. This inset shows an emission spectrum,recorded from a series of small windows formed inthe top contact ofa2mmcavity device, forI=1:3Ith.In addition to the expected broad spontaneous emis- sion, lasing modes are observed, consistent with theleakage of some inter-cavity stimulated emissionthrough the cladding layers. The mechanism responsi-ble for light leakage and the resultant gain modulationin the present device hence remains unclear. Possible2In longer cavity devices the number of mode groups can beconsiderably larger e.g.10fora2mmcavity (Ref. [4]).3Recent measurements on devices with di erent substrate thick-nesses demonstrate a good agreement between the measured lasingmode group spacing and the predictions of the theoretical model(P.M. Smowton, private communication).
D.J. Mowbray et al./Physica E 7 (2000) 489{493493explanations include lasing on a higher-
order trans-verse mode, which would exhibit a greater penetra-tion through the cladding layers than the fundamentalmode, or leakage enhanced by scattering from therandom spatial distribution of QDs.In conclusion, two di erent techniques have beenapplied to study the gain characteristics of QD lasers.The maximum ground state modal gain for a sin-gle layer of dots is shown to be relatively low. Aquasi-periodicmodulationofthebelow-thresholdgainis shown to determine the form of the subsequent las-ing spectra.We wish to thank M. Al-Khafaji for the structuralmeasurements and P.N. Robson, P.M. Smowton, E.P.O'Reilly,E.A.AvrutinandA.I.Onischenkoforusefuldiscussions. This work is supported by the Engineer-ingandPhysicalSciencesResearchCouncil(EPSRC)UK Grant Numbers GR=L95489 and GR=L28821.References[1] Y. Arakawa, H. Sakaki, Appl. Phys. Lett. 40 (1982) 939.[2] D. Bimberg, M. Grundmann, N.N. Ledetsov, Quantum DotHeterostructures, Wiley, Chichester, 1998.[3] M.J. Steer, D.J. Mowbray, W.R. Tribe, M.S. Skolnick,M.D. Sturge, M. Hopkinson, A.G. Cullis, C.R. Whitehouse,R. Murray, Phys. Rev. B 54 (1996) 17738.[4] L. Harris, D.J. Mowbray, M.S. Skolnick, M. Hopkinson,G. Hill, Appl. Phys. Lett. 73 (1998) 969.[5] M. Grundmann, D. Bimberg, Jpn. J. Appl. Phys. 36 (1997)4181.[6] L.A. Coldren, S.W. Corzine, Diode Lasers and PhotonicIntegrated Circuits, Wiley, New York, 1995.[7] A.E. Zhukov, A.R. Kovsh, V.M. Ustinov, A.Yu. Egorov,N.N. Ledentsov, A.F. Tsatsul'nikov, M.V. Maximov, Yu.M.Shernyakov, V.I. Kopchatov, A.V. Lunev, P.S. Kop'ev, D.Bimberg, Zh.I. Alferov, Semicond. Sci. Technol. 14 (1999)118.[8] N. Kirstaedter, O.G. Schmidt, N.N. Ledentsov, D. Bimberg,V.M. Ustinov, A.Yu. Egorov, A.E. Zhukov, M.V. Maximov,P.S. Kop'ev, Zh.I. Alferov, Appl. Phys. Lett. 69 (1996)1226.[9] M. Grundmann, D. Bimberg, Phys. Rev. B 55 (1997) 9740.[10] H. Shoji, Y. Nakata, K. Mukai, Y. Sugiyama, M. Sugawara,N. Yokoyama, H. Ishikawa, IEEE J. Quantum. Electron. 3(1997) 188.[11] B.W. Hakki, T.L. Paoli, J. Appl. Phys. 44 (1973) 4113.[12]E.V. Arzhanov, A.P. Bogatov, V.P. Konyaev, O.M. Nikitian,V.I. Shvekin, Quantum Electron. 24 (1994) 581.[13] E.P. O'Reilly, A.I. Onischenko, E.A. Avrutin, D.Bhattacharyya, J.H. Marsh, Electron. Lett. 34 (1998)2035.