Package for micro light emitting diodes for chip to chip communication

Abstract

A micro-LED based optical chip-to-chip interconnect can optically couple chips in various ways. The micro-LEDs can be positioned within a waveguide, and the interconnects can be arranged as direct connections, bus topologies, or repeaters.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to chip-to-chip optical interconnects, and more generally to chip-to-chip optical interconnects including micro light emitting diodes (microLEDs). BACKGROUND

[0002] Logic and memory capabilities can be limited by electrical interconnects. As the requirements of data flow increase, particularly in machine learning and other high performance applications, chips can not be packaged individually, but rather combined into multi-chip modules. Within these multi-chip modules, which in some cases can be considered SiPs (system in a package), multiple chips are connected using very dense wiring. For example, logic and memory can be mounted on a silicon interposer using micro bumps. However, even such interposers can not be sufficient to connect the desired number of ICs together at sufficiently high data rates. Connections within these packages can be limited by the parasitic resistance, inductance, and capacitance of the wiring, and chips can need to be mounted very close to each other to reduce the effects. This can limit the number of ICs that can be combined in a multi-chip module. Additionally, wiring density also presents a constraint. If the number of lanes for data communication is reduced by increasing the speed per lane, the parasitic effects become worse, and additional power is typically consumed in the SERDES to multiplex the data to the higher rate. SUMMARY

[0003] Some embodiments provide optical links between integrated circuits (ICs) using light from microLEDs. In some embodiments, the microLEDs are packaged to improve light transmission characteristics.

[0004] Some embodiments provide a chip-to-chip optical interconnect including microLEDs, comprising: a first semiconductor chip electrically coupled to an interposer; a second semiconductor chip electrically coupled to the interposer; the interposer including electrical signal paths electrically coupling the first semiconductor chip and the second semiconductor chip; a first microLED; first circuitry electrically coupled to the first semiconductor chip for driving the first microLED based on data from the first semiconductor chip; a first encapsulant substantially encapsulating the first microLED; a first photodetector; first amplification circuitry for amplifying a signal from the first photodetector, the first amplification circuitry electrically coupled to provide an electrical signal to the second semiconductor chip; a first waveguide optically coupling the first microLED and the first photodetector, the encapsulated first microLED being within a material of the first waveguide.

[0005] Some embodiments further provide: a second micro-LED; second circuitry electrically coupled to the second semiconductor chip for driving the first micro-LED based on data from the second semiconductor chip; a second encapsulant substantially encapsulating the second micro-LED; a second photodetector; second amplification circuitry for amplifying a signal from the second photodetector, the second amplification circuitry electrically coupled to provide an electrical signal to the first semiconductor chip; the first waveguide optically coupling the second micro-LED and the second photodetector, the encapsulated second micro-LED within a material of the first waveguide.

[0006] Some embodiments alternatively or additionally further provide: a second photodetector; and second amplification circuitry for amplifying a signal from the second photodetector, the second amplification circuitry electrically coupled to provide an electrical signal to a third semiconductor chip; and wherein the first waveguide optically couples the first micro-LED and the second photodetector.

[0007] Some embodiments alternatively or additionally further provide: a second micro-LED; second circuitry electrically coupled to the first amplification circuitry for driving the second micro-LED based on data from the first photodetector; and a second optical waveguide optically coupling the second micro-LED and a third photodetector.

[0008] These and other aspects of the application are more fully understood in conjunction with the following written description. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 An example of two chips in data communication according to aspects of the application is shown.

[0010] Figure 2 Another example of two chips in data communication according to aspects of the application is shown.

[0011] Figure 3 An example basic architecture of using micro-LEDs for chip-to-chip interconnects in 2D format according to aspects of the application is shown.

[0012] Figure 4A A top view of a structure for using encapsulation and reflectors for increasing coupling of light from micro-LEDs into a multimode waveguide according to aspects of the application is illustrated.

[0013] Figure 4B DESCRIPTION Figure 4A A side cross-sectional view of the structure of

[0014] Figure 5A A micro-LED pillar mounted on a silicon substrate with metallization layers above, below, and to the sides according to aspects of the application is shown.

[0015] Figure 5B This is a 3D illustration of a miniature LED column according to an aspect of the invention, showing a rear reflector wrapped around the side to deflect light in the forward direction.

[0016] Figure 6 A side view illustrating the structure of a reflector with a spherical reflective geometry in the vertical direction, according to an aspect of the invention, for increasing the coupling of light from a micro-LED into a multimode waveguide.

[0017] Figure 7A A perspective view showing a transmitter and receiver based on a half-duplex micro-LED coupled to a solder ball according to an aspect of the present invention.

[0018] Figure 7B yes Figure 7A The device is partially unfolded, with the waveguide not shown for clarity.

[0019] Figure 8 Description of aspects of the present invention Figure 7A Multiple optical interconnects.

[0020] Figure 9 This invention demonstrates a micro-LED-based TX coupled to a waveguide, having a splitter to multiple RXs.

[0021] Figure 10 An embodiment of three nodes (each node having TX and RX) according to an aspect of the invention is shown, all nodes being connected via a star topology.

[0022] Figure 11 A side view of an example of an optical-to-electrical-to-optical (OEO) repeater based on a micro-LED, according to an aspect of the invention, is shown.

[0023] Figure 12 A top view showing an example of an interconnect based on a bidirectional repeater according to an aspect of the present invention.

[0024] Figure 13 An example of the internal logic of a repeater according to an aspect of the invention is shown.

[0025] Figure 14 A half-schematic representation of an example embodiment of an array of drivers and receivers between an FPGA and an HBM stack, according to aspects of the present invention.

[0026] Figure 15 An example of integrating a FET driver with a micro LED according to an aspect of the present invention is described.

[0027] Figure 16An example of a detector integrated with an amplifier according to aspects of the application is shown. DETAILED DESCRIPTION

[0028] Some embodiments provide optical links between integrated circuits (ICs), and micro-LEDs are particularly useful for this application. By using very dense arrays of optical links running at slower speeds, serialiser-deserialiser (SERDES) can be reduced or not needed, and high data rates can be achieved with lower power consumption. There are also typically no parasitic circuit limitations or electrical crosstalk in optical links. Such optical interconnects can be implemented in 2D format in waveguides, or in 3D format using lens arrays, holograms or 3D waveguides. Light from micro-LEDs is quite different from that of lasers in emission pattern, spectrum and other characteristics, and so different packaging architectures can be preferred for using micro-LEDs for chip-to-chip communication.

[0029] Optically enhanced interposer

[0030] Figure 1 Two chips 111, 113 in data communication are shown. In some embodiments, the chips are micro-chips. In some embodiments, the chips are part of the same multi-chip module. In some embodiments, the chips are in the same package. In some embodiments, the chips are on the same substrate, or coupled to the same substrate. The two chips are interconnected together, but instead of a conventional electrical link, there are a number of transmitter (Tx) 115 and receiver (Rx) 117 blocks, which convert electrical signals to optical signals and back again. The link 119 between the Tx blocks and the micro-chip is electrical, as is the link 121 between the Rx blocks and the micro-chip. But the connection 123 between the Tx blocks and the Rx blocks is optical. The Tx blocks include micro-LEDs for generating optical signals based on electrical signals, and the Rx blocks include optical detectors for generating electrical signals based on received optical signals. Figure 1 Individual Tx and Rx blocks are shown, but multiple Tx and Rx blocks can be integrated together on a single integrated circuit.

[0031] Typically, both the input and output of a micro-chip use the same electrical pin, typically half-duplex mode, where the wire is either transmitting or receiving. Various handshaking routines can be used to transmit and receive using the same line. This can also apply to the optical domain, as Figure 2Here, each wire is connected to a Tx block 115 and an Rx block 117. In some embodiments, logic 211 connects the two together, such that when transmitting data the Tx block is activated, and when receiving the data the Rx block is activated. In some embodiments, a separate optical wire or waveguide is used for transmission, and another optical wire or waveguide is used for reception, as shown in the figure. Alternatively, in some embodiments, a single waveguide or optical connection is used to carry data in both directions. The latter option can provide higher density, as fewer waveguides can be used. However, there can be some loss or "blockage" associated with using the same wire for transmission and reception.

[0032] Figure 3 A basic architecture for using micro-LEDs for chip-to-chip interconnects is shown in a 2D format. Various chips 311, 313 to be interconnected can be micro-bumped onto an optical interposer assembly that includes drivers 363, micro-LEDs 361, optical propagation media such as waveguides 323, photodetectors 371, and amplifiers 373. In some embodiments, the chips can be silicon processors. In some embodiments, the chips can include CPUs, GPUs, and / or memory. The chips 311, 313 are mounted to the interposer, for example, using solder balls and in some embodiments, additional items. In some embodiments, the chips can be mounted to the interposer using flip-chip technology. Figure 3 In the figure, chip 311 can provide data to the driver 363. The driver activates the micro-LEDs 361 to produce light that encodes the data, where the light enters a first end of the waveguide 323 and travels through the waveguide to a second end of the waveguide. The photodetector 371 is located at the second end of the waveguide, and produces an electrical signal based on the received light. The amplifier amplifies the electrical signal from the photodetector, where the data of the amplified signal is provided to the second chip. Although more complex than a simple "wire," the link can generally operate at lower power, as the capacitance at both the source and destination chip ends is generally greatly reduced relative to an electrical connection. The interposer 351 can still have basic electrical wires (e.g., power for the driver and receiver, ground wires, and other control signals that are relatively slow). Hybrid implementations are also possible, where some high-speed wires (e.g., some of the high-speed wires between chips 311, 313) are electrical, and some are optical. For many power and signal wires, there can be through-holes (e.g., through-chip vias TCVs or through-substrate vias TSVs) through the interposer substrate. There can be electrical connections through solder bumps (e.g., controlled collapse chip connections or "C4" bumps) from the back of the chips to the interposer, and through wirebonds from the top of the chips to the interposer. Thus, the optics can augment the electrical interposer, rather than completely replace all electrical wires.

[0033] Optimizing micro-LEDs with waveguides for optical interconnects

[0034] In contrast to emission from lasers, which are typically confined to a relatively small number of spatial modes and have a relatively directional emission, emission from micro-LEDs is typically Lambertian or omnidirectional. Light in a micro-LED is generated inside a high-index medium (typically a III-V semiconductor), and thus can suffer significant total internal reflection, and is difficult to extract. In fact, without additional modifications, total internal reflection in a high-index material typically limits extraction efficiency to only a few percent. For illumination applications, the LED surface can be roughened to reduce total internal reflection losses. The LED can also be placed on a reflective surface to take advantage of light that is sent to the back of the device. In small devices, a reflector can be placed on the sides to similarly direct emitted light laterally to the front of the device. Since it is difficult to mold LEDs into a spherical geometry, it can also be very useful to increase extraction efficiency using a high-index medium (such as a polymer or epoxy) to encapsulate the LED into a spherical mold, as the total internal reflection criteria is reduced.

[0035] In the following discussion, a transmitter (TX) includes a micro-LED and a micro-LED driver circuit that drives the micro-LED. A receiver (RX) includes a photodetector followed by a receiver circuit, where a typical receiver circuit includes a transimpedance amplifier (TIA) followed by a limiting amplifier (LA). This TX and RX can also be considered as examples of a TX block and a RX block of Figure 1 and 2 .

[0036] A micro-LED is made of a p-n junction of a direct bandgap semiconductor material. Micro-LEDs can be distinguished from semiconductor lasers (SLs) in the following ways: (1) micro-LEDs do not have an optical resonant structure; (2) optical output from micro-LEDs is almost entirely spontaneous emission, while output from SLs is primarily stimulated emission; (3) optical output from micro-LEDs is temporally and spatially incoherent, while optical output from SLs has significant temporal and spatial coherence; (4) micro-LEDs are designed to operate down to zero minimum current, while SLs are designed to operate above a minimum threshold current, which is typically at least 1 mA.

[0037] Micro-LEDs can be distinguished from standard LEDs in the following ways: (1) have an emission area that is less than 100 pm x 100 pm; (2) typically have positive and negative contacts on the top and bottom surfaces, while standard LEDs typically have both positive and negative contacts on a single surface; (3) are typically used in large arrays for display and interconnect applications.

[0038] Figure 4A and 4B Structures are described for increasing coupling of light from a micro-LED 411 into a multimode waveguide 415. Figure 4A and 4BAn example using a polymer waveguide. In various embodiments, different waveguide materials can be used, such as a germanium-doped silica waveguide, a silicon nitride / silicon oxide waveguide, or other materials. The cladding can be a lower index material, air, an oxidized surface of a silicon wafer, or even a high-reflectivity mirror, such as an anti-resonant reflecting optical waveguide (ARROW). In some embodiments, the top and side cladding can be air, with a bottom polymer cladding.

[0039] In some embodiments, the micro-LEDs can be encapsulated in an encapsulant 413, which can be a high-index material, or formed in the waveguide itself. For example, a polymer waveguide can be formed in a polymer with an opening. The micro-LEDs can be bonded into this hole. The hole can then be filled with an encapsulant, such as a silicone elastomer material. The encapsulant can increase the extraction efficiency by reducing the total internal reflection constraints.

[0040] In some embodiments, reflectors are placed under and around the micro-LEDs to direct light into the waveguide. The bottom surface 451a on which the micro-LEDs are bonded can be reflective, for example with a metal such as silver or aluminum. The top surface of the micro-LEDs can also be coated with a reflective contact 451b. Without these reflectors, light that is emitted vertically up or down from the micro-LEDs can often be lost. These reflectors send the light back into the structure, where it can be scattered into the waveguide, or in the case of being absorbed, create electron-hole pairs, which in turn can be re-emitted into the desired optical mode. In either case, the reflectors can help turn light that is emitted in an undesired direction into a direction that can propagate along the waveguide, and thus increase the efficiency of coupling into the waveguide. The back edge of the waveguide, around the opposite end from the light detector 419, can also contain a metallized mirror 417, for example with a parabolic shape. The light detector can have electrical connections 421a,b.

[0041] The metallization around the micro-LED itself can be very effective. Figure 5A A micro-LED pillar 511 mounted on a silicon substrate 561 is shown (indicated in Figure 5B ), with metallization above, below, and to the sides. The micro-LED contains an N-type GaN layer 553 on an n-metal contact 519, and a P-type GaN layer 551 under a p-metal contact 513. The P-type GaN layer and N-type GaN layer sandwich an intrinsic GaN region 555 with InGaN quantum wells. The p-metal contact contains a portion 515 that extends along the micro-LED to provide a portion 517 on a silicon dioxide layer 559, on which the n-metal contact also sits. The silicon dioxide layer is on top of the silicon substrate. A passivation layer 557 around the micro-LED can prevent shorts between the p-metal and n-metal. A window on the front allows light to escape in the correct direction (as viewed in FIG. 5). Figure 5B is, for example Figure 5AA 3D illustration of the device of the device showing how the back reflector 515 (e.g. p-metal) can wrap around the sides, deflecting the light in the forward direction. The selective metallization layer on the back of the micro-LED can be implemented using angled evaporation or other lithography techniques. The passivation can be AlN, SiN, Si02, or other dielectric material, and can be completely around the pillar and to the back, as the typical thin films are transparent at the wavelengths of interest. The proper thickness (1 / 4 wave) can also be used as an anti-reflective coating to further enhance the optical extraction efficiency.

[0042] In some embodiments, the back of the waveguide is coated with a reflective layer. Ideally, the back surface can form an approximate parabolic mirror, with the micro-LED at the focal point, and the reflective coating 417 placed to the back of the waveguide, as shown in Figure 4A These reflectors redirect the light traveling backwards to the desired forward direction.

[0043] The light detector on the other end of the waveguide can be butt-coupled to the waveguide, with the detection area covering the waveguide cross-section. Alternatively, it can be placed under or on the side of the waveguide, as all the beams impinge on the sides as well as the back of the waveguide. This is further explained in subsequent figures.

[0044] It should be noted that Figure 4A The curvature and reflective geometry in can also be applied in the vertical direction in combination with the proper shaping of the waveguide. In some embodiments, an approximate parabolic reflector is formed in the top and even the bottom surface as well as the sides of the waveguide. The encapsulant can also be optimally shaped for maximum extraction efficiency. Ideally, the micro-LED is spherical in shape, but in addition to that, the encapsulant or the properly shaped reflector(s) and waveguide can optimize the coupling into the waveguide mode. Figure 6 It is shown how the spherical reflective geometry can be used in the vertical direction. In Figure 6 In, the micro-LED 611 is encapsulated by an encapsulant 613. The micro-LED and the encapsulant are within and near the first end of the waveguide 615, which can be a polymer waveguide. The first end of the waveguide has a spherical geometry in at least the vertical direction (with the waveguide length from the first end to the second end in the horizontal direction), with a metallized mirror 617 formed on the first end of the waveguide. The light detector 619 is located at the second end of the waveguide, where in Figure 6 In, the light detector is butt-coupled to the second end of the waveguide. The waveguide is on a Si02 layer 659 on top of a silicon substrate 661. The Si02 layer acts as a lower cladding layer for the waveguide. Metal traces 652 on the Si02 layer electrically connect the micro-LED to a micro-LED driver 612 also on the Si02 layer. Similarly, metal traces 621 connect the light detector to a transimpedance amplifier 622 also on the Si02 layer.

[0045] Figure 7AA 3D view of a half-duplex micro-LED based transmitter and receiver connected to solder balls. In Figure 7A In the middle, the optical waveguide 715 extends over a portion of the RX / TX chip 771 and the SiO2 layer 759 on the silicon wafer 761 on which the RX / TX chip is located. The RX / TX chip is electrically connected to the solder balls by first wires 746, where the RX / TX chip is also connected to a positive voltage line 742 and a ground line 744, all on the SiO2 layer. In some aspects, embodiments in accordance with Figure 7A The large substrate in the figure can be a silicon or glass wafer, coated with a thick layer of silicon dioxide. In some embodiments, the silicon dioxide can serve as both a lower cladding for the waveguide and as an electrical insulator. The solder balls, which can be similar to balls connected to logic or memory ICs, are electrically connected to RX / TX micro-chips with RX and TX functionality. This micro-chip can be transferred onto the substrate, or even fabricated in the substrate itself. One or more photolithography steps connect this micro-chip to the power lines, ground, and connect the signal lines to the solder balls. There can be other electrical connections to the chip, for example, a clock or signal to equalize or enhance the transmitter or receiver signal. The optical waveguide is formed on the substrate and coupled to the micro-chip.

[0046] Figure 7B is Figure 7A part of the device of Figure 4A and 4BThere is a micro-LED 711 mounted on the microchip, which has back and top metallization layers to emit light in the forward (along the waveguide) direction. This micro-LED is electrically connected to a driver circuit 712 on the microchip. This driver can be a very simple transistor that changes the signal level from the source into the appropriate voltage or current to drive the micro-LED. It can also have other functions, such as the way to set the bias and modulate the voltage, the ability to enhance higher frequencies, or drive the reverse bias to sweep out the carrier to enhance the speed. There is also a photodetector 719 on the microchip to receive the optical signal in the waveguide. There are various geometrical structures of photodetectors, such as vertical p-i-n detectors, metal Schottky diode detectors, or lateral metal-semiconductor-metal detectors. In this case, Figure 7 shows the interleaved fingers of n+ and p- regions formed in a low-doped semiconductor to form a lateral p-i-n region. The two electrical connections from the detector are biased, and the photocurrent is measured by a transimpedance amplifier 722 formed in the microchip. Since the detector is integrated with the transimpedance amplifier, the capacitance in the connection between the two is very small, and low power consumption and high speed performance can be obtained. There is also some logic on the microchip, which in some embodiments can be as simple as a diode that determines whether the microchip is acting as a transmitter that activates the micro-LED or as a receiver that takes the signal from the photodetector.

[0047] In Figure 7A In embodiments, the waveguide is significantly larger than the micro-LED, and the entire micro-LED is contained within the waveguide. As a transmitter, the micro-LED sends data along the waveguide. The micro-LED is followed by a detector on the bottom side. As a receiver, most of the light is passed by the micro-LED, reflected on different surfaces of the waveguide, and eventually absorbed in the lower layer between the interleaved fingers and generates photocurrent. There is some loss since part of the incident light is blocked by the micro-LED, but a proper bit error rate can be obtained if enough light is received to achieve the desired signal-to-noise ratio.

[0048] There are many variations on this. For example, some embodiments have two separate waveguides, one for transmitting and one for receiving the signal. The micro-LED will be connected to one waveguide, and the detector will be connected to the other waveguide.

[0049] The micro-LED itself can use a detector, where a reverse bias is applied to the diode to generate photocurrent, or other photodetector structures can be implemented.

[0050] Figure 8 Description Figure 7Amultiple optical interconnects. The logic or memory chip 811 will typically have multiple rows of micro bumps or solder balls 813 connected to the silicon interposer below. In this case, each signal is routed to a Tx / Rx chip 815 and optical waveguide 817. The waveguides from all of these bumps transport the signal from one area to another.

[0051] Diverse micro-LED based optical interconnects

[0052] Once the light is in the waveguide, many optical functions can be performed optically rather than electrically, and this can have many advantages. For example, in the electrical domain, the function of a splitter is often difficult to implement because of impedance discontinuities, splitting ratios, time delays, and crosstalk to other lines can be problematic. Micro-LEDs can be coupled to a waveguide, followed by an optical splitter for connection to multiple detectors. These splitters can use star couplers or other structures. Figure 9 A micro-LED based TX coupled to a waveguide is shown with splitters to multiple RXs. In Figure 9 In the micro-LED based TX coupled to a waveguide, electrical connections 913 are coupled to micro-LED drivers 912. The micro-LED drivers drive micro-LEDs 911 in the waveguide 915. A small fraction of the optical power can be split off from the main waveguide by an optical tap and sent to an Rx. Multiple taps / RXs can be cascaded to implement a tapped bus. In Figure 9 In the micro-LED based TX coupled to a waveguide, multiple taps are shown, where each tap terminates in a photodetector 919a-d, coupled to a transimpedance amplifier 922a-d, in turn coupled to an electrical connection 924a-d. Alternatively, a single 1-to-N optical splitter can be used to connect one TX to multiple RXs. In some embodiments, multiple RXs are used for multiple logic units, for example as can be found in AI / machine learning / neural network applications. In some embodiments, multiple RXs are used for multiple multiplication units, for example for matrix multiplication, where each number can be multiplied by N other numbers and the results summed. By having splitters in the optical domain, the same signal can be sent precisely to multiple receivers. This can also work in the analog domain, where optical signals are divided, subtracted, or conversely, two signals can be added. Optical splitters can be very precise and split into tens, hundreds, or thousands of waveguides, maintaining signal integrity by splitting.

[0053] Although Figure 9 A micro-LED at the end of one waveguide and a detector in the other location is shown to implement a simplex 1-to-N connectivity, but Figure 9Links in the above can operate in duplex mode (full or half) by placing a micro-LED and detector at the end of each waveguide. Since the waveguide is typically multi-mode, a portion of the waveguide can be connected to the source and another portion can be connected to the receiver. For memory access and some other applications, the waveguide can operate in half duplex, where only one source is transmitting at a given time.

[0054] Figure 10 An implementation is shown of three nodes (each node having a TX and RX) 1011a-c, all connected in a star topology by waveguides 1015, so any node can transmit to the other nodes and any node can receive from any other node. Each node is coupled to a micro-bump 1013 for passing electrical signals to and from the chip. Each node includes a micro-LED 1011 and a photodetector 1019, with a driver 1012 for driving the micro-LED based on a signal received via the micro-bump, and a transimpedance amplifier 1022 for amplifying the signal from the photodetector. When operating in half duplex mode, each micro-bump provides an input electrical signal to the TX or obtains an electrical output signal from the RX, depending on whether the node is transmitting or receiving. The TIA and driver can be powered using voltages available on the interposer. In this architecture, there are some losses associated with the splitter, but a micro-LED powerful enough can provide enough optical power to overcome the additional losses in the link, and achieve a useful received signal-to-noise ratio (SNR) and / or bit error rate (BER). While Figure 10 Only three nodes are shown, but the architecture can be scaled to a larger number of nodes, primarily constrained by the ability to obtain a proper SNR / BER at each RX.

[0055] Repeater in micro-LED based optical interconnect

[0056] When the splitter becomes large, or the number of connections becomes too many, one or more repeaters of some type can be used to maintain sufficient signal amplitude and SNR. One application is the connection between logic to multiple memory modules. In a typical computer architecture, the processor data lines are shared between multiple peripherals and memory, and chip select lines are used to turn on various modules. For dense memory such as high bandwidth memory (HBM), this can not be preferred, as electrical links are limited to short lengths (typically <10 mm), and splitting one signal electrically to multiple modules can result in too much signal degradation.

[0057] Some embodiments provide an optical method with a micro-LED based optical-to-electrical-to-optical (OEO) repeater, such as substantially in Figure 11 is described in Figure 11In this circuit, logic chip 111 provides signals to the RX / TX chip via solder balls 1153 to drive light source 1161a. The light source can be a miniature LED. The light source is coupled to waveguide 1115a and terminates at the first module 1113a. Figure 11 The photodetector 1119a is located at the HBM stack. The signal can be provided to the HBM stack via solder balls 1153b. At the same termination point, the signal is regenerated by the RX / TX chip 1150a to another microLED 1161b coupled to the second waveguide 1115b, and reaches the photodetector 1119b of the RX / TX chip 1150b coupled to the second HBM stack 1113. The signal can be provided to the second HBM stack via solder balls 1153c. In this way, multiple memories or other types of modules can be cascaded. Figure 11 In the image, the RX / TX chip, micro LED, waveguide, and photodetector are displayed on a SiO2 layer on top of a silicon substrate.

[0058] Figure 12 Demonstrating interconnects based on bidirectional repeaters (e.g., Figure 11 A top view of an example of interconnecting components. Figure 12 In this configuration, each of the logic chip 1211, the first HBM stack 1213a, and the second HBM stack 1213b has a repeater, wherein the repeaters for the logic chip and the first HBM stack are coupled by waveguides 1215a and 1215b, and the repeaters for the first HBM stack and the second HBM stack are coupled by waveguides 1215c and 1215d. For example, a microLED 1261a is provided at the logic chip end of waveguide 1215b, and a photodetector 1219a is provided at the first HBM stack end of waveguide 1215b, while a microLED 1261b is provided at the first HBM stack end of waveguide 1215a, and a photodetector 1219b is provided at the logic chip end of waveguide 1215a. The bidirectional interconnect uses two repeaters, each of which has an input waveguide optically coupled to an RX, which is electrically connected to a microLED-based TX, and the microLED-based TX is optically coupled to an output waveguide. Figure 12 This demonstrates the use of a separate waveguide in each direction of a bidirectional link. However, the connection can also be implemented using only one waveguide for both directions of the link.

[0059] Figure 13An example of the internal logic of a display repeater. As data moves from left to right, the leftmost receiver 1319c and TIA are active, and the leftmost light emitter 1317a (micro-LED) and driver are off, and the rightmost receiver 1319b and rightmost emitter 1317b are the reverse. When light moves from right to left, the logic inverts, as shown in the table. If there is some additional logic 1313 incorporated in the receiver, then the signal from either direction can simply be repeated, the incoming signal ignored and terminated and a new signal generated, or simply blocked. This becomes equivalent to a more powerful version of "tri-state" logic, where the chip can control the bus, disconnect from the bus, or even just terminate the bus.

[0060] Electronic integration

[0061] In some embodiments, the drivers and receivers are monolithically integrated into an array on an IC. Figure 14 A semi-schematic diagram showing an array implementation between an FPGA 1411 and an HBM stack 1419, with a single IC used on either end. Signals to and from the FPGA are provided on a bus to a first TX / RX ASIC 1413, which drives the LEDs and processes signals from the photodetectors. Similarly, signals to and from the HBM stack are provided on a bus to a second TX / RX ASIC 1417, which drives the LEDs and processes signals from the photodetectors. A plurality of waveguides 1415 couple the corresponding LEDs and photodetectors associated with the ASICs 1413, 1417. An advantage of an array on an IC is that timing signals can be shared with other control signals, such as control signals for bias, clock, pre-emphasis, and equalization.

[0062] The drivers can also be monolithically integrated with the micro-LEDs themselves. As Figure 15 A combined device of a micro-LED with an enhancement mode FET is illustrated. The combination device is shown schematically on the left, and in cross-section on the right. A small voltage on the gate creates a connection in the accumulation region, and connects the source to the drain. This drives current through the micro-LED, and turns on the micro-LED.

[0063] There are multiple ways to integrate the FET with the micro-LED. One implementation is shown in Figure 15 The micro-LED includes an n-GaN layer 1512 and a p-GaN layer 1510, which sandwich an intrinsic InGaN region that will include quantum wells 1514. The FET is integrated on the p-GaN layer 1510, and includes a gate 1518, a source 1516, and a drain 1512. The FET is an enhancement mode FET, and the gate 1518 is connected to the p-GaN layer 1510. The source 1516 is connected to the n-GaN layer 1512, and the drain 1512 is connected to the p-GaN layer 1510. The FET is connected to the micro-LED, and the p-GaN layer 1510 is connected to the n-GaN layer 1512. Figure 15In this design, the LED is inverted, with a p-GaN layer on a silicon substrate 1530. A 2D accumulation region is created in the p-type GaN by depositing AlGaN on top. Some of this material is removed for the gate 1522. The AlGaN has the effect of creating an n-type accumulation region underneath, which is connected to the source 1526 and drain 1524. A positive voltage on the gate forms a connection between the two accumulation regions, thereby grounding the cathode of the microLED and turning it on.

[0064] Figure 16 This demonstrates another example of a detector integrated with an amplifier. Figure 16 In this embodiment, a silicon microchip 1619 is situated on a silicon dioxide layer 1613 of a silicon substrate 1611. An optical waveguide extends over at least a portion of the silicon substrate and the silicon microchip, said portion serving as a photodetector. Figure 16 The description states that a lightly doped silicon layer is formed in the microchip 1619, with one portion of the microchip acting as a detector 1617 and another portion acting as a FET 1621 that amplifies the photocurrent. Many possible circuit configurations exist that can be monolithically integrated.

[0065] Although the invention has been discussed with respect to various embodiments, it should be recognized that the invention includes novel and non-obvious claims supported by this disclosure.

Claims

1. A chip-to-chip optical interconnect comprising micro-LEDs, comprising: a first semiconductor chip electrically coupled to an interposer; a second semiconductor chip electrically coupled to the interposer; the interposer comprising electrical signal paths electrically coupling the first semiconductor chip and the second semiconductor chip; a first micro-LED; first circuitry electrically coupled to the first semiconductor chip for driving the first micro-LED based on data from the first semiconductor chip; a first encapsulant substantially encapsulating the first micro-LED; a first photodetector; first amplification circuitry for amplifying a signal from the first photodetector, the first amplification circuitry electrically coupled to provide an electrical signal to the second semiconductor chip; and a first optical waveguide made of a polymer material and optically coupling the first micro-LED and the first photodetector, the encapsulated first micro-LED being within a cavity of the polymer material of the first optical waveguide; wherein the first optical waveguide comprises a back surface that is optically in an opposite direction from the first photodetector relative to a location of the first micro-LED, the back surface comprising a reflector to reflect light toward the first photodetector, and at least some of the polymer material of the first optical waveguide is located between the first micro-LED and the back surface of the first optical waveguide.

2. The chip-to-chip optical interconnect of claim 1, wherein the first optical waveguide is on a surface of the interposer.

3. The chip-to-chip optical interconnect of claim 1, further comprising: a second micro-LED; second circuitry electrically coupled to the second semiconductor chip for driving the second micro-LED based on data from the second semiconductor chip; a second encapsulant substantially encapsulating the second micro-LED; a second photodetector; second amplification circuitry for amplifying a signal from the second photodetector, the second amplification circuitry electrically coupled to provide an electrical signal to the first semiconductor chip; the first optical waveguide optically coupling the second micro-LED and the second photodetector, the encapsulated second micro-LED being within a second cavity of the polymer material of the first optical waveguide.

4. The chip-to-chip optical interconnect of claim 3, wherein the first micro-LED is positioned in an optical path from the second micro-LED to the second photodetector, and the second micro-LED is positioned in an optical path from the first micro-LED to the first photodetector.

5. The chip-to-chip optical interconnect of claim 1, wherein the first semiconductor chip comprises a processor, and the second semiconductor chip is a memory chip.

6. The chip-to-chip optical interconnect of claim 1, further comprising: a second photodetector; and a second optical waveguide made of a polymer material and optically coupling the second micro-LED and the second photodetector, the encapsulated second micro-LED being within a cavity of the polymer material of the second optical waveguide. ​ second amplification circuitry for amplifying a signal from the second photodetector, the second amplification circuitry electrically coupled to provide an electrical signal to a third semiconductor chip; and wherein the first optical waveguide optically couples the first micro-LED and the second photodetector.

7. The chip-to-chip optical interconnect of claim 1, further comprising: a second micro-LED; second circuitry electrically coupled to the first amplification circuitry for driving the second micro-LED based on data from the first photodetector; and a second optical waveguide optically coupling the second micro-LED and a third photodetector.

8. The chip-to-chip optical interconnect of claim 1, wherein the first circuitry is monolithically integrated with the first micro-LED.

9. The chip-to-chip optical interconnect of claim 1, wherein the reflector of the back surface is parabolic, and the first micro-LED is located at a focal point of the parabolic reflector.

Images