Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
; :~ wo 92/11517 2 ~ 9 ~ ~ 3 ~ Pc~r/uS9l/0963
PLANAR WAVEGUIDE SPECIROGRAPH
Cross Reference to Related A~lications
This application is a continuation-in-part of application Senal No Ql,898.. : '
filed December21, 1990, entitled PLANAR WAVEGUIDE SPECTROGRAPH,
which is a continuation-in-part of application Serial No. 325,249, filed
March l7, 1989, now U.S. Patent 4,999,489.
Field of the Invention
The present invention relates to spectrographs and, in particular, to a
compact spectrograph that uses a planar waveguide. The spectrograph of the
invention has particular utility in wavelength division multiplexed optical sensing
systems.
Background of the Invention --
A number of systems have been developed for multiplexing optical fiber-
coupled transducers. Such systems include optical time division multiplexing,
coherence multiplexing, and wavelength division multiplexing (WDM). Although
all three systems have been demonstrated, WDM appears to offer the most promise
for near term implementation in aircraft and other complex systems. In a WDM
system, discrete spectral wavelength ranges or bands propagating along a fiber bus
- are modulated by one or more of the sensors or transducers. A crucial component
2û of a WDM system is a demultiplexer/detector, capable of receiving a broadband
optical signal. and of detecting the optical enérgy in different wavelength bands.
Summarv of the Invention
The present invention provides a spectrograph that is well suited to serve as
a demultiplexer/detector in a WDM system. The spectrograph receives an optical
~5 input signal. and detects the optical energy of the input signal in a plurality of
different wavelength ranges.
:
:
~, ,, .,, . . , , . .... , , , , , . ... ., , ,, . . " . . . .. . .
wO 92/11517 ~ ~ 9 8 ~ 3 ~ pcr/~s9l/o963
-2 -
In a preferred embodiment, the spectrograph of the present invention
comprises a planar waveguide and a detector array. The planar waveguide has a
plurality of side edges extending between upper and lower faces. The side edges
include a dispersive edge having an inwardly concave shape, an input edge, and a
5 straight outpu~ edge. The dispersive edge has a reflective diffraction gratingformed on it, the grating comprising a plurality of lines with a variable line
spacing. The line spacing and the positions of the input and output edges are
selected such that when the optical input signal is introduced into the waveguide at
the input edge, the input signal travels through the waveguide and strikes the
10 grating. The grating focuses the optical energy in each of the wavelength ranges of
the input signal at a focal spot at the output edge, with the position of each focal
spot being a function of wavelength.
The detector array comprises a plurality of photodetectors positioned along
a straight line. The detector array is positioned such that the photodetectors are
15 positioned at the respective focal spots, so that each photodetector (or group of
adjacent photodetectors) detects the optical energy in a corresponding one of the
wavelength ranges. In a preferred embodiment, the waveguide comprises a unitary
sheet of material, and the grating comprises grooves formed on the dispersive
edge, The grooves may be formed mechanically, by mechanical replication, or by
20 selectively etching the dispersive edge using radiation from a pair of coherent
illumination points.
In a second preferred embodiment, a plurality of waveguides of the type
described above are stacked one upon the other to form a multi-channel
spectrograph. The multi-channel spectrograph receives a plurality of optical input
25 signals, and detects the optical energy of each input signa~ in a p]urality of
different wavelength ranges. The output edges are preferably positioned in a
plane, so that the stack of waveguides can be directly interfaced to a tWO-
dimensional detector array such as a CCD array. Adjacent waveguides may be
isolated from one another by plates, or by thin films deposited on the upper and30 lower waveguide surfaces, to maintain total interna~ reflection. Total internal
reflection may also be maintained by lowering the refractive index of the
waveguide near its surfaces, for example, by ion-diffusion.
Brief Description of the l:~rawing$
FIGURES lA and lB schematically illustrate the operation of Rowland
35 speetrometers;
FIGURE 2 illustrates a prior art demultiplexer using a slab waveguide;
!. . WO 92/1 1~17 2 0 ~ ~ ~ 3 ~ pcr/us9l/og634
~,~.~, 3
FIGURE 3 is a second schematic view of the demultiplexer of FIGURE 2;
FIGURE 4 is a perspective drawing of the spectrograph of the present
invention;
FIGURE5 is a graph illustrating the geometry of the spectrograph of
S FIGURE 4;
FIGURE 6 illustrates a first preferred method for optically forming the
grating;
FIGURE 7 illustrates a second preferred method for optically forming the
grating;
FIGURE 8 is a schematic cross-sectional view showing multimode
propagation within a waveguide;
FIGURE 9 is a graph showing the line shape produced by a monochromatic
input;
FIGURES lOa-lOc illust~ate the use of separate cylindrical lenses to
collimate the input light;
FIGURES 11a-llb illustrate the use of cylindrical lens features directly
attached to the edge of the waveguide;
FIGURE 11c illustrates the use of a slab lens integral with the edge of the
waveguide;
FIGURE 12 illustrates the line shape with input collimation;
FIGURE 13 illustrates a multi-channel spectrograph embodiment, of the
invention;
FIGURE 14 illustrates the detector array of the embodiment of
FIGURE 13: and
FIGURE lS illustrates the assembly process for a two channel
spectrograph.
Detailed Descri~?tion of the Preferred Embodiment
The s~ectrograph of the present invention is derived from the classical
Rowland spectrometer that is schematically illustrated in FIGURE lA. The
Rowland spectrometer makes use of a concave, reflective diffrac~ion grating lO as
a dispersion element. Grating 10 is formed on substrate 12, and has a concave
spherical shape, with a radius of curvature equal to R. The grating comprises lines
that are equa~ly spaced from one another along a chord of the concave grating.
For such a concave grating, there is an associated Rowland circle 14 that
has a radius of R/2 and that is tangent to the midpoint of the grating. The
sigsuficance of the Rowland circle is that if a monochromatic point source 16 is
,, ~ . .
..
, . - .: -: - .
. `
wo 92/1151~ 2 0 9 8 ~ 3 ~ PCT/~'S91/0963~
positioned on the Rowland circle, then grating lO will produce a focused image 18
of source 16 at another position on the Rowland circle. The position of image 18is a function of the wavelength of the light. Thus, if source 16 comprises threedifferent spectral ranges or bands, then three focused images 18, 18', and 18" will
S be produced on the Rowland circle. If source 16 is a broadband source, then a
continuous spectrum will be imaged on the Rowland circle. E~carnples of prior art
Rowland spectrometers are shown in U.S. Patents Nos. 3,532,429 and 4,030,828.
With the advent of high powered, coherent optical sources, it has become
possible to produce diffraction gratings with complex, variable line spacings, using
lO what are loosely referred to as holographic techniques. Such techniques are
described in U.S. Patent No. 3,973,850, and in the paper by N.K. Pavlycheva,
entitled "Design of Flat-Field Spectrograph Employing a Holographic Grating,"
8 Sov. J. Opt. Technol. (U~) 46, 394-6 (1979).
FIGURE lB illustrates one of the principal advantages that can be obtained
15 through the use of variable line spacings. Diffraction grating 20 has been
produced holographically, and the spacing of the lines is no longer constant along a
chord as in the conventional Rowland spectrometer. As a result, 2 point source 24
having three wavelength components will be imaged at points 26, 26', and 26",
with the source and image points now no longer located on Rowland circle 22. In
20 general, radiation emitted by source 24 will be imaged along focal line 28, at
positions that depend upon the wavelengths emitted by the source. As further
described below, the location of focal line 28 can be controlled by adjustment of
the location of source 24 and of the variable line spacing of the grating.
In recent years, "planar" Rowland spectrometers have been developed as
25 uavelength division multiplexers and demultiplexers for telecommunication
applications, with both inputs and outputs being optical fibers. An example of aprior art demultiplexer based on a classical Rowland spectrometer is shown in
FIGURES 2 and 3. The device comprises a body 30 forrned by epoxying a thin
cover glass 32 between two microscope slides 34, to thereby form a thin planar
30 waveguide 36 in the cover glass layer. Body 30 has a cylindrical end face 38
having a radius of curvature R, and a cylindrical front face 40 having a radius of
curvature of R/2. A diffraction grating is forrned on end face 38 by wrapping a
gold foil with a linear grating on it around the end face. The distance between the
center points of the front and end faces is equal to R, the radius of curvature of the
35 grating.
WO 92/11517 2 B 9 ~ ~ 3 7 PCr/US91/0963
;.~ -5 -
Input optical fiber 44, containing a signal to be demultiplexed, is butted up
against waveguide 36 at front face 40. The input signal on optical fiber 44
includes a plurality of different wavelength components. Light introduced into
waveguide 36 from optical fiber 44 is reflected and diffracted by grating 42, at an
5 angle that depends upon the wavelength of the light. Output optical fibers 46 are
also butted up against waveguide 36 at the same front face 40, such that each
output optical fiber receives light of a different wavelength.
The operation of the demultiplexer shown in FIGURES 2 and 3 is generally .
similar to that of a conventional Rowland spectrometer, except that the light path is
two-dimensional and confined within the planar waveguide. Confinement of the
light within the waveguide degrades the resolution to about 1% of the input
wavelength for an input optical fiber having a numerical aperture of 0.2, due tomultimode propagation within the waveguide. Since this intrinsic degradation is
comparable to the resolution limit due to the optical fiber diameter, small devices
lS (e.g., l cm. square) are practical where fiber diameter effects will still dominate.
A factor that significantly degrades the performance of the demultiplexer is the fact
that the grating is linear along the circumference of the Rowland circle, and not
along a chord as required by the Rowland theory.
A first preferred embodiment of the spectrograph of the present invention is
illustrated in FIGURE4. The spectrograph comprises planar waveguide 50,
photodetector array 52, and measurement circuit 54. Waveguide 50 comprises a
thin layer of a material that is transparent at the wavelengths of interest, andincludes an inwardly concave surface 56 on which diffraction grating 60 is formed.
Waveguide 50 also includes input surface 62 and straight (planar) oueput
surface 64. Optical fiber 66 is butted up against input surface 62, while
photodetector array 52 is butted up against output surface 64.
An input optical signal on optical fiber 66 enters waveguide 50 at entry
point 68, spreads laterally at a rate that depends upon the numerical aperture of the
optical fiber, and strikes concave diffraction grating 60. Grating 60 disperses the
different wavelength ranges or components of the optical input signal, and images
such components onto straight output surface 64, with the position of each
component being a function of its wavelength. Photodetector array 52 comprises alinear array of individually addressable photodetectors, e.g., a CCD array. In such
an array, each photodetector converts incident optical energy into a quantity ofelectrical cha~ge, which charge is then converted into a voltage signal that may be
read and recorded by measurement circuit54. The measllrement circuit may
. . ,,: .
: . . .... ~ : .. . . .
.- . . . .. . . - .
.
- . .
WO 92/1 l~l' 2 Q ~ g ~ 3 ~ PCr/US91/09631~j
~;
-6-
thereby determine the amount of optical energy of the input signal in each
wavelength range.
An important feature of the present invention is that the spectrograph
focuses different wavelength components along a straight line (i.e., output
S surface 64). This feature permits the focused light to be directly interfaced to the
photodetector array. Flat field spectrographs differ from the classical Rowland
spectrometer in two respects. First, the source point is located off the Rowlandcircle. Second, the grating line spacing is asymmetrically varied about the center
of the grating. A geometric plan view of such a spectrograph is shown in
10 FIGURE 5. Concave grating 60 is cylindrical in shape, -and has a circular cross
section with center of curvature 70, radius of curvature R, and center point 72.The corresponding Rowland circle of diameterR is indicated by reference
numeral 74.
X and Y axes are constructed as shown, with axisX being normal to
15 grating 60 at center point 72, and axis Y being tangential to the grating at center
psint 72. The location of the optical source is designated by A, and the image
position (for a particular wavelength) is designated by B. Point A corresponds to
entry point 68 in FIGURE 4.
For a given source point location and a given ~ine spacing for grating 60,
20 the focused image will in general be located along a focus line 80. with the
position of the image along focus line 80 being a function of wavelength. It is
desired to design the line spacing of grating 60 such that focus line 80 includes a
region 82 along which the focus line is substantially a straight line.
The first order equations that describe the positions of points A and B as a
25 function of wavelength are:
sin cr + sin,B = mA/(a~g) (1)
cos2 crlra + cos2 ,B/rb - 1 (cos cY + cos ,~) = (mA/~g,) (~Lm H200/i~o) (2)
Equations (1) and (2) are derived from H. Noda, T. Namioka, and M. Seya,
"Geometric Theory of the Grating," J. Opt. Soc. Am. 64, pp. 1031-1036 (1974).
In these equations, A (ra, ~Y) is the source point, B (rb, ,B) is the image point, m is
the diffraction order, o is the line spacing at the center of the grating, H200 is the
35 azimuthal focusing constant that defines the variability and degree of the line
spacing variation, R is the radius of curvature of the diffraction grating (i.e., the
. : . , . , - ................... . : .
-., . ,: :. . " . :
..
3.~
~:~ WO92/11517 PCrtUS91/0963
-7-
diameter of the Rowland circle), ~\ is the wavelength of the radiation, Ao is the
wavelength of the light used to create the grating (further described below), ~g is
the refractive index of waveguide 50, and ~m is a refractive index that is further
described below. In Equations (l) and (2), the astigmatic focus conditions have
5 been ignored, because of the planar geometry.
In FIG~JRE 5, focus line 80 indicates the position of image point B as a
function of wavelength A for a given source position A and for selected values of
oand H200. Combinations of these parameters can be found for which focus
line 80 approximates a straight line over a desired wavelength range. Once a
l0 particular design has been selected, the parameters o and H200 may be determined
by Equations (l) and (2).
In general, grating 60 may be created either optically or mechanically by
replication. In an optical process, a thin uniform layer of photoresist is deposited
on the curved edge of the waveguide on which the diffracdon grating is to be
formed, for example, on curved edge 56 of waveguide 50 shown in FIGURE 4.
The photoresist is then exposed to coherent light from two slits or pin holes
positioned precisely at a pair of illumination points. These points are indicatea by
the letters C and D in FIGURE 5. The values for the parameters a and H200 are
used to fix the locations of the two coherent illumination points C (rc, ~) and
20 D (rd, ~) according to the following equations, which again are derived from
Noda et al.:
sin ~ - sin ~ = ~o/(~m~)
cos ~/rc - cos ~lrd - l/R (cos y - cos ~) = H200 (4)
~m is the refractive index of the medium through which the light from sources C
and D will pass during exposure of the photoresist. This medium may either be
waveguide 50 or air, as further described below. Equations (3) and (4) indicate
that there are infinite number of pairs of positions for the two sources. Preferably,
30 the source positions are chosen to minimize the spectral line width by setting the
coma-like third order aberration term to zero. The positions of sources C and D
that minimize coma and hence line width are then selected.
A spectrograph of the type described above was produced, and was found
to perform exceptionally well as an optical demultiplexer. The specifications of35 the spectrograph were as follows:
, . . .:
.
2~g8~
wo 92/11517 Pcr/~ss
-8-
a l500 lines per millimeter
~g 1.51
R l.0 (normalized)
ra l.08 (nor nalized)
a~ 12
H200 0-305
~o 488 nanometers
The thickness of the planar waveguide was 0.05 inches. The spectrograph
operated over wavelength range of 700 900 nanometers, and had a focal field
length along output edge 64 of 8 millimeters.
FIGURES 6 and 7 illustrate two methods for positioning sources C' and D'
for creating grating 60 on an edge of waveguide 50. In the method shown in
FIGURE 6, slits are located at illumination positions C' and D', and diverging
radiation from thé slits passes through waveguide 50, and falls on a portion of
curved edge 56 to form grating 60. Radiation from source C' enters waveguide 50
via lateral edge 78, while radiation from point D' enters waveguide 50 through
input surface 62. In the assembled spectrograph, source point A will be located
along input surface 62, while image point B will be located along output
surface64, as indicated in FIGURE4 Since it is impossible to locate the
illumination points within waveguide 50, the light rays pass first through air and
then through waveguide 50 in reaching surface 56. Such light is therefore
refracted at the air/waveguide interface, and such refraction must be taken intoaccount Illumination points C' and D' appear to be located at points C and D
-~ which would apply if the waveguide was extended away from curved edge 56.
A second and generally more convenient method for creating the grating is
illustrated in FIGURE 7. In this arrangement, slits are located at illumination
; points C' and D', and the radiation from such points is converted to converging
- radiation by lenses 84 and 86, respectively. The illumination points C' and D' and
lenses 84 and 86 are selected such that the radiation strilcing edge 56 appears to be
coming from virtual sources C and D. Virtual sources C and D are then positionedas in the arrangement of FIGURE 6.
Interferometrically controlled ruling engines now permit ruled master
diffraction gratings to be fabncated with variable groove spacing. The ruling
en~ine is computer controlled, and the groove spacing is typically specified by a
polynomial or an analytical expression. After a spectroglaph has been designed
, ~ and developed using the holographic method described above, it may be desirable
j .
b~ .
..
2U9~ ?~
` .~ wo 92/1 1517 Pcr/~S9t/0963
to transfer the design to a ruled master, so that spectro~raphs can be mass
produced by replication rather than by individual holographic exposure. The
efficiency is enhanced by putting a blaze on the grating.
An analytic expression for the line frequency f (lines per millimeter)
5 projected onto a chord as a function of the locations of the two illumination
positions C and D is as follows:
1 ~R~/l-w~/R2 +(Yd w) R ~ w~~/-R2 +(Yc--w)
~ L ~(Xd z~ + (Yd -- W) V~(Xc --Z) + (YC -- W) _l
(5)
For this equation, the X and Y axes are located as in FIGURE 5, w is the
distance along the chord parallel to the Y axis, and z = R-R [l-w21R2] '~'. It has
been determined that the blaze angle for maximum efficiency is e~ual to (~ + ,B)I2
where ~ is the source angle, and ~ is the dif~raction angle for the wavelength at the
middle of the range that is imaged onto the linear portion of focal curve 80
15 (FIGURE 5) Replica gratings formed in gold coated epoxy reflecting layers have
been produced, and have identical focusing properties to the ori~inal
holographically-produced spectrograph.
In FIGURE 5, for a monochromatic source A, the width (along line 82) of
the focused and diffracted image B is affected by multimode light propagation
0 within planar waveguideS0. Referring to FIGURE 8, light rays enterin~
waveguide 50 from optical fiber 66 form a cone 90 of light. Inside the waveguide.
the rays therefore have a propagation component in the direction parallel to thediffraction grating rulings, i.e., normal to the plane of the waveguide. Following
the deriva~ion of the general form of the grating equation described in
25 G.W. Stroke, "Diffraction Gratings," Encyclopedia of Phvsics, Vol. XXIX Optical
Instruments, Ed. S. Flugge, published by Springer-Verlag, pp. 426, 754 (1967),
the diffraction equation for rays propagating within the waveguide may be
determined to be
m~ . ~1 2m~ sini(l-cos~)~
( d ) ¦ d (md~' -sini) J (6)
- . - , - : ::; . . . ,-. .. . ~
, .. . ,- .
. . .
: ::: - , , ,
wo 92/l lsl7 2 ~ 9 8 8 3 2 PCI/-S91/0963~i
-10-
In equation 6, i and r are the angles of incidence and reflection at the gratingsurface in the plane of the waveguide, d is the local grating spacing, m is the
diffraction order, and ~ is the angle between the ray and the plane of the
5 waveguide. As ~ increases for a constant value of A, the diffraction angle
increases. Since the cone of rays 90 emerging from the optical fiber into the
waveguide has a distribution of ~ values up to ~max~ asymmetric broadening of
the diffracted image towards longer wavelengths occurs. A typical line shape for a
monochromatic source is shown by curve 94 in FIGURE 9. It can be seen that
10 curve 94 includes a long wavelength tail 96. The line width at the 10% point
is 4.5 tirnes greater than that which would result from broadening due to the width
of the optical fiber source alone. The discontinuous nature of the graph in
FIGURE 9 results from the finite pixel siæ of the detector array.
The line broadening illustrated in FIGURE 9 may be minimized by ;,
15 reducing the angle ~max shown in FIGURE 8. In principle, this could be
achieved by reducing the numerical aperture of the optical fiber. However, for
many applications, this would lead to reduced light being coupled into the
waveguide, and would therefore be counterproductive. In accordance with the
invention, the line broadening is minimized by collimating the light emerging from
20 the optical fiber in the plane normal to the plane of the waveguide. In a first
preferred embodiment, such collimation is achieved as shown in FIGURE lOa, by
placing cylindrical lens 100 between optical fiber 66 and input surface 62 of
waveguide 50. The cylindrical lens may comprise a cylindrical rod lens 102 as
shown in cross section in FIGURE lOb, or a plano-convex cylindrical lens lOl as
25 shown in cross section in FIGURE lOc. In either case, the rays emerging from
optical fiber 66 are substantially collimated along the direction normal to the
waveguide, thereby reducing the angle ~max and minimizing line broadening.
Three further preferred methods for collimating the input light are
illustrated in FIGURES Ila-llc. In FIGURE lla, cylindrical lens 106 is directly
30 attached to waveguide 50, for exa nple using an optical cement. ~his arrangement
has the advantage of compactness and of easier alignment. FIGURE 1 lb shows a
variation. in which input surface 62' of waveguide 50 is shaped so as to form a
cylindrical lens that is integral with the waveguide. Finally, in FIGURE llc,
graded index slab lens 108 may be attached directly to input surface 62 of
35 waveguide S0, with optical fiber 66 being butted up directly against the graded
index lens. ~uitable graded index slab lenses are available from Hova Optics Inc.
: , . ..
.; ., . . . . :. ....... . .
~: ;
~ ~ , . , . -
- ,, ~ -. . .. - . .... . .
:' . . : : . .. . '` ` . : :: . : .
` ' '::: . : ` ~ . ,, : . ,
"~ WO92/11511 2~ PCr/US9l/0963
The advantage produced by collimating the input light may be seen by
comparing FIGURE 9 (discussed above) with FIGURE 12. FIGURE 12 illustrates
the spectral line 120 obtained at the same wavelength as FIGURE 9, by collimating
the input beam using a cylindrical lens as in FIGURE lOa. Collimation reduces
the width at half maximum from 3.8 nanometers to 2.9 nanometers, and reduces
the 10% line width from 10.7 nanometers to 5.9 nanometers. Line 120 can be
seen to be nearly symmetrical, whereas line 94 obtained using an uncollimated
beam includes a very distinct long wavelength tail 96. The narrow line width
obtained using collimated input beams means that the channel spacing that can beobtained is about half that which can be obtained using uncollimated input.
The planar waveguide profile of the spectrograph of the present invention
permits a plurality of spectrographs to be stacked in a highly compact fashion. For
example, FIGURE 13 illustrates a side view of a stack 140 of six planar
waveguides 142(1)-142(6). Waveguides 142 are positioned such that their ou~put
surfaces lie in a plane, so that the waveguide output surfaces can be directly
interfaced to a two-dimensional photodetector array 144. In this confi~uration, the
six spectrographs do not occupy significantly greater space than a single
spectrograph. As shown in FIGURE 14, the face of array 144 contains a
rectangular, planar array of individual photodetectors 146, arranged in six rows.
Array 144 may be a two-dimensional CCD array, such as those used in television
cameras. In one arrangement, each row of the array receives light from only one
spectrograph. while each column of the array receives light of a particular
wavelength. Alternatively, each spectrograph may illuminate one or more rows of
photodetectors .
In one preferred approach, a~ter fabrication of the stacl~ 140, a singie
diffraction grating is formed extending across the entire thickness of the s~ack,
such that an identical grating is created on each waveguide, and such that the
gratings on the different waveguides have their grooves aligned with one another.
In an alternate approach, different gratings are first formed on the individual
waveguides, and the waveguides are then assembled into a stack. In this second
approach, the mapping of wavelengths to the photodetectors in a row can be varied
from one charmel to another, for increased versatility.
It is important that rays propagating in one waveguide do not significantiy
leak into the adjacent waveguides. This means that all the rays propagating within
each waveguide should be totally internally reflected, and that the evanescent tails
of the higher order modes propagating within a given waveguide should not leak
.
;
,
,
2~9~8~2
wo 92/11517 PCr/US91/0963
-12-
into an adjacent waveguide or into any spacer material used. This may be
accomplished in a number of ways.
In a preferred embodiment, a relatively lower refractive index material is
deposited in the form of a thin (e.g., two micron) film on the upper and lower
5 surfaces of each waveguide. A preferred film material is amorphous silica, which
forms a robust thin film via ion-assisted electron beam evaporation or sputtering
Silica also has an extremely low optical absorption in the near infrared, so
radiation loss of such light propagating in the film is negligible. For the case in
which the waveguide is fabricated from BK-7 glass with a refractive index of 1.52,
10 the refractive index of amorphous silica (1.46) is sufficiently lower to provide the
required confinement, and to produce a relatively high numerical aperture of about
0.38, to provide efficient coupling of light from the optical fiber.
In general, amorphous silica will not function optimally if heating is
required, and other techniques for providing isolation between the waveguides
15 should be used for such cases. Such techniques include the use of a lower
refractive index polymeric coating and/or a lower refractive index spacer material.
An example of a lower refractive index polymeric coating is
poly(methyl-methacrylate) with a refractive index of 1.49. An example of a lowerrefractive index spacer material is fused quartz (refractive index of 1.46). A thin
20 region of lower refractive index may also be created by ion-diffusion in the
wave~uide surface. For some cases, the use of a spacer having a rough (i.e., non-
optically fJat) surface will provide enough of a gap to obviate the light leakage
problem.
A preferred technique for fabrication of a two-channel spectrograph in
''5 accordance with the present invention is outlined in FIGURE 15. The process
begins with a pair of waveguides 150, 152. Each waveguide may comprise a disl;
of BK-7 glass. 0.0lS inches thick, with two micron thick films of amorphous silica
on each of its sides. Waveguides 150 and 152 are formed into a stack l60 using
plates 154-156. A suitable material for the plates is a ceramic material available
30 from Con~ing under the trademark MACOR. Each plate surface to which a
waveguide will be joined includes a pattern of grooves 162 in which solder
glass 164, or another suitable adhesive, is deposited. The use of grooves provides
precise separation of the waveguides, to ensure effective registration of an array of
input optieal fibers to the waveguides. The adhesive may optionally be deposited35 directly onto the surfaces of waveguides lS0 and lS'. respeetively.
, :, :, .. ,. : : :: ,:, . , . ~ -.. , - . . . . . . . . , .. .. ~ . .. . ..
2 ~ 3 ~
~;;,~ wo 92/t 1517 PCT/~!S91/0963
-13-
The illustrated layers are then formed into stack 160 in a suitable jig, and
heated and/or pressed to cause the adhesive to flow, to join the waveguides and
plates in place into a precisely spaced and rigid stack. A preferred adhesive
material is Ferro 1160HZ solder glass, because it has a coefficient of thermal
5 expansion and optical finishing characteristics that closely match BK-7 glass and
MACOR ceranic. Other suitable adhesives include inorganic alumina and
zirconium based ceramic adhesives such as a Cotronics 903HP, polymer based
adhesives (e.g., silicones, alkaloids and epoxies), and metallic alloys such as low
melting indium based solders.
Once stack 160 has been formed, its perimeter is subjected to conventional
optical finishing and polishing, and the stack is then divided into a pair of
spectrograph blanks 166 and 168. The blanks are then funher cut and polished toform spectrograph bodies 170 and 172. A diffraction grating is then formed on
each body, using either holographic, replication, or ion milling techniques. For15 each body, the diffraction grating is preferably formed in a single step, such that
identical gratings are formed on the two waveguides contained within the bodv.
This assures accurate registration of the gratings with one another. Forming andfinishing of the waveguide edges after assembly of the waveguides into a sandwich
structure provides precise end face coplanarity, and enables radial alignment of the
20 gratings.
While the preferred embodimenls of the invention have been described
variations will be apparent to those skilled in the art. Accordingly, the scope of
the invention is to be determined by reference to the following claims.
;'
~, . . .. ... . .. .. . .
