length is 980nm and the half-width of the line spectrum is -4nm. From the slope of the top output power characteristics the differ- ential quantum efficiency is obtained as qd = 0.63. Thermal roll- over occurs at 180mW maximum output power. Threshold voltage = 1.8V and, as inferred from the current-voltage characteristics, the differential resistance is pretty much constant at 3.8Q above threshold. The corresponding conversion or wallplug efficiency shows a broad maximum at q = 20% for an output power of 160mW. It should be noted that the output power is measured with a calibrated optical power meter (Newport 1830-C). From the observed emission wavelength shift with dissipated power All AP = 6.76nmA;v and the increase of resonance wavelength with temperature AUAT = 0.069nmiK the thermal resistance is inferred as Rlh = ATIAP = 98KIW.

0 100 200 300 400 laser current,rnA 1895121

Fig. 2 Output power, wall plug efJi:ciency and voltage against driving current for a device of 1 4 6 ~ active diameter

wall plug efficiency

~ 5 O r 10001 I200

I O E 20 40 60 80 100 120 140 160 activediarneter rn

Fig. 3 Maximum output power, maximum Mialb?ug eff icienFand d$ ferential resistance for VCSELs of dgferent active diameters

We have investigated output characteristics of broad area VCSELs of D = 16, 20, 24, 46, 96 and 1 4 6 ~ active diameter. Results for size dependent thermal resistances as well as maximum wallplug efficiencies and maximum output powers are plotted in Fig. 3. Thermal resistance decreases with increasing active diame- ter, in good accordance with the simple fomula [8] R,,, = (2hJ-I valid for sufficiently small devices, where 1, denotes thennal con- ductivity. The solid line in Fig. 3 fitted to the experimental data is calculated for 1, = 0.4W/Kcm, which is close to the GaAs ther- mal conductivity of 0.44W/K.cm. Maximum output power roughly increases linearly with growing diameter [9] at a character- istic slope of 1.25W/mm, which is quantitatively understood from simplified theory if radiation is taken into account. Maximum wallplug efficiency is as high as 40% for devices up to 4 6 ~ . The decrease in efficiency for larger devices is due to laterally inhomo- geneous carrier injection into the active layer. Current crowding effects favour lasing near the perimeter, resulting in ring shaped near field emission patterns with a characteristic ring thickness of - 2 5 p .

It is expected that more efficient heat sinking will result in a fur- ther considerably improved performance of the top emitting devices studied. Impairments due to inhomogeneous carrier injec- tion may be suppressed by employing noncircular VCSEL geometries. Alternatively, bottom emitting devices might be even more advantageous also with respect to efficient heat sinking.

Conclusion: We have studied size-dependent thermal resistance, wallplug efficiency and maximum output power of oxide confined top emitting broad area VCSELs. In devices with 1 4 6 p diameter we obtain 160mW output power at 20% conversion efficiency and record a high maximum output power of 180mW with non-opti- mised heat sinking.

Acknowledgment: We acknowledge support from DFG, BMBF and NEDO. P

5 July 1996 \ 0 IEE 1996 Electronics Letters Online No: 19961155 M. Grabherr, B. Weigl, G. Reiner, R. Michalzik, M. Miller and K.J. Ebeling (University of Ulm, Department of Optoelectronics, D- 89069 Ulnz, Germany)

References

HUFFAKER, D.L., DEPPE, D.G, KUMAR, K., and ROGERS, T.J.: ‘Native- oxide defined ring contact for low threshold vertical-cavity lasers’, Appl. Plzys. Lett., 1994, 65, ( I ) , pp. 97-99

GEIB. K M : ’Selectively oxidised vertical-cavity surface-emitting lasers with 50% power conversion efficiency’, Electron. Lett., 1995, 31: ( 3 ) , pp. 208-209 WEIGL. B., GRABHERR. M , REINER, c., and EBELING, K.J.: ‘High efficiency selectively oxidised MBE grown vertical-cavity surface- emitting lasers’, Electron. Lett., 1996, 32, (6) , pp. 557-558 CHOQUEME, K . D , SCHNEIDER. R P , .IT, LEAR, K.L., KILCOYNE, s.P., and GEIB. K.M.: ‘Low threshold voltage vertical-cavity lasers fabricated by selective oxidation’, Electron. Lett., 1994, 30, (24), pp. 2043- 2044

FIGIEL, J.J., and ZOLPER, J c : ‘Vertical-cavity surface-emitting lasers with 21% efficiency by metalorganic vapor phase epitaxy’, IEEE Photonics Technol. Lett., 1994, 6 , (9), pp. 1053-1055 PETERS. F.H , PETERS, M C , YOUNG, D B., SCOTT, J W., THJBEAULT, B.J., CORLITE. s.w.. and COLDREN, L.A.: ‘High-power vertical-cavity surface-emitting lasers’, Electron. Lett., 1993, 29, (2) , pp. 200-201 REIVER. G , ZEEB. E , MOLLER, B , RIES, M., and EBELINC, K.J.: ‘Optimisation of planar Be-doped InGaAs VCSELs with two sided output‘, IEEE Photonics Technol. Lett., 1995, 7, (7), pp. 730-732 UAKWASKI. w., and OsiNsKI, M.: ‘Thermal resistance of top-surface- emitting vertical-cavity semiconductor lasers and monolithic two- dimensional arrays’, Electron. Lett., 1992, 28, (6) , pp. 572-574 WIPIEJEWSKI, I , PETERS, M.G., THIBEAULT, B.J, YOUNG, D.B., and COLDREN, L A : ‘Size-dependent output power saturation of vertical- cavity surface-emitting laser diodes’, IEEE Photonics Technol. Lett., 1996, 8. (l) , pp. 10-12

LEAR. K L., CHOQUETTE. K.D., SCHNEIDER, R P Jr., KILCOYNE, S P , and

LEAR, K L., SCHNEIDER, R.P., CHOQUETTE, K.D , KILCOYNE, S.P.,

Hydrogenation of GaAs MlSFETs with Al2O3 as the gate insulator

P.A. Parikh, S.S. Shi, J. Ibettson, E.L. Hu and U.K. Mishra

Indexing terms: MISFETs, Gallium arsenide

A GaAs MISFET with AI,O, formed by the wet oxidation of AlAs as the gate oxide is reported. It is observed that hydrogenation treatment proves to be effective in reducing the state density at the Al,O,/GaAs interface due to removal of excess arsenic, which is a possible cause of interface stales in this system.

Introduction: Low gate leakage in FETs is required for several applications, most notably low power high speed ICs and high reliability power amplifiers. Furthermore, enhancement-mode FETs with the low gate-leakage necessary for dense low power ICs are currently extremely difficult to achieve. An insulator gate would be a natural choice for high speed GaAs FETs with mini- mal gate leakage. However, GaAs based electronic devices still suffer from the lack of a suitable stable oxide. Recently, there has been increased interest in the wet oxidation of AlAs, mainly for current confinement applications in optical devices [l, 21. A GaAs

1724 ELECTRONICS LETTERS 29th August 1996 Vol. 32 No. 18

based FET with Al,O, as the gate insulator was recently reported [3]. We report here a GaAs MISFET with Al,O, formed by the wet oxidation of AlAs as the gate oxide, and in particular the effect of hydrogenation in improving the device characteristics.

Fig. 1 Layer structure and first stage of process for A1203 M I S F E T

In the first stage SiO, is deposited and patterned, and the AlAs is exposed by etching

Fig. 2 Second stage of process

Wet oxidation of AlAs

Fig. 3 Third stage of process C1, based RIE etch to buffer layer, nL GaAs selectively regrown by MOCVD to form source and drain contacts

Fig. 4 Final stage of process

Device isolation accomplished my mesa etching, SiO, removed, tung- sten defines source, drain and gate currents

Experimental: The layer structure of the depletion-mode MISFET is shown in Fig. 1. The GaAs cap is to prevent the oxidation of AlAs in the atmosphere. First, SiO, is deposited and patterned to open the sourceidrain regions (i.e the gate mesa is defined). Then we etch down to the channel using C1, based RIE (wet etch is avoided to prevent any undesirable undercutting), thereby expos- ing the AlAs from the side (Fig. 1). The nominal gate mesa size is 4 p . This is followed by the wet oxidation of AlAs (Fig. 2). The steam oxidation is carried out in a single zone quartz furnace at 45OoC, fed by a bubbler maintained at 85°C. The oxidation time is -5min, which is sufficient to completely oxidise the AlAs layer under the gate laterally. The next step is a Cl, based RIE etch to the buffer layer (Fig. 3). After this, n+ GaAs is selectively regrown by MOCVD to form the source and drain contacts (Fig. 3). The SiO, deposited initially serves as the mask for the selective regrowth. Next, device isolation is accomplished by mesa etching, again with C1,-based RIE. The SiO, is removed and tungsten is

ELECTRONICS LETTERS 29th August 1996 Vol. 32

used to define the source, drain and gate electrodes (Fig. 4). The 200A GaAs cap is etched off just prior to depositing the gate metal, which then sits on the AI,,O, forming the MISFET. The use of a refractory metal in conjunction with the MOCVD regrowth serves to form temperature stable contacts. In future, these would also enable us to carry out annealing studies, once the device fabrication is completed. Finally the devices were also sub- jected to a room temperature hydrogenation treatment with hydrogen ions at 400eV for 30nnin. The sample was tilted to ena- ble the penetration of hydrogen 1.0 the gate oxide. The details of the hydrogenation system are described elsewhere [4].

0 1 2 3 4 51 6 7 8

Vd<;,v Fig. 5 I - V (haracteiistics of 150pn wide MISFET befove hydrogena- tion V,, = 0 to -9V

Results and discussion: The I-V chlaracteristics of the MISFET are shown in Fig. 5. The FETs have a maximum gate current of -SOmA/mm, and maximum transconductance g,, of -7mS/mm. The knee voltage is higher because the soiirce and drain contacts were not perfectly ohmic. This 'was because W metal was depos- ited directly on n- GaAs, which evidently gives a Schottky type contact. Grading the regrowth to n+ InAs would solve this prob- lem. The modulation of charge i n the FET requires that the Fermi level move freely through the gap and the presence of interface

6

5

4 Q

E. 3 U

2

1

0

L---- 4 1 J/ ---

0 2 4 6 8 10 Lbs. v /83616/

Fig. 6 I -V characteristics of 150pn wide MISFET (of Fig. 5 ) after , hydrogenution V,, = 0 to -6V

states can be readily observed via g,, compression at particular bias voltages. The g, compression is evident in Fig. 5. Fig. 6 shows the 1-V characteristics of tbe FET after hydrogenation. It is clear that the g, compression is considerably reduced, and the pin- choff characteristics have also improved. The gm of the device after hydrogenation is enhanced and is -1 5mSimm. This suggests that the interface state density at the oxideiGaAs interface is reduced by hydrogenation. We hypothesise that the interface states in this system arise mainly due to the e m s arsenic left during the oxida- tion of U s . This is supported b,y our DLTS study of the oxide formed by wet oxidation [5], which yielded a single dominant trap level at 0.74eV, very close to the commonly observed ASG, antisite. It has been shown that hiydrogenation is helpful in remov- ing the As as ASH, [4].

No. 18 1725

Conclusion:, We have demonstrated the use of Al,O, as a gate insu- lator for GaAs based MISFETs. More importantly, we show that hydrogenation could be a useful tool in reducing the interface state density in this system. Forming a defect free oxide on GaAs is an extremely valuable pursuit. It will reduce leakage current in FETs at a minimum, leading to improved reliability power ampli- fiers and low phase noise oscillators. Finally, with reduction in the interface state densities to levels that would allow inversion, Al,O, formed by the wet oxidation of AlAs is a very promising insulator for GaAs based MOS and CMOS electronics.

0 IEE 1996 Electronics Letters Online No: 19961158 P.A. Parikh, S.S. Shi, J. Ibettson, E.L. Hu and U.K. Mishra (Department of Electrical and Coniputer Engineering, University of California, Santa Barbara, CA 93106, U S A )

I July I996

References

1 DALLESASSE, J.M , HOLONYAK. N . , SUGG, A R , RICHARD. T A , and EL- Z E I N . N : ‘Hydrolization oxidation of A1,Ga,~,/As-A1As-GaAs quantum heterostructures and superlattices’, Appl. Phys. Lett., 1990, 57, (26), pp. 2844-2846

2 DALLESASSE, J M . and HOLONYAK. N.: ‘Native-oxide stripe-geometry Al,Ga, .As-GaAs quantum well heterostructure lasers’, Appl Phys. Lett., 1991, 58, (4), pp. 394396

3 CHEN. E.I., HOLONYAK, N., and MARANOWSKI, S.A : ‘AI,Ga, ,As-GaAs metal-oxide semiconductor field effect transistors formed by lateral water vapor oxidation of AlAs’, Appl. Phys. L e t t , 1995, 66, (20), pp. 2688-2690

4 SHI. s , and H U , E.L : ‘Photoluminescence study of hydrogenated aluminum oxide semiconductor interface’. Electronic Materials Conf., Santa Barbara, 1996

5 PARIKH, P , JAIN. S , IBBETSON, J , MCCARTHY. L., CHAVARKAR, P , CHAMPLAIN, J , SHI, S , HL. EL. , and MISHRA, U K : ‘DLTS study O f

GaAs MOS capacitors with aluminum oxide as the gate insulator‘. Electronic Materials Conf., Santa Barbara, 1996

Normal incident infrared absorption from InGaAs/GaAs quantum dot superlattice

D. Pan, Y.P. Zeng, M.Y. Kong, J. Wu, Y.Q. Zhu, C.H. Zhang, J.M. Li and C.Y. Wang

Indexinx terms: Infiuved detectors, Semiconductor quantuni dots, Semiconductor superlattices

The authors report for the first time, normal incident infrared absorption around the wavelength of 13-15pm from a 20 period InGaAs/GaAs quantum dot superlattice (QDS). The structure of a QDS has been confirmed by cross-section transmission electron microscopy (TEM) and by a photoluminescence spectrum (PL). This opens the way to high performance 8 - 1 4 p quantum dot infrared detectors.

Zero-dimensional (OD) quantum dot (QD) structures have attracted much interest in recent years due to their 6 function-like density-of-states, strong carrier localisation, increased excitation binding energies, and enhanced oscillator strength [l - 71. High quality QDs can be self-formed in situ, in the Stranski-Krastanow growth-mode, without any substrate patterning process [2 - 41.

The potential for quantum dot device applications has also been of much interest. It was shown that the discrete levels in QDs hinder carriers relaxing into the ground state, this is known as the ‘phonon bottleneck‘ [5 ~ 71. Although it is inappropriate for laser application, slow carrier relaxation is very useful for application to intersub-band absorption infrared photodetection. Unlike the con- ventional GaAsiAlGaAs quantum well infrared photodetector (QWIP) that requires a coupler (e.g. grating) to couple the normal incident radiation [8], the intersub-band transition in quantum dot structures can be induced by the normal incident radiation due to the localised state in quantum dots. Hence, the intersubband absorption quantum dot infrared photodetector (QDIP) has a great advantage over the conventional GaAsiAIGaAs QWIP.

This Letter describes an experiment in growing high quality 20 period InGaAsiGaAs quantum dot supperlattice (QDS) with a standard QWIP structure. The normal incident infrared absorp- tion can be observed around the 13-15pn point.

Expevimental details: The InGaAslGaAs QDSs were grown by molecular beam epitaxy Riber-32P on a semi-insulting GaAs (100) substrate. The structure of InGaAsiGaAs QDS is very similar to that of the conventional QWIP. The layers consisted of, from the substrate side, a 1 . 0 ~ buffer layer, a 1 . 0 ~ n+ contact layer, a 20 period InGaAsJGaAs quantum dot arr;ay,. and a top contact layer consisting of 0 . 5 ~ GaAs and 2000A AlGaAs (x = 0.04). The InGaAs layers and the contact layers were Si-doped with concen- trations of 1 x 1018cm-3, and &doping for the InGaAs layer. The quantum dots were self formed by their coherent relaxation into islands of tens ofo monolayers of In, ,Ga, ,As between undoped GaAs layers -300A thick. The actual amount of indium incorpo- rated in the dots is difficult to measure or calculate due to the complex dynamics of the adatoms during island formation.

Fig. 1 Cross-sectional vieiv of quantum dot superlattice

Lo c

I . ,

1 2 1 4 16 1 8 energy , eV

Fig. 2 Loii trinperature photoluminescence spectrum of sample

10K. 10mW. InGaAdGaAs FWHM = 53meV. 6?iil\. = 4 2%

Result and discussion: Fig. 1 shows the cross-section TEM micro- graph of the quantum dot array. A 20 period defect free quantum dot array is clearly shown. It is estimated that the average size of quantum dots is 300A (diameter) and 6CL70A (height). It is worth noting that the actual size of a quantum dot is less than that meas- ured by TEM due to the distributions measured by TEM being affected by strain, which tends to overestimate the size. Fig. 2 shows the low temperature (10 K) Photoluminescence (PL) spec- trum of QDS. A strong luminescence peak around 1.25eV is noted. The dot peaks have a full width at half maximum of -53meV, which is due to the size fluctuation of quantum dots.

1726 ELECTRONICS LETTERS 29th August 1996 Vol. 32 No. 18