-tt558 OPTICAL UONLINEMITIES IN OAAS/GAAL*5 NULTIPLE GUiTIM I/,WELLS FABRICATED B (U)UIJNIVERSITY OF SOUTHEERNCALIFORNIA LOS ANGELES DEPT OF ELECTRI
UNCLASSIFIED E GAMIREET AL DEC 87 AFOSR-TR-88-SII4 F/G 24/12 M
- . ", 1r32b
1.0
01-1
..0.-.
Op..
'p....
.O 0 0 S 0 0 0 0 ~
nTlI FILE COPI Form Approved
UMENTATION PAGE OMB No. 0704-0188AD-A 191 558 lb RESTRICTIVE MARKINGS
2a. SECURITY CLASSIFICATION AUTHUO IT 3 DISTRIBUTION /AVAILABILITY OF REPORTApproved for public release,
2b. ECLASSIFICATION' DOWNGRADING SCHEDULE diztribu ion unlimited
4. PERFORMING ORGANIZATiON REPORT NUMBER(S) 5 MONITORING ORGANIZATION REPORT NUMBER(S)
AFOSR -T1(- 8 8 - 0 1 3 0
6a. NAME OF PERFORMING ORGANIZATION 6b OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATION(if applicable)
Univ of California AFOSR/Ne
6c. ADDRESS (City, State, and ZIP Code) 7b ADDRESS (City, State, and ZIP Code)
Los Angeles, CA 90089-0483 Bldg 410Bolling AFB, Bolling AFB, DC 20332-6448
8a. NAME OF FUNDING/SPONSORING 8b OFFICE SYMBOL 9 PROCUREMENT INSTRUMENT IDENTIFICATION NUMBERORGANIZATION (If applicable)
AFOSR NE AFOSR-85-0297
5c. ADDRESS (City, State, and ZIP Code) 10 SOURCE OF FUNDING NUMBERSPROGRAM PROJECT • SK IWORK UNITBldg 410 ELEMENT NO NO NTOoE ACCESSIN NOBolling AFB, DC 20332-6448 61102F 2305/ Bi
11 TITLE (Include Security Classification)OPTICAL NONLINEARITIES IN GaAs/GaA1As Multiple Quantum WELLS FABRICATED BYMETALORGANIC CHEMICAL VAPOR DEPOSITION FOR USE IN OPTICAL SIGNAL PROCESSING
'2 PERSONAL AUTHOR(S)GARMIRE
34 TYPE OF REPORT 113b TIME COVERED 114. DATE OF REPORT (Year, Month, Day) 115 PAGE COUNTFinal FROM 01/08/8SrOL106Z81
'6 StPPLEMENTARY NOTATION
'7 COSATI CODES 'B S.BECT TERMS (Continue on reverse if necessary and identify by block number)rELD GROUP SuB-GROUP
'9 ABSR-AC (Continue on reverse if necessary ard denttfy by block number)
Preparation of high qua:ty GaAsiGaAlAs Multiple quantum wells (MQW)grow-i by Metalorganic Cem.:al Vapor Deposition on GaAs substratesand measurement of non>:near saturation has been completed in this18 month contract. h :-sults show materials which rivals the highestquality MQW's grown i_ r.y technique. Preparation of GaAs/GaA1As.!QW's on GaP subst-'i; .7! a -easurement of nonlinear saturation has)een completed. , that this material has high cw saturation:ntensity and, it .: - ".>near Fabry-Perot, would be usefulOnly in pulsed exact. . .: .aterial looks ideal for hybrid devices.
20 D'S RI I ON AVAILAB'LTV OP ABS- ,. ° 2' ABSTRACT SECURITY CLASSIFICATION - 2 '.0 .%CASS,;'ED UNL'M'-D S .CS3 SA%'N S.FRS r4F
22a NAME OF RESPONSBLE NDi'D',AL 22b TELEPHONE (Include Area Code) 22c OFFIC BOL.
i ( 'l) '7A7_O'Z Mr;x
DO Form 1473, JUN 86 . i,. ir. n sr ire obsolete SECURITY CLASSiFICATION OF )L ( -AI.(E
%I
AFOSR-T. 88- 0 130
Final Report
Optical Nonlinearities in GaAs/GaAIAs Multiple Quantum Wellsfabricated by Metalorganic Chemical Vapor Deposition
for use in Optical Signal Processing f "
Submitted toDr. Lee Giles
AFOSR Accession ForBoiling Air Force Base NTIS GRA&I
Washington, D. C. DTIC TABUnannounced
Justlficationby
By
E. Garmire and P. D. Dapkus Distribution/Department of Electrical Engineering Availability Codes
University of Southern California AvalI and/orLos Angeles, CA 90089-0483 ;Dist Spocial
December, 1987Abstract
Preparation of high quality GaAs/GaAIAs Multiple quantum wells (MQW) grown
by Metalorganic Chemical Vapor Depositiun on GaAs substrates andmeasurement of nonlinear saturation has been completed in this 18-month
contract. The results show material which rivals the highest quality MQW'sgrown by any technique. Preparation of GaAs/GaAIAs MOW"s on GaP
substrates and measurement of nonlinear saturation has also been completed.
It was shown that this material has high cw saturation intensity and, if used in a
nonlinear Fabry-Perot, would be useful only in pulsed experiments. The
material looks ideal for hybrid devices, however.
Modelling of the nonlinear Fabry-Perot based on the experimental
determination of the saturable absorption and the corresponding nonlinear
refractive index has shown that operation on reflection, very near the exciton
resonance, will give low threshold bistability (
GaAs/GaAIAs Mutiple Quantum Wells Fabricated by MetalorganicChemical Vapor Deposition for use in Optical Signal Processing
IntroductionThe research carried out on this contract had three main goals, all of which
-. were met. The first was to prepare high quality mutiple quantum wells (MQW)by metalorganic chemical vapor deposition (MOCVD) and to measure thenonlinear refractive index in these materials. This was done by measuring thesaturation in the absorption and using Kramers Kronig relationships todetermine the nonlinear refractive index.
• The second goal was to fabricate and test MOW's grown on transparentsubstrates, chosen to be GaP. The result of this effort was single crystalmaterial with carrier lifetimes short enough that the saturation intensity washigher than for MOW's fabricated on GaAs. Projections for the use of thismaterial are in hybrid devices.
The third goal was to fabricate distributed Bragg reflector (DBR) layers on asubstrate prior to the growth of MQW's. Such reflectors can provide nonlinearFabry-Perots and (NLFP) and devices which may be used in reflection withoutremoving the substrate. High reflectance multi-layers were successfully grown
by MOCVD.
This report is divided in the following sections: Section II contains the paperwhich was published in Applied Physics Letters, describing the firstmeasurements of nonlinear effects in MOCVD-grown materials. These resultsdemonstrate the lowest saturation intensities reported by any technique.
O Further documentation of various portions of the work are found in the theses ofH. C. Lee (PhD) and M. Kawase (MS).
Section III contains an invited paper submittted to Journal of QuantumElectronics discussing our experimental results and analysis of the structuredependence of MOW nonlinear optical properties.
* MOWs for Optical Signal Processing Dapkus and Garmire1C,, ."
.-, . ,., . .. .. . .,,. . . _. . .. ',. .... .. ,,, . -.. -, ,- - ..- .C. % , ,.-,
Section IV shows recent results obtained in MQW structures on GaP substrates.
In these we demonstrate the electroabsorption properties of these structures.Also discussed is our data on the growth of high reflectivity DBR reflectors.
Section V contains discussions of our work relevant to optical computingapplications. In particular, we make projections about the use of NLFP devicesin parallel digital systems.
II. The first demonstration of nonlinear absorption in MOCVD grownmaterials.The following paper describes our initial investigations into the controlling
growth parameters for achieving strong excitonic resonances and and low" threshold nonlinear absorption. It was published in Applied Physics Letters
50,1182 (1987).
-°.
.°
MOWS for Optical Signal Processing Dapkus and Garrn're
p2.
% %
el&oe
* Nonlinear absorption in AIGaAs/GaAs multiple quantum well structures- grown by metalorganic chemical vapor deposition
V. a '--rz. and ' - arm
I A :o -i -1ar rA
I i r- I 1) 11 L 12c1 ' r I r Y
m% ., ... , . , . .* "' ... I r , " .'! ':.,,;- ... .t :"::.''r 2' ,1\", ." >1 ..- ' " ,, ... , '
-". . 'k '.-" • ( ''2 "" ,,"'', ,l,.":,'-,4 ",
* I- T \l\
• -. .'-. ... .. ,, 'tI )..,: b . .] ', . :' bt d I\ ub ,:.. o - }:'!rm ' .
T oo... '-I.'
.I f F 1 11 LI I 'I' \ t
i l r C IXCL II a C1 1 \IJ± n 1 Llt 1i 01 1 111 )FX I lr :, i L' I'.1
-\ 1 t \ 1 7 i 7. . I I I I I r L ! t ' o :I * 7 / t jI t A1 N I I 2I niJ " rI .0 r J .Mi. I i i h i t . 'c m r c : o Ir uif- '''** I' I i t!.n' \7 1fl ', n1 iI ri i!) cc! .,iXoh: i 'C cX 111-1 U' .1
I. . . u. Ari -- ' ' lu: 1 1 1 h2 X I AIx' .1 Ii CI.
MI ,W'~ :y'>XXII f II I( \ j) :1i 1ui 11 11 11 1 1~ic c 0 iPn ; A u c - ''
aJ:
p I I1
:I 1. . . . .7 1 1 1 :1 1
fo. -.
. . . . . ...-.- :.. ':. '-:C 'C: : ;_ '7'.. :N:: . ., %\' :: . ,, \,!i.h~ r '...: :: : : ,v, :: :N.: ':: :':: ,-.'.:.'4, " . :_:/ :': ' :: ,_:Z K: : =...: : . .:•, w]A' : ,." : : :
/:z::::: :: : ::: • ::At ".''.':;.: : ;: I ' : : :.:: ::
a~1 . 111 !1 - - . -- - .
* .It
*l . a
':1: -X::t t;0 1
7:l, 1271 rC0 I, L I ' h c .,. . . . .
~ic: .n' rin arc .)[1ih rh dx
.\Nipl i .irniihJcn.hLr.:t:'.2.. t
.5V..
d'
M " _A - p-L1 III As*,\ _ A
aIt III[erts ItI I hs Icts that I rL:, tircd to sat urate: t ite L-,citon re~o-'A .o ,:.?sl:>.L:. ite : should be nitoed that the purit\ of the trimth IxI -
hum sorcnateriali also has an, im~pact: ott t he puntl o t heNIWsmlsant! the presence- of oxcitonic Feaitures,. Thus,.
sour1ce partt 'Aill atetthe irowth temlper-ature ait 'A hich~ S~~eea~tsit th a:c cit tZileaes a h ber d Fitall. itotilitear ah-
0510T2at iu1-fti idulator to \,ir% t N Il\X MN os, r'O h. \)l1 )C VD A 'tt 7on )T 1 iittl'Ot\ 1:1 11 It iicne1! 1-%attt1. I lomith incid ci t anrd t ransm i t ted Wcrhshci iauc it- ontptild miaterials. 1h1I I,
* ~ ~~~~~ ... -s~tuoe tttsm time. Thre be!am size- was co:- irh h es eotdxLtito " 1 I- urn. Bulk sanple: lltat~itg %kas 1hL zito r> A like, t) lihalk Dr WX. H-. St.r!ieriai
.. 1% h I iLLtnIII th, incident beam w~ith a lowk duty Dr. J. Ku. or for tuhnea yeuso i ad N. Wat ,at-'- rItI" ittr\ hetine w as estimaited tob hnmc,- ai V. Mri 1 oi atc-c>t a: Unptbifhed wkork. Thi' 'Aor k has
-~ C:. 0 ii t k t'. 'A Ied. In Fiuz. 4 the blcaching ot 17, Ii p l supp ,'tied h Air Fcrce Otice of cineRe-- -~.i~tuP'> rpt ot tsc. rv sen it title absorption spectra o 1 "n-t 'Ind~ un .1102'. NASA zr-awi NA\G _
tuna- in~Li \ itoT he tsso s ltarp and wll-resolved Of o ~ SouL:I' hirt Ca lif,)T 'rtt Joinit 1Cers icshee ot1, tt' pasosese under low. e.'ieitattoit iflvltitte i ProeriJT1. the: Arms eea (M)iLce. anTd litheNt' t
a: t 1'.rt'ted to L,.\ itotms associated ss ith heatv hiti v anid Si'e ouuti
hI es repIi clv. TheseL are consistent with the! exd-ta 't ,Ct'raitt. T he ditinct excitoitic absorption enhantce-
IaIturaited at low\ inltensities and merges wkith titehac eotiidahbsorpt onl as .,hoNtn in Fig. 4. Further satura-t)< it -'1A5: F2."tO at1 .IU'CJ ht tite effect ot band filling and requireLs much sci.IF.r311
* tutetuer escitaiioi itieiiies. The hizfhest irtte itit data hate f) S 1) .t:i t ' . . A '*~ ~ '".*' INorrecte to acott tor Fil,-% rot effcts. t ut-
d . !, Ii:: 4 iind troni a nteastiirei tent of the: ittetusits de:- i - PAI t1 t' \ t 40
c , t 5: It \1 :1' Ti I
'it
- r .0),. t IN t 111 t l 'i T(~i III- reallt it th j't' -K K.It It :,7 Pt t144i>
k-i ;t -fi \ kI
%A
% 14
0N
1III. The structure dependence of the nonlinearities in AIGaAs/GaAsMQW's.The following paper has been submitted as an invited contribution to the 1988Special Issue on the Physics of Quantum Wells. The paper describes
• .'2, measurements of absorption saturation for various well widths and coupledwith minority carrier lifetime measurements allows one to determine thefundamental properties of quantum wells that control the saturation intensity.
0
P.
, .1
... ..
The Nonlinear Absorption Properties of AIGaAs/GaAs MultipleQuantum Wells Grown by Metalorganic Chemical Vapor Deposition
I1. C. Leet, A. Kost*, M. Kawase*, A. Harizt, P. D. Dapkust and E. Garmire,Department of Electrical Engineering
tCenter for Photonic Technology*Center for Laser Studies
University of Southern CaliforniaLos Angeles, CA 90089
Abstract
The nonlinear absorption properties of the excitonic resonances associated withmultiple quantum wells in AlGaAs/GaAs grown by metalorganic chemical vaporSdposition are reported. The dependence of the saturation properties on .ro\\ Whparameters. especially growth temperature, and the well width are described. It isfound that the minimum measured saturation intensity for these materials is of theorder 250 Wicm-. When corrected for lateral diffusion effects and the measuredminority carrier lifetime, the data suggest that saturation intensities as low as 2.3\V'-cm- can be achieved in this system. The growth of MQW structures on transparentGaP substrates is demonstrated and the electroabsorption properties of these
structures are reviewed.
:-A
'"
i1" o
I. Introduction
Multiple quantum wells (MQW) have become an important class of materialsfor optoelectronic devices. In particular. a great deal of work has been devotedto exploring their use as nonlinear optical materials in bistable optical devices.nonlinear waveguide switches and optoelectronic electroabsorption switchs.
.. The purpose of the work reported in this paper is to explore the limits to theutilization of NIQW's in these applications by determining the dependence ofthe nonlinear optical absorption on the materials properties and the structural"design of the cuantum wells. In addition, potential solutions to a fundamental
technological issue in AlGaAs,'GaAs MQW structures - the opacity of the. substrate to near band edge radiation - are explored and presented.
The "u"tn cn-mt of elect.rons and holes in quantum wells gives rse toan absorption spectrum that is fundamentally different from bulk materials .
1.First. the eacr2%v distributions of the conduction and valence band states arealtered to a step-like distribution characteristic of a quasi-two-dimensioi.aisolid. The positions of these density steps are associated with the allowedenergies of the "square well potential" and are a strong function of thedimensions of the well and the composition of the barrier. In addition, theconfinement increases the oscillator strenoth and the binding energy of freeexcitons in these, lairs resulting in strong excitonic resonances at each of thesteps in the linear absorption spectra of these materials 2. These resonances canbe saturated by optical excitation 3; an effect that has been used to fabricatenonlinear wa-veguide elements 4 and two dimensional bistable switchingarrays 5. The resonances have also been observed to persist in presence oflarge electric fields applied perpendicular to the layers 6 and to broaden andshift to lower energies with increasing field strength. This effect has beencalled the quantum confined Stark effect (QCSE) 6. The QCSE has been
o. exploited by Miller and co-workers 7 to develop optoelectronic switches suchas the self-enhanced electroptic device (SEED). Application of the switchingsdevices has been limited by the excessive power requirements of the nonlinearoptical devices 8 and the complex processing required for the optoelectroncsw% itches. In this paper we will address the minimum optical power required..-fr the sauration of the sorption in quantum wells based on direct
mea urement of the controlling properties. We will also demonstrate a newNonlinear Ahsorption of MQW's Lee, ef. al.
. .. . . . . . . . . .,.et. .al.
• ° - . .?/ ,"," -i.;-;....; . ..- ,.. ..., .. .-.. "."."..... .. .. ........ ...- ...-4.. . : 5 *: ; :.. . . . . . . . . . . ... . ,.
low %Wq
3
technology for the growth of quantum wells on transparent substrates that maymediate the fabrication difficulties inherent in optoelectronic switches.
,.- The AIGaAs/GaAs MQW materials used in the studies presented here havebeen grown by metalorganic chemical vapor deposition (MOCVD) 9. Thematerials properties and their dependence on the growth parameters willhopefully provide guidelines for the application of this technolgy to thefabrication of MQW nonlinear optical devices. The paper is organized so thatthe experimental methods of crystal growth and characterization are describedfirst (Section II). The linear optical and optoelectronic properties of thematerials structures are then presented followed by the nonlinear opticalcharacteristics (Sections III, IV, and V). Section VI presents analysis of theseproperties that allows the extraction of the fundamental factors controlling thenonlinear properties of these materials. Section VII describes alternatestruc.tures for nonlinear and optoelectronic switch arrays and Section VIIIsummarizes the results and implications of these studies.
Si1. Experimental Techniques
A. Crvstal GrowthAll samples were _grown by atmospheric pressure MOCVD 10,11. The reactoris designed to minimize the interfacial transition width in the NIQW's 12 It
employs differential pressure transducers in a vent/run configuration todynamically balance the pressure and high flow rates (9 slpm) to reduce thegas retention time. Trimethylgallium and trimethylaluminum were used for Gaand Al sources respectively. Arsine gas was purified by an in-line molecularsieve. The operation of the manifold switching and critical flow rates are
*':- computer controlled. All MQW layers discussed in this paper were nominally
undoped. The samples grown for this study were grown in the temperatureO. range 650°C to 750'C. The background carrier concentration in undoped GaAs
is observed to vary with source quality and growth temperature. Thisparameter will be discussed later.The typical sample employed for the studies reported is as shown in Fig.l. It
consists of 1.0 - 2.0 pm of MQW material surrounded by thicker confinementlayers of Al 0 3 2Ga 0 6 8As. The growth is initiated by the growth of a thin
Nonlinear Absorption of MQ1V's Lee, el. al.
Ao
4
.\ 1'( BvYhvtox h\ U . 5 t m A IG iAs Ivcr w i h hi 1h A]1a :~in e : ; t op L\ ecr fv substrate re rnoval. The subsrt
'A s -'.peI ()I c h. Tizt il B ridumcrani - crowkn GaAs. The MIQW
ma l vi teJ : :,,m e rods of the al te mat inc strucuet i d03PM 01 o Ga.V (1\,NM11 c'iu'tum w lls.Te spacer layers were 100OA in thIckeness,,
itar b ut th iA ]Aei> n this case the barrier thickness was 1 _':OA to
nuni~e e tu nnonovr due to tunrin- betweeen the wells. TPheC,arIerI rs crc :\ l(;,C,) W)As. The top and bottom confiinment lavers of
Itu~ue used fSL,
nsec. The wavelength of the laser was 7800,A and its output power was
variable up to 1W peak. The light from the sample was dispersed by a one
meter monochronomator and was detected using photon counting techniques
with a GaAs photocathode photomultiplier. The transient response of the
luminescence was measured by the delayed coincidence technique 15. The total
system response time was limited by the laser fall time.
D. Linear and Nonlinear Absorption Measurements
The low level absorption properties of the MQW samples were measured in
transmission with a commercial UV-VIS spectrometer or with the attenuatedoutput of a CW dye laser. In the latter case care was taken to insure that the
excitation level was low enough to avoid the nonlinear absorption regime.
The nonlinear absorption properties of the MQW materials were measured b%
using an argon ion laser-pumped styrl-9 CW dye laser. The output power of" -] the laser was attenuated by the use of an acousto-optic modulator. The first
order diffracted beam was selected with an aperature. This system provided a2000:1 dynamic range. The modulator output was pulsed with a 2 usec
I duration for high level excitation or a 200 jisec duration for low level- excitation. The repetition period was chosen to be 10 msec. The incident.
reflected and transmitted beams were measured with calibrated photodiodes
through the use of a beam splitter assembly and suitable neutral density filters.The spot was focussed to an area of a few hundred square microns by a longworking distance lens assembly.
*, III. Linear Optical Properties
Early studies indicated that the major growth parameter controlling the
nonlinear optical properties of MQW's grown by MOCVD was growth
temperature 16. The dependence of the nonlinear absorption properties on thequantum well width is also of importance as a fundamental issue. In addition.
this variation may depend upon the growth process and was. in any case,.
unknown. As a result, a study to measure the variation of relevant properties as% a function of both parameters was undertaken. In Fig.2' we show the
dcpcndncL, on growth tempei-atur2- of the room temperature PI, of three-1
Nonlinear Ahvorption of MQW's Lee, et. al.
'p.-
6
samples with IOOA quantum wells grown at various temperatures ranging from650'C to 750'C. The structural and electrical characteristics of these samplesare listed in Fable 1. The key points to be made are that the PL spectra of allsamples shows the presence of the lowest energy (n=l) confined statetransitions terminating in the heavy hole (hh) and light hole (lh) valence bandstates and that the width of the spectra increases slightly with increasinggrowth temperature. The increasing spectral width reflects, we believe, theincreased concentration of free carriers in the well owing to an increase in theback2round carrier concentration with increased growth temperature. Thistrend is consistent with the electrical properties of bulk GaAs samples grownunder same conditions as shown in Table 1 17. This is further supported by thefact that the linewidth of the high purity samples increases by several meV athigher excitation levels and thus higher carrier concentrations.
hle absorption properties of samples grown at various temperature also showthe effect of increasing impurity concentration at higher growth temperatures asshow, n in the low resolution spectra of Fig. 3. The sample grovwn at 650'Cexhibits a well defined excitonic resonance characteristic of MQW's while the
4 resonance for the sample grown at 7000C is somewhat broadened and reducedin amplitude. The third sample shown was grown at 750°C with an impure Gasource. It has relatively high background carrier concentration and exhibits no
excitonic absorption resonance.
The PI, properties of samples grown at 700'C with various quantum wellwidths are shown in Fig 4. The properties of the samples are shown in Table
-2.The wells range from 54A to 193A in width as judged by the peak positionof the n=l, hh PL transition and the energy differences between the othertransitions. In the narrow ells, the light and heavy hole transitions are well
resolved owing to the larger energy separation of the terminal states. Thespectra of the thicker wells show peaks due to transitions from the n=2 andn=3 electron states. The occupation of these states is made possible by thedecreased separation from the n= I state as the well width is increased.
The room temperature linear absorption spectra of the same samples is shownin Fig.5. Note the step-like nature of the absorption and the presence ofresonances associated with the n=l transitions. Both of these features areNonlinear Absorption of MQIW's Lee, et. al.
'S.
7
unique to these quasi two dimensional systems. Note that the spectral width ofthe transmission spectra is comparable to the PL spectral width in thesesamples. Because the PL laser excitation hardly penetrates the top claddinglaver we expect that only the first few wells are excited by the carriers thatdiffuse from the cladding layer. The close correspondence between theemission and absorption linewidths suggests that the well width is extremelyuniform from well to well through the samples.
IV. Minority Carrier Lifetime
The minority carrier lifetime of quantum well samples was measured bymonitoring the PL decay time under the pulsed excitation. For opticaltransitions involving free carriers or weakly bound excitations such as freeexcitons. the PL decay time accurately reflects the minority carrier lifetime. We
.* have chosen to use photon energies in the excitation that are directly absorbedby the quantum well layers to avoid possible ambiguities involving nonunifromexcitation or carrier capture by the wells. The PL time decays of threepreviously discussed samples grown at various temperatures are shown in Fig.6. The measurements were performed at the highest excitation intensity
. . allowed by experimental conditions ( 50 W/cm 2 ) to simulate as close aspossible the high level conditions involved in absorption saturation. This limit
was set by the power of the laser and the need to maintain a spot size largeenough to avoid any lateral diffusion effects on the measurement. The timedecay of each of the samples contains a dominant component that isexponential over 2-3 orders of magnitude. This component of the decay timealso increases with increasing growth temperature. This is consistent with the
. trend observed in the PL efficiency. The samples grown at the highertemperatures exhibit an initial decay that is somewhat shorter than thedominant decay mode and a transition to a shorter decay at longer times. Thelatter effect is the result of emptying of saturated recombination centers in the
.. material at longer times, which increases the rate of minority carrierrecombination. We have studied these effects in detail and will report them in aseparate publication. We have observed, however, that samples with longerminority carrier lifetimes at low levels exhibit strong increases in the lifetimewith excitation, presumably because there are fewer recombination centers inthe material and they are more easily saturated as a result. The shorter initial
A Nonlinear Absorption of MQIV's Lee, et. al.
0 % d " " " " ." "o. . . '" " " %"" " " " .- , . . . " " " " % " " td % % ' ' %
8
decay component in the samples with long minority carrier lifetimes is theresult of the short pulse excitation used in these experiments . It may beregarded as the time for the carriers to establish steady state conditions.
* Table 3 lists the PL decay times of the absorption samples with different wellwidths. The decay time is observed to decrease with decreasing well width.The longest decay times we have observed exceed 1 [tsec. This, we believe, is
a reflection of the high quality of our samples. The dependence of lifetime onthe well width reflects increased penetration of the electron and hole wavefunction into the lower quality barrier layers where the nonradiativerecombination rate is higher. This observation is consistent with the data ofFig. 6 because the dominant effect of growth temperature on the PL efficiencyis to improve the quality of the A1GaAs barrier layers.
* V. Nonlinear Absorption Properties
Saturation IntensityThe transmission spectra of all samples exhibit Fabry-Perot resonance peaksdue to the residual reflectivity of the front and back surfaces. In Fig 7 a weshow the transmission spectra of the sample with the 72A well width. The lowintensity spectrum shows transmission minima at 829 nm and 837 nm thatcorrespond to the n=1 light and heavy hole absorption resonances. The localminima at 822 nm and 855 nm correspond to destructive interference of thereflections from the front and back surfaces. In all cases the wavelengthseparation of these resonances agreed with the expected separation based uponthe known thickness of the samples. These minima must be treated to
* accurately determine the absorption coefficient. The Fabry - Perot resonanceswere removed from the spectra by analyzing the spectra according to thefollowing relationship:
T.-. (1-Rf) ( -Rb)e-L 1_ _(1- Reff )2 1 + F sin 2 (8/2)
where Rf and Rb are the front and back surface reflectivities, respectively, (X is
the absorption coefficient, L is the integrated quantum well thickness,
v Nonlinear Absorption of A!MQW's Lee, et. al.
........... v "......."..............-....... .. .
9
:" Rcff = e-aL (Rf Rb)" 2 is the effective mean reflectance, F 4 Reff/(l-Reff) isthe finesse and 5 is the wavelength dependent phase that depends on the optical
length in the cavity. The absorption coefficient can be determined from (I) andthe experimental data. We used Rf= 0.08 and Rb = 0.1.8(X) was determined
from the resonances in the spectra at high excitation. Fig.7b shows the
absorption spectra of this sample at several intensities. Note that the Fabry -Perot effects have been completely removed from the spectra."The light and heavy hole excitonic resonances in the absorption spectra of Fig.
7 are saturable at moderate intensities. In Fig. Sa we show the excitationdependence of the absorption coefficient of sample V158. Note that the heavyhole exciton resonance decreases in amplitude at a rather modest input
., intensity. It also decreases more rapidly with excitation than the light holeresonance or the continuum. We have reported a saturation intens v of 250VV cm- for this sample which is the lowest reporteL. -'alue for the
AIGaAs GaAs materials svstem 3. At 752 W/'cm 2 no well resolved exciton.resonances are discermible in the spectrum of Fig. 8. The continuum spectrum
is saturated at 7520 W,"cm-. The excitation dependence of the peak absorptionfeature at about 8500A is shown in Fig 8b. As will be described in the nex:section, the saturation behavior can be described by two separate mechanisms
p 3 that are due to the exciton bleaching and to bandfilling of the continuumstates IS. The excitation behavior of these two mechanisms are indicated by thesolid curves below the data. The solid line through the data is the sum of thesetwo mechanisms. The saturation intensity of sample V136 grown at the the
.- highest temperature (750'C) was comparable to the band to band saturation
intensity of sample V158. This is to be expected since the high carrierconcentration in sample V136 resulted in screening of the exciton resonance.
The dependence of the n=1 heavy hole excitonic absorption coefficient underresonant excitation for the MQW samples with different well widths is shownin Fig. 9. The saturation mechanism in each case can described by a two
component saturation similar to the data of Fig. 8. Tne magnitude of the
absorption attributed to the excitonic resonance is observed to decrease withincreasing well width. The background continuum absorption is observed todecrease somewhat more slowly than the excitonic component with increasing
. well width. These trends are the result of the decreasing density of available
Nonlinear Absorption of AIQIW's Lee, et. al.
V6..'
[wI
10
band edge states associated with the first allowed level and the decreasedbinding energv of the exciton as the well width is increased. The measuredsaturation intensity for the excitonic and background component is listed inTable 3. Proper interpretation of these measurements must take into accountthe minority carrier lifetime of the samples. This will be addressed in the nextsection.
The measured saturation behavior of the li2ht hole resonance and the higherenergy (Ehh +50 meV) continuum band edge states for the sample with 54 A
wells is shown in Fig.10 a&b. A two component saturation is seen for the lighthole resonance whereas the continuum band edge absorption showed only onecomponent as expected. This behavior is observed for the 72A and 102A wellsamples but the light hole resonance could not be resolved for samples withthicker wells. The saturation intensities for the resolvable light hole resonances
"4 and the band to band transitions are shown in Table 3. The saturation intensityof the light hole resonance is uniformly higher than the heavy hole saturationintensity. This is to be expected from the ratio of the effective masses.
a. VI. Analysis of Nonlinear Absorption PropertiesSaturation IntensityThe mechanism for saturation of the excitonic absorption features in multiplequantum wells has been studied in some detail. Resonant excitation withpicosecond pulses has conclusively shown that the exciton formed in theabsorption process is thermalized in a few hundred femtoseconds leading tothe formation of free electrons and holes 19 The reduction in oscillatorstrength then results from the interaction of these free carriers with free excitonstates. The primary interactions responsible for bleaching of the excitonicresonance are thought to be screening of the Coulombic interaction and phasespace filling 18. The continuum states that overlap the exciton resonances aresaturated by band fillng of the available continuum states . To properlyinterpret and analyze the data shown in previous section we adopt aphenomonological model for these mechanisms proposed by Chemla and co-workers 20. The dependence of the absorption coefficient upon excitation isexpressed as
I
Nonlinear Absorption of MQW's Lee, et. al.
t
i1*
0
,tx (2)'exCCXc = 1+ (I / Is) (2)
0%,where c&xc is the low intensity value of the absorption coefficient and Is is the
intensity at which the absorption coefficient is reduced by half, the saturationintensity, given by:
0Is -- hi)/ 2 L~z c(-exc T[ Aj) (3)
In (3), L, is the well width. T is the minority carrier lifetime, and A, is the
effective area of the exciton. Note that the saturation intensity decreases with-increasing well width and minoritv carrier lifetime. The exciton area has been
calculated to be a function of the well width 21. We will compare the valuedetermined from our measurements with the theory.
Since continuum ab.-orption also contributes to the absorption measured at theexciton resonance, the total absorption is the sum of two components both ofwhich we model with a functional form similar to (1).
0) 0"fcxc Ocon
I + (I / s) + 1 + (I / is2) (4)0
The background continuum absorption, Ucon, is a function of carrier
' concentration and crystalline perfection. In addition, there may be anunsaturable background absorption detrimental to device operation. In most
instances that we have examined Isl ll the effective absorbance is given by the second term of (4). then
1(c,,n and Is2 can be determined from the slope and intercept of a plot of I / ototas a function of I as shown in Fig. I Ia for the sample with 102A quantum
wells.
Nonlinear Absorption of MfQW's le e, c1. a!.
.% %.
In the low in1tens"It% r1,el,, the absorption coefficient cain ble
renr sete b the firstI term Ot 0) Plus an 1iin,ti 1independent contIuum.
(7, 0:8 fI t P 0( 1 fC - c41 V[U5 I \icic's the values at 'cf\C and I,1
1 h h:. ~ ~2.i\ >s 1o th saple t Pe. a. \\e h al curvelo
1,; Le. s aev ecre used to calcuILa .he t LoreIca uvs
I hee\~mio&dpendent absorption coefficient for the samples of Fi2 L. 9. Theexccllen-t aereemenit between experiment and data exIbie In th 1f111re
Vl es ample justification tar)I the procedure used and the em pirical saturation
h rul iv en bv (4).
;\ mila prceduire was used to calculate the satura tion intensities aInd peak
absorptions for the light hole exciton resonance and the hicher enerevcontiuum tts. In the latter ease there was often somne residual excitonic
*absorption present due to the continuum States Of the exciton1. The1se values ate
also listed in Table 3.
The peak a-bsorption coefficient for the n= 1 heavy, hole exciton resonance is
f ph tted, !S a functilon of the well width in Fl,,~. 12. Note that the peak absorption
increases monotonicallv with decreasing well width. This is to be expectedsince the exciton binding enere y inCceases With decreasin2 well wi'dth. Ilie
ha.2 nrlcant inuum absor-ption at the exciton res onance also increases w ith
deceain2wel idth. Since the density of' continuumu ban d d2C stateS
\Nnlinzear A4bsorption of MlQ11"s Lee, ct. al.
6%
___ ~ ~ %~~ % -:-. ~ - ~ 1/40
13
increases as the well width decreases, we would expect that any band tailsassociated with these states would also increase. However, this component isfound to be more erratic and its variation may reflect a varying perfection of thesamples used in this study.
Minoritv Carrier ILifetimc and Diffusion Effects"The minority carrier lifetime, x. controls the concentration of free carriers that
are present at a given excitation. The denisty of free carriers in a uniformlyexcited materials is pr, prtional to I-t and the lateral diffusion in a locally excitedmaterial is controlled by the diffusion length, L = (D T)1/ 2 . The saturationmechanisms of interest in this work are dependent upon the concentration offree carriers. As a result it is of interest to understand the dependence of -c upon
ecitation and the structural properties of the MQW materials. We constructhere a simplified model for the minority carrier lifetime that includes both
0 radiative and saturable non-radiative processes to assess the effects ofexcitation on An, the excess free carrier concentration.
VC consider steady, state conditions to coincide with the quasi-CW excitationused in our saturation experiments. The relevant kinetic equations for the free
.U electron concentration and the saturable non-radiative processes have beensolved to yield the following relationship for the minority carrier lifetime:
Bp + 1 1T Tns tc I + (On An) (Op p)
The first term is the raditive lifetime, where B is the radiative coeffecient and pis the free hole concentration. The second term is the non-saturable non-
* radiative lifetime due to deep recombination centers. The final term is a
saturable nonradiative lifetime due to recombination centers that can be filled bycapturin- excess electrons (i.e. On >> Gp) where the O's are the capture crosssections for electrons and holes. This equation has been written assuming that
* electrons are the minority carriers. An analogous equation for holes could bewritten in the case electrons are the majority carriers by reversing the n an p.For lightly doped samples as we have here, the radiative term is usuallynegzlicible at low excitation and the lifetime is the reciprocal sum of tns and rc.
0 At hiher excitation the final terma saturates and the lifetime becomes Tns.
Nonlinear Absorption of MQIV's Lee, et. al.
SA"
,j'. .- -. . - - .-o o.* . - . , . - . . . . - - . . - - - . . ' ,. - . % ' " 1°
,° h% ,% . "" % % ,
14
becomes ins. Under conductivitv modulation (p=Ap An), the lifetime becomesthe reciprocal sum of (BAn) - 1 and Tns. We have observed this behavior in the
time decay of the samples used in this study and discussed the manifestationsof it in Section IV. The task we face here is to decide which reaime to use toanalyze the saturation data of Table 3. The decays of samples with relativelyshort minority carrier lifetimes showed no evidence of a fast initial component.The lifetimes of all samples were completely saturated at the maximumintensitv available in the lifetime measurement and were independent of
intensity over a factor of four in excitation below that. We assume thatmeasured decay time in these cases is the appropriate lifetime to use in thisanalvsis. ve recognize that the radiative component may become important athiaher excitations near the point of absorption saturation. We estimate that fora free carrier concentration of An= 1017 cm- 3 the radiative lifetime will be of theorder 100-200 riser. This is based on our observation of the excitation
dependence of the time decay measured on verv long lifetime samples andother data in the literature 22. Our lifetimes were measured with an impulse
"--- excl-itton that generates approximately 1016 cm -3 carriers. We are clearly in the
re ime where radiative processes may be important in the determination of theU saturation intensitv. In fact, for very efficient materials (i.e. "ns >> (BAn)- 1 ).
photon reabsorption may artificially lengthen the lifetime and lower the
saturation intensity.
The long minority carrier lifetimes we measure in this study result in diffusionlengths comparable to the effective spot diameters used in the saturation
,. measurements. Recent work by Olsson et al. 23 suogests that the effective spotsize must be corrected to include the outdiffusion of carriers. The effective
.. saturation intensity is then calculated from:
lef Ps / (4Dc + Aspot)()
, where P, is the saturation power. This correction has the effect of increasing
the spot size dramatically for the sample with the longest lifetime. We haveused a diffusion constant ) = 17 cm 2/sec 23 to account for ambipolar diffusionof the carriers. The corrected saturation intensities are shown in Table 3. These
, data show that very low saturation intensities are possible at the peak of the
Nonzlinear A bsorption of MQIV's Lee, et. al.
I0" . " - " . ".,- ° - ° . " - " , % % " .% " • '% " % ,% , % % ' , % % " % " " % " % " " " . "V..t, tl..f la tl ,~W. lg., 'tl i lelal Lr im ,l~ J Jl la ~lt e j l~~j j -- _ :
15,
• ." exciton resonance it suitable carrier confinement can be included in the device
st ut ure.
Siturntion Densitv and Exciton RadiusBased on the measured saturation intensities, measured lifetimes and measuredabsorption coefficients we can calculate the saturation density of carriersderined by"
IsT cX(Is)ns (9)
hv
nhe saturation density ns is the density of photogenerated carriers that reduces
thc xcitonic component of the absorption coefficient to one half of its lowitensity value. Fig. 13a shows the dependence of ns upon well width. Thesaturation densitv shows a broad maximum in the range Lz =100A. We
-,.-ie the apparrent maximum is shifted to laroer well widths by the data point:, 3.:. This sample is believed to have spuriously high saturation intensitv. Itis the only sample that exhibited a smaller minority carrier lifetime in the- at_-icated absorption sample than in the bulk sample. In addition. the measuredli.etime of the absorption sample does not follow the trend observed in the
other saumples. On the other hand no visbie physical damage is present and wehave included the sample data for completeness. In view of these-.cOnsiderations, we believe that the maximum saturation density is near- =75A. This can be understood if we realize that:
AexLz I/2ns (10)
Te exctton radius calculated from (10) by assuming rexc (Aexc/,7) 1 2 Is* plotted in Fig. 13b. Note that rec, reaches a minimum near Lz =75.A and
increases on either side. This behavior is a reflection of the increased
Sconlinement of the exciton as the well is decreased up to the point that the
S, e functions of the carriers begin to significantly overlap the barrier regions.lis is expected to occur in the range L= 75A. It must be pointed out that
Slihot, h the exciton is not spherically symmetric 24, its dimension along thelvers also contracts as the transverse dimension is increasinglv confined. TheC- uan.itv ACX increases with decreasing well width for samples with thes :vllct wlCl widths while I, decreases. As a result, the saturation density, n
Nonlinear A .orplion of MiQ "s Lee, et. al.
..} . • -. -. - "". . - .'' ..,* ' ... ' .-. _ .'_ ..-' . .- '- '_'..'.'.*. .,. ,- .-",. , * .- - ., .,'. ..',,',.*, ' *, ,
16
proportional to (.\ - is ak- dependent upon the well width. For the.- idest wells, A,, - 1. I- - 1. and n, are all small.
VII. Alternate Structure for Photonic S%%itches
A major technological difficulty in the fabrication of arrays of photonicswitches in the AlGa:\s GaAs system is the opacity of the GaAs substrate.Jewel et al. 5 and Boyd et al. -have demonstrated photonic switches thatcontain internal mirrors based on epitaxial Bragg reflectors. Fig 14 shows thereflection spectrum of one such reflector grown by MOCVD. The layerthickness of this structure was chosen to maximize the reflectance in thespectral region near the excitonic resonances in GaAs quantum wells. Note thatthis reflector has a 600A bandwidth and a maximum reflectivitv of 0.96. The
1 .,,contairc. of .l\s.\1yG:b] s each laVer being a quarte-*,11 ! c .eLenth thick. The L:,C o Cu. a hiLh refiectivitW mirTor isolates the dei *,ce.. rom te su strate an r -1'--roi the susrt and proids a double pass throuch the active re2ion. Thi,"-s been put to use in reflective modulator devices to improve the contrast ratio
i. The malor difficulty with this approach is the resultant requirement to usethe device in a reflective mode.
\Ve have recently demonstrated an alternate approach to photonic switchtechnoloy by the growvth of AIGaAsi GaAs MQW structures on transparentGaP substrates 27 The large lattice mismatch of the substrate and the activelayers has been accommodated by interposing a GaAs 0, 6P0 .4 buffer layer
between the active re(cion and the GaP transparent substrate. The resultant\,IQW structures show a hi2h degreee of structural perfection. smooth laver
* morphology and a \well resolved excitonic resonance. These structures exhibitvery hiah saturation intensities and are likely not suitable for use as nonlinear
.(-. -. optical materials. owever, they can be incorporated into PIN
clectroabsorption modulators that utilize the QCSE 2 7. The absorptionSOspectrum of one such modulator is shown in Fig 15 at various applied
voltaces. Note the shift of the resonance to lower enerev that is characteristicof the quantum confined Stark effect 7. The energv shifts observed in this
modulator compare favorably wiith results obtain on GaAs substrates.
VIII )iscussion and Conclusions,oflitcar A bsorption of MQ11"s Lee, 0i. al.
. .. ....... .. .. .. ... .. .
17
\Vc have reported the dependence of the nonlinear absorption coefficient inAIGaAsGaAs multiple quanitumn wells grown by MOCVD. The resultspresented clearly show that NIOCVD is a suitable technique for the u owththese nonlinear materials, In fact the low saturation Intensities k~e have
*Observed promise to open new -applications for these materials in nonlinear11'aecic Switches that do not rely upon a high density of switches. In
patiuarw have demonstrated that the saturation intensity% Of _MOCD growknmnaterials are currently controlled bytemnrt crirlftme i the Lat1n7,an d to some extent by' the quantum well structure chosen. Even thouah thesaturation Intensities for the hea-vy hole resonance- of A1GaAs, GaAs MQW' >demonstrated in this paper- are among the- lowest eve.- reported f'or tis mnater'als
systm, he vsteat~ deendece f tis quantity as -a function of the well%vldih suessth-at fine, tuning of the quant-um well desi2n can irn prove the
0r prmrnan c of thes e mmnaa La. I n pa rticulIar. i rP-rs tha te atrnw
inite2n s ity i n t heo s-amnple cs i nvL-st gat1e d to d atLe d Lcease S~ it SIIn'1 e7sna werll Nwidth lariz-ly ecu thc mi'ncritv carrier lifeti'Me inra 11-Tedeede
01 01 w l idh isnttoet t be L dmet~btrte eutC h
quality Cf the AlG-AS barrier layers pr-scntd vaiia : In thli context a R, sperhaps not Surprising that lnP.'inGa.-\s MQ\V sam ple-r P-s have exhibitedsomewvhat lower saturaton VItensities 8. Tw o key Points In asseSS!,ne- the1inimu posil sauainit ity are rltdto thec minority rre
lifetlime. First, lonzoer minority carrier lifetime causes more lateral diffusion ofthe photoexcite-d carriers that affects the apparenit saturation inest.In quasi-
% CXV applications this suL-ests that Some form of lateral car-rier- confinine2st'cur wI be necessary to obtain optimum performance. For shrt Rise
0. applications the- re-levant quant.1itv is the sat urtineer Jan.1,:.TINquanti ty is plotted in Fig. 16. Note that the, safirtinenrv uCn-ce h_"maLximum near a we-ll v. idth of 75.A.The exclionriu P, aimnmm n.a
% resutI t, the saturation density i*s a maXimumM iM Li reeon. It i> t 1us advisato desien nonlinear materials wih 1-7 n frT" 1niism\ 1 ~i u~app1licat ions.
Th Second point to be ma-de isthat ridiive cpro-eses have bec-,n Larcii~nored in the analysis of satura.lon inte- nsi;:e As the qu i-\ A l~j A"
Nonlinezar A bsorpflonz of MQ 1's Lee, ca. al.
I -I nioa e .
-impiores. 1Q .are less likely to be limited by ihe AIGaAs quality and theminority caU-rier liiftime ,,will ultimately be limited bv the radiative processes. Atpresent little is know ri experimently of the radiative recombination coefficient.13. in MQW structures. As the quality of these struictures improves knowledgeof this parameter increases in importance.
The technolov of photonic switch arrays has been complicated by thenecessity of using opaque GaAs substrates on which to grow the MQWepitaxial structures. Selective removal of the substrate requires the use of af raIle. thinned substrate of an alreadv brittle material. This. in turn. willultin-ately limit the size of arrays usable in photonic switches and increase theircost. The use of a transparent substrate such as GaP or sapphire may alleviatethis difficulty. h1e encouraging electroabsorption results we have observed in"lQ\V structure grown on GaP strongly sugcst that this approach will play an
0 important role should the technology of AlGaAsiGaAs MQW optoelectronicswitc.hes be incorporated into practical systems. The s o i aI -ems. euccess of this approacl-
and the availability of high reflectivitv Brag u reflectors Lives the svste,,.- designer a greater measure of flexibility in the design of the systerconfiguration.
This work was supported in part by funding from the Air Force Office of..7 Scientific Research, the National Science Foundation, the Arm, Research
Office, and NASA Lewis Research Center.
Nonlinear Abhsorption of AIQW's Lee, et. al.
L 'v .'2
19
" -" R F IT R 11- N (T I-S
11 R. IDirigic, "Confined Carrier Quantum States in Ultrathin Serniconduc:orI ~erostrac~ure:C, in L"tkpprobmmb e,. (Advances in Solid State Physics), editedby 11. J. QucissCr (Pergamon, New York 1975), vol. XV, pp. 21-48.
'-.- 2 D. S. _hc-nla and D. A. B. Miller, "Room-temperature excitonic nonlinear opticaleffcts in semiconductor quantum-well structures," J. Opt. Soc. Amer.. vol. B2.pp. 1155-1173, 1985.
1 D. A. 3. Mi Her, D. S. Chemla, D. J. Eilenber,,er. P. W. Smith, A. C. Gossardand . . TSana. "Large Room-Temperature Optical Nonlinearity in GaAs,GaI-..x-lAs Multiple Quantum Well Structures," Appl. Phys. Lert.. vol. 41, pp. 679-681, 1982.
"] J. S. Weiner. D. S. Chemla. D. A. B. Miller. H. A. Haus. A. C. Gossard. XV.Wiemann, and C. A. Burrus, "Highly anisotropic optical properties of single--
2 0)
-[- tc. Lee. A. flari:., P. D. Dapkus. A. Ko.t. 1. Ka ase, and E. Garmire,
:\Ga\s GaAs multiple quantum \XCl nonlinear optical materials grovvn by"uctaloreanc chcmical vapor deposition." 11-6. Electronic Natcrials Confe-rence.Santa Barbara, CA, June, 1987.
•-.n R. A. Loan and F. K. Reinhart, "Optical Waveuides in Ga.s-AlGaAs EpitLxial
Layers., I. Appl. Phvs., vol. 44, pp. 4172-4176, 1973.
1 5' 1 . F Fou'uet and R. D. Burnham. "Recombination dynamics in GaAs A.-!GA,1S.4 cuantum wekl structures." 1ET J. Qutianun Eit'c:ron., vol. QE-22, pp 1799-IS 10.
19S6.
1101 H. C. Lee, A. Ilariz, P. D. Dapkus, A. Kost, M. Kawase. and E. Gar,3,uire,"Nonlinear Absorption in AlGaAs. Ga.s Multiple Quantum Well Structures Crownby NMeaorganic Chemical Vapor Deposition," ApPr. Phys . 'u.. sol. 50. 70.IiS2-1134. 1987.
1 ,- P. D. Dapkus. H. M. Manasev it, K. L_. Hess. T'. S. Low, and G. E. Stii!lmar."iih purity GaAs prepared from trime:hvlailium and arv rnc" J. of Cr.':.Growrh. vol. 55. pp. 10-23. 1981.
S1[ i. Hauc and S. Schmitt-Rink. "Basic M, dcani>ms of tee opt:cdil nonliincarities ofsemiconductors near the bandedee," J. O.. Soc. Amer. vol. B2. pp. 1135-. .-2.19S5.
1 191 W. H. Knox. R. F. Fork, N. C. Dokncr. -). AB. MlNl!r. D. S. Cherua, ane- C.V. Shank. "Femtosecond dynamics of rc,oriantlv excitcd excitons in room
-1 temperature GaAs quantum ells," PKv. A''. l.w., vol. 54, pp. 1306-1309.1985.
.201 D. S. Chemla. D. A. B. Miller. P.W. Smith. A. C. Gossard, and W. Vie,.raun."Room Temperature Excitonic Nonlinear Ahsorption and Reflection inGaAsi'AIGaAs Multiple Quantum Wcl Structures," IEE J. Q:antur : L!c'c:r.?.,vol. QE-20, pp. 265-275,1984.
[211 A. Kost. M. Kawase, E. Garmire, 11. C. L.ee. A. lar:., and P. 1. Dapkus."Contributions to optical absorption in Ga..' AlGaAs multiplc quantum wells." inTechnical Digest. Optical Society of Arn.rica 1987 .- nnual M\cc:im (O"tical Soc-etyof America. Washington, D. C.. 1987) p. 4fl.
'221 See for example : B. .esko\ar. C. C. I.. P. Iartie. and K Saer." "Phoncounting system for subnanowecond hlurcecc :itetime reent
21
19~ S 7.
G.ti D. Bovd. D. A. B. Miller. D. S. Chemla, S. L. McCall, A. C. Gossard, and J.Hi. Engclish. "Multiple quantum well reflection modulator," Appi. Phvls. Lett., vol.45op 19-1121. 1987.
* IT} IH. C. Lee, K. M. Dzurko, P. D. Dapkus, and E. Garmfire, "Electroabsorption in* AlGaAsGaAs multiple quantum well structures grown on GaP transparent* subst~ate.' Api. Phys. Let., vol. 51, (Nov. 16. 1987).
M~.N. Fox. A. C. Mdciel. MI. G. Shorthose. J. F. Ry an, M. D. Scott. J. 1. Davies,* . a:-d J. R. Riffat, "Nonlinear excionic optical absorption in GalnAs'lnP quantum
wells." Ap!nL. Phv, s. Letn.. vol. 51. pp. 30-32. 1987.
N\otllfiit'ar A1b.wirptiott of 4'Q IV 's lee, el. al.
22
Table 1 -Characteristics of MQW Structures
S-mple ; V158 V159 V160 V136
Gruwth Temp (C) 650 700 750 750
B:ukaround Carrier..Concentration (cm- 3 ) 6x 10, 4x10" 6
SNumber of Periods 61 61 61 51
We' Thickness (A) 100 100 100 100",Brrier Thickness (A) 100 100 100 100
.\ .Comvosition 0.32 0.32 0.32 0.32
0 E h (e\) 1.4571 1.4553 1.4588 1.458s
* Ei - C, V) 1.4659 1.4663
FWH\1 (meV) 15.4 18.8 20.8 >20
Nonliniear Properties of A!Q)IV's Lee, et. al.%%
.0 -I -I., j . *2 .I. Cq~, :'e:-_ ?
23
Table 2 -Characteristics of MNQW Structures with Different WellWidths.
'1apl 38 237 240 236 239
- ell Width (A) 54 72102 133 193
*-of Periods 100 67 50 33 20
GaiAs Active LayerThickness (i im) 0.54 0.48 0 .51 0.44 0.39
Al Composition 0.321 0.32 0. 32 0.32 0.32
D-rrier Thickness (A) 15 0 100 100 100 100
S..
24
Table 3 Summary of Nonlinear Optical Properties of MQWsamples with Different Well Widths.
Sample . 238 237 240 236 2 V
\cll Thickness (A) 54 72 102 13 193
Minority CarrierLifeime (nse) 59.5 68.7 205 63.8 024
"hh (nm) 825 839 854 So2.5 870-lh (nm) 815 829.5& (nm79 810 831 860 830
... ex -1 1.3Ie h ( ain 1.74 1.41 1 0.77 0.t
*" - 1.16 1.- 1.1 1.14 S 640. %,q (.n1 -,• . ex
CtIh (.m ') 0.767 1.06
a, (LLr- 1.90 1.65
Sex (jl m 1 ) 0.0 .15 0.31 0.35 0.24
a (Lirn ) 2.35 1.96 0.71 1.24 1.2S
Saturation Intensities below measured in k\ icm-.
ex
sathh (corr.) 0.062 0.074 0.020 0.128 0.0023
b (7h (corr.) 3.11 26.4 9.00 10.1 1.50
e x(corr.) 0.065 0.197
bI 'h (corr.) 3.19 19.3
O.x. la~con (corr.) 0.012 0.10 0.1) 0.02
kI.d(C."
1.V b (corr.) 10.3 26.4 3.88 9 67 1.54
Nonlinear Properties of AQW's Lee, et. al.@
..V ' . .-. . . a 5. . .
List of Figures
vi:: 1 d SJ 1nit dray. 2 c typi Cal IQ I )Vrin In1Car opti Cal 10.pl
_,th rowkn at b0i) C. 700:C, and 7"50 C'. No'(tc th rsne ttepw.1aen iint nd heav% hiole transistioOTS.
l~. L~rceolution roori temnperature absoi-ption spectra of three MIQ\V saimnpc, %%L"C IX. d. eli wits ':rown at di fMeCren t 'Le, I'erTa t ure S. Sami \ 1 C o wa 7rv 1 '
a lowk pufltvt meh yalu source.
[1.4Roor m t'r~ r'notoiumTilnescenice spcira of five MQW samples "k11-wellI withs. The various confined state transitions are labeled.
[h.5Room -emncrature a nrearabsorption spectra of fie IQW Aatie i:n
12. 6 Room Wtmpe1ratUIe pnotolurinescencc time deaso h he \Q zine-li. Te decay.s were recorded with an excitat on :nten"*!%- of ;0 W cm-. 1
eCua:,.tton source yXaS eItiin at 54 v.
(d). I le transmlssion snee:tra of tihe sample wkith 72.\ we! wI ICth at uW2ine2.uerten iter te )[ot the prese)nce Of lkibrV-Per7C, resonnceT1: T11:n1!mL d"
S22 nn and S55 Tim. The s uta ere tajken at itntesta)2.b)>c~>an~d d) 16.500 NV. cTT,.
Tb Ihe ab, orpt;CnI spe Ct11ra correspondilnei tLo (a) after corr7:n fr thec reti c:::':Lof front and back surt"accs and Fabry-Perot effec:ts.
[12-. 8 (a) Te absorption spectrum of sample Vi158 at various excita: on intensities'.(b) The excitation dependence of the heavy hole resonance absorpti fcr ape
S 5. The solid curves shw he excitation dependneoh xio~ nback _ rouncd components of the absorption. The saturation IIntelit ofte ec::o~component is 250 Wir M
T g 1he experimental behav ior of the heavy, hole resonance absorption coefficient as atfunction of excitation intensity for four MIQW' samples with various well wv'iths.The solid curves are the theoretical fts to the experimental data showine, tw\%ocomnponents of absorption. The horizontal axis display Is the logarithm of ir''""'-::units of kN-crn2 and the vertical absorption axis is in units of~in
I . 'I ino absorption coeff"icienit (a) at the light hole resonance an.d (b) in the cont*:TiaUn;1state,, as a function of excitation for an \IQW sample wvith 54A\ wvell width1s.
the procedure for determn ite curxe fit paramecters. forte
dan,,orption Satturation carves). Thec pro cdure is Illustrated for the NIQ\V samt'!lc %k:10A ide wls
1K epndeceof the uin-Kitoratedl heavy hole cxci ton tibscrpti on coceflic icr:aa[aet nf the wel w 1 t I 'I lhe solid Curve show.s a I, .j.depndence.
ANonlinvar A/)sorptionl of [1!QW's Lee, et. al.
26
-~I I:cr ofcee the (a1) satura -t on earner dens tN'.n,, and (b) exc.,on radius, rC\ . as atuco.of the well w idth.
:2 4 kVc2..i[cespctrunm of a 30 period Al0 :Guo .SAs:,AlAs 113ag- reflector grovvn b.,
1:2 ir::\~i~or~spectruin of .lGaAs GiaAs NIQW electroabsorption inodulator -,row'Ton GaP ait vaous applied votges.
1>c ~:uurio enrgyflueice Lr. as a functon of the wel wIdt f:ia\.iI 'N' nsc ,., 'u n e,1,-l Od h o \, i:sC L.\
6Q
Nounlinear Abhsorption of AIQ W's Lee, et. a!.
-0 0-40 It L
-low
If))
6c
6L
Is
S..S..
6NSN
8503A1.4584 eV
FL at 3000K
8520 A1.4553eV
C
FWHM=20.8meVV 160 - 750 0 C
z
FWHM=18.8meVV 159 -700 0C
FWHM= 15.4meVV 158-6500 C
8000 8200 8400 8600 8800
00
WA E EN T (A)K-:K--/.'\-
LC)
C.)'
U 0
U) 0 000 PO 0o 0 0u)r rf C
CC) L)-
too
0Ltfl
4-n
0n-o
aNO J0 0Oi8O9
b .".1U"I.-'
300 0 K n=l
n2
S3
j 0
-5 L-193A
- .H->._ n=i
00z L= 133A
"- F-
,;-.:: zIh0 n=i
-L L102A
hhh0~0
So,-. L 72A n =
... '': L =52__ hh
• I I 1 I
1 790 810 830 850 870 890WAVELENGTH (nm)
S::
S
. .. 1
n=4iL= 19hhh3n=
nzll
Ih hh
* 70 750800 50 9WVLnGT5 (nm)n
0L 133
n= n
0i
0 In 2. *.. -
(\j
0
00LC)
Q) 0
o C)
0 U)0 10
00r;7 w
Q) o n0K) C'j
FU-)
2 0IC)j
*n 0 0
o\IN~~ 11AIIV I
V 1 ---
0
co
co-QU
cq 0
00
J(\
-Io
0CC
0LC)
0C
-cC0
Lf-
0
0LL
0
(WTI) iN]I31JII0OD NOlII&JOS8V
AI2e A
-7-7 T-t TVT'" 's qp w W T' -1 W. -M
000oc
>0
CLO
(9
cOD>
00
0 600*~C~ LOT/) INICi]D OjdO8
0 o
0 00 N *C
3 c5 d
0 0
0
-0
0 0 - FDz
0-
0
Z-11)
54~ 102A
f il iI I j I ji I I1(4 Il MI I Iili i I I'IT f ifilII Ili i I
c d
2.m13 1934
3 -32-1 0 1 2 -3 -2 -10 1 2 3
-~2-
31
b
2
LU
*INTENSITY (kW/cm 2)
1.5a
Lz-102 A hh-854 nm (OW)
E 1.0-----
0.0
0 10 20 30 40INTENSITY (kWi/cm 2 )
2.0b
Lz~l102 A hh-854 nm (OW)
E 1.5-
O0 .0
~ 0. 0.1f 0c= .62.473 3
INTENSITY (kW/cM 2 )
CD
LOW*~
CD*C
*~~~~~C (LL)IiIO FV OJX
SC
Su
- . . . . . . .. . . . - - ' . - . . . N -
°-?~ 10O8
E I0 a
'0 U
U)
Li€m
o101. 0 50 100 150 200 250
WELL THICKNESS (A)
150b
D 100
.: z50 -0
00
0 50 100 150 200 250
IWELL THICKNESS (A)II,
' I X !
I 0
00
I -
0 0
*(juaojd) ]ONVi9T-l~i
9 0
""Oa,
'00'.0 I~/0
.1001,/tl 'o/
/l
S C3
zz
E
80-06
0 40-
40i
o 20W0 50 100 150 200 250
WELL THICKNESS (A)
, 0 , - , , 4 , w . ,- ,, ,, , . . . , , - - . . , , , , . . ~ . ,4 . , , d d .c
0. ."
I;:' 1'
------------------. *
IV. Growth of high reflectivity Bragg reflectors and the
electroabsorption of MQWs grown on GaP substrates.
To cbvi.ate the need for removinc the GaAs substrate when implementina arrays*. ofoo:ocnlc s,,:c>-es, we have investigated the use of novel materials struc ures
tr: rely upon the unicue characteristics of MOCVD. The first of these is the DBR
* c- or. Th:s is a multilayer stack of quarter wavelength layers that ailernate in
c osi;ton anc, thus, refractive index. Thrcugh appropriate choice of materials
-. - o moo-'t'c it is possibie to crow multiiayer structure with exceedirgly h gh
-.. ;vty a kno,.In spectral band that is also transparent to 'he reflected light.To mat... to waveincths f Icht suitable for MOW appicat-n we have chosen
to ccnstr, c, our reflectors from AlAs and A 0 2 Ga 0 8 As mu'tilayer stacks. Fig. 1
s rcws a typical reflectivity sc, ectrum from one of these sam.les. Several have
.-'.2 c-ocr which show ccmrarable results and device structures have been
.r'wn on too of them to verify that the surface morphology of the layers was
...-. .... ......, These structures wih allow us to f-bricate -e -:,c .- arrays u:, ,r.
1'. LFP or hybrid devices.
"drsorss~ve cevices presenoty require the removal of the GaAs substrate. Wehave recently demonstrated the growth of high quality MOW structures on
transparent GaP substrates. We have measured the nonlinear absorption in
ese structures and found them to be exceedingly high (10000 Wicm2) ands* -. unsuitable for NLFP devices. However we believe that these structures will be
* of great benefit in the fabrication of SEED and SEPT device structures. The
"-' ,electroabsorption properties of thes structure has been measured and is
described in the following paper, published in Applied Physics Letters 51,1582
, .. (1987).
A
M,.>/ f:''/',- h:r Optil. ,S .n:'rjI Prcces.,•' fl.?pk~s ,ar'd , :, ?m'
- - l.,i~ liliJiml'.M |I .,.. . . . . . ..l i l - - . .
I - - - --~---------------------------------------------
mc
*~ 0-
_ (co
_ _ _ _ -
_ (-5-GO- _l QO Zn"i
___ __ _ _ Z-
Eiectroabsorption in AIGaAs/GaAs multiple quantum well structure grown-~ on a GaP transparent substrate
i-i . Kr M.Lr~ PC D~ '~
I i:tt' I~c!I 11M ti'~ of ~ " Icrt 0,1, r i il 1 tt\' (qii Hii itim 'III ,1111i11 itt T r
\'!t9I l I I I icI~ r , -, \\t m \ it %Il ( I ill I 'nt rrlLdti (C acr bet o:L:Ii
l, c ., 111 111, 1~ 1u h: 'iHlrl tc ( [ ' . t*11.c .. orpuh oI, )i\ a nd .i hr Ipt IlI CI-1i''.L 11x c
i I 12 h! . h \ k~ It]h c-tt t ~I cI V t -i I :L!, oN Ill ~ c A ~ r 5111CV -2f tTI Ih,,o IIIOII rCdii'
2''
iii
- I .t 71 XC
Ic I I cri.............
it ii I
I,t tr i tha t hr. m io i I( \ict to rcm( : . t 1
11, :,l 111 1' ' It P ' trn t f kn' I'Ic'c I('It
I ir. Ii nc 1 i ht r fil l, nit:,% I~\ itc c fu rt hcr (IClT11iMI-I , lti ,h',r lt M2 h(~ (Ml~tu ti I C Kh ')S I 11T 1
* i1 iiiit t ~iiii~~..It t- Ii it
ec : - .: L tmc ia, '-Allimc u,
7 -i 15' r 0i 1p -a '- 1 V b
1"
r7 IL 7t)'r j. nii r-llfl l i i: r % i o a c n t e t T1C J ~ ii - Ia t
*~~1): at 11r a:.c 2 t '- - ' I82>2. A..:-ire ti
- ~.-.I.:-Li ' 7 :e,., -. ~..-.. A22 ute i - 5>\A m-curo x rrl i 4)\ V
- -w' .. F W.v. ~ ' -rV
:rorrr: ac1. I > is .h , ii F .D. A Jcrae i i 'n
hi~hc 'eltceN r it >n'r'icr'~14ci ll,
\IQ~k tma Pv tt~r' ic t ,ub Irate %%it .l sirtere >cp
6 P2 YI hutlCrF ' 1r I' " r'a r it 011n abhser' cH Inttli
11t-r11at1x to P me currui f '.)dinc 2t r'o fl ' me (aA .U
sr a t c Lr c a ra _ t;I rrr:I Lt 7:i tot i\. .s e i a Ie
,5ICui: tU1ic- Rxearch 1Lmd, th Lh'.ert Souher
ai:rr'Joint Se;r e> Lt::rottN j r !"IF rnai
siurpor: 'Ct~his \x ark.
- . ... -FeT-i r reerse maid
2 1- _.:_ 2 3 I:.:,.4iL. .7
....................-. u
2' -NJ..-rc
-rr P 44.'1' 1
M .. '7 Z - wl . h r ' ~imj ,Itt.D V B %ilc.A Ci,7 _:''s- % f hcktes - i i ne .>i a r - i mann . \ -iPhL ett:.: 4D2, t:-. k...i .. J..f.h3.T
.4~~~~~ ~~ N .l ii y at r I 4m *~i P~.L:: .tAV* { Ien NI WS . .rla... (. % ~ict wr. 2rr~ p.
u~-~ I '>.4)p a~ :; o m a ii z e . _ r n C . .% 4 4 6 ~
-w -V1 _V
- .V. Implications of our research for optical computing applications.
The use of exciton resonance in MQW for an optical logic or switching devices
means that a large nonlinear refractive index occurs because of a large amount
of saturable absorption. We have found that at modest power levels there
remains a relatively large amount of unsaturated absorption. It is the purpose of
this section to consider the trade-off in performance between the size of the
index change and the unsaturated absorption. We will design an nonlinear
Fabry-Perot interferometer working on the exciton resonance, where the
switching intensities are the lowest. Our design is based on the measured
performance of MOCVD quantum wells. We will show that operation in
reflection provides high contrast, low threshold, low loss operation. Conversely,
operation in transmission is prohibitively lossy in the low threshold regime. This
* is a new result and has profound implications in design of future devices.
A. Material performance.
Figure 2a shows the change in measured absorption with varying input
intensities as a function of wavelength within a MQW of well-width 100 A.
Figure 2b shows the related change in refractive index at the same intensity
levels as a fuction of wavelength, This curve was generated by using the
Kramer's Kronig relation between the change in index of refraction An(E) and
the corresponding absorption change Act(E'):
•ch P Aox(E')dE'An (E) = ( ) PE E2 (1)
where E' denotes the photon energy at which the absorption change Aa is
measured, E represents the photon energy at which the change in refractive
index An is determined and P indicates that the integral is a Cauchy Principal
-Y Valve.
. MOW's for Optical Si;gnal Processing Dapkus and Garmire
S%
-9-
- To use this relation, we found it necessary to convert a discrete set of measuredpoints .\t(E'i) into a continuous set by linear interpolation to avoid problems of
convergence. We were then able to analytically integrate between theS. measured points. Using this method, the Kramers-Kronig relation becomes
n j"."A n(Ei)An(E) EX h Em2{ _ E (2)
i=0
where m = i+l - E'i and bi=Auo- miE'im, E'i~l- E1i
In this expression, E is the photon energy at which we calculate An, AoX is the ith
- - absorption change measurement, made at photon energy E'i, and mi and bi arethe slope and horizontal intercept respectively of the line between Aui and
A.i +1.
The integral can be evaluated analytically to obtain:
I 0in(E) = C)~ In I ~ " I~h 1 E(~>I
,sr i E~ 2 _ E2 bE (Ei-l E)(Ei - E)
The parameter which is of physical importance in device design for the
nonlinear Fabry-Perot (NLFP) is the ratio of the index change to absorption
change, as will be shown later. This ratio, calculated from experimental results,
is shown in Figure 3.
B. Existence of Two Saturation Intensities
What is notable to the understanding of the use of the MQW in a NLFP is the factO. that there are two satuaration intensities. This can be inferred from Figures 2
and 3. At incident power levels as low as 250 W/cm 2, it can be seen that the"- .exciton saturates. That is, its absorption changes from 2.8x10 4 cm - 1 to 2.1x10 4
cm- 1 . This 30% change in absorption is not sufficient to cause any usefulO. devices based on absorption alone, however, at this same power level of 250
W/cm 2 , the induced refractive index change is as large as 0.038. This induced
MOW's for Optical Sgnal Processing Dapkus and Garmire
- T t7 T'-..7=
File: ABS158Abaopton vs Energy. 158-acw.'O 1 06
:'2.__... . .. . . - i -
I __-___ A 'i 2IZ _ _-_ . . . . . .
* _ _ if __j I i_ _ _ _ _ _ _ _ I_ _ _ _ _ _ _ _ _ _ _
/ -, ,, .. . ..-
,- ,-C it c m . - . .; - i -l - -
______5 __ _ _ 2 c:7
-'-7 I. .. . .... .. . . . . .__ .. . .
1 43 145 1 4--1 _ -_ . n c.
. .(I e nf
-:. 7 I
L. . ... _ _ _ "% . .. __.. .. I _ -:.. .
S1 3 1245 147 149 151
• Energy e!l
>:'uro .a. Akborption as a futnction of energw: for di fferent levels
. of incident intensity for 100d A well.
-0O3
:". 005 2. c- - - "2_--_ _-
00 • _ _ ,i-i-. ..-:) -- tt -I - !-001-;--- -C- _ _- i
-002 1 \, - . . ..> 0042 ~ _ .. . . .. . . 1, !__ : ... .- I.
0 it a ( T I ,, , . ~~ , 1i"-A 752 . - Jl n !. .
%/
- % % -e7(•-. * I ' i,-- . 1 II:..,14.3 145 147 14'9 1 51
i'.' " " :' l l'* ' K -..I -:', , r r , f r i c t i v * i n t (' ,, ; a ; ; f il l p t i o n1 of l ih i t ln
r .- ('utr t,,' a; a fun n t til on f it n t 'n t I tn t on t y#
,,,:c nI a ;t '1 f ron V ra;::,rg Frot;i rl t m'1a 1on;.
S) :
* "t1 "" - @ ' " " w, ., . . - . # .. ," . " i
" 4 " . " .%, . " . . ' . P .
,- . .. '' .. '''" • ". ." ",". . ,= ' . ' . . "' . -' . % . i% • 5 - % . " % " .
index change cannot make a switching device because of the large unsaturated
absorption due to the band-to-band tails. However, we will show that it is not
necessary to complete saturate the band-tails to induce switching. In fact, wepredict switching with intensities of about 1000 W/'cm 2 when operating right on
the exciton resonance. By operating a NLFP in reflection, we find that excellent
-.' performance for optical computers can be attained.
First, however, we decided to investigate how to engineer a MOW with the
minimum band tail absorption in the presence of saturated exciton resonance.
This requires understanding the relative contributions to absortpion from the two
saturation mechanisms. In Figure 4 we show a careful study of the intensity
dependence of the saturation energy, measured on the peak of the excitonresonance of a IooA well sample. It can be seen that the absorption is
- composed of two compnents, that from the exciton resonance and that from the* band tail. The absorption decreases from each mechanism, but with a different
saturation intensity. It can also be seen that the contribution from eachmechanism is initially comparable. At an intensity of 1000 W/cm 2 the excitonl-,as almost completely bleached, while there has been very little reduction in
the band edge absorption.
C Dependence on Well Width
1 The question which must be answered for device applications is whether there
is an optimum well size or an optimum wavelength at which the bandedge
absorption is minimized in the presence of saturation exciton. For this reasonwe undertook a study of various well sizes and various excitation wavelengths.
* The results are summarized in Figure 5, which shows linear absorption
measurements of these various wells and experimental data points whichindicate the relative contribution of bandedge absorption and the exciton
resonance.
The results of this extensive study were that for all well widths, the contributions
to saturable absorption from exciton bleaching and band edge absorption are
comparable. We also found, not surprisingly, that the ratio was largest directly
* on the exciton resonance.
.
voWs for Optical Sqnal Frcessing Dapkus and Garmire
0
* - C - .- ,. A.-... *..44
A s,
c r
0.
~0.8
~0.6-0
cc)Q.5
0i0.
H 0 .3LU
L0.2
0.1 EXIO LAFN
0.01 34 5I 10 1 10, 10 10
-*lr A', rp 1tiui ;i, I ftitnct ior of inton!;i ty , f or the waivelengthri 5,11t oil (,X( i toll Ii l Pr in t ing~ tim( COH t r ib l 1)11on(C f rol-
- Xl I t (1t1o 1) 1 (1 li C i, h I p~ l, ~~ f 1 1 1 1 H n O I ar Cn( cI t
0I
'V -,
The bottom line of this study is that at exciton bleaching intensities (-1000
MW cm 2 ) there remains roughly 500' unsaturated absorption. However, if the
intensity is increased to 10,000 W/cm 2 , even the band edge absorption can be
bleached.
D The Nonlinear Fabry-Perot Threshold
Because of the lower saturation intensity of the exciton feature, we were
interested in exploring the operation of a NLFP in the presence of an overall
unsaturated absorption. This loss will alter the performance of NLFP, which
must be properly designed. It is the purpose of the following sections to show
that the MQW's we have measured will provided a low threshold NLFP,
operating in reflection with excellent characteristics for optical computing.
The transmission of a FP in the presence of loss can be written as [for example,
see Craig Poole's Thesis, USC, 1984]
BT I +Fsin 2 o (3)
where-" 4 1e e-aB _ L(l RB) (1-RF)Li F- [IR. and B = 1R)2
1 Rej' (1 Re)(4)
with Re (RFRB) 1'2 exp(- aL).
(5)
Note that neither F nor B depends on the order of the two reflectivities; that is
whether, the large or small refractive index comes first. Note also that F does
not depend on whether the two relfectivities are equal. It depends only on the
effective reflectivity, Re. In the absence of loss, B = 1 and the on-resonance
trasmission is 100%. When loss is present, the on-resonance transmission
drops. The introduction of loss also decreases the fraction F through the
effective reflectivity,
MOWs for Optical Signal Processing Dapkus and Garmire
* * *- .. . . . .I:S "- -~.-. ... ....- ,.. -:..-..-* . *. ..... ,.. .... ...-...-. : .. .... ...
The sw;tching threshold in the presence of loss can be most easily seen by-"considering the graphical demonstration of the threshold for nonlinear switching
(Figure 5), which plots the relation between the input and output light intensities,
I OUt/I in = T (o), (6)
where o is half the round-trip phase change within the cavity. For arbitraryintensity-indued refractive index change, An, we write.
.Ao = An (Icav)dz. (7)
The index change An is expressed in terms of the intensity inside the Fabry-Perot cavity, lca v , which is, in general, a function of positon z within the FP,which we will measure from the output facet. The cavity intensity is related to
O- the output intensity through,
'ca v (Z) = C(z)I out, (8)where
C (Z E?(z + RB e-(Iz1-HB
* . The expression for C was fouund by adding forward and backward travellingwaves inside the cavity and ignoring standing wave effects. When there is noloss, C = (1 + RE)!(1-4B).
When the phase depends linearly on itensity and An - n21, the round-trip phasechange can be determined by a simple average, and
. = n2 (Icav)kL DI out, (9)
lw h e re
"-'e(, L- 1 +RR(1 -e -(iL) ( eaL R3-L ~ aD n2 k n2 ke.L)(Rgc-CL+l)eUL
(,-.(1- RB)B)OxWhen the propagation loss a-L is much less than the reflection loss, 1-RB,
O,
V O W's for Opt cal Signal Processing Dapkus and Garmire
"S ": ''"''''',¢,'' " - "'"". € . ' . . '''., '".. " - "."€ " . ..
0
°7
~AB SURPIYII (JF AIQ2\V'S
"° - I
2 u*
00
.70.75 0.8 0.85 0.9
iiv~i~t ~ii
m ultiple q u,-n turn w cls. The (lots reprosent the fraction*of t I e n o r p) t I o n dI u e t o L he ( ha c k< F, r-n un (I un d c, r no va t h1 t h1
fx c It on f e at ur e, mea!;ured b Ittn o h itiaiabhsonr pt ion.
-So
'..
('caV) (1+Rp): lout (1-R[) (10)
. Figure 6 shows graphical plots of equation (6), which the right hand side is the
transmission as a function of o and the left hand side is a straight line of slope
. 1."Dlin. Figure 5a shows four examples: a) at low intensity (large slope), in
which there is only one intersection; b)at critical intensity, so that the slope of
the straight line matches that of the FP transmission; c)at higher intensity, in
which there are two stable intersections (the region of bistability); d) at highest
critical intensity, which is the threshold for the switching from the low to hightransmisson state. This critical intensity is the threshold for optical bistability.
1 E. Requirements to achieve nonlinear switching.
_ There is a minimum critical intensity below which switching is impossible. This
occurs at a particular round-trip phase bias, OB, as can be seen geometrically in
Figure 5b. This bias ensures that the slope of the straight line intersects the
transmissicn curve at the point of its maximum slope. From the figure it can be
4 see that to have nonlinear switching, we must be able to change the round tripphase change Aoc by an amount roughly equal to the half-width of the
transmission resonance. When the finesse is reasonably large, this width is
given by
2f f (11)
2ntwhere f is the fineese, defined by f -F. This will occur at a zero-intensity bias
phase for the FP of oB o m,- - .
The critical phase change can be re-expressed in terms of a critical average
index change within the etalon, cc - (An)c kL. We write1 .
, (An)c = kLN\F. (12)
Thus we minimize the required index change by maximizing
L IF =2 .,rR e 1,*L\ F = -- T -
AMOWs for Optical Signal Processing Dapkus and Garmire
I%,.
i. . % . * * . .il* * "l.4. .. . -- . .
Clearly this is maximized by maximizing Re. From equ. 4, it can te seen a .
is maximized for maximum RFFB and minimum loss. Defining RF = 1 -TF ar-.c F>11-TB, and expanding square roots and exponentials in Taylor seres
(assuming small loss),
1 - Re (TB+TF)2+ c L.
Defin:nc a -a.rameter u such that (TB + TF) 2 LdtL. we may write
2:-:.:.(Z[1 -,-u)
P uL"- t-'e ecua.-c-n
(RF -R,.E = RR 1 -Re),
It can be seen that when E =0, the minimum ref lectivity =0. This will keep as
m,,uch intensity inside the FP as possible and available for sw.itching. T's
means that for the lowest input switching intensity we require.
RF =R,' or RF =RB exp (-2ctL).
* - In order +o calculate the required incident intensity, we notice from Figure 5 ta3B
te transmission at minimumn switchina th reshold equals 4i-- This mean that
* 'Cut. M 1 i n-m
This expressi; allows s to re'ate the input intensity to the Phase chargxv
within the cavity at switchir- :Iogh equation (9):
3B1in m 11)o D ---- ~- which we set =F 14 1'
This meanis I inmr 4 1;
2u (1 e)e ,LC) l'in, ML
3n~k
-. aC
-:1
-, r p i lI o s r c o or tlI S i i o a n n n
Fa r -P r t 1%utiol for arbitr I n tIaI d Lun n B
.) itt l r cr ca nitI Ieti
%~
0
(1 -R;:e-2(L 1 2
.Assuming R,- is given, we set the differential of this expression equal to zero !o
-obtain a condition for a minimum of incident intensity. The result is
-- In the limit of R -1 1,this given e IL ~1. Thus, write R 1 -C. Then
3 (1-8
Expanding in a Taylor series, in the limit of small 8, A 1 -7- That Is. e (4L
-8. 2, or 2cfL = 1 -RB. Thus, an optimum design has RB 1-2fL RF = RBe -2
1 - 4ufL. This means that the loss in a round-trip through the medium equals :he
loss in the back mirror. The loss is the front mirror is two times the round-!:-p
W loss,- Inserting this cavity design into the required threshold intenity, Eq 3, Lurdc-
* -. condition of low-loss,
The incident theshold is minimized by increasing the back reflect ity as murk* as possible. To compare this to the data on .An(I) which we dete--m'->c
experimentally, we write(i'm 4?.
I inm=- (-R[3)-
* ~where t rn is the intensity at which .\n is measum, dU2rqt'*-0 2.im, and the fact that 4'/,'7 1 lhim, a reflectiv *y cf R5 7 v. r:
0
That is, if we assume An is proportional to I, the incident intensity can be as
n-ismail as 4c of the intensity at which the value of .\n (t. = 0.2 !.,m was measure.
C Desiq" of ootimum NLFB based on MOW
We have found that the minimum Anu required to achieve bistability occurs for
m ratio of reflection loss to round-trip transmission loss) = 0. However, we haveseen that the minimum incident intensity occurs for p = 3. Optimium peformancein a given system will require a trade-off between these two extremes of cavity
design.
When p- 3 (the condition for minimum incident intensity) the reqauired An/L
(from equ. 13) is An/cL = J2= 0.26urn. However, from the data of Figure 1 and 2.
. the nonlinear index is not linear with intensity. This means that even high cavity* intensities will not necessarily achieve the required n2/k. To ensure bistability,
it may be necessary to operate in the regime that t
I in, rn = (1) (23)3oxL
where Ic is the intensity required to attain An/a = /2, 0.13. Plugging innumbers from the experimental data for the 75A sample, and extrapolating
between measurements at 400 and 2500 W/cm2, we determine that Ic 1000W/cm2, and we predict a threshold of between 100 and 200 W/cm 2.
Table I. Modelling for a NLFP
Case I (High Reflectivity Back Mirror)
R = 0.98, AL = 0.05, RF 0.05, RF 9.92 lin,m = 0.151
linm = 100 W/cm 2 , Pm = 0.6mW. L 0.08 .m = 800 A.
Case II (Lower Reflectivity Back Mirror)
RB= 0.95, c.L =0.1, RF =0.85. 1= 011.lin, m = 200 W/cm 2 , PM = 0.12mW. L 0. 167 urn = 1670A.
For a minimum spot size of 25 p.m X 25 l.m (in the absence of diffusion), thepredicted switching intensity gives a switching power of 0.6mW per pixel, or two
orders of magnitude lower than has previously been observed.
The favorable numbers obtained in this optimization suggest that we operate inthis regime, where An/a has its largest value. This requires the use of very thin
* cavities, determined from the small values of L. For our estimate of ox = 0.6 l.m1 at switching threshold, this means MOW material thickness of only 1000 A.This length indicated here is that of the quantum wells only, each typically 75 A
*-.- thick, so that the optimized NLFP will contain only ten to twenty quantum wells.• .When the thickness of 100 A cladding per quantum well is included, the overall
thickness of NLFP will be on the order of 2000 A to 4000 A. Why bistability hasnot been seen previously at such low power levels as we predict here is due to
' the fact that thin enough samples have not been investigated. Growth on. reflective or transparent substrates will allow such thin samples to be prepared
easily.
&MOW's for Optical Signal Processing Dapkus and Garmire
% %
/%
H. Performance of NLFP
We will show that when optimized for switching with low threshold, the
performance of the NLFP in reflection is much than in transmission. This is
because of the existence of substantial losss in the transmission device. We
will also show that the NLFP in reflection has an output of almost 100% and an
off-state of zero. This is ideal for an optical computer.
The transmissive FP is characterized by its maximum and minimum
transmission (at o = 0 and o = 90-):Tm.' BT max -= B T min =(,1 +-F)
When there is no loss, B = 1 and Tmax =1. Otherwise, when RF = Re,B = (1-RB) e-aL (24)B 1-He
For small loss, and under the conditions that .= 1, we may write
B g- (25)2ctL
This means that, for our two cases considered above, B has values of 0.2 and
*0.25; at least 3/4 of the light will be lost to absorption , when the FP has its
maximum transmission. Any attempt to reduce the loss will increase the
threshold unacceptably. Clearly the MOW NLFP in transmission is not a useful
device.
However, consider the NLFP in reflection. In this case the relevant requations
are:
Rmin E, R max 1+F (26)
In our previous analysis we have chosen RF = Re so that E =0. This means that
the minimum reflectivity is zero. The maximum reflectivity is
F 1,-.-R ma~ix -(1+F) [1 +1/F]'
• As L: fore, when the loss is small, and under the optimum condiition, 1-Re =.,,.-,2otL:
MOWs for Optical Signal Processing Dapkus and Garmire
% % %
1 [1-Rett] 2 (L) 2
F 4Heff
Thus Rmax - [1 +(cL) 2 ] 1.
The result of this analysis is that a NLFP used in reflection has a maximumreflectivity of almost one and a minimum reflectivity of zero. This is thereforemuch more practical than an NLFP used in transmission.
I. Conclusions
It can be seen that the experimental data from our study of the MOW can te
used to design a NLFP with a threshold for switching of only 0.6 mW threshclc* by designirng a thin structure only ten quantum well thick and operating on the
peak of n,'ux. Under these conditions the reflective NLFP is much more useful for
optical computing than transmissive NLFP's. We can achieve large contras:S.' ratio with high output in the high state as well as low threshold. If three-pcr
devices are required using a relfective device, these may include two- polarizations, or the use of light incident at small angels. What is required in the
next phase of this study is a more complete systems study of the use of NLFP's
in reflection.
V1. Summary and Recommendations
The work described in this report represents the first systematic investigation cf
* the nonlinear optical properties of MOW structures. The conclusions of section V
show that the structures investigated here will be useful primarily as reflective., modulators with extremely high contrast ratio. As transmissive medium the
MQW's are most easily used as the active element in SEED or other hybr'd* optoelectronic devices.
The study we have completed shows also that MOCVD, under the apprpriategrowth conditions, is a most suitabl
