Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
1~81536
FABRICATION OF OPI'IC~L WAVEGUIDES
The present invention relates to the fabrication
of optieal waveguides, and in particular to the fabrieation
of optieal waveguides in integrated optical circuits.
In integrated optical circuits it is necessary
to provide op-tical waveguides of controlled refraetive
index to provide optical connection between the various
optieal components.
Such op-tical waveguides have been made by flame
lo hydrolysis, as described in US patent no 3,806,2~3 of
Keck et al. In this method fine glass particles are produced
as a soot Erom the fl.ame of a gas burner which is fed
with SiC14. The soot is deposi.ted on a substrate, sueh
as fused silica, having the appropriate optical and mechanieal
properties. A less heavily doped silica layer is then
formed over the layer of doped soot. Finally, the structure
is sintered at about 1500C to vitrify and densify the
so~ty layers.
UK published patent application no 2066805A
teaehes the use of a furnaee in plaee of a gas burner
to hydrolyse halides of Si and Ti, B, P or Ge, together
with oxygen or steam to procluee fine glass partieles on
a substrate heated to 600-11.00C. Vitrifieation involved
heating the substrate to 1300-1.600C.
The disadvantage o both these teehniques is
that they recluire that the substrate be exposed to very
high temperatures, partieularly duri.ng the vitrifieation
step, whieh limits the ehoiee of substrate materials (effee-
tively just to high temperature glasses: semieondue-tors
eannot be used) and dopant materials (no volatile speeies
sueh as arsenie or phosphorus).
~g
~281536
Moreover, differential thermal expansion be-tween the thin
deposited films and the substrate is more detrimental
in cooling from high temperatures. Such extreme heating
can also lead to poor flatness, rendering subsequent micro
lithography dif:Eicult.
An alternative approach has been taken by Stutius
and Streifer, Applied Optics, Vol 16, No 12, December
1977, pages 3218-3222. They experimented with chemical
vapour deposition (CVD) of silicon nitride onto thermally
lo oxidised silicon, and favoured low pressure CVD in preference
to atmospheric pressure CVD and plasma enhanced CVD, both
of which produced high loss films which were subject to
cracking. No details of reaction conditions are given,
but low pressure CVD of silicon nitride is usually performed
a-t 800-900C using dichlorosilane and ammonia.
The advantages of using CVD instead of flame
deposition are that the process is carried out at a lower
temperature, and by virtue of the reaction r.lechanism CVD
gives rise di.rectly to a film with better coverage and
greater integrity (with flame deposition the :Ei.lm is not
formed until the sooty layer has been vitrified). In
particular, the vitrification step, wi-th its very high
temperatures, is avoided.
D K W Lam, ~pplied Optics, Vol. 23, No 16, August
1984, page 2744 to 2746, proposes the use oE plasma enhanced
chemical vapour deposition (PECVD) as a means o depositing
silicon oxynitride (SiXOyN ) from silane (SiH4) and nitrous
oxide (N2O) at a very low temperature of 200C. Because
of the low temperature used, the process is said to be
suitable for use over group III-V semiconductor compounds
such as GaAs and InP which
536
decompose at 500C, unwanted drive-in diffusion of dopants
already in the semiconductor substrate is also avoided.
In spite of the advanta~es inherent in CVD
processes, disadvantages remain with the approaches of
both Stutius and Lam. The use of silicon ni-tride (n 2.01)
presents problems in coupling to optical fibres. With
Stutius differential thermal expansion can be expected to
be problematic. In the Lam process. although the use of
plasma enhancement allows the temperature to be dropped to
200C, energy from the plasma also leads to the formation
of Si-Si bonds, extending -the UV absorption edge. Indeed
in the Stutius and Streifer paper, PECVD i9 rejected
because the films produced in that way contained excess
silicon, and no guided mode could be launched at 6328A.
Additionally, N-H and 0-H bonds may be Eormed adversely
affecting absorption in the near infra-red. Also,
silicon-oxynitride, produced by CVD, has an unacceptable
degree of surface roughness which must be reduced by
reflowing at high temperature to reduce scatter loss.
Internal defects also occur in CVD silicon nitride and
oxynitride, and these too can give rise -to high losses
unless annealed out ~a C02 laser was uYed for this by
Lam~.
Thu~ it can be seen that there exists a need for
a wave~uide fabrication process which does not re~uire the
use of plaYmas or excessively hig}l temperatures. It would
also be desirable if the fabrication process produced a
smooth surface, and hence avoided the need for laser
annealing or high temperature baking to reduce scatter
loss to an acceptable level.
According to one aspect of the present invention
there is provided an optical waveguide comprising a
guiding layer of arsenosilicate glass consisting
essentially of arsenic, silicon and oxygen or a doped
silica glass in which arsenic is -the principal dopant, the
arsenosilicate glass containing arsenic, the arsenic
content of the glass bein~ at most 17 mole %, and the
guiding layer being formed on and in direct contact with a
~,~
1~:8~S36
cladding layer having a refractive index lower than tha-t
of the ~uiding layer.
Ano-ther aspect of -the invention provides a
method of Eabricating an optical waveguide comprising the
step of forming a guiding layer of arsenosilicate glass
consisting essentially of arsenic silicon and oxygen or a
doped silica glass in which arsenic is the principal
dopant the arsenosilica-te glass contairling arsenic -the
arsenic content of the glass being a-t most 1~ mole % on
and in direct contact with a cladding layer having a
-efractive index lower than that of the guiding layer.
The invention will be fur-ther described by way
of example only with reference to the accompanying
drawings in which:
Figure 1 is a diagrammatic representation of
apparatus suitable for use in depositing arsenosilicate
glass;
Figure 2 is a graph showir1g how changes in the
arsine and silane flow rates affect the deposition rate
with temperature;
Figure 3 is a graph of deposi-tion rate against
temperature for the silane-oxygen system and the silane-
arsine-oxygen system;
Figure 4 i8 a micrograph showirlg surface
roughness typical of converltiorlally deposited low
temperature CVD oxide;
Figure 5 is a contrastirlg micrograph showing the
smooth surface attainable with arsenosilicate glass
produced as the result of a heterogeneous reaction:
Figure 6 is a micrograph showing a step covered
by a conformal coating of arsenosilicate glass produced as
the result of a heterogeneous reaction;
Figure ~ is a contrasting micrograph showing
poor step coverage with phosphosilicate glass as in Figure
5;
Figure ~ is a graph of deposition rate against
temperature for the silane-arsine-oxygen system;
1~815~
Figure 9 is a micrograph showing the effect of reflowing
arsenosili.cate glass at ~00C;
Figure 10 is a similar micrograph showing the effect of
reflowing at 900C;
Figure 11 shows the RBS spectra of as-deposited and densified
films of arsenosilicate glass;
Figure 12 shows the compositional changes experienced
by arsenosilica-te glass subjected to a typical
fabrication sequence;0 Figure 13 is a depth profile of a 0.6 ~m densified filmof arsenosilicate glass;
Fi.gure 14 is a graph showing the relationshi.p between
refracti.ve index and arsenic content in arsenosilicate
glass;5 Figure 15 is a micrograph showing the smooth surface of
an arsenosilicate glass waveguide;
Figure 16 is a micrograph showing -the ripple-:Eree sides
oE an arsenosi.licate glass waveguide.
We have discovered that arsenosilicate gl.ass
(ASG) is suitable for use as a waveguide, and, moreover,
offers surprising~y low loss in the near inra red, despite
the fact that the glass is made from hydrides (SiH4 and
AsH3) which would be expected to give rise to ]arge losses
in the near infra-red. The refractive i.ndex of the as
deposited film varies with arsenic content, between about
1.53 for 12% As to about 1.45 to 1.7% As. This is a very
convenient range, making ASG suitable for use as a guiding
layer with silica (N 1.46) cladding, and also enabling
the production of waveguides in which the cladding and
guiding layers are each made of ASG. Arsenosilicate glass
is thus preferable to silicon nitride (n 2.01) for use
with SiO2, since a large refractive
F P ~ 3 ~J ~ J ' -' ' ' '
~Z8i536
1ndex dlfference necess1tates the use of very thln
waveguldes 1f only low order modes are to be supported.
For asymmetrlc wavegu1des, where the gu1d~ng layer, w1th
refract1ve 1ndex n2 1s bounded by two layers whose
refractlve 1ndexes dlffer w1dely ~n3~nl) the
followlng equat10n allows calculatlon of the th1ckness, t,
requ1red for wavegu1dlng of any part1cular mode, m ~ O, 1,
2,...... ,:-
/on ~ n2 - n3 > (2m ~ I)2 A20
132n2 t )
where ~o 1s the vacuum wavelength ~see chapters 2 and 3 of
Hlntegrated Opt1cs : Theory and Technology" by R G
Hunsperger, publlshed by Sprlnger-Verlag). A d1sadvantage
~5 of very th1n waveguldes ls that they are d1ff1cult to
couple to optlcal fibres.
Arsenosll1cate glass 1s an arsen1c doped form of
sll1con d10x1de, and ls conventlonally produced by
react1ng s11ane (S1H4), arslne (AsH3) and oxygen 1n a
CVD reactor. However, the react10n ls a homogeneous gas
phase reactlon and hence the ASG 1s depos~ted wtth a rough
surface whlch needs to be reflowed to reduce scatter loss
to an acceptably low level. Arsenlc reduces the melttng
po1nt of slltcon dloxlde, the meltlng po1nt decreas1ng
progresslvely wlth an tncreas1ng concentrat10n of dopant,
so that ASG wlth more than about 10/o As can be
reflowed at temperatures as low as 800/900-C I tn steam~.
Arsen k 1s present ln ASG ln the form of arsenlc
tr10xlde (As203) wh k h has a slgnlflcant vapour
pressure over ASG and 1t can readlly be lost by
evaporatlon when the glass ls heated. Because of th1s,
the arsen~c content of the glass drops dur1ng reflowlng
.
F F. ~ M r" ~ ~ ~3 r~ n ~ 1 I! t. . 1~ J ~
~8~S36
as a result of whlch the refract1ve 1ndex also falls. The
loss of arsen~c 1s not un1form throughout the thlckness of
the film; the greatest amount ls lost from the surface
layers, as wlll be expla1ned ln greater detall below.
Although wavegu1des can be Made by depos1t~ng ASG
uslng conventlonal processes, 1t 1s preferably deposlted
accord1ng to the method descrlbed ln our copendlng
European appl1cat10n number 85300172.5 flled
10 January 1985, and publlshed under the number 0150088.
o In that appllcatlon we descrlbe how, by us~ng the
appropr~ate reactlon cond1tlons, lt ls posslble to modlfy
the react10n mechan~sm so that the ASG 1s formed as a
conformal coatlng as the result of a heterogeneous
react~on. The reactlon can be carr1ed out at temperatures
down to below 400 C w1thout the use of a plasma, and by
vlrtue of the react10n mechan1sm, the ASG ls depos1ted
w1th a much smoother surface than 1s ach1eved wlth
convent10nal processes. the as depos~ted dens~ty 1s also
lmproved compared to that of conventlonal ASG. However,
should lt be reqùlred, the ASG can be baked at 600-900 C
to further dens1fy and reflow the layer.
The advantages of uslng ASG depos1ted as the result of
the mod1fled, heterogeneous react10n are that the
substrate need not be exposed to h1gh temperatures, no
plasma 1s needed; and the smooth surface produced by the
reactlon g1ves rlse to low scatter loss, even w1thout
belng reflowed.
The process by whlch ASG can be deposlted conformally
w111 now be descrlbed, and examples w~ll be glven of the
reactlon condlt10ns used to produce ASG films of v~r10us
composlt~ons.
The ASG ls produced ln a chem~cal vapour deposltlon
1~81536
~CVD) process such as ray be carr1ed out ln a commerc~al
CYD mach1ne. Mach~nes des1gned for the s11ane-oxygen
react10n for CVD of slllcon dloxlde, such as the PYROX
Reactor produced by Tempress-Xynet1cs, are partlcularly
sultab1é for carrylng out the ASG depos1t~on, although
other machlnes may also be sultable. For the purposes of
descrlpt10n ~t w~11 be assumed that d PYROX 216 Reactor ls
to be used, and such a reactor 1s shown d1agrammatlcally
Flgure 1.
The PYROX Reactor, whlch prov~des for batch processlng
of wafers, has a water cooled 102 reactor head 100 wlthin
~h1ch there ~s a rotatable c~rcular table 101 upon wh1ch
are placed wafers 103 to be treated. The table 101, wh1ch
supports a graph1te wafer carrier 104, ls heated from
underneath dur1ng processlng, the temperature of the
table 101 and hence of the wafers belng measured by means
of a thermocouple. ~n the experlments to be reported,
three-lnch wafers were used. The wafers were held on
s111con carblde coated graphlte succeptors 99, arranged ln
a c1rcle of twelve around the outer rlng of an elghteen
wafer carr~er.
The reactor head cons~sts of four concentrlc zones
whlch, movlng out from the centre, are termed A. 8, C, and
D. The gas flow to each of these zones can be adJusted to
vary the condltlons wlth1n the reactor head. Separate
flow control valves 105, 106 and 107 and pressure gauges
108, 109, and 110 are provlded for zones A, B, and C; flow
to Zone D 1s not lndependently controllable. The
compos~t10n of the gas fed to the reactor head can be
ad~usted by flow control valves 111-116 ln each of slx
flowl1nes, the flowrates ln each of the flow11nes be~ng
mon1tored by means of rotameters 117-122, conta1n1ng
floats 123-128. In the present case, only flve flowl~nes
are requ1red:
* trademark
F P O PI r~ J 81 IJ
3LZ 8~L5 3 6
FlowllneGas Rotameter Float
Ident1tyCompos1tlon Porter Model Type
Humber
Main N2 B250-8 Stain7ess Steel
Hitrogen
Sllane 5/oS1H4 B125-40
~n N2
Dopant ~ 1/oAsH3 B125^40
ln N2
Oxygen 2 ~125-40 u
D11ution
Nitrogen N2 B250-8 n
rhroughout the experiments the zone pressures were
maintAined at values rout~rely used when deposlting USG or
PSG:
ZONE A 13 psi
ZONE ~ 11 pS1
ZONE C 12 psl
~ONE D - not dlrectly measurable
Results for reactions carried out at plate Iwafer)
temperatures between 400 and 450 degrees C are shown ln
Figure 2. This figure shows how temperature affects the
th~n fllm deposlt~on rate for flve dlfferent tot~l hydrlde
flow r~tes 119,29,75,110,130 cc/mlnute) with the oxygen
flow rate held constant at 2500 cc/mlnute, and with the
ma~n nitrogen and dllut~on nltrogen flow rates each held
at 3~ litres/mlnute.
It is instructlve to compare the deposition vs
temperature curves obta1ned ~ith sllane-arsine-oxygen wlth
those obtalned wlth the silane-oxygen system. In Flgure 3
examples of each are compared. The curve for
~ 8~536
sllane-arslne-oxygen ~for 75cc total hydrlde flow, SlH4:
Arslne ratlo = 61:14) shows the two reglons whlch
characterlse lt as a heterogeneous reactlon. In the low
temperature reglon, where there ls k~netlc control, the
deposltlon ra~e ls reactlon rate llmlted and shows the
exponentlal rlse wlth temperature predlcted by the
Arrhenlus rate equat10n:-
D . Ae - /oE/RT
wlth some temperature varlatlon of A (as predlcted by
the Eyring rate equatlon). In the second reglon (the mass
transport llmlted region) the deposltlon rate 15 llmlted
by the dlffuslon rate of the reactants through a very thin
lS depleted zone near the surface whlch wlll follow the
contours of the surface. By comparlson, the sllane and
oxygen system shows a very small dependence of deposltlon
rate on temperature (ln the example lllustrated lt ~s
practlcally constant at 9A per C, whlch ls small when
compared to the 29A per C to 63 A per C for the
arslne-sllane-oxygen example shown) and the lack of any
dlffuslon llmlt lndlcates that lt ls homogeneous gas phase
reactlon.
The reactlon mechanlsm determlnes the type of specles
that wlll arrlve at the surface. If the reactlon ls
homogeneous, the oxygen and sllane react to form slllcon
dloxlde, or a slmllar specles, ln the gas phase. These
molecules may condense ln the gas phase to form colloldal
partlcles, The slllcon dloxlde wlll arrlve at the surface
as partlcles ranglng 1n slze from the monomer to collo~dal
partlcles, glv~ng rlse to the characterlst~c rough pebble
- llke texture of low temperature CVD ox~des, as shown ln
1536
11
Figure 4. The reaction parameters such as pressure and
gas composition will control the particle size distribution
and hence the surface tex-ture. The mobility of these
particles will be small and decrease with increasing particle
size.
The significance of the heterogeneous reaction
is that the deposition rate is controlled by the surface
temperature and not by the geometry of the surface, and
hence one may expect such a reaction to give conformal
lo oxide coatings. Moreover in a heterogeneous reac~ion
the silane and oxygen are absorbed onto a surface where
they subsequently react to form silicon dioxide. ~s these
absorbed species will be very mobile, good step coverage
and smooth surfaces should result. Figure 5 shows the
smooth surface produced as a result of heterogeneous re-
action, and should be compared with Figure 4 which shows
a typical equlvalent (PSG) deposited from a homogeneous
reaction. Figure 6 shows a conformal coating 50 of ~SG
over a 1 ~m high aluminium track 51 wlth near vertical
slde walls 52. This should be compared with Fiyure 7
which shows a typical non-conEormal coating 60 produced
as a result of a homogeneous reaction between silane and
oxygen. The results oE the homogeneous reaction can be
seen as overhangs 61 at the sides of the track 51; such
overhangs are typical of the non-conformal deposition
which characterises homogeneous reactions.
The conformal coating of ASG as shown in Figure
6 was produced with the instrument settings given in the
following example:
I:F.UII ~J 1 3~:Y IJ~ J .:~, . 11~.. 1~. 1 1l: ',~:
~ 8 1~;3 6
EXAMPLE
Gas Rotameter read1ng
helght ln mm
Maln N2 56
5/o SlH41nN2 40
1 /AsH31nN2 44
D11ut10n N2 56
Oxygen 9S
Thls equals 61 cc/m~nute of pure S~4
14 cc/mlnute of pure ASH3
2500 cc/m1nute of pure 2
The zone pressures were malnta1ned as above at 13
ps1 Zone A; 11 ps1 Zone B; 12 ps~ ~one C Plate temperRture
- 450 C. 3-1nch s111con wafers placed 1n outer c1rcle of
an 18 wafer plate. Th1n f11m depos1tlon of rate of A
575 A/mlnute.
Glass depos1ted under these condlt10ns was found to have
an lntrlnslc stress of 5 x 10 Dynes cm tens11e.
~he glass had a compslt10n of 12 mo1 /o AS203,
88 mol /o S102.
Sat1sf~ctory conformal coatlngs have been prcduced
wlth sllane:ars1ne rat10s between about 3.8:1 and 11.7:1.
A deposltlon rate versus temperature curve for a s11ane
flow rate of 60 cc m1nute~l and an arslne flow rate of
lO cc m1n~1 (w1th 2500 cc m1n~l02, ma1n N2 ; 38.
L1tres m1n 1, and d11ution N2 ~ 38 L1tres mln~ ) 1s
shown 1n Flgure 8.
It has been found that 1n general an 1ncrease 1n
oxygen and/or sllane concentrat10n favours a homogeneous
react10n and an 1ncrease ln ars1ne concentrat10n favours
heterogeneous react10n.
F p l l l l u 1 .; ~:: IJ IJ .~
1'~3~;3~;
~3
The followlng gas m1xtures, used under the cond1t10ns
set out above, have been found to 91ve the reactlon type
1nd kated:
Ars1ne S11ane Oxygen H1trogen
Runml/m1nml/m1n 1/m1n 1/m1n React10n
A 6.5 61 1.4 76 heterogeneous
B 3 .1 3~ 1. 4 76 homogeneous
C 6.5 76 1.4 76 heterogeneous
D 6.5 113 1.4 76 homogeneous
E 6.5 148 1.4 76 homogeneous
The As203 content of the glasses produced under
heterogeneous react~on cond1tions were as follows:
Run A 6/o when depos1ted at 400 C, 4/o when
depos1ted at 450-C
Run C 3/o when deposlted at 400 C, 2/o when
depos1ted at 450 C
the f11ms w111 reflow 1n 2' steam and
POC13/02. ~e have found th~t the f11m flow ls
greatest 1n steam. To ach1Qve the same degree of
planar1zat10n f11ms that ~re reflowed 1n steam requlre
temperatures about 100 C lower than those reflowed ln
POC13 /2'
Wlth 17 mole /o As203 (not produced as a
conformal fllm), complete reflow has been ach1eved ln 3
mlnutes at 800C ln steam. Flgures 9 and lO show the
cross sectlon o~ ASG fllms 18000A 17 001e /oAs203)
that have been deposlted on polys111con steps (5000A hlgh
wlth plasma etched near vertlcal walls) and reflo~ed ln
steam for 15 m1nutes. The fllm ln Flgure 9 was flowed at
800 C and that ln Flgure 10 at 900 C.
~Z8~S36
14
Following re~l~wing! in steam a bake in oxygen
at 600C is advisable to remove water from the film.
Films deposited directly onto silicon should
not be baked in pure nitrogen or argon, as this may lead
to the ormation of elemental arsenic at the ASG/silicon
interface.
As noted above, arsenic trioxide (As2O3) has
a significant vapour pressure over ~SG and it can readil.y
be lost by evaporation when the glass is heated. Since
the melting point of the glass decreases progressively
with an increasing concentration of dopant oxide, it should
be possible to flow a glass containing a high proportion
of As2O3 at a relatively low temperature and in the same
process remove some of the As2Q3.
Many analytical techniques have been used to
determine the composition of doped oxide films, but infra
red spectrophotometry (IRS) is the popular technique for
rapid routine determinations. A calibration graph for
ASG films, which relates absorption peak intensiti.es to
film composition has been generated by Wong and Ghezzo
(J Electrochem Soc V118, No 9, pl540, 1.971) who used the
X-ray microprobe anal.ysis o:E the Ei].ms and the Elame spectro-
photometric analysis of the dissol.ved :Eilms as calibration
standards.
In a subsequent investigation of densified ASG
films Wong :Eound that I~ analysis, which indicated that
the As2O3 content had decreased during densification,
did not agree with the X-ray microprobe analysis, which
indicated that no change had occurred (J Electrochem Soc
30 V120 p 122 1973). He concluded that it was a physical
change in the glass matrix and not a compositional change
that was causing the change to the IR spectrum.
1Z8~536
To investigate the matrix e.ffect for our ASG
layers we deposited films of ASG (2000A, 10 Mole % As2O3)
on silicon wafers (1000 ohm cm p type~ some of which
had a 400A layer of oxide grown on the surface to prevent
the diffusion of arsenic (As) into the silicon. The films
were analysed by IR and Rutherford Backscattering (RBS)
both before and a:Eter ~ensification (980C, 15 min, 10
2 in N2) and -the results compared.
The RBS spectra of a film deposited on a grown
oxide is shown in Figure 11. Although a brief inspection
of the spectra, which were taken before and after densification,
shows that As was .lost during this processing more information
can be obtained from the interpretation of all the spectra
tabulated below:
15 S~BSTRATE MOLE % As O
2 3
AS DEPOSITED AFTER DENSIFICATION
IR RBS IR RBS
Silicon 10.5 - 5.7 6.9
Grown Oxide 10.5 12.7 6.2 7.5
The IR results were calcul.ated using Wong and
Ghezzo's calibration chart or as-deposited ASG fllms.
It can be seen that in each case where the IR
result is compared to the RBS result the RBS result gives
a value that is consistently about 20% higher. The further
conclusi.ons from these results are:-
1 The changes in peak intensi-ties observed
in the IR spectrum are due to compositional changes and
not physical changes in the glass matrix so that IR can
3 be used to monitor As2O3 content during the densification
process.
~81536
16
2 During densiEication As2O3 leaves the film
by evaporation Erom the surface and if the densification
is performed on a bare silicon wafer As will diffuse into
the substra-te, but the quantity will be small compared
to that lost by evaporation.
To confirm these results a second series of
samples was prepared which had a thicker film of deposited
oxide (6000 At. The experiments were repeated, but in
this case the films were analysed by IR and atomic absorption
lo spectroscopy of the dissolved films. The conclusions
from the previous experiments were confirmed.
FILM DENSIFICATION
To investigate the changes that occur during
densification a series of 6000 A ASG films, each with
a different composition, were deposited onto silicon substrates
with a 400 A oxide film grown on the surface. The composition
of the deposi-ted films varied from 1.6 to 17 mole % ~s2O3.
The samples were analyzed by IR beEore and aEter a typical
fabrication sequence that would be used between ~SG deposi-
tion and first metal deposition. The sequence includeda back gettering step using POC13 at 980C.
The results from these e~periments are plo-tted
on the graph in Flgure 12. It can be seen from the graph
that if the inltial concentratLon oE ~s2O3 was below 6.5
mole % tllen about 25~ of it was lost during processing.
If the lnitial concentration was greater than 6.5 mole
% then the final concentratlon always fell to a plateau
level of 4.7 mole % (this limit only applies to the particular
processing sequence used here: further heating would
lead to greater loss). This result indicated that there
were two mechanisms by which the
~X81~;36
17
As2O3 left the fi]m. The first mechanism which was kinetically
fast only occurred if the initial concentration was greater
than 6.5 mole %, and should be interpreted as a super-
saturated solution expelling the dopant oxide until -the
saturation limit was reached. The second, slower, mechanism
appeared to be the diffusion of the As2O3 to the surface
where it was lost by evaporation to the ambient atmosphere.
To verify the latter mechanism a depth profile
of the As2O3 concentration was required. As it has been
shown that errors can occur when Auger electron spectroscopy/depth
profiling is used to evaluate doped oxide films it was
decided to obtain an IR depth profile. Once the IR spectrum
of a densified film has been recorded it was etched in
a solution of 5% HF to remove a layer 500 to 1000 A thick
from the surEace of the film. The IR spectrum of the
thinner remaining film was recorded and by comparing the
two spectra the composition of -the dlssolved layer could
be calculated. By repeating this sequence of etching
and recording -the IR spectrum a depth profile of the whole
film could be obtained.
A typical result from a film that contained
an initial concentration of 13 mole % ~s2O3 is shown in
Figure 13. The graph clearly shows the concentration
gradient caused by the dif~usion of As2O3 From the surface.
The P2O5 generated during the back gettering step can
be seen difusing from the surace into the film creating
a near-surface layer of arsenophosphosilicate glass which
should give the ilm better gettering properties.
Although loss of arsenic occurs on heating the
film, there need not be a corresponding fall in refractive
F l ' U I I '~ J , ~ ~
~153
1ndex. The reason for th~s ~s apparently that the
denslflcat10n wh1ch occurs on heat1ng ~at least by 600 C)
ralses the refractlve lndex, offsettlng the fall due to
arsenlc loss. Other workers have observed slmllar
lncreases ln refractlve 1ndex on heatlng and denslfylng
Sl02 For example, Pllskln and Lehman, ln J Electrochem
Soc Vol 112, No 10, pages 1013-1019, report an lncrease 1n
refract1ve lndex from 1.43 to 1.46 for S102 heated 1n
steam at 850 C for lS m1nutes. It should be reallsed
however that once the f~lm has been dens~fied, whlch can
be done at 600 C, contlnued or repeated heat1ng wlll dr~ve
off more arsenlc, and the refractlve 1ndex would be
expected to fall.
Arsenosll~cate glass contaln1ng 12 mole /o arsenlc
was depos~ted, under heterogeneous reaction condltlons
(mass transport llm~ted reglon) on a 3u sll~ca wafer Itype
Q2 3W55.10.C, made by the Uoya corporatlon) to a depth of
2~,m at a temperature of 450 C, at a rate of 400A per
mlnute. A reference slllcon wafer was also ooated at the
same tlmel and the ASG thlckness measured on that by
optlcal lnterference technlques. The thtckness was found
to vary by about ~ 3/o across the 3u wafer. A pr1sm
coupler ~nd hellum neon (6328A) laser were used to
determlne the fll~th~ckness ~nd refractlve lndex of the
fllm on the slllca wafer. The thlckness measurements
agreed wlth that c~rrled out on the sll~con wafer. The
f~lm was found to be b~modal tat 63Z8A). The effect1ve
~ndex of the zero order TE and TM modes was found to be
1.502, and that of the flrst order TE and tM modes 1.417.
the bulk refractlve lndex of the fllm calculated to be
about 1.53, about 0.07 above that of slllca, and agrees
well wlth that measured on the s1mllar fllm on s~llcon.
F P ~ r~ 3 ~ J
lX81536
19
The decay of the scattered llght (6328A) along the
propagat1ng beam was measured and losses of between 0.3
and 0.5dB cm~l for the fundamental mode, and 0.8 to
1.2dB cm~l for the flrst order mode were deter~ned.
Exam1nat1On of the uncoupled reflected llght beam 1ndlcate
a very low level of scatter loss.
EXAMPLE 2
Slm11ar samples were baked at 600 C for 15 mlnutes 1n
an o~ygen atmosphere. The va1ues for effect1ve refract~ve
IU 1ndex were: zero order, l.5087; f1rst order, 1.479. The
very close agreement between these f1gures and those
obtalned on the as deposlted samples are, w1th1n the
11m1ts of exper1mental error, the same. Th1s result 1s a
11ttle surpr1s1ng 1n that the fall 1n 1ndex due to arsen1c
loss was apparently closely balanced by the 1ncrease ln
lndex due to denslflcatlon~
The substant1ally 11near relatlonsh1p between
refractlve 1ndex and arsenlc content for as deposlted
f11ms ls shown ln Flgure 14.
A s1mple wavegulde can be made w1th ~ust a guld1ng
layer of ASG on a s11lca or other su1t~ble substrate.
However, ASG ls also well su1ted to manufacture of bur1ed
wavegu1de dev1ces. Because surface scatter loss 1s
dependent upon the refract1ve 1ndex d1fference across the
surface, bur~ed waveguldes~ whlch reduce that d1fference
can have reduced scatter loss. As can be seen from
F1gure 14, lt 1s posslble to vary the refractlve lndex of
ASG between about 1.45 and 1.53 by varylng the arsenlc
content between about 2 and 12/o,
EXAMPLE 3
A2.5~ layer of ASG conta1nlng 10 ~ole percent
AS203 (n 1.52) was deposited on a s111ca sllce. A
~8~;36
500 A layer of aluminium was evaporated onto the back
of the substrate to facili-tate mask align~ent. Nex-t a
1 ~m layer of positive resist was spun on-to the ASG layer.
The waveguide pattern was then printed using a contact
mask, and the resist developed to define the pattern.
The ASG layer was plasma etched in a c2F6/cllF3/E~e atmosphere
(oxygen being excluded to prevent erosion of the resist
mask). The remaining resist was stripped. Finally the
aluminium was removed from the rear side. Waveguides
produced by this technique are shown in Figures 15 and
16. of particular note are the waveguide's very smooth
top and its freedom from edge ripple. Figure 16 clearly
shows the waveguide's smooth sides.
The increased loss of arsenic from the surface
layers on heating ASG, gives rise to the possibility of
producing a pseudo buried waveguide s-tructure. The more
heavily doped region adjacent the subs-trate would form
the guiding layer, while the depleted surface layers would
produce a more gradual index change between the guiding
layer and air, reducing scatter loss.
~ dopting a conventiona~ approach to the fabrication
of a buried waveguide structure, two separate ASG layers
may be provided. A heavily doped .layer, with 10-.12 mole
~ arsenic, 2 ~m thick is depos:Lted For the guiding ]ayer.
Using standard microlithographic techniques (coat with
photo~electron beam resist, expose, develop resist to
define pattern, wet or plasma etch) a passive device structure
such as a ring resonator, beam splitter, or coupler etc.
can be formed in the first layer. Next, a second layer,
with 2-4~ arsenic, 3 ~Im thick is deposited to bury the
first.
If plasma etching is used it may be necessary
to briefly reflow the first ASG layer to remove wall rough-
ness before depositing the second layer.
Depending upon the desired arsenic content,
the reaction temperature can be reduced to about 390C.
However, to ensure uniform thickness it is advisable to
~81S3~,
21
operate in the mass transport limited region - where
deposition rate is ]ess sensitive to temperature ehanges.
Although the heterogeneous reaction produces
ASG films of low stress, some cracking has been observed
with films of 5 llm thickness. Where more than one layer
is deposited, reflowing earlier layers should reduce the
incidence of cracks in later layers, enabling grea-ter
overall thicknesses to be built up.
Since the ASG ean be deposited at temperatures
between about 390 and 450C, provided no reflowing is
required it ean be used in conjunction with group III-
V semiconductors. With a suitable buffer layer (which
eould be a low arsenic ASG layer) between the ASG III-
V compound, ASG may be deposited on a substrate of group
III-V material. Sueh an approach would enable monolithic
optical integrated eireuits to be fabricated, using active
(light generating) components sueh as lasers fabricated
from group III-V compounds.
It would also be possible -to integrate ASG with
a lower index eladding o an opt:ieally aetive mater:La,
e.g., an optically aetive organie materia]..
~, , . . j.
