EP1417713A4 - Laser arrays for high power fiber amplifier pumps - Google Patents

Abstract

A pump laser capable of delivering at least a specified amount of output power is described. The pump laser has an array of N semiconductor lasers each having a first wavelength and an individual available output power (P) such that the product of N times p is equal to or greater than the specified amount of output power. The pump laser also has a coupler configured to couple light emitted by the individual lasers in the array to an individual optical fiber.

Description

TITLE

LASER ARRAYS FOR HIGH POWER FIBER AMPLIFIER PUMPS

CROSS REFERENCE TO RELATED APPLICATIONS This application claims priority under 35 USC 119(e)(1) of United States Provisional

Patent Application Serial No. 60/302,600, filed June 29, 2001, United States Provisional Patent Application Serial No. 60/365,996, filed March 19, 2002, and United States Provisional Patent Application Serial No. 60/365,489, filed March 18, 2002.

This application is also a continuation in part of commonly assigned U.S. Patent Application Serial No. 09/896,189, U.S. Patent Application Serial No. 09/897,160, U.S.

Patent Application Serial No. 09/896,983, U.S. Patent Application Serial No. 09/897,158, and U.S. Patent Application Serial No. 09/896,797, all filed June 29, 2002.

FIELD OF THE INVENTION This invention relates to lasers and, more particularly, to lasers used to amplify optical signals.

BACKGROUND OF THE INVENTION

Optical fiber amplifiers are often used in optical fiber systems to prevent data from being attenuated while traveling through an optical fiber system. Data attenuation between transmitter and receiver is a problem in optical fiber systems. This is because, if the data is attenuated during its travel from the transmission end to the receiving end, when the data arrives at the receiving end the data may not be readable.

To address the above problem, optical fiber systems make use of optical fiber amplifiers that add energy to the signal via a pump laser. However, optical fiber amplifiers require a large input power, in excess of several watts, and must be reliable over a long period of time. As a result, creating a reliable pumping source that also has high power has proven difficult.

The use of multiple pump lasers that have a suitable collective power has also been tried. Although such multiple pump lasers exist, they only use a maximum of four discrete lasers. With a small number of lasers, a failure of one or more lasers substantially affects the output power, thereby reducing reliability. Good pump lasers are also highly temperature sensitive, so there is a high cost associated with maintaining a pump laser within a narrow temperature band. Thus, there is a need for a way to create a cheap reliable pumping laser. SUMMARY OF THE INVENTION

In accordance with the invention, creation of a pump laser that does not suffer from the problems of the prior art is made possible. One aspect of the invention involves a pump laser capable of delivering at least a specified amount of output power. The pump laser has an array of N semiconductor lasers, each having a first wavelength and an individual available output power (P). The individual available output power is such that the product of N times P is equal to or greater than the specified amount of output power. The pump laser also has a coupler configured to couple light emitted by the individual lasers in the array to an individual optical fiber.

The advantages and features described herein are a few of the many advantages and features available from representative embodiments and are presented only to assist in understanding the invention. It should be understood that they are not to be considered limitations on the invention as defined by the claims, or limitations on equivalents to the claims. For instance, some of these advantages are mutually contradictory, in that they cannot be simultaneously present in a single embodiment. Similarly, some advantages are applicable to one aspect of the invention, and inapplicable to others. Thus, this summary of features and advantages should not be considered dispositive in determining equivalence. Additional features and advantages of the invention will become apparent in the following description, from the drawings, and from the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a an example of a laser array according to one variant of the invention; FIG. 2 illustrates a lens array approach to coupling a laser array to a common fiber; FIG. 3 illustrates an alternate approach to the approach of FIG. 2 using diffractive optical elements;

FIG. 4. is a graph showing measurements taken from a 980nm laser array constructed according to the principles of the invention;

FIG. 5. is a table showing coupling efficiency, output power, and chip size for different sized laser arrays in accordance with the invention;

FIG. 6 illustrates approaches that have been used in the prior art to attach multiple bottom emitting devices to form an integrated electro-optical chip ;

FIG. 7 illustrates approaches that have been used in the prior art to attach multiple bottom emitting devices to form an integrated electro-optical chip; FIG. 8 illustrates a single optical device with contact pads placed in the position specified by its manufacturer and a portion of an electronic wafer with contact pads placed in the position specified by its manufacturer;

FIG. 9 illustrates a single optical device with contact pads placed in the position specified by its manufacturer and a portion of an electronic wafer with contact pads placed in the position specified by its manufacturer of which each will not be aligned;

FIG. 10 illustrates in simplified high level overview, one example approach according to the teachings of the invention;

FIG. 11 and 12 illustrates several different access way variant examples; FIG. 13 illustrates an optical array in which fibers are supported by the substrate;

FIG. 14 illustrates an optical array that accommodates an array of microlenses;

FIG. 15 illustrates one example process for creating an electro-optical chip variant according to the techniques described;

FIG. 16 illustrates one example process for creating an electro-optical chip variant according to the techniques described;

FIG. 17 illustrates one example process for creating an electro-optical chip variant according to the techniques described;

FIG. 18 illustrates one example process for creating an electro-optical chip variant according to the techniques described; FIG. 19 illustrates another opto-electronic device being created in a manner similar to the devices of FIGS. 15-17;

FIG. 20 illustrates a process usable for bottom active devices;

FIG. 21 A illustrates a process usable for topside active devices;

FIG. 21B illustrates the process where the contact holes are coated, but not filled, and can assist in alignment;

FIG. 21 C shows an optical chip with its contacts rerouted by patterning traces on the substrate to match the contacts on another chip;

FIG. 2 ID shows the contacts on an electronic chip rerouted by patterning traces on the substrate to match the contacts on an optical chip; FIG. 22 illustrates a process similar to that shown in Figure 16A except that a carrier is not used;

FIG. 23 illustrates a connection chip or adapter chip used to connect different devices;

FIG. 24 illustrates another alternative implementation, which is a further variant of the adapter or connection chip variant, usable for topside active devices; FIG. 25 A illustrates the stacking of two or more devices using one of the techniques according to the invention;

FIG. 25B illustrates a modulator stacked on top of a laser using one of the techniques according to the invention; FIG. 26 illustrates an array of, for example, one hundred lasers created using one of the techniques according to the invention;

FIG. 27 illustrates the steps in creating an array for a DWDM application using one of the techniques according to the invention;

FIG. 28 illustrates the process of FIG. 21 from a top view; FIG. 29 shows a redundant laser pair from an array in accordance with the invention;

FIG. 30A shows a group of four redundant lasers from an array according to the invention;

FIG. 3 OB functionally shows contacts for the group of FIG. 30 A;

FIG. 31 shows the functional components of an opto-electronic chip suitable for use in accordance with the invention;

FIG. 32 shows the chip of FIG. 31 employing pairs of redundant lasers according to the invention;

PIG. 33 shows an alternative variant to the chip of FIG. 32;

FIG. 34 shows the chip of FIG. 31 employing groups of four redundant lasers according to the invention;

FIG. 35 shows the chip of FIG. 31 employing pairs of redundant photodetectors according to the invention;

FIG. 36 shows a device of FIG. 31 employing groups of four redundant photo detectors according to the invention; FIG. 37A shows one functional example of circuitry for selecting from among two or more redundant devices according to the invention;

FIG. 37B shows another functional example of circuitry from among two or more redundant devices according to the invention;

FIG. 38 functionally shows an opto-electronic transceiver incorporating the invention; and

FIG. 39 is a functional block diagram of example automatic failover circuitry for a group of two devices. DETAILED DESCRIPTION

U. S. Provisional Patent Application Serial No. 60/302,600, filed June 29, 2001, U. S. Provisional Patent Application Serial No. 60/365,996, filed March 19, 2002, and U. S. Provisional Patent Application Serial No. 60/365,489, filed March 18, 2002, U.S. Patent Application Serial No. 09/896,189, filed June 29, 2002, U.S. Patent Application Serial No. 09/897,160, filed June 29, 2002, U.S. Patent Application Serial No. 09/896,983, filed June 29, 2002, U.S. Patent Application Serial No. 09/897,158, filed June 29, 2002, and U.S. Patent Application Serial No. 09/896,797, filed June 29, 2002, are all incorporated by reference herein in their entirety. FIG. 1 shows one example of a laser array constructed in accordance with the invention. The laser array is made up of either many discrete lasers, multiple individual lasers each having two or more active regions, or some combination thereof (interchangeably referred to herein as "lasers" or "laser elements"). Depending upon the particular application in which the pump laser is to be used, the lasers will have a wavelength of 968nm, 980nm, 14380nm, "14XX"nm, or some combination thereof.

As shown in FIG. 1, the lasers are coupled to a fiber by a lens array. The lens array is enlarged to illustrate how it couples light from the array's individual laser elements into the optical fiber. As shown in FIG. 1, for purposes of example, the array is a 13 X 13 array of lasers (i.e. 169 laser elements) with each being capable of outputting at least 1/169 of the output required of a pumping laser. Of course, depending upon the particular implementation, the size of the array can be increased or decreased to meet specific requirements of the intended application and the lasers used.

Additionally, for coupling the laser output to the fiber, lens arrays of different sizes or types (including single and compound lens arrays) can be used depending upon the particular application, either alone or in conjunction with other optical elements. For example, if an extremely large laser array is used, for example a 100 X 100 array, a larger lens array, a combination of lenses and arrays or additional elements such as faceplates, collimators, waveguides, etc. may be needed to properly focus the beam onto the optical fiber.

FIG. 2 shows one variant of the invention involving a lens array approach. Each laser element in the array releases a light beam that passes through a corresponding lens element in the lens array. The lens array directs the beams toward the opening of the optical fiber.

Depending upon the lenses and fiber, the light goes directly into the fiber or goes from the lens array through one or more lenses. The lens then focuses the light into a single beam that travels down the optical fiber. Although the FIG. 2 shows only one bi-convex lens to focus the beams from the laser elements into the fiber, multiple lenses can be used for large arrays. Additionally, in some cases the light may need to be focussed more narrowly, for example, if the optical fiber is a single mode fiber (i.e. it is extremely small). In such cases, more lenses or a more complex arrangement of elements can be used. FIG. 3 shows an alternate variant of the invention. In this variant, each laser element of the laser array emits a light beam into a diffractive optical element (DOE). The DOE defracts the light toward a first tier of bi-convex lenses. The layer of bi-convex lenses collect light from the DOE array and focuses the light toward one or more additional lenses on a tier between the first tier and the optical fiber. A lens, near the fiber, focuses the incidental light from the laser array into the optical fiber.

Although FIG. 3 shows two tiers of lenses, additional tiers can be added (with or without other elements) for different applications. In addition, the diffraction gradient could, in fact, be made up of multiple tiers depending upon, for example, the wavelength of emitted light, the size of the laser array, etc... FIG. 4 shows experimental measurements made by a vertical cavity surface emitting pumping laser, constructed according to the teachings of the invention, with 32 laser elements each having wavelengths of 968nm. As the current to the lasers increases from 0 to 10mA the output voltage of the laser increases from 1.25N to nearly 3N. The output power increases from OmW to nearly 6.25mW. The conversion efficiency of the laser array varies from 0% to 30% as the laser current is increased. Maximum efficiency is achieved at approximately 3.5mA.

FIG. 5 is a pump power analysis table for several different sized arrays according to the teaching of the invention. For each array size, the coupling efficiency was about 70%. As expected, the output power is larger for arrays containing more elements. Specifically, the power increases linearly with respect to the number of laser elements in the array. For example, with 72 lasers the power is 0.504mW while with 1024 lasers, the power is 7.168mW.

As shown in FIG. 5, although increasing the number of laser elements means that the size of the optical component chip and/or die must be increased, increasing the size of the array from 144 lasers, which has a size of 1.5mm X 1.5 mm, to 576 lasers (a 4X increase in number of lasers) does not increase the optical component chip size at all and only increases the die size to 3mm X 3mm ( an increase of only 1.5mm per side). Similarly, an increase to 1024 lasers from 144 lasers adds 880 lasers but increases the optical component chip size by only 0.5mm per side while increasing the die size by only 2.5mm per side to accommodate the 880 additional connection points. Thus, an extremely large array can be constructed within a small area.

Although the table only lists four array sizes, it illustrates an advantage achievable at low cost, namely scalability since the number of laser elements can be readily increased to thousands of laser elements.

Another advantage achievable in some implementations is compact size. This is accomplished by integrating the lasers into a custom-made integrated circuit ("IC") using wafer scale technology, for example using a technique described in the commonly assigned United States Patent Applications entitled "Opto-Electronic Device Integration", Serial Nos. 09/896,189, 09/897,160, 09/896,983, 09/897,158, the specification of which is reproduced below in the section of the same title. The IC is scalable and can be built with thousands of array elements for different applications. The cost of mass producing the laser array future modifications and/or variations of the design are low relative to the cost of pumping lasers in the prior art. Another advantage of this invention is that it can be used for different applications or with different devices. For example, although 980nm lasers were used for the experimental data of FIG. 4, this invention can be used with lasers of different wavelengths including lasers having the typical 968nm, 980nm, 1480nm, or 14XXnm wavelengths, to identify a few. Moreover, the ability to change the output power of the entire array by adjusting the power level of the individual laser elements by small incremental amounts means that the array can readily be used in systems that require different amounts of signal boast.

Another advantage achievable by applying the teachings of the invention is longer device life. There are at least three different mechanisms for ensuring a long lifetime for the pumping lasers. The first mechanism is passive in that, with laser arrays according to the invention, if a single laser or, in the case of a very large array even a significant number of lasers, fails to operate properly, the overall power of the array will not be significantly affected. For example, if one laser element in an array of 1024 fails, the total output power decreases by only 0.098%. For many applications the reduction in power is too small to adversely affect performance. Moreover, since increasing the size of the array corresponds to a decrease in individual power to each laser element, the affect of an element failing decreases for larger arrays. Therefore, the efficiency of the passive mechanism increases as larger arrays are used. The second mechanism used by this invention to achieve a long lifetime is redundancy. Redundancy is achieved by providing each laser with at least one backup such as shown in the commonly assigned United States Patent Application Serial No. 09/896,797 entitled "Redundant Optical Device Array" the entire disclosure of which is incorporated herein by reference and the specification of which is reproduced below in the section of the same title. The lasers are integrated onto an integrated circuit (IC) with the ability to automatically switch to backup devices if necessary. If a laser fails, the backup can be manually or automatically turned on to take the place of the original. This prevents the total output power of the laser array from decreasing as the device progresses through its life cycle. For example, a 144-element laser array could be configured to have 72 operating elements and 72 backups. If a laser element ceases to function, its backup element can be used in its place and adds to the power of the remaining lasers to achieve the same total output power as the original 72 lasers. The same array could also be configured to have 136 operating elements and only 12 backups for applications where the life cycle is expected to be significantly shorter than the mean time between failures ("MTBF") associated with 12 lasers failing.

The third mechanism used to ensure a long lifetime is achieved by altering the output power of the laser elements. For this mechanism, the array is configured so that if a laser element fails, the power to the remaining laser elements will be increased to make up the difference. By using a large array, if an element fails, the fractional increase in power to the remaining functioning elements is small and does not put a strain on the remaining laser elements. For example, if one laser fails in an array of 72, the power to the remaining 71 lasers need only be increased by 1/72 of the total power before the failure.

Another advantageous aspect is manufacturing yield. In accordance with the present invention, because most variants will have identical lasers, the incremental cost of manufacturing a laser array having 1000 lasers instead of 100 (1 OX as many) is minimal. As a result, one can readily manufacture an array that is large enough so that the presence of any laser that is defective, does not work or operates in an anomalous or detrimental fashion will have little to no effect on yield. Building upon the above, it should be appreciated that, by adding a few additional elements in a straightforward manner, the array can be made to function in an intelligent manner. For example, in some implementations the array is constructed to also include one or more photodetectors that are used for feedback purposes. Depending upon the particular implementation, the photodetectors are used to measure or sample the overall output power of the pump laser. In this manner, if the output power deviates from the desired level by more than a specified amount, the laser drive circuitry can be signaled to cause a change in the output of the individual lasers to compensate for the fluctuation.

Similarly, in some implementations, one or more temperature sensors can be incorporated into the array to allow for laser output fluctuations or drift caused by temperature changes. In those implementations, as temperature changes affect pump laser output power, adjusting the individual lasers as described above makes it possible to compensate.

It is expected that, for some implementations, control and compensation will occur using some form of programmed control. For example, in the simplest cases, a state machine can be used. In other more complex cases, programmed intelligence in the form of a computer, microprocessor, etc. (all hereafter interchangeably referred to as a "processor") will be used.

In the case of a processor, a program is used to process whatever feedback is available, for example, feedback from one or more photodetectors, temperature sensors, etc. and determine what changes, if any, need to be made. Depending upon the particular implementation, this may involve conversion formulas, the use of look up tables or both. If a change needs to be made, the processor, operating under program control will send the appropriate signals to the drive circuitry for the lasers to bring about the desired change.

In more sophisticated implementations, the processor may also be configured to control individual lasers. In this manner, compensation can be controlled on a more granular scale. In addition, this type of arrangement also allows for compensation through bringing individual lasers on and offline, either alone or in conjunction with the controlling of individual laser output as described herein.

Opto-Electronic Device integration

FIGS. 6 and 7 illustrate approaches that have been used in the prior art to attach multiple bottom emitting (or detecting) (also referred to as "backside emitting (or detecting)") devices to form an integrated electro-optical chip.

According to the approach of FIG. 6, multiple lasers, are formed on a wafer substrate 6-102 in a conventional manner, as are multiple detectors (interchangeably referred to herein as photodetectors) on their own or on a wafer substrate in common with the lasers. Typically, the portion 6-104 of the substrate 6-102 closest to the junction between the optical devices 6- 106, 6-108 and the substrate 6-102 is made of a material which is optically transparent at the wavelength at which the optical devices operate. The devices 6-106, 6-108 are then processed using conventional techniques such as wet or dry etching to form trenches 6-112 among the devices 6-106, 6-108 which separate them into a series of discrete individual lasers 6-106 or detector 6-108 devices . Depending upon the particular technique used, the etched trenches 6-112 may stop prior to reaching the substrates 6-102 or extend partly into the substrates 6-102. Following etching, the substrates 6-102 and their associated devices are inverted, aligned to the proper location over a Silicon (Si) electronic wafer 6-114, and bonded to the Si electronic wafer 6-114 using conventional flip-chip bonding techniques. Following bonding, the entirety of the substrates 6-102 are thinned extremely thin, by conventional mechanical polishing methods, conventional etch techniques or some combination thereof, to on the order of about 5 microns or less to allow for close optical access to the devices and create an integrated electro-optical wafer 6-116. Optionally, the integrated electro-optical wafer 6-116 is then patterned, using conventional techniques, to protect the individual lasers and the individual detectors are coated with an anti-reflection (AR) coating 6-118.

A related alternative approach to the technique of FIG. 6 is shown in FIG. 7. In this approach, lasers and detectors are formed as described above. However, when the technique of FIG. 7 used, the trenches 7-112 are etched into the substrates 7-102. The substrates 7-102 and their associated devices are then inverted, aligned to the proper location over a Silicon (Si) electronic wafer 7-114, and bonded to the Si electronic wafer 7-114 using conventional flip-chip bonding techniques. Following bonding, the substrates 7-102 are then wholly removed, by conventional mechanical polishing methods, conventional etch techniques or some combination thereof, to allow for close optical access to the devices and create an integrated electro-optical wafer 7-116.

Optionally, the integrated electro-optical wafer 7-116 is then patterned to protect the individual lasers and the individual detectors are coated with an anti-reflection (AR) coating. The techniques of both FIG. 6 and FIG. 7 make it possible to get optical fibers or optical lenses close enough to the devices to capture the appropriate light without allowing light coming from, or going to, adjacent devices to affect any of those adjacent devices, a problem known as "crosstalk". Typically, this requires that the separation distance between a device and an optical fiber or optical microlens be less than 100 microns. Additionally, both techniques ensure that there are no significant absorbing layers over the active region of the devices that will prevent light from escaping since the thinning technique of FIG. 6 reduces the thickness of the entire substrate 6-102 to about 5 microns or less and the approach of FIG. 7 removes the substrate 7-102 entirely, leaving multiple wholly independent optical devices.

Both of these techniques however, characteristically create opto-electronic chips that have heat dissipation problems during use and leave the individual devices more sensitive to thermal and mechanical stresses produced during the manufacturing process, thereby reducing individual device lifetimes and, accordingly, decreasing yields and overall chip life. Moreover, for the approach of both FIG. 6 (where the substrate is extremely thin) and

FIG. 7 (where the substrate is completely removed), stresses experienced by the devices are primarily transferred to the very thin optical device layer which is the structurally weakest part of the device.

Thus, there is a need for a way to create an integrated opto-electronic chip that is not as sensitive to the thermal and or structural stresses resulting from processing and/or use.

In addition, a manufacturer of opto-electronic devices has two avenues for obtaining the optical and electronic wafer - they can manufacture either or both themselves, or they can obtain one or both from a third party. By manufacturing both the optical devices (interchangeably referred to for simplicity as an "optical chip") and the electronic wafer (interchangeably referred to for simplicity as an "electronic chip"), the manufacturer can take measures to ensure that the pads on each are properly placed so as to align with each other when the optical chip is positioned over the electronic chip. However, typically electrical and optical chips are not designed concurrently, even if they are designed and fabricated within the same organization. Thus, even with a single manufacturer, unless there is close coordination within the organization with regard to both the optical and electronic chip design, a lack of correspondence between contact pads on each can easily occur - particularly where one or both are also designed with sales to third parties in mind or integration with devices from other sources is contemplated. Moreover, subsequent improvements or changes in the design of either may necessitate altering the location of the contact pads, thereby introducing a pad misalignment where none previously existed.

Even worse, if the electronic chip is designed to be used with a variety of different optical chips, but the optical chips are commodity stock obtained from third parties (for example, chips containing: topside emitting vertical cavity lasers, bottom emitting vertical cavity lasers, distributed feedback (DFB) or distributed Bragg reflector (DBR) lasers (which each have better chirp and linewidth characteristics for long distance applications), topside receiving detectors or bottom receiving detectors) that are mass manufactured for distribution to multiple unrelated users, it is unlikely that the pads on the optical devices will all be located in the same place, even if they are otherwise compatible with the electronic chip. For example, as shown in FIG. 8, a single optical device 8-300 has contact pads 8-

302, 8-304 placed in the position specified by its manufacturer. A portion of an electronic wafer 8-306 also has contact pads 8-308, 8-310, onto which an optical device can be connected, placed in the position specified by its manufacturer. If the optical device is flipped over, for flip-chip type bonding with the electronic wafer, the contact pads 8-302, 8- 304, 8-308, 8-310, of each will not be aligned, as shown in FIG. 9.

This presents a problem in that it limits the ability to "mix-and match" devices. Moreover, if a chip is designed with connection to a particular other chip in mind, and subsequent events create a need to use a different device with a different contact placement, all the planning and coordination done for the original device will be irrelevant to the new device.

Thus, there is a further need for a process that facilitates the ability to mix and match devices without there being any coordination between the designers of either or the use of a standard or common contact placement scheme.

In addition, in some cases it is sometimes desirable to coat some of the devices, specifically the detectors, with an AR coating.

An AR coating prevents light from hitting the top of a detector device and being reflected at the detector-air interface due to the differences in the indexes of refraction. This is important for detectors because reflected light is light that does not enter the detector itself and hence can not be converted into electrical signals (i.e. it is 'lost light' from a system point of view). Thus an AR coating optimizes the collection efficiency of the detector because it prevents light from being reflected at that interface.

Lasers however, require a top mirror of very high in reflectivity in order to operate. AR coating on a laser changes the reflectivity of the top mirror. As a result, at a minimum it will detrimentally affect the lasing action of the laser, if not prevent it from lasing altogether. If a wafer has both lasers and detectors in an array, in order to AR coat only the detectors, conventional wisdom would mandate that special patterning of the wafer be performed to protect the lasers during the AR coating deposition phase to ensure that those laser devices were not covered by the AR coating. The protection or disparate treatment of the various different devices on the wafer requires extra processing steps, which costs time, and hence increases the cost of processing. It also introduces the possibility of damaging the protected devices. Finally, it forces the electrical contact pads to be protected as well. In addition disparate treatment of devices causes other processing problems when the processing must be performed on a chip having electrical contact pads in the same area. For example, if a chip has electrical contacts near the devices and electroplating, electroless plating, thermal evaporates, e-beam evaporated or sputtering techniques are used to place solder on the contact pads, the height of the resulting solder bumps, renders it difficult to pattern areas to protect lasers from AR coating because the solder bumps are much taller than the optical devices.

Prior art lacks a way to eliminate the need to pattern a protective layer over the lasers while allowing the entire wafer (i.e. lasers and detectors) to be AR coated.

Thus, there is a further need for a way to permit integration of multiple types of devices on an electronic chip so that any additional processing steps, such as anti-reflection coating, can be done on the whole wafer at one time and without special patterning after integration.

We create opto-electronic chips which, in some variants, provides one or more of the following advantages: allows use of a lower operating current, thereby reducing power consumption and heat generation; provides better dissipation of heat that is generated, allowing the lasers to run at lower temperatures thereby increasing their usable life and/or providing better wavelength control; and/or having a higher structural integrity resulting in fewer defects and increased device lifetime.

We have further devised a way to integrate optical and electronic chips to create an integrated opto-electronic device, irrespective of whether the component devices are manufactured in a coordinated manner or have compatibly matching electrical contact points.

Still further, we have devised a way to create an integrated opto-electronic device that allows for an entire wafer having disparate devices to be AR coated, without special processing to protect the lasers or affecting their ability to lase. When integrating optical devices intimately with electronic chips, four attributes are desirable to create reliable integrated optical devices.

First, it must be possible to get optical fibers or optical lenses close enough to capture the light without crosstalk. Second, there must be no absorbing layers above the active region of the devices that would prevent light from escaping or entering the particular devices. Third, there should be a large enough thermal mass attached to the devices to allow for efficient heat dissipation. Fourth, the structural integrity of the devices should be maintained during processing so that stresses or strains experienced by the devices do not impact device performance. As noted above, the approaches of FIG. 6 and FIG. 7 can satisfy the first two attributes however, neither of those approaches satisfies the third or fourth since neither approach results in a large thermal mass attached to the devices (i.e. the substrate of the devices) or reduces stresses on the devices.

Although applicants are unaware of any such case existing in the prior art or otherwise, the approach of FIG. 6 could potentially be made to satisfy the fourth attribute by leaving a thicker layer of substrate on the device. However, this could likely only be accomplished if the operating wavelength of the particular devices were very transparent to the wavelength at which the devices operated. Moreover, for many cases, this would reduce, if not destroy, the ability to satisfy the first attribute and would likely also detrimentally impact the operation of a laser device unless the laser were redesigned to emit into, for example, a semiconductor material rather than being designed to emit into air. In addition, if thicker substrates were left, it would be necessary to AR coat the structure to prevent optical feedback into the laser. In addition, such an approach would likely also foreclose the use of commercially purchasable prefabricated semiconductor optical devices, such as most third party offered Nertical Cavity Surface Emitting Lasers (NCSELs), Distributed Feed Back (DFB) lasers or Distributed Bragg Reflector (DBR) lasers.

In sum, we have devised a way to closely integrate optical devices and an electronic chip to create an opto-electronic chip that can satisfy all four attributes. Moreover we can do so using devices acquired from third parties when desired. Still further, we offer advantages over the prior art in terms of lower cost to produce, higher yield and improved operating life. FIG. 10 shows, in simplified high level overview, one example approach according to the teachings of the invention. This approach overcomes shortcomings of the prior art while permitting close optical access, removing absorbing regions, providing a higher structural integrity, and having better thermal dissipation characteristics. In the approach of FIG. 10, a laser wafer 10-502 (made up of lasers integrated with a substrate 10-102) and a detector wafer 10-504 (made up of detectors integrated with a substrate 10-102) is obtained, for example, by manufacturing them using a conventional technique or by purchase from an appropriate third party. Alternatively, a hybrid wafer made up of both lasers and detectors integrated with a common substrate, for example, in some alternating pattern or other grouping, is manufactured or obtained.

Trenches 10-506 are etched to process a wafer into individual devices (by etching into the substrate) or, in some cases, into appropriate groups of devices, for example, as shown in a commonly assigned application entitled Redundant Device Array filed concurrently herewith (and which is incorporated herein by reference) by etching into the substrate in some places while stopping the etch prior to it reaching the substrate in others.

Alternatively, since the invention is not the creation of the optical chip itself, per se (i.e, the creation of the wafer, growth of the devices, or etching to created discrete devices), the above would be skipped entirely if the optical device wafer was purchased instead of made.

The optical device wafer is then inverted and aligned over an electronic wafer 10-508 and bonded to the electronic wafer 10-508 using, for example, conventional flip-chip bonding techniques or some other appropriate proprietary technique that accomplishes bonding of the optical wafer to the electronic wafer in a suitable and reliable manner.

Alternatively, and advantageously in some cases, further processing of the substrate 10-102 can be accomplished, as described immediately below, either prior to bonding an optical wafer to the electronic wafer or after bonding, so long as it is done before cycling the devices over operational temperature extremes by device operation if done after. Such processing is unsuitable for the prior art techniques described above in connection with FIGS. 6 and 7 because, if used, it would dramatically increase the cost of producing devices by requiring individual bonding of each discrete device if the substrate were completely removed or dramatically reduce the yield, due to stress and/or strain problems when the substrate is very thin. Depending upon the particular wafer(s) and optical devices used, different processing variants are now possible.

In a first variant, the substrate is thinned down to a thickness in excess of 50 microns, typically to within a range of between about 50 microns to about the 100 micron thickness typically required for close optical access. In a second variant, the substrate is thinned to a thickness of between about 100 microns and about a thickness corresponding to the thickness of the optical device portion of the wafer.

In a third variant, the substrate is thinned to between about 20 microns and about 50 microns. In a fourth variant, where the thickness of the substrate is about equal to the thickness of the optical device portion of the wafer, thinning is not required.

In a fifth variant, the substrate is thinned down to a thickness about equal to the thickness of the optical device portion of the wafer. As will be apparent from the description below, in accordance with the invention, the thickness of the overall substrate could also be kept larger that the thickness necessary for close optical access, for example, where access ways are constructed (as described below) to allow for insertion of an optical fiber or microlens into the access way to a separation spacing from the device within the close optical access range. However, it is expected that such a case will be atypical.

An access way 10-510, in the form of a trench or hole is also etched or drilled in the substrate over the portion of an optical device where light is emitted or detected, for example, using conventional etching or drilling techniques, while preferably leaving some of the remaining substrate intact. Depending upon the particular substrate and device(s) different techniques can be used including laser drilling, etching or some combination thereof. In addition, depending upon the particular technique used, the access ways may have straight sidewalls, sloped sidewalls or some combination thereof.

For example, in order to produce an access way 10-510, having initially straight sidewalls near the substrate outer surface and sloped sidewalls near where the substrate meets the device, in a Gallium Arsinate (GaAs) substrate with an (Aluminum Gallium Arsinate) AlGaAs stop layer (supporting optical devices such as NCSELs and/or photodetectors (interchangeably referred to herein as detectors)) hybridized to an ASIC (collectively referred to as the "Sample"), the following approach can be used:

First, the access ways 10-510 are resist patterned on the substrate. Then the sample is loaded into a 13.56 MHz parallel plate reactive ion etcher (RLE) and evacuated to a pressure below about 3X10^"5 Torr before introduction of the precess gasses to reduce or eliminate residual water. Once this base pressure is reached, the first part of the etch is initiated at the process conditions of Table 1.

Table 1 This produces a straight sidewall extending from the surface of the substrate into the substrate for a distance towards the device.

The process conditions are then optimized to produce the portion of the access ways 10-510 having sloped sidewalls with, in this example case, GaAs to AlGaAs selectivity near infinity with minimaldevice damage. In particular the process conditions are shown in Table 2.

Table 2 Then the process conditions are optimized to getter the residual Cl from the AlGaAs stoplayer. This is to prevent further formation of HC1 (i.e. performing a wet etching) after the Sample is unloaded from the processing chamber. The process conditions for this portion of the process are set forth in Table 3.

Table 3

In the simplest case, the access way will be as small as possible, so as to maximize the amount of substrate left on the device. The remaining substrate provides a rigid framework which prevents the individual devices from undergoing stresses, for example, during attachment to the electronic wafer. Depending upon the particular devices and substrate used however additional removal of substrate may further be performed, for example, at the time the access way is created, or by patterning the substrate at some point, for example, following attachment to the electronic wafer.

It should be noted however, that if removal of additional subsfrate is not properly planned, as more substrate is removed, the thermal dissipation advantage may be reduced or even eliminated. Moreover, depending upon how much, and/or from where, additional substrate is removed, the ability to withstand stress and strain may also be decreased. However, it can be appreciated, that, in some cases, by selective removal of substrate thermal dissipation can be improved by increasing the overall surface area of the substrate without sacrificing much, if any, of the structural advantages. Thus, it should be understood that the important aspect of the substrate removal is that sufficient substrate is left on the devices to ensure the desired thermal and structural characteristics are achieved.

Moreover, depending upon the particular technique used, provision of the access ways may advantageously be, in some cases, performed before or after bonding is performed, for example, before, after, or while the trenches separating the individual devices are etched.

Optionally, an AR coating can be applied to the detectors, if desired.

Depending upon which of the three immediately preceding variants above are used, different processing will occur. FIGS.11 and 12 show several different accessways variant examples. For example, if the first variant was used, the access ways may extend entirely through the substrate (as shown in FIGS. 11a, lib, 12a, 12c, 12e). Alternatively, they may extend from the outer surface of the substrate to a depth where the substrate remaining directly over the portion of an optical device where light is emitted or detected is reduced but not completely removed, for example, as shown in FIGS, lie, lid, 12b, 12d, 12f). In general, the substrate remaining directly over the portion of the optical device where light is emitted or detected will be reduced to a thickness of about 100 microns or less to enable close optical access to the device. In other cases, the thickness may be reduced to about 50 microns or less, and in some cases 20 microns or less, although typically the thickness will be within the range of about 20 microns to about 50 microns. Additionally, depending upon the particular access way created, the access way may further be advantageously used to accommodate an optical fiber, for example, as shown in FIGS. 1 la, 1 lc, 12b or a microlens, for example, as shown in FIGS. 1 lb, lid, 12a, 12c.

Thus, by employing one of the above approaches, an optical array in which ends of fibers are supported by the substrate can be created (such as shown in FIG. 13), an optical array that accommodates one or more individually placed microlenses supported by the substrate can be created (such as shown in FIGS. 1 lb, lid, 12a, 12c, 12e), or an optical array that accommodates an array of microlenses can be created (such as shown in FIG. 14).

As noted above, the substrate can also be patterned to roughen the surface of the substrate and increase the exposed surface area for better thermal dissipation. It should be appreciated that, by using the techniques described herein, i.e. leaving substrate attached, stresses will primarily not propagate to optical devices, but rather will be taken up by the connecting medium or the electronic chip, both of which are better able to withstand such stresses. FIGS. 15-18 are each example illustrations of the process of creating electro-optical chip variants according to the techniques described above.

•FIG. 15a is a simplified view of a single bottom surface emitting laser device 15-1002 that is part of an array of laser devices, the rest of which are not shown. The device 15-1002 is isolated from its neighbors by isolating trenches 15-1004 and is supported on a substrate 15-1006 made of an appropriate material, for example, Silicon (Si), Silicon-Germanium (SiGe), Gallium-Arsenide (GaAs) or Indium-Phosphate (InP). Although the particular material used for the substrate will likely be determined by factors independent of the invention, it is worth noting that stresses due to thermal factors can be reduced by matching the coefficients of expansion of the optical device substrate and the electronic wafer as closely as possible. Ideally, the two should be of the same material, so that the coefficients of expansion of both are the same.

Electrical contacts 15-1008, 15-1010 used for laser excitation and control are each mounted on a stand 15-1012, 15-1014 for support. One end 15-1016, 15-1018 of each electrical contact acts as an electrode for the laser device and the other end of each is a pad 15-1020, 15-1022 onto which an electrically conductive material 15-1024, such as a solder, is deposited for bonding the device 15-1002 to an electronic wafer.

FIG. 15b shows the laser device 15-1002 of FIG. 15a after the laser array has been inverted and positioned over corresponding pads 15-1026, 15-1028 of an electronic wafer 15- 1030.

FIG. 15c shows the laser device 15-1002 after it has been attached to the electronic wafer 15-1030 via a solder bond 15-1032 between the respective pads 15-1020, 15-1022, 15- 1026, 15-1028.

FIG. 15d shows the laser device after the substrate 15-1006 has been thinned to between about 20 microns and about 50 microns.

FIG. 15e shows the device after the access way 15-1034 has been created in the substrate 15-1006, in this case via etching instead of drilling. Note that in this example, the access way extends from the surface of the substrate 15-1036 to the device cladding layer 15- 1038. FIG. 15f shows the device of FIG. 15e after an optional thermally conductive material

15-1040 has been applied to the device such as, for example, a low viscosity (so it flows well for good coverage) thermal epoxy having good thermal conductivity when cured. Although the above was illustrated with reference to a laser device, the process would be that same for a detector type device, except that the detector device may also be AR coated.

FIGS. 16a-16f show another opto-electronic device being created in a manner similar to the one shown in FIGS. 16a-16f except that this laser device uses the semiconductor material of the device as the stands 16-1102, 16-1104.

FIGS. 17a-17f show another opto-electronic device being created in a manner similar to the preceding devices. As shown, this device is of the type where the device semiconductor material is not used for the stands. Additionally, the lasers of this opto- electronic device are grouped so that they can be used in a redundant fashion. As noted above, the creation of an array having redundant lasers is described in the section entitled Redundant Optical Device Array. Specifically, FIG. 18, shows two adjacent lasers in the array where, in addition to creating an access way 18-1034, grouping trenches 18-1302, 18- 1304 are etched in the remaining substrate 18-1006 using known etching techniques, to a depth that connects the grouping trenches 18-1302, 18-1304 with some of the isolating trenches 18-1004. In this manner, two or more lasers can be arranged to share a common fiber with one or more serving as a back-up laser, such as described in the section entitled Redundant Optical Device Array.

One advantage arising from grouping the lasers in this manner is that yield for a single wafer is increased because, for example, with a pair of grouped lasers, if one laser is damaged, the other can be used in its place. Another potential advantage to doing so can be an increased lifetime for the opto-electronic device. For example, when one laser of the pair finally dies, if the lasers are externally, independently selectable, the second laser can be selected and brought on line in place of the bad one. Yet another achievable advantage is reduced cost to achieve one or both of the immediately preceding two advantages. Since the incremental cost of increasing the number of lasers on a wafer is negligible, the improved yield and/or reliability/extended life is virtually free.

FIG. 18 also shows a functional representation of an example array 18-1306 produced using the technique of FIGS. 14a-14f. The array 18-1306 is illustrated from the top of the device so that the access way 18-1034 and remaining substrate 18-1006 over each laser is clearly visible. As shown in FIG. 18, the lasers are grouped in fours, a group 18-1308 being defined by the grouping trenches 18-1302, 18-1304 which ensure that there is no current path between adjacent lasers in the group 18-1308 via the substrate 18-1006 which is electrically conducting. For purposes of illustration, some of the isolating trenches 18-1004 are shown although none would actually be visible from this vantage point.

FIGS. 19a-19f, show another opto-electronic device being created in a manner similar to the devices of FIGS. 15 through 17. As shown, this device is of the type where the device semiconductor material is used for the stands 19-1402, 19-1404. Additionally, the lasers of this opto-electronic device are also grouped in the manner of FIGS. 17 and 18 except in pairs (one of which is not shown), as is evident from the grouping trenches.

As noted above, a manufacturer of opto-electronic devices of the type described above has two avenues for obtaining the optical devices - they can manufacture them themselves, or they can obtain them from a third party. By manufacturing the optical devices (referred to hereafter for simplicity as an "optical chip") and the electronic wafer (referred to hereafter for simplicity as an "electronic chip"), the manufacturer can take measures to ensure that the pads on each are properly placed so as to align with each other when the optical chip is positioned over the electronic chip. However, typically electrical and optical chips are not designed concurrently, even if they are designed and fabricated within the same organization. Thus, even with a single manufacturer, unless there is close coordination within the organization with regard to both the optical and electronic chip design, a lack of correspondence between contact pads on each can easily occur - particularly where one or both are also designed with sales to third parties or integration with devices from other sources is contemplated. Moreover, subsequent improvements or changes in the design of either may necessitate altering the location of the contact pads, thereby introducing a pad misalignment where none previously existed, even within the same organization.

Even worse, if the electronic chip is designed to be used with a variety of different optical chips, but the optical chips are commodity stock obtained from third parties (for example, chips containing: topside emitting cavity lasers, bottom emitting cavity lasers, DFB or DBR lasers, topside receiving detectors or bottom receiving detectors) that are mass manufactured for distribution to multiple unrelated users, it is unlikely that the pads on the optical devices will all be located in the same place, if they are otherwise compatible with the elecfronic chip. For example, as shown above in connection with FIG. 8, a single optical device has contact pads placed in the position specified by its manufacturer and an electronic wafer also has contact pads, onto which an optical device can be connected, placed in the position specified by its manufacturer. When the optical device is flipped over, for flip-chip type bonding with the electronic wafer, the contact pads of each will not be aligned. Nevertheless, by altering the technique described above, the invention can be employed with lasers other than the bottom emitting lasers referred to in the examples up until now, as well as with bottom emitting lasers having different contact pad alignments, top or bottom receiving detectors. Advantageously, this allows for the selection arid use of the "best-of-breed" chips having the best individual performance for the application and avoids eliminating such vendors merely because they can not, or will not, meet an electrical contact placement requirement or standard.

In general, two different processes are used, depending upon whether the optical devices are bottom emitting/receiving or topside emitting/receiving.

For ease of explanation, the term "bottom active" will be used to refer to both bottom emitting devices (lasers) and bottom receiving devices (detectors). Similarly, "top active" or "topside active" will refer to both top emitting lasers and top receiying detectors.

Bottom Active Device Process The process as usable for bottom emitting/receiving devices (i.e. bottom active devices) will now be explained, with reference to FIG. 20. To facilitate explanation, it should be presumed that the optical wafer 20-1502 was processed into an optical chip 20-1504 as discussed above. Alternatively, the optical chip 20-1504, can have been obtained from some third party. First, an insulating layer 20-1506 is added to the surface of the optical chip 20-1504 using known techniques.

Then openings or vias 20-1508 are created in the insulating layer 20-1506 to allow access to the contact pads of the optical chip. This is again done by laser drilling or etching, for example in the manner used for creating through holes in wafers described in commonly assigned application Serial No. 09/896,513 entitled "Multi-Piece Fiber Optic Component And Manufacturing Technique" filed June 29, 2001 and incorporated herein by reference.

Alternatively, the openings or vias 20-1508 can be pre-formed in the insulating layer prior to attachment, for example, if the contact pad locations are known in advance.

Then, the openings or vias 20-1508 are made electrically conductive by applying an electrically conductive material 20-1510 to the sidewalls of the openings or vias (which may optionally have been previously coated with an insulator) or filling the openings or vias with the material 20-1510.

Advantageously, if the openings or vias are not fully filled, they can be used to aid alignment. This can be done if the openings or vias are wide enough to allow the solder bumps on the other chip to "slot" into the holes, thereby providing an initial alignment between the two. Moreover, in some cases, capillary action will cause the solder to be partly drawn into the openings or vias as it melts causing a better connection and further aiding in alignment. Optionally, and alternatively, if the openings or vias were pre-formed prior to attachment, the coating or filling of the openings or vias (as desired) can also be performed prior to attaching the insulating layer to the optical chip.

Next, electrical traces 20-1512 are patterned on the exposed side of the insulator to create a conductive path from the (now coated or filled) opening or via to the location(s) on the insulator surface that will align with the placement of the contact pads on the electrical wafer. Optionally, if several different alignments are possible, depending upon the particular electronic chip the optical chip will be mated with, a single trace can create two or more alternative connection points or create a connection region if the contacts to be mated with are offset from each other slightly, but within a manageable defined area. In a variant of the above, if the chip to which the optical chip will be joined is an electronic chip (as opposed to another optical chip, such as a modμlator, or another laser to which the optical chip is optically transparent) the electrical traces could be patterned on the electronic chip since, in general, most electronic chips already come with an insulating layer that can be used for contact rerouting. Once this is accomplished, the process proceeds as described above, with the joining of the two chips 20-1514 (in this example, using flip-chip techniques) followed by, in the particular case, thinning of the substrate, removal of the substrate entirely, or leaving of the substrate at the thickness it is. Thereafter, creation of access ways 20-1516, patterning of the chip substrate, flowing of a thermal conductor, or application of AR coating can be accomplished as desired or needed.

Topside Active Device Process

The process as usable for topside emitting/receiving devices (i.e. topside active devices) will now be explained, with reference to FIG. 21. To facilitate explanation, it should be presumed that the optical chip was obtained from some third party, the process of creating the optical chip itself being independent of the invention.

In addition, either or both of two optional steps can be performed prior to starting the process. The first, attaches a carrier by the top-side surface of the optical chip. This carrier can be made of any material and is merely used for rigidity and holding the optical chip during the rest of the processing. The second, involves thinning the optical chip substrate. This reduces the amount of material that must be etched or drilled through to access the contacts present on the front of the optical chip.

At this point, the process proceeds in an analogous manner to the process of FIG. 20 as follows. Holes or vias are either etched or drilled through the optical chip substrate to the contacts on the front of the optical chip.

The holes or vias are then coated or filled with, an electrically conductive material (which may be under layered by an insulator coating) to bring the contacts out to the back of the optical chip. Alternatively, for example, if the contacts are located such that access directly from the back of the chip through the substrate would damage the chip or present some other problem, the holes or vias are etched or drilled in a suitable location and an electrical conductor can be added to the front side to connect the contact pad with the conductor coating or filling the vias or holes. Advantageously, if the openings or vias are not fully filled, they can be used to aid alignment. This can be done if the openings or vias are wide enough to allow the solder bumps on the other chip to "slot" into the holes (FIG. 2 IB), thereby providing an initial alignment between the two. Moreover, in some cases, capillary action will cause the solder to be partly drawn into the openings or vias as it melts causing a better connection and further aiding in alignment. Or, if the vias or holes can be located so as to coincide with the proper location for aligned mating with the electronic chip, that can also be done, and the vias or holes can be connected to the contact pads on the front side using conventional techniques.

As with the backside emitting/receiving device integration process, if the vias or holes do not coincide with the contact pads of the electronic chip, electrical traces are patterned on the substrate of the optical wafer FIG. 21C or the other chip FIG. 21D, in this case the electronic chip, to provide a connection between the vias or holes and the contact locations on the other chip.

At this point, the chips can be brought together and connected as described above. If the optional step of adding the carrier was performed, the carrier can now be removed. If the carrier is so thick as to cause optical access problems or has an incompatible complex refractive index which would adversely affect transmission of laser light through the carrier, it should be removed. In alternative variants, the carrier can be left on, even if it would cause optical access problems or has an incompatible complex refractive index, by patterning access ways or through holes in the carrier, preferably prior to attachment to the optical chip.

In addition, if desired, one or more additional optical elements, such as microlenses or waveguides, can be put on top of the carrier. FIG. 22 shows a process similar to that shown in FIG. 21 except a carrier is not used.

Connection or Adapter Chip Alternative

In an alternative variant usable, for example, when both the optical chip and the other chip are purchased from different parties or two or more different chips are under consideration and they have different contact pad placements, but the contact pad placement on each is known, an "adapter" or connection chip can be readily fabricated by employing the teachings herein in a straightforward manner, thus allowing design and/or manufacture to proceed nevertheless.

Referring now to FIG. 23 which shows a connection chip or adapter chip used to connect different devices, the top side 23-1802 and bottom side 23-1804 of a common wafer 23-1800 is patterned so as to create traces 23-1806, 23-1808, 23-1810 on each side from the specified contact pad locations 23-1812, 23-1814, 23-1816, 23-1818 for each chip to some common point for each.

Through holes are then created and crated or filled with a conductive material so as to join corresponding pairs, e.g., one contact on the top with its appropriate contact on the bottom when the two are brought together.

FIG. 24 shows another alternative implementation, which is a further variant of the adapter or connection chip variant, usable for topside active devices. As shown, the adapter or connection chip 24-1902 has electrical contacts 24-1904 on only one side for direct connection to the optical chip 24-1906 via connection pads 24-1908 and connection to the electronic chip 24-1910 via, for example, standoffs 24-1912, jumpers, wires, ribbons or other known attachment devices. In this arrangement, because the devices are top emitting/receiving and the adapter is located on the top side and, "optical vias" 24-1914 are also provided in the adapter to allow access to the optical light.

Then the optical chip can be placed on top of the electronic chip and the connection chip can be placed on top of both chips to provide connectivity between the optical and electronic chips.

As a side note, although described in connection with mating optical chips with electronic chips, the same basic process (i.e. use of a connection chip or appropriately patterned insulating layer or substrate to account for pad mismatch) can be adapted in a straightforward manner to account for a pad misalignment between any combination of optical, electrical, electronic, or electro-mechanical wafers.

Further Variants

As also noted above, in some cases it is sometimes desirable to coat some of the devices, specifically the detectors, with an AR coating. However, the opto-electronic chips described above are made up of two (or potentially more) dissimilar types of optical devices. And it is undesirable to have the AR coating detrimentally affect the lasers.

Advantageously, in a further optional variant of the above processes, the devices that need to be AR coated do not have to be distinguished from those that ordinarily would not be AR coated.

The process largely follows the process flows described above in connection with FIG. 10 where the laser wafers and detector wafers are created, flipped over and attached to the electronic chip via flip-chip bonding techniques.

The substrates are thinned, but as to the laser substrate, only to the point where the substrate could still be considered thick relative to the thickness of the laser cavity. Although different types of laser devices will require a different specific thickness, the thickness of the substrate should be at least several times as large as the thickness of the laser cavity, in the case of DFBs and DBRs and the distance between the mirrors, in the case of NCSELs. Since the precise distance will vary from device to device, a good rule of thumb is to use a factor of 10X the thickness of the laser cavity. However, if the thickness can be controlled precisely, it can be less than the 10X factor, the particular minimum thickness being empirically ascertainable as the minimum thickness where the AR coating does not affect the laser's ability to lase.

An analogous approach can be used for topside active lasers. In the case of topside active lasers, a substrate (which can be the carrier noted above, a separate substrate applied after carrier removal, or, if contact rerouting is not necessary or performed on the other chip, instead of a carrier) is attached to the topside of the lasers. The substrate is either thinned, after application, to a thickness as noted above, thinned to such thickness prior to application.

Once this is achieved, the lasers and detectors can be anti-reflection coated at the same time. Thus, there is no need for special patterning or otherwise distinguishing between the lasers and detectors during the AR coating process.

Thus, it should be understood that the above processes can be applied to various different devices. For example, using the teachings of the invention, stacking of modulators on top of lasers in an array compatible format can be done. In fact, it can be done when the modulators are on top of or below the laser. Moreover, it can be done whether or not the two (or more) devices are created in a single epitaxial step. Similarly, stacking of topside active devices on top of either topside or backside active devices can be performed as can stacking of backside active devices on top of either topside or backside active devices such as shown in FIGS. 25 A and in greater detail in for modulator mounted on a backside emitting laser 25B.

Devices that have a lattice mismatch can similarly be stacked regardless of the functions the individual devices perform.

In a further application, devices from different epitaxial wafers can be integrated together on a common chip on a wafer scale level. Thus, lasers of different wavelengths can be intermixed for dual wavelength division multiplexing (DWDM) or multiple wavelength division multiplexing (MWDM) applications, such as shown in FIG. 26.

FIG. 26 shows an array of one hundred different wavelength lasers all integrated on a common chip on a wafer scale. By doing so, and making each laser selectable, a specific wavelength (or combination of wavelengths can be selected. Thereby eliminating the need for tunable lasers which rely on analog movements of physical pieces or show thermal changes or effects and where speed is limited to microseconds and accuracy is limited.

In addition, wavelengths can be switched at the same rate that data is sent, thereby making construction of a system that multiplexes various data streams at different wavelengths at the bit rate. Thus, switching can be achieved in about 100 picoseconds (10s of gigabits/sec).

Moreover, different devices, of different types (i.e. different types of lasers, lasers and detectors, etc.) can be intermixed such as shown in FIG. 27 from a cutaway side view.

As shown in FIG. 27, strips of two different wavelength lasers 27-2202, 27-2206 are created, as are two different strips of complementary wavelength photodetectors 27-2204, 27- 2208. The strips of the first devices (illustratively lasers 27-2202 (λ_\)) are attached using the processes described herein. The strips of the next devices (illustratively detectors 27-2204 (Δ_λi)) are attached in similar fashion. Next the strips of the third devices (illustratively lasers 27-2206 (λ₂)) are attached in similar fashion. Finally, the strips of the last devices (illustratively detectors 27-2208 (Aχ₂)) are attached in similar fashion.

Depending upon the particular case, the substrate or carrier can be removed or thinned from all the devices at once, for example if they did not interfere with the integration of the next devices, or they can be removed or thinned after each set of devices is attached. FIG. 28 shows the integration of the devices of FIG. 27 from a top view. As shown, all the first wavelength lasers are attached. Then, all the first wavelength photodetectors are attached. Then all the second wavelength lasers are attached, followed by all the second wavelength photodetectors so that the end result is a fully integrated dual wavelength fransceiver chip, a portion of which is shown in enlarged form on the right side of FIG. 28.

Of course, while the immediately preceding example used two lasers and two detectors, the process would be essentially the same irrespective of the number of different devices, whether they are top or bottom active, grouped, all lasers, all detectors, etc., since an advantage of the process is the ability to mix and match - particularly on a wafer scale. In these cases, the integration can readily be performed on an individual device (or device type) basis or can be done, for example, in strips (as shown) or by groups, with the substrate left on defining the strip 28-2202, 28-2204, 28-2206, 28-2208 or group.

Still further, by integrating groups of redundant lasers of one wavelength with those of other wavelengths, an extremely reliable DWDM or MWDM module can be produced at low cost.

Thus, since single device, integrated transmitter arrays for DWDM systems are not available in the prior art, by integrating large numbers of lasers on a single chip, packaging size can be reduced. By integrating arrays often or more lasers, of two or more different wavelengths, onto a single chip and coupling a set of them into a single fiber, for example, using a fiber based combiner/inverse splitter, a holographic lense array, or the techniques of the incorporated by reference applications entitled Multi-Piece Fiber Optic Component And Manufacturing Technique, the multiplexing of multiple wavelengths can be achieved in the output fiber, in some cases without the need for an opto-mechanical or electro-optical element to do the switching (optical crossconnect). In a further application of the techniques, a large array can be constructed that can serve as both a pumping laser and a communications laser, either at different times or concurrently.

Redundant Optical Device Array

We have also created a way to deploy large numbers of optical devices in a manner which provides a higher overall yield and greater reliability. Depending upon the particular implementation, further advantages such as reparability after deployment, and performance optimization can be achieved.

One aspect of he invention involves an optical module has multiple optical devices. At least two of the multiple optical devices are a group. Each of the optical devices in the group are individually selectable relative to the others. The optical module also has a controller, coupled to the devices such that the controller can select which of the devices in the group will be active at a given time.

Another aspect of the invention involves a method of creating an optical chip, having redundant devices, for use in an opto-electronic unit involves growing active portions of multiple optical devices on a wafer, processing the wafer to create complete optical devices, creating individual optical devices, grouping the devices; and connecting the devices in agroup to a control circuit such that, common data can be received by any of the devices in the group but the common data will only be handled by the device in the group that is active. Yet another aspect of the invention involves a communications network that has a first transmitter having a number of usable channels, a first receiver, and optical fibers connecting the first transmitter to the first receiver. The first transmitter has multiple lasers, at least some being selectable as either active or backup lasers. The multiple lasers are controllable such that, if a specific channel is in use by an active laser and a laser failure occurs for that channel, a redundant laser can be substituted for the active laser and, after the substitution, the specific channel can be used using the redundant laser.

FIG. 29 shows a portion 29-300 of a two dimensional array of lasers 29-302 created according to the principles of the invention. The portion shows two individual laser devices 29-302 bonded via contact pads 29-304 to an electronic chip 29-306. As shown, the devices 29-302 are bottom emitting laser devices that have been flip chipped bonded to the electronic chip, although as will be apparent from the description herein, bottom receiving, top emitting or top receiving devices can be used as well.

Because the substrates 29-308 have not been removed or excessively thinned, emissions of the lasers occur via access ways 29-310 created in the substrate 29-308 on which the laser devices were supported to allow for close optical access to the devices. The spacing between the access ways, i.e. the pitch, is such that each of the lasers 29-302 can be either directly coupled to a single optical fiber, or directed into a common optical fiber, for example, by focusing the light output using a micro lens or guiding the light using an optical waveguide. Thus, depending upon the particular lasers and fibers used, the pitch between the two lasers can be as small as 5-10 microns for direct lasing into a single mode fiber or 50-100 microns for direct lasing into a multimode fiber. Alternatively, if an optical wave guide or focussing lens is used, the inter-device pitch becomes less important and may be as much as a 100 microns or more as needed. During device creation the lasers are separated into individual devices by patterning the laser wafer prior to bonding with the electronic chip. Additionally, the devices are patterned with grouping trenches 29-312 which physically and electrically define groups by creating boundaries separating individual groups 29-314 of redundant devices. The grouping trenches 29-312 ensure isolation among the individual groups while allowing for carrier movement among the devices within the group via the electrically conductive substrate 29- 308.

All the devices in a group 29-314 share a common connection (either the positive or negative contact) so that any signal to be sent or received by any of the devices can be sent or received by any other of the devices in the group irrespective of which one is selected as being active using the contacts. In other words, if three lasers constitute a group in an optical module, such as an optical transmitter, they will be coupled to a single optical fiber, all have one contact in common and all have individual opposite polarity contacts. If the transmitter were to send data through the optical fiber, the same signal would be sent irrespective of which laser was active at the time. Moreover, from the perspective of the functions of any prior art optical transmitter, the transmitter incorporating the invention can be oblivious to which laser in the group is active. Advantageously, the purchaser or user of the transmitter, or any other device employing the invention, need not know it contains device redundancy. The features and elements that allow selection of the particular active laser can be wholly transparent to anyone other than the manufacturer or can be made accessible to third parties to varying degrees.

FIG. 30A shows a portion of a laser array employing groups 30-400 of four lasers 30- 402 as a redundant group. As shown, the individual devices have been separated through patterning of separation trenches 30-404 which isolate the individual device contacts 30-406, and groups 30-400 have been created by patterning of grouping trenches 30-408 which isolate the common contact 30-410 from the common contact(s) of other neighboring groups. As with FIG. 29, access ways 29-310 are provided through the substrate to provide for close optical access to the lasers. FIG. 30B is a functional representation of the group of FIG. 30A but showing the discrete contacts 30-406 for each laser and the 29-308, which is used as the common contact.

Advantageously, by grouping the lasers in fours, significant flexibility can be obtained. For example, the best two of the four lasers can be used as a redundant pair with or without the remaining two lasers serving as back up devices for either laser in the pair, the best of the four lasers can be employed as a primary laser with each of the remaining three being available should the primary laser fail, or should any individual laser in the group be bad, it can be disregarded entirely.

FIG. 31 shows the functional components of an opto-electronic device 31-500 suitable for employing the principles of the invention. Functionally, the device includes a laser portion 31-502 which contains an array of individual lasers. The device also includes a detector portion 31-504 which includes an array of individual photodetectors. A control portion 31-506 is provided which contains the control electronics for accessing the individual lasers and/or detectors. Additionally, a storage portion 31-508 can optionally be provided, as will be described in greater detail below. Finally, the device includes an interface portion 31- 510 through which the opto-electronic chip may be electrically or programmatically connected to other devices or control electronics. Depending upon the particular implementation, the interface portion 31-510 may be functionally located between the control portion 31-506 (and/or the storage 31-508 if this option is used) and the devices 31-502, 31- 504, for example where the control 31-506 and/or the storage 31-508 can be provided by a third party. In other variants, the interface 31-510 may provide a way to bypass or override either or both of the control portion 31-506 and/or storage 31-508 if either or both are present.

Functionally, the control portion 31-506 is, in whole or part, the "brains" of the optoelectronic chip 31-500. At least, it is the brains with respect to the redundancy feature. The control portion 31-506 is physically made up of the hardware used to activate the individual devices based upon, for example, information stored in the storage, and/or to specify, update and/or change the stored information to initialize the chip or reprogram it following a failure. Depending upon the particular implementation, the control portion will be a processor, for example, a microprocessor operating under program control, a state machine, hard wired circuitry or some combination thereof.

Depending upon the particular implementation, the storage will be in the form of static random access memory (SRAM), dynamic random access memory (DRAM or RAM), or some form of read only memory (ROM) which maybe, for example, a device such as a programmable read only memory (PROM), an electronically programmable read only memory (EPROM), an electronically erasable programmable read only memory (EEPROM), a programmable logic device (PLD), etc. to name a few.

The storage 31-508 is accessible by the control portion 31-506 and is configured to allow the active device in each group to be specified. Optionally, the storage 31-508 can be further configured to keep track of redundant (i.e. back-up) devices and, as a further option, can be configured to specify the hierarchy or ordering for bringing on-line the remaining devices in the group if needed.

For example, Table 4 shows a simple table that can be employed for groups of device pairs. Each pair has a group address or identifier that uniquely identifies, directly or indirectly, each discrete group. A single bit is used to designate the active device, for example, with a binary 0 representing the first device in the group and a binary 1 representing the second device in the group.

Table 5 shows an alternative arrangement for identifying the active device in the storage. As with Table 4, an address or identifier uniquely identifies the particular group. Associated with that address is a two-bit binary number, where each bit corresponds to one device in the group and is used to signify whether that device is to be active.

For example, a bit pattern of 00 would specify that neither device is active. Bit patterns of 01 or 10 would indicate that one or the other device in the pair is active. Depending upon the particular implementation, a bit pattern of 11 could, for example, be used to activate both devices for some special case or could simply be an invalid state.

Table 6 shows a similar arrangement for a chip having groups made up of four devices. In this case, a similar two bit binary number is used except, the actual number in binary is used to indicate the active device.

For example, a 00 would indicate that the first device in the group is active, a 01 would indicate the second device in the group is active. A 10 would indicate that the third device is active and a ll would indicate that the fourth device is active.

Table 7 shows a more complex arrangement for keeping track of the active devices in a particular array having individual four device groups. As shown Table 7 includes an address as noted above. In addition, an eight-bit binary number (X1.X₀A1A₀B1B₀C1C0) is used to identify the particular laser device in the group that is the primary (i.e. active) device as well as a hierarchy for the remaining devices in the group.

For example, for a particular address, an entry of 01110010 would indicate that the second device (01) is active. In the event that device was unusable or failed, the next device to be brought on-line is the fourth device. If that device were to fail, the next devices brought on-line thereafter would be, in order, the first followed by the third.

As can be appreciated, there are numerous ways other ways to identify active devices and/or specify alternative devices, either by employing some variant or combination of the above examples, or creating some other methodology, for example, by designating each laser with a unique address (irrespective of its group) and maintaining a list of the addresses for the lasers in each group in the order they are to be brought on line or providing space for settings for each laser, such as bias and modulation, and filling the setting information in for active lasers and/or setting the bias and/or modulation settings to zero and/or an invalid value to deactivate a laser.

In an alternative implementation, involving no storage for device selection, the devices incorporate fusible links that can be used to bring a device on- or off-line. For example, each device may incorporate two fusible links. Initially, neither link is blown so the device is inactive but available. To bring a device on line, circuitry is activated that causes a particular link to blow and renders the device active, hi the event that device dies some time in the future, other circuitry can be enabled to blow the remaining link, rendering the device inactive. A redundant device in the group can then be brought on-line by blowing the first link for that redundant device in a similar manner. Still other alternative implementations use a combination of storage and hard wiring or fusible links to accomplish the functions of the control and/or storage.

FIG. 32 shows an opto-electronic device of the type shown in FIG. 31 in greater detail and constructed according to the principles of the invention. As shown, the detector portion 32-604 is made up of 36 individual detectors and the laser portion 32-602 is made up of 36 pairs of redundant lasers. As shown, the individual lasers 32-606, 32-608 in a group 32-610 are separated by device trenches 32-612 and the groups are separated from each other by grouping trenches 32-614.

In addition, there are available areas 32-616 between adjacent rows of the paired redundant lasers. Depending upon the particular implementation, those areas may be wholly unused, may be occupied by lasers of other wavelengths than those of the redundant pair, or may represent additional lasers of the same type as the redundant pairs which have been disabled for one reason or another.

FIG. 33 shows an opto-electronic chip 33-700 similar to that of FIG. 10 except that the array has been patterned as if four discrete devices were present to make up a group 33- 702. However, each group contains only two lasers 33-704, 33-706.

FIG. 34 shows a chip 34-800 similar to the chip of FIG. 32 except that each individual group 34-802 is now made up of four individual lasers 34-804, 34-806, 34-808, 34-810.

FIG. 35 shows a chip 35-900 like the device of FIG. 31 but having pairs 35-902 of redundant photodetectors. As shown, the photodetectors are grouped, like the lasers of FIG. 10, by grouping trenches 35-904 and individual photo detectors 35-906, 35-908 within a group are separated by device trenches 35-910.

It is important to note in connection with redundant detectors, that the use of redundant detectors will require that either some additional device be used to redirect the incident light from one detector to the other detector in order to switch between them. Alternatively, the light can be defocused or detracted so as to be incident on all pertinent devices on both (in this case) as required. As should be apparent however, if redundant detectors are used and no light redirection is provided the system must be capable of accepting the losses due to such defocusing or diffracting because the amount of incident light will be reduced exponentially as it is defocused to a larger and larger area to accommodate a larger number of redundant devices or a large pitch among them.

FIG. 36 shows a chip 36-1000 having an array 36-1002 similar to the array of FIG. 35 except that the array of FIG. 36 incorporates four redundant detectors 36-1004, 36-1006, 36- 1008, 36-1010 per group. Having shown a number of functional variants according to the invention, some examples of aspects usable for specific implementations will now be provided.

FIG. 37A shows one functional example of a circuitry arrangement for selecting from among two or more redundant devices according to the invention, h variants according to this example, a common data signal source 37 A- 1102 is connected to all of the lasers 37A- 1104 in a group. As shown two or more lasers are in the group. A multiplexor 37 A- 1106 (for 1 -to- 1 connections) or a selector (for 1 -to- 1 or more connections) is inserted between the power source 37A-1108 for the lasers and the lasers themselves. The control information (whether bit based or bias/modulation based) is used by the control portion 37A-1110 to select which laser receives power. Alternatively, in some variants, the multiplexor can be replaced with a selector that can select any one or more of the lasers.

FIG. 37B shows another functional example of a circuitry arrangement from among two or more redundant devices according to the invention. In variants according to this example, a signal source 37B-1112 is amplified by an amplifier 37B-1114 and connected to the lasers 37B-1106 via a multiplexor (for 1-to-l connections) or a selector (for 1-to-l or more connections). The multiplexor 37B-1106 or selector is controlled in a similar manner to FIG. 37A.

FIG. 38 functionally shows a communication system including an opto-electronic transceiver 38-1200 incorporating the invention. As shown, the transceiver 38-1200 includes a chip 38-1202 incorporating redundant lasers 38-1204 in accordance with the invention. The transceiver 38-1200 is arranged so that each pair of lasers 38-1204 is coupled to a common fiber 38-1206. As shown, optical waveguides 38-1208 shaped like a "Y", are used to guide laser light from either laser 38-1210 in the pair to a common fiber 38-1206. In other variants, other forms of waveguides, or microlenses, gratings, fused fibers, etc., are used to couple the two or more lasers to a common fiber, the particular coupling method used being irrelevant to understanding the invention.

The transceiver 38-1200 also includes an electronic interface 38-1212 through which electrical signals, for example digital data can be received and sent. Depending upon the particular set up, the transceiver 38-1200 may be constructed to convert received digital signals into optical signals to be transmitted over one or more fibers using the lasers, to a receiver 38-1214, which may be a standalone unit or be part of another transceiver, having photodetectors 38-1216. Additionally or alternatively, the transceiver 38-1200 may use those digital signals as control signals and/or receive the signals for use as in any conventional electro-optical fransceiver. Similarly, the transceiver 38-1200 is constructed to detect incident light received on its detectors 38-1218 and convert that light to digital signals that are then output via the electronic interface in a conventional manner.

Advantageously, further variants can be constructed for automatic failover. FIG. 39 is a functional block diagram of one example way to integrate automatic failover. As shown, a group 39-1300 is made up of two lasers 39-1302, 39-1304 coupled to a common fiber, for example, a "cone" or "funnel" shaped waveguide 39-1305, that is common to both lasers 39- 1302, 39-1304. The controller 39-1306 selects which laser is active by outputting a logical one or zero. A sensor 39-1308 monitors the output of the active laser, for example via sampling the output power of the laser when in use, and feeds the result back to a failover controller 39-1310, which may or may not be part of the controller 39-1306 but is functionally shown separately for purposes of understanding. The failover controller 39-1310 is used to determine if the active laser should be switched out in favor of another laser in the group based upon some value related to the performance of the laser - in this case output power. Depending upon the particular implementation, any of the many different known techniques for determining if a value is at a limit or within a range can be used. For example, a comparator may be used to directly or logically compare the sample to a threshold value, a trigger can be set to fire when the sample falls below a threshold, etc. . .

If, at some point, the laser power falls below the specified limit or goes outside the desired range, that laser will be deactivated in favor of another laser in the group using one of the techniques noted above. For example, as shown, the failover controller 39-1310 is connected to the storage 39-1312 so that if a failover for a laser is required, the failover controller 39-1310 changes the value in the storage 39-1312. That causes the controller 39- 1306 to de-activate the one laser 39-1302 in favor of the backup laser 39-1304.

Depending upon the particular implementation, it may be desirable include circuitry or stored information such that, if a substitution of one device for another has occurred

(whether automatically or manually) the "bad" device can be designated as such to prevent a switch back to the bad device if the backup device fails.

It should be understood that, although described largely in connection with an optical transceiver, the invention may be straightforwardly employed in an optical transmitter module or an optical receiver module, there being no need for any particular implementation to have two different types of devices (i.e. transmitters and receivers) to be present in the same unit to use the invention. Moreover, it should be understood that the invention may be straightforwardly employed with any type of laser device, i.e. surface emitting lasers, distributed feedback (DFB) lasers, Distributed Bragg Reflector (DBR) lasers and/or any type of photodetectors.

It should therefore be understood that the above description is only representative of illustrative embodiments. For the convenience of the reader, the above description has focused on a representative sample of all possible embodiments, a sample that teaches the principles of the invention. The description has not attempted to exhaustively enumerate all possible variations. That alternate embodiments may not have been presented for a specific portion of the invention, or that further undescribed alternate embodiments may be available for a portion, is not to be considered a disclaimer of those alternate embodiments. One of ordinary skill will appreciate that many of those undescribed embodiments incorporate the same principles of the invention and others are equivalent.

Claims

What is claimed is:

1. A pump laser capable of delivering at least a specified amount of output power, the pump laser comprising: an array of N semiconductor lasers each having a first wavelength and an individual available output power (P) such that the product of N times P is equal to or greater than the specified amount of output power; and a coupler configured to couple light emitted by the individual lasers in the array to an individual optical fiber.

2. The pump laser of claim 1 , wherein the first wavelength is 968nm.

3. The pump laser of claim 1 , wherein the first wavelength is 980nm.

4. The pump laser of claim 1, wherein the first wavelength is 1480nm.

5. The pump laser of claim 1 , wherein the first wavelength is 968nm.

6. The pump laser of claim 1, wherein N is at least 100.

7. The pump laser of claim 1, wherein N is at least 1000.

8. The pump laser of claim 1 , wherein N is a multiple of 2^m where m is at least 4.

9. A laser array comprising: multiple individual lasers of a single specified wavelength, each capable of delivering at least a fraction of a desired aggregate output for the array; at least one photodetector configured to, when N of the individual lasers are operating, sample an aggregate output of the operating lasers so that, when the aggregate output of the operating lasers differs from the desired aggregate output for the array, an output power adjustment can be made to the operating lasers by adjusting each of the operating laser's output power by 1/Nth of the difference between the aggregate output and the desired aggregate output.

10. The laser array of claim 9 further comprising a temperature sensor.

11. The laser array of claim 10 further comprising a processor programmmed to control laser drive circuitry based upon feedback provided by at least one of the photodetector or the temperature sensor.

12. The laser array of claim 9 further comprising a processor prograrnmmed to control laser drive circuitry based upon feedback provided by the photodetector.