Method of forming an integrated optical circuit

Information

  • Patent Application
  • 20060040190
  • Publication Number
    20060040190
  • Date Filed
    August 19, 2004
    20 years ago
  • Date Published
    February 23, 2006
    18 years ago
Abstract
A method for forming an optical circuit on a substrate. The method comprising a plurality of mask images to define the optical circuit image in photoresist. Each of the mask images defining parts of the optical circuit and the totality of all mask images substantially defining all of the optical circuit. A photolithography system globally aligns and exposes the mask images in photoresist. The resultant composite image being substantially indistinguishable from a single image of the entire optical circuit. Different images for each of the parts can be substituted by other images, thereby exponentially increasing the number of circuit permutations from a predetermined number of images. The method is also applicable to generating a unique circuit from a pre-existing library of reticle images. The images are printed in predetermined locations on a substrate to define the desired optical circuit.
Description
BACKGROUND OF THE INVENTION

1. Field of the Invention


The present invention relates to the field of integrate optics, and more specifically to a method of producing an integrated optical circuit using lithographic techniques.


2. Description of Related Art


Integrated optical devices are optical devices that are realized on a substrate. FIG. 1 is a cross section through the substrate of an integrated optical circuit. Referring to FIG. 1, the substrate 100 includes a carrier 101 and a materials stack 102 comprising thin films of dielectric, semiconductor, or polymer materials grown or deposited on one side of the carrier, here called the top side 103. Common substrate carriers include silicon, quartz, and gallium arsenide. Common dielectrics include silica, germanium doped silica, silicon, silicon nitride, silicon oxynitride, silicon oxycarbide, and LiNbO. Common semiconductors include doped silicon, GaAsInP, and other III-V compounds. Channels are formed in the dielectrics and these channels guide light and perform some sort of optical signal processing function.


In an often used approach for the fabrication of optical devices, a photolithographic mask (also called a photomask or simply mask) containing an image of the optical circuit design is used to print the circuit. FIG. 2 shows a cross sectional view of a photoresist of radiation through a mask. Referring to FIG. 2, the circuit is printed by exposing a photoresist covered substrate 201 to a source of irradiance 202 through a mask 203. The mask 203 contains a pattern of the circuit to be printed. The mask gets its name from the fact that the circuit pattern on the mask selectively blocks (masks) or transmits (unmasks) the source of irradiance. The remaining irradiance 204 that reaches the photoresist after passing through the mask defines the optical circuit pattern. The photoresist 205 on the substrate is subsequently chemically developed leaving a patterned copy of the optical circuit image in photoresist on the surface of the substrate. The photoresist pattern can be either positive or negative. A positive image means that photoresist remains after development where irradiance is masked, while a negative image means photoresist remains where irradiance is unmasked. The polarity (positive or negative) depends on the mask pattern and the type of photoresist used. The photoresist pattern is transferred into the substrate by etching techniques. A general description of techniques for the fabrication of optical integrated circuits can be found in Hiroshi Nishihara et. al. “Optical Integrated Circuits”, McGraw-Hill, New York, 1987, and is incorporated herein by reference.


The technique of exposing a mask image onto a photoresist covered substrate is described by the term “photolithography”. The three major types of photolithography are contact printing, proximity printing, and projection printing. In contact printing, the mask is placed directly on a photoresist covered substrate. In proximity printing the mask and photoresist covered substrate are separated by a small distance and there is no optical imaging between the two. In projection printing, lens elements or mirrors are used to focus the mask image onto the photoresist covered substrate, which is spaced from the mask by a large distance. The mask used in projection lithography is commonly called a reticle. (We will use the terms reticle and mask interchangeably). Projection printing is commonly used in semiconductor fabrication where substrates are called wafers, and many technologies have been developed, including scanners and step and repeat systems, or “steppers”. Steppers project an image only onto a portion of the substrate. The maximum area printed by a stepper in one exposure is called the stepper field. Multiple images of the reticle, or different reticles, are stepped and repeated over the entire substrate area. Reticle images are typically one to ten times (1× to 10×) the size of the image projected onto the substrate, with reduction provided by the lens system. The enlarged master image on the reticle results in finer resolution on the actual substrate image. Compared with contact or proximity systems, steppers and scanners are equipped to more precisely align a mask image to a specific position on a substrate.


Conventionally, the entire optical circuit image resides on a single mask. Therefore, in a single exposure of the mask, the entire circuit is printed on a portion of the wafer. If the actual circuit to be realized is larger than that which the reticle size can accommodate, then the circuit image must be distributed across several individual reticles. The individual reticle images are then printed sequentially in a mosaic-like approach. Some means for the alignment of the images must be provided since misalignment will compromise performance. Printing reticles in this manner is called “stitching”, or “field stitching”. An example of an optical circuit that stitches together numerous images was disclosed in U.S. Pat. No. 6,517,997 to S. W. Roberts. Technically this is “inter-field stitching”, since numerous fields are stitched together in order to realize a circuit that can not be accommodated within a single field. The term “stitching boundary” is used to describe the boarder separating two exposed images in the photoresist. In U.S. Pat. No. 6,517,997 there are no waveguide channels traversing across any stitching boundary. An example of stitching for electronic integrated circuits is disclosed in U.S. Pat. No. 6,030,752 to H. J. Fulford.


Field stitching using stepper lithography is typically required for conventional planar lightwave circuits that use low core-to-cladding refractive index contrasts (contrasts below 1%), because such devices are invariably larger than the size of a single stepper reticle. The size of an integrated optical circuit is proportionately related to the core-to-cladding refractive index contrast (which we will refer to here simply as “index contrast”), and in the last decade the use of high index contrast materials have enable the reduction of the size of devices to the point where they can fit entirely within the field of a reticle. For example articles by Little in Optical Fiber Conference (“A VLSI Photonics Platform”, Proceeding of the Optical Fiber Conference vol. 2, pp. 444-445, 2003), and in Little et. al. in Photonics Technology Letters, (“Ultra-Compact Si—SiO2 microring resonator optical channel dropping filters,” Photonic Tech. Lett. vol. 10, pp. 549-551, 1998) both incorporated herein by reference, describe high index contrast material systems and devices for realizing micro-circuits. U.S. Pat. No. 6,614,997 and U.S. Pat. No. 6,771,868 both to F. Johnson et. al. discloses an ideal low loss, high index contrast material system, and is incorporated herein by reference.


Each unique optical circuit design requires a unique mask image. Optical circuit design, mask design, and substrate processing (micro-fabrication) are all interrelated. In optimizing an optical circuit for production, many iterations of the design and micro-fabrication cycle need to be carried out. This incurs a significant cost and time commitment because commercial grade masks are time consuming to produce and expensive. Indeed, the performance of the final optical circuit is usually dictated by the budget used to procure masks, and the time allowed to accomplish a certain number of iterations.


An integrated optical circuit is comprised of a number optical sub-elements and waveguides. The sub-elements and waveguides are defined geometrically and have a number of parameters that affect their performance. During the development phase of an optical circuit, the optical sub-elements must be optimized both individually and as an aggregate within the circuit, in order to optimize the performance of the entire optical circuit. For example, consider an optical circuit that contains three critical sub-elements. We wish to investigate five different designs for each critical element, and we wish to do this in a complete circuit configuration. This would lead to 5×5×5 or 125 unique circuit permutations requiring 125 mask images. Each unique optical circuit, characterized by a unique combination of sub-elements and waveguides, requires its own unique mask image. Even if one parameter on one sub-element is modified, the entire optical circuit requires a new mask image. Reticles are expensive and time consuming to produce. The optimization of an optical circuit therefore can be costly and time consuming since many reticles need to be procured. Methods to reduce the number of reticles without sacrificing on the number of variables have hitherto not been disclosed.


Further, it often occurs that different products require optical circuits that although unique, are nonetheless very similar in the majority of their circuitry and architecture. For instance, a circuit may comprise twenty five optical sub-elements, twenty three of which are identical for each custom circuit. Each of these circuits requires its own reticle or series of reticles. It would be preferable if minor customization could be achieved by only changing portions of a circuit and re-using the common portion, rather than procuring an entirely custom reticle set for each product. No method has hitherto been disclosed which realizes unique circuits without using unique mask sets.


3. Objects and Advantages


The present invention provides a method for optical circuit generation which uses a “pick-and-place” technique. The invention overcomes the limitations of the prior art, all of which require unique mask (or reticle) images for every unique optical circuit. Using pick-and-place circuit generation, it is a first objective of this invention to provide a method to develop and optimize an optical circuit which requires fewer photolithographic reticles or masks than the conventional approach of having a unique image for every unique circuit configuration. Equivalently, it is a second objective of this invention to allow a more complete optimization of an optical circuit using the same number of photolithographic reticles or masks that would otherwise be used to conventionally generate a given number of unique circuits. It is a third objective of this invention to provide a method to reuse reticle or mask images across products that although unique, share common optical sub-elements or building blocks. It is a forth objective of this invention to allow rapid customization of products that could share common optical sub-elements. It is a fifth objective of this invention to allow an existing circuit to be updated with improved sub-elements without the need to generate an entirely new circuit, but only requires regeneration of the sub-elements that are updated. It is a sixth objective of this invention to provide a method to generate a unique optical circuit using pre-existing optical sub-elements that are stitched together in unique ways. All the foregoing objectives of the invention result in a better circuit, faster turn around times in development, and lower costs.


Further objects and advantages of the invention will become apparent from a consideration of the drawings and ensuing description.


SUMMARY OF THE INVENTION

An optical circuit is sub-divided into elements. Each element is designed as an independent image residing on a mask or reticle. The element images are printed sequentially on a wafer by exposing each image onto only its corresponding portion of the circuit or circuits on a photoresist covered substrate. Means are provided to align exposed images so that they define a contiguous resultant circuit image. Elements can take any shape and orientation with respect to each other, and images can be printed within the boundaries of other images.


Any particular element can have numerous variations, all of which are interchangeable within the overall circuit thereby generating variations and customizations of the overall circuit by only changing one or a few elements. Elements can be reused across different products or circuit configurations.


The ability to generate circuits using the pick-and-place technique requires a means to align abutting elements to one another forming continuous and uninterrupted pathways for light to travel through the optical circuit. Alignment can be done on an element by element basis, or by global reference marks. Stepper lithography systems for example, can use global reference marks placed on the wafer to align all subsequent images.




BRIEF DESCRIPTION OF THE DRAWINGS


FIG. 1 shows a cross section through the substrate of an optical circuit.



FIG. 2 shows an optical circuit being printed.



FIG. 3 shows an optical circuit that is subdivided into elements.



FIG. 4 shows an optical circuit that is subdivided into alternate elements.



FIG. 5 shows the independent sub-element images of an optical circuit.



FIG. 6 shows an arrangement of sub-elements on a reticle.



FIG. 7 shows a sequence of steps in printing an optical circuit image by pick-and-place method.



FIG. 8 shows the alignment of abutting sub-elements.



FIG. 9 shows a variation of the optical circuit in FIG. 3 having several common elements.



FIG. 10 shows how more primitive sub-elements might be defined and used.



FIG. 11 shows a system for generating optical circuits from pick-and-place images stored in an image library.




DETAILED DESCRIPTION OF THE INVENTION

We describe the pick-and-place method by referring to a specific optical circuit. FIG. 3 shows an image of an optical circuit 301. The optical circuit is comprised of waveguides and other optical elements. The circuit can be conceptually and physically subdivided into elements. Referring to the circuit in FIG. 3, the circuit has been subdivided into four elements. The elements are: a circuit comprising waveguides, element 1302, a switch, element 2303, and two ring resonator type wavelength filters, element 3304, and element 4305. Dashed lines around a particular element are used to demark the boundaries of a particular element. The shape and size of these boundaries are the circuit designer's choice.


The optical circuit described in conjunction with FIG. 3 is for illustrative purposes of the pick and place method. The concepts apply to any arbitrary optical circuit with an arbitrary number of elements. The elements can be arbitrarily ascribed, and will depend on the optical circuit designer's objectives. For example, for the same circuit of FIG. 3301, the two ring resonator wavelength filters may be lumped together as one element as shown in FIG. 4A401. Alternatively, the elements can be demarcated as shown in FIG. 4B where the elements now include portions of the output ports 410, 411.



FIG. 5 shows how the optical circuit of FIG. 3 is decomposed into four independent element images. These images can all reside on a single reticle or mask, or they may be distributed on various masks. FIG. 5A shows element 1501. It is comprised of the optical waveguide portion of the circuit. It has locations to accommodate the other elements, shown by the hatched areas 502, 503, 504. We call these locations “drop-in-boxes” because they are the spots where other elements are effectively dropped into during the printing sequence. The “drop-in” connotation is similar to that used in electronic circuit assembly industry wherein chip ICs (integrated circuits), are dropped into a socket on a printed circuit board. The electronics assembler has the option to drop in any brand of IC that serves the same type of functionality and that fits in the same socket. The drop-in boxes of FIGS. 3 and 5A can have arbitrary shapes, however the elements to be dropped into that location must have similar shape. FIG. 5B shows element 2520, which in this case is an optical switch. FIG. 5C shows element 3530, which is a ring resonator based wavelength filter. FIG. 5D shows element 4540, which is also a ring resonator based wavelength filter. The two ring resonator filters might have different characteristics resulting in different wavelength being dropped, or they might have different spectral lineshapes.



FIG. 6 shows how the independent element images might be laid out on a reticle or mask. FIG. 6 shows a top down view of a mask plate 600. In this particular case there is one image for element 1601, but there are four different images for each of the other three elements of the circuit. For instance, images 610, 611, 612, and 613 are all optical switch elements 2. They may each have a slightly different design, yet all can be dropped in to location 502 on element 1 in FIG. 5A. Likewise, images 620, 621, 622, 623 are all optical filter elements 3. They might each have a different design to filter a different wavelength for instance, yet all can be dropped in to location 502 on element 1 in FIG. 5A. Images 630, 631, 632, 633 are all optical filter elements 4. Lithography systems, such as stepper lithography, have the ability to print portions of a mask. Therefore, any image on the reticle shown in FIG. 6 can be printed independently of all other images, and can be printed anywhere on a wafer with high precision.


By picking one image from the four possible designs for each of the elements 2, 3, and 4 in FIG. 6, and dropping them into corresponding locations in element 1 of FIG. 6, there can be a total of 4×4×4=64 unique optical circuit configurations. These 64 circuit configurations are a result of 13 independent elements, all residing on a single mask. Note that typically elements of a circuit are much smaller than the circuit itself. Therefore, many independent elements can be accommodated on a single mask as demonstrated in FIG. 6. The flexibility of the Pick-and-Place approach is apparent, as a comparable conventional approach where a complete hard-wired circuit image is put on a mask can only accommodate one circuit variation per mask.



FIGS. 7A to 7D show one possible sequence of printing one variation of the optical circuit depicted in FIG. 3. FIGS. 7A to 7D show the sequences of images that would be present in the photoresist after each step is completed. In FIG. 7A, element 1 from FIG. 6A (601 in FIG. 6A) is first printed. The hatched areas are three unexposed photoresist areas for the three other elements to be subsequently printed. These drop-in boxes are for element 702, element 3703, and element 4704. In FIG. 7B, element 2 is printed in the drop-in box location 702 designated for that image. Next, in FIG. 7C element 3 is printed in the drop-in box location 703 designated for its image. Finally, in FIG. 7D element 4 is printed in its corresponding position 704. The resultant image after step D in FIG. 7D is the complete circuit image existing in photoresist. By printing, we imply that an image on a mask is exposed onto a photoresist covered wafer. During printing of an image, it is important that areas outside an image are not exposed. Exposure of areas outside an image could result in multiply exposing pre-existing features and washing them out. For instance, during the exposure of element 2 in FIG. 7B, the rest of the circuit area should see no irradiance because this would affect the image that was previously printed in step A of FIG. 7A, or affect subsequent images in FIGS. 7C and 7D. Protection of the remainder of the circuit image can be done in photolithographic systems such as stepper lithography systems by using the shutter blades of the stepper. Using such systems typically confine the definition of a element to reside in a rectangular area. On the other hand, rectangular shaped elements have the benefit of widest compatibility and interchangeability among elements. Multiple exposures of areas of the circuit where there are no optical elements or waveguides might be acceptable, as these regions do not affect optical performance of the circuit. The sequence in FIG. 7 is arbitrary, and the elements can be printed in any order, since the printing of one does not affect the others.


Note that in the sequence in FIGS. 7A-7D, elements 2702, 3703, and 4704, are all printed within the boundaries of element 1701. This is intra-field stitching. Inter-field stitching on the other hand is used to create an image that is larger than that which could fit entirely within one mask or stepper field. The invention described here applies to both intra-field and inter-field stitching for optical circuit generation.


Elements must be accurately aligned to one another, else the performance of the optical circuit will be degraded. FIG. 8 shows an expanded view of the optical circuit in FIG. 3 around the location of element 3 in FIG. 3 (304 in FIG. 3). The optical waveguides 801 comprising the circuit have a certain width w 802. Light is confined to the optical waveguide region. When the element 3 (for example) is printed it must register to, or be aligned to, the rest of the circuit 804. If there is an offset between the waveguides across an element boundary, such as the lateral offset shown in FIG. 8805, there will be scattering of the optical signal which results in an increase in loss. If the offset is larger than the waveguide width, most of the signal will be lost in traversing the boundary between those elements. Means must therefore be provided to align elements with respect to each other.


Stepper lithography systems have mechanisms to align reticle images. Reference marks on the reticle are used to align to reference marks on the wafer or on the wafer handling chuck. An example are DFAS (Dark Field Alignment System) marks. The geometry of the marks and the mechanics used for alignment differ among stepper manufactures. In contact lithography alignment marks are also used, however the alignment mechanisms are more primitive and include manual positioning the mask to reference marks on the wafer or wafer holder.


The Pick-and-Place method of circuit generation is versatile for custom circuit generation. For example, consider the optical circuit 901 shown in FIG. 9. This circuit is similar to, but not identical to the circuit 301 in FIG. 3. This circuit has a switch in the location of element 3902, as opposed to a ring resonator filter as in FIG. 3 (304 in FIG. 3). In order to turn the circuit of FIG. 3 into the circuit of FIG. 9, only element 3 needs to change. Further, if the design and layout is compatible, the switch image that is used for element 2 (303 in FIG. 3, 903 in FIG. 9) might also be re-used in the element 3 location.


Elements may be arbitrarily defined, and can be as primitive as desired. For example, elements might comprise wholly of straight and bent waveguides that comprise all other elements. For instance ring resonator element 3530 in FIG. 5C may be itself subdivided into subelements as shown in FIG. 10A. Referring to FIG. 10A, element 3530 has been subdivided into eight subelements consisting of straight and curved arc waveguide sections as shown by the dashed lines. FIG. 10B shows how the subelements in FIG. 10A would appear as independent images on a mask or series of masks. Although dividing an optical circuit into smaller and smaller subelements increases the total number of drop-in boxes, those more primitive images can be reused in more locations. For instance, an image of a straight waveguide can be used in all locations in the circuit that have a straight waveguide section.


Using a library of pre-existing elements or primitive subelements, a wide variety of unique optical circuits can be generated. If design rules are used for the design of the elements, such as common sizes, elements can be reused and shared across a wide variety of circuits. Considerable time and cost savings would be expected. FIG. 11 shows the system concept of arbitrary optical circuit generation using a reticle library of images. The optical circuit designer chooses elements of the library, and decides how they are to be interconnected in order to satisfy his or her circuit requirements. The size of the optical circuits are typically small, and many independent circuits can fit on a wafer. FIG. 12 shows a schematic of a wafer that has over 75 optical circuits. Each circuit might be unique.


The pick and place method is also useful for updating products with improved circuit elements or designs. For instance, if a new design with improved performance for one of the elements is discovered, it can replace the prior version of that element without needing to regenerate and re-optimize the entire circuit. Product upgrades can be done with little disruption.


In the foregoing discussions on the pick-and-place method for fabricating an optical circuit, photoresist was the preferred method of assimilating all the independent images into a final compound image. Photoresist is widely used in semiconductor manufacturing. It is possible to print images directly into an optical glass or polymer if the glass or polymer is photorefractive. Exposure to irradiance directly affects the refractive index creating waveguides and optical elements without the use of pattern transfer using photoresist. The pick-and-place method described herein applies to the printing of optical circuit in such glasses and polymers, and the procedure is essentially the same as discussed in this invention.

Claims
  • 1. A method for forming an integrated optical circuit comprising: (a) providing a substrate with a photoresist coating; (b) providing a plurality of mask images, said mask images defining parts of said optical circuit and the totality of all mask images substantially defining all of said optical circuit; (c) providing a printing means for exposing each of said images into said photoresist coating; (d) providing means for mutually aligning each said image when printed into said photoresist coating; (e) aligning and printing each of said images sequentially into said photoresist coating resulting in a complete image of said optical circuit in said coating; and (e) means for transferring said complete optical circuit image formed from the plurality of mask images from photoresist into said substrate, whereby an optical circuit is formed that is essentially indistinguishable from an optical circuit that is printed by the application of a single image of the circuit.
  • 2. The method of claim 1 wherein said substrate includes a carrier composed of one of silicon, quartz, indium Phosphide or gallium arsenide.
  • 3. The method of claim 1 wherein said printing means is one of contact printing, stepper lithography, or proximity printing.
  • 4. The method of claim 1 said means for transferring optical circuit image includes reactive ion etching of the photoresist covered substrate.
  • 5. A method for forming an integrated optical circuit from a set of mask images comprising: (a) providing a reticle image library, said image library comprising a plurality of optical images defining optical building blocks, including all building blocks sufficient to define the optical circuit; (b) providing a lithographic printing and image alignment means; (c) providing a substrate with a photoresist coating; (d) printing a predetermined subset of said building block images at predetermined locations on a said photoresist coated substrate resulting in an image of said optical circuit; and (e) means for transferring said optical circuit image into said substrate.
  • 6. The method of claim 5 wherein said lithographic printing and image alignment means includes one of stepper lithography, proximity printing, or contact printing.
  • 7. The method of claim 5 wherein said substrate includes a carrier composed of one of silicon, quartz, Indium Phosphide or Gallium Arsenide.
  • 8. The method of claim 5 wherein means for transferring optical circuit image includes reactive ion etching of the photoresist covered substrate.