High-speed and multi-contact light-emitting diodes for data communications
By introducing base-emitter junction and MOS structure in micro-LEDs and optimizing carrier injection and sweeping, the signal delay and modulation speed limitations in chip-to-chip communication are solved, and efficient short-distance data transmission is achieved.
Patent Information
- Application Number
- CN202080088700.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-02-07
- Filing Date
- 2020-11-18
- Publication Date
- 2025-10-03
- Estimated Expiration
- 2040-11-18
AI Technical Summary
Existing lasers have difficulties in very short-distance optical communications, especially in chip-to-chip communication, and traditional micro-LEDs are limited by carrier lifetime and capacitance at high-speed modulation.
A micro-LED structure with additional contacts is adopted, including base-emitter junction and MOS structure, and carrier injection and sweeping are controlled by base-collector voltage, optimizing carrier recombination time to achieve high-speed modulation.
It achieves efficient data transmission in short-distance communication, reduces signal delay, improves modulation speed, reduces power consumption, and maintains high quantum efficiency.
Smart Images

Figure CN114868262B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates generally to light emitting diodes (LEDs), and more particularly to high-speed LEDs for data communications. Background Art
[0002] Lasers have become dominant in optical communications due to their narrow linewidth, single spatial mode output, and high speed. The narrow linewidth of lasers means that high-speed signals can travel long distances through dispersive media without pulse broadening. Long-haul fiber links are often dispersion-limited, and narrow-linewidth lasers are therefore essential for these links. The single spatial mode of lasers is also relatively easy to couple into single-mode fibers.
[0003] Stimulated emission of laser light can also allow for high modulation speeds. Directly modulated optical links can easily operate at 25 Gb / s and potentially carry 50 Gb / s of information using PAM4 modulation.
[0004] However, the use of lasers can present difficulties for very short-distance optical communications, such as chip-to-chip communications. Summary of the Invention
[0005] In some embodiments, an optical communication system for transmitting information provided by a processor to another area of the processor or another module in a multi-chip module includes: an LED associated with the processor; an LED driver for activating the LED to generate light based on data provided from the processor to the LED driver; a detector for performing photoelectric conversion using the light; and an optical waveguide for optically coupling the LED and the detector; wherein the LED includes a bipolar junction transistor (BJT) having a base including a quantum well. In some embodiments, the emitter of the BJT includes AlGaN.
[0006] In some embodiments, an optical communication system for transmitting information provided by a processor to another area of the processor or another module in a multi-chip module, the optical communication system comprising: an LED associated with the processor; an LED driver for modulating the output optical power of the LED so that the LED will generate light based on data provided from the processor to the LED driver; a detector for performing photoelectric conversion using the light, the detector having, for example, an electrical output modulated by the optical power incident on the detector; and an optical waveguide optically coupling the LED with the detector; wherein the LED comprises: a pn junction having a metal oxide semiconductor (MOS) structure.
[0007] These and other aspects of the invention will be more fully understood after a review of this disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0008] Figure 1A showing an electrically controllable LED manufactured with additional contacts according to aspects of the invention, Figure 1B a band diagram showing the conduction state of an LED, and Figure 1C A strip diagram showing the off state of an LED.
[0009] Figure 2 An electrically controllable LED fabricated with additional contacts and an AlGaN emitter is shown according to aspects of the invention.
[0010] Figure 3 An LED structure according to aspects of the invention is shown with an additional contact combining a pn junction with a MOS structure that can sweep out minority carriers but collect them in an accumulation region.
[0011] Figure 4 An LED structure with additional contacts including a lateral accumulation region according to aspects of the invention is shown.
[0012] Figure 5 Examples using LEDs variously discussed herein are presented according to aspects of the present invention. DETAILED DESCRIPTION
[0013] The properties of lasers may be less important for very short distance optical communications (e.g., chip-to-chip communications). In some embodiments, micro-LEDs (especially structures optimized for high modulation speeds) are used to couple light into waveguides, e.g., as described with respect to Figure 5 As discussed. In some embodiments, microLEDs are used to provide highly parallel communication between chips, for example, on an interposer or through 3D optical structures, such as those that include optical waveguides and / or free-space optical propagation using optical elements such as lenses and holograms. GaN-based microLEDs have been developed for display applications, and a packaging ecosystem has been developed for mounting such devices on silicon CMOS or polysilicon-on-glass backplanes. With relatively minor modifications, components of this packaging ecosystem can be used to interconnect ICs for chip-to-chip communication.
[0014] In some embodiments, microLEDs differ from semiconductor lasers (SLs) in that: (1) microLEDs do not have an optical resonator structure; (2) the optical output from a microLED is almost entirely spontaneous emission, while the output from an SL is primarily stimulated emission; (3) the optical output from a microLED is temporally and spatially incoherent, while the optical output from an SL is significantly temporally and spatially coherent; (4) microLEDs are designed to drive a minimum current as low as zero, while SLs are designed to drive a minimum threshold current, which is typically at least 1 mA. In some embodiments, microLEDs differ from standard LEDs in that: (1) they have an emitting area less than 100 microns by 100 microns (less than 100um x 100um in some embodiments); (2) they 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; and (3) they are typically used in large arrays for display and interconnect applications.
[0015] In some embodiments, for chip-to-chip communication, the distance is so short that dispersion is not necessarily an issue. A simple calculation indicates that for a GaN LED with a central wavelength in the range of 400nm to 450nm and a spectral width of 20nm, if the LED is modulated at 4Gb / s and propagated through a doped SiO2 waveguide or fiber, the waveguide or fiber can be up to 5 meters long with less than 2dB of dispersion power loss. Since chip-to-chip communication distances within a multi-chip module (MCM) or across a PC board are typically less than tens of centimeters, the LED's wider spectral width may not be an issue. Furthermore, one can even use highly multimode waveguides into which the output light from the LED is relatively easy to couple. Due to the short distance, modal dispersion of the multimode waveguide may also not be an issue. At a signal rate of 4Gb / s, even in a waveguide with a 10% core-cladding refractive index step and an NA of 0.67, the waveguide length can be as long as 85cm with minimal dispersion power loss; smaller core-cladding refractive index steps generally provide longer reach. Therefore, in many embodiments, a broad spectrum LED and a multimode waveguide are sufficient for chip-to-chip communication.
[0016] Furthermore, in various embodiments, the microLEDs are manufactured in very small sizes, with device sizes of less than 2um. This small mode has very high intrinsic brightness (i.e., low elongation) and can typically be easily coupled to a multimode waveguide. Although the output is typically Lambertian, with appropriate use of reflectors, in some embodiments microlenses, and in some embodiments embedding the microLEDs in the waveguide, the coupling efficiency can be 30% or higher. MicroLEDs typically have high quantum efficiencies similar to or even exceeding those of lasers. Since at short distances, not much loss is suffered even at blue or green wavelengths, not much emission power is required, and in some embodiments, a small microLED running at around 10uA may be sufficient.
[0017] In general, microLEDs are limited by carrier lifetime (and, if they are too large, by capacitance) and generally cannot achieve the modulation speeds of high-speed lasers. However, clock speeds in microprocessors and logic appear to be limited to a few Gb / s, and data in and out of ICs is typically accelerated using serializers / deserializers (SERDES) to produce a smaller number of high-speed channels. For example, commercially available switch ICs currently operate at clock speeds of several GHz, but communicate with 256 or 512 channels at 50 Gb / s or 100 Gb / s per channel, with each channel having an associated SERDES. These SERDES consume a significant amount of electrical power and could be eliminated if the switch IC instead used more low-speed channels. Optics allow for this parallelism and achieve higher throughput by having a greater number of channels, even at slower channel speeds. However, LEDs operating at the highest possible modulation speed may be preferred.
[0018] Furthermore, a significant advantage of GaN micro-LEDs over lasers is their lack of a significant threshold current. While quantum efficiency is a function of drive current, micro-LEDs do not have a discrete threshold level and can operate at currents far lower than lasers. Given their usefulness in displays, a substantial infrastructure exists for mounting, connecting, and testing micro-LEDs on a variety of substrates. Furthermore, GaN micro-LEDs generally offer high-temperature performance and reliability far superior to semiconductor lasers.
[0019] Typically, GaN micro-LEDs optimized for display applications include a cylindrical or quasi-cylindrical structure with a pin doping profile. The LED is turned on by forward biasing the diode and injecting electrons from the n-region and holes from the p-region into the middle intrinsic region containing the InGaN quantum well. The p-contact is on one side of the structure, and the n-contact is on the other side. In many applications, the device is mounted on a chip, with the "bottom" side making electrical contact with the chip and the "top" side making contact with a common lead (such as a ground or power lead). The top-side contact can be a transparent conductor such as indium tin oxide (ITO). In micro-LEDs, vertical structures are generally preferred, but there are also lateral structures, or n-contacts adjacent to the p-contacts. In any case, this structure is not optimized for speed, simply because displays typically run at 60Hz or 120Hz frame rates, not at Gb / s.
[0020] Changes can be made to optimize the structure for speed. Generally speaking, micro-LEDs are limited by the LED's capacitance and carrier recombination time (or diffusion capacitance). This capacitance forms an RC circuit with the driver output impedance, causing roll-off at high frequencies. Carrier lifetime means that it takes time for the LED to turn off, because even after the electrical pulse ends, we must wait for the few injected carriers to recombine and stop emitting light. Due to the small size of micro-LEDs, their capacitance (typically only a few femtofarads) does not significantly limit the device's modulation speed; instead, modulation speed is generally limited by carrier lifetime.
[0021] In general, the speed of a micro-LED increases with the injected current density. Carriers can recombine in an LED in three ways. At low current levels, recombination is mediated by traps (SRH recombination). At higher current densities, these traps become saturated and the quantum efficiency of the LED improves because radiative recombination dominates. As the carrier density increases, this radiative recombination rate accelerates, thereby increasing the radiative efficiency and reducing the carrier lifetime. Therefore, the harder the micro-LED is to drive (for example, the higher the current density), the faster it operates. At higher current densities, nonlinear non-radiative mechanisms such as Auger recombination further reduce the carrier lifetime, but these non-radiative mechanisms also reduce the radiative quantum efficiency. For small-diameter fast micro-LEDs driven with high current densities, the traps are relatively unimportant because they are saturated, and the relative importance of the nonlinear non-radiative recombination rate to the radiative recombination rate determines the quantum efficiency.
[0022] Structural changes can be made to the microLEDs to improve speed. In some embodiments, the electrically controllable LEDs can be fabricated with additional contacts. One possible configuration is to have light emission occur in the base of an LED in the general form of a bipolar junction transistor, having base, emitter, and collector regions, each with associated electrical contacts. The base-emitter junction can be forward biased, where the emitter injects carriers into the base and recombine, causing light to be emitted from the base region. For light emission, the transistor can be biased into saturation, with the collector unbiased. To quickly turn off the LED, the collector is reverse biased, sweeping minority carriers out of the base. Figure 1A Show this structure and Figure 1B and 1C The band diagrams for the on-state and the off-state are shown in FIG.
[0023] like Figure 1A As shown, in some embodiments, the LED includes an n+ GaN emitter 111 and an n- GaN collector. A p- GaN base region 113 is located between the emitter and collector. The base region includes a quantum well.
[0024] Figure 1B The band diagram is for the on state Figure 1A LEDs. Figure 1B The band diagram of shows a conduction band 131a above a valence band 133a. The bands extend across the emitter region 121, the base region 123, and the collector region 125, respectively. The band gap between the conduction band and the valence band is generally constant across each region, with the energy level generally increasing in the recombination region between the emitter and base regions and the energy level dropping slightly in the collector region. In the "on" state, electrons are injected 137a from the emitter region into the base region, where most of the electrons recombine; the remaining unrecombined electrons are swept out of the base region 139a into the collector region, as shown in FIG. Figure 1B The relative sizes are indicated by arrows.
[0025] Figure 1C The band diagram is for the "off" state Figure 1A LED. Figure 1B Like the picture with Figure 1C The band diagram of FIG shows a conduction band 131b above the valence band 133b, which extends across the emitter region 121, the base region 123, and the collector region 125. In the "off" state, electrons are injected 137b from the emitter region into the base region. However, unlike in the "on" state, in the "off" state, a large portion of the electrons are swept out of the base region 139b into the collector region, similarly as in the "on" state. Figure 1C The relative sizes are indicated by arrows.
[0026] A further enhancement would be to make the emitter region out of AlGaN to enable better injection into the base region. In some embodiments, the AlGaN acts as a barrier on the n-region to further enhance the injection of carriers into the p-doped base region and prevent hole injection into the n-type region. In 3D, the structure can be Figure 2 The form shown in the schematic cross section.
[0027] Figure 2 An n+GaN buffer and subcollector layer 213 are shown on top of substrate 211. The substrate can be silicon, GaN, or, most commonly, sapphire. An n-GaN collector layer 215 is located on top of a first portion of the subcollector layer, while a collector electrical contact 227 is located on top of a second portion of the subcollector layer. A p-base region layer 217 is located on top of the collector layer. In many embodiments, the p-base region layer includes one or more quantum wells. In operation, light can be emitted from the base region layer. An AlGaN n+ emitter layer 219 is located on top of a first portion of the base region layer, while a base electrical contact 225 is located on top of a second portion of the base layer. The AlGaN n+ emitter layer separates the base region layer from an n+GaN contact layer 221 located on top of the AlGaN n+ emitter layer. An emitter electrical contact 223 is located on top of a portion of the n+GaN contact layer.
[0028] The base-emitter junction can be forward biased by applying voltage 229, thereby injecting holes into the base. However, the electron concentration in the base is controlled by the base-collector voltage, where a negative bias sweeps out carriers. Therefore, the base-collector junction will be reverse biased, and a modulation signal 231 will be applied. This structure will be much faster than an LED because the turn-off time is no longer limited by the intrinsic carrier lifetime, but is determined by the speed at which the electric field in the base-collector removes carriers from the base region. However, this speed-up can come at the expense of reduced quantum efficiency, because carriers removed from the base region cannot recombine to produce photons.
[0029] An alternative structure is to combine a pn junction with a MOS structure that can sweep minority carriers from the base region into the accumulation region. Laterally, if fabricated on a flat surface, the structure can be Figure 3 The form shown in the schematic cross section of Figure 3 In FIG, an n-type GaN buffer layer 313 is shown on a substrate 311. A p-base region layer 315 is located on top of a first portion of the n-type GaN layer, with an electrical contact 325 located on top of a second portion of the n-type GaN layer. The n-type GaN (potentially with an AlGaN barrier layer between it and the base layer in some embodiments) injects electrons into the p-base region. These minority electrons recombine with a local population of holes in the base region to generate light.
[0030] A p+ GaN contact layer is located on top of a first portion of the p-base region layer, and a metal oxide semiconductor (MOS) structure with an electrical contact 321 provided on top by a dielectric (e.g., aluminum nitride (AlN) 317) is located on top of a second portion of the p-base region layer. The MOS structure is close to a recombination region, where a positive bias on the gate will pull the injected carriers into the accumulation region 319 under the MOS dielectric and turn off light emission. Removing this voltage or making it slightly negative will force the carriers back to the base region, where they will recombine and emit light. Therefore, the injected electrons are not removed from the system, but are "recycled" to achieve a higher quantum efficiency than a transistor structure.
[0031] Figure 3 The structure in [1] typically has an accumulation region close to the base region, since the modulation speed of the device is related to the speed at which carriers can sweep between the two regions. In fact, the response time can often depend on the separation distance between the base and accumulation regions, as well as the electric field (mobility) or saturation velocity, v sat In GaN, the saturation velocity of electrons is 1.4×10 7 cm / s, so we can achieve a modulation speed of at least 4 GHz using a pitch less than 10 μm. This is easily achieved using micro-LEDs with a diameter in the range of 2 μm.
[0032] Figure 4 An alternative MOS device geometry is shown, in which a MOS dielectric layer 425 is formed around the circumference of the microLED, potentially leading to higher speeds. In the schematic, the LED is bonded to a metal substrate 411, which provides an n-contact. An n-GaN buffer layer 313 is located on the metal substrate. An insulating layer 415, also bonded to the metal substrate, is located at least on the opposite side of the buffer layer. The insulating layer extends slightly above the top of the buffer layer, on which the p-GaN base region layer 315 is located. A p+GaN layer 320 is located on top of the base layer. A p-contact 423 (e.g., a transparent p-contact) is located on top of the p+GaN layer. MOS dielectric layers are formed around the sides of the base region layer and also around the sides of the p+GaN layer, with metallization 427 surrounding the outside of the MOS dielectric layer. An accumulation region 419 is located in the portion of the base region adjacent to the AlN. The insulating layer 415 acts as a placement stop, for example, by depositing the MOS dielectric and metallization around the n-GaN buffer layer. The lateral deposition of dielectric layers and metallization around the sides can be considered a MOS structure.
[0033] Figure 5 An example of using an LED is shown, which may be a micro-LED 515, as discussed in various ways herein. Figure 5In the embodiment of the present invention, the silicon processor 511 performs various operations on or with the data. For example, the silicon processor may perform calculations on the data, may perform switching functions, or may perform other functions. The silicon processor provides at least some of the data to the LED driver 513, where the LED driver activates the micro-LEDs to generate light having a power level modulated by at least some of the data. The generated light is provided to the optical coupler 517, which transmits the light to the optical propagation medium 519. In some embodiments, the optical propagation medium, which may be, for example, a waveguide, may be used to transfer light from one area of the silicon processor to another area of the silicon processor. In other embodiments, the optical propagation medium may be used to transfer light from the silicon processor to, for example, a multi-chip module ( Figure 5 521, which in turn passes the light to a detector 523 (e.g., a photodiode) for photoelectric conversion. The electrical signal from the detector output, which includes at least some data, can be amplified by an amplifier 525 and provided to the silicon processor (or another chip in the multi-chip module). In some embodiments, the micro-LEDs and detectors can be coupled to the waveguide individually, and / or in some embodiments, they can be coupled in parallel as an array. In addition to transferring light and data from one location to another, the optical waveguide can also split the light into two or more outputs, allowing data fanning out. The optical waveguide or medium can also perform some switching to direct the output from one receiver to another.
[0034] The micro-LEDs and detectors may be coupled to the waveguide individually, or may be coupled to the waveguide in parallel as an array.
[0035] While the invention has been described with respect to various embodiments, it will be recognized that the invention encompasses novel and non-obvious claims supported by the invention.
Claims
1. An optical communication system for transmitting information provided by a processor to another area of the processor or another module in a multi-chip module, comprising: an LED associated with the processor; an LED driver for activating the LED to produce light modulated based on data provided from the processor to the LED driver; a detector for performing photoelectric conversion using the light; and an optical waveguide optically coupling the LED and the detector; The LEDs include: a pn junction having a metal oxide semiconductor (MOS) structure; The pn junction having a MOS structure includes: n-doped GaN buffer layer; a p-base layer located on at least a portion of the n-doped GaN buffer layer; a p+GaN contact layer on the first portion of the p-base layer; a dielectric layer in contact with the second portion of the p-base layer; and A metal layer contacts the dielectric layer. 2 . The system of claim 1 , wherein the dielectric layer is located on the second portion of the p-base layer. 3 . The system of claim 1 , wherein the second portion of the p-base layer is located on at least one side of the p-base layer.
4. The system of claim 1 , further comprising: another LED associated with the other region of the processor or another chip in the multi-chip module; another LED driver for activating the another LED to produce light modulated based on data provided to the another LED driver from the another area of the processor or another chip in the multi-chip module; and another detector for performing photoelectric conversion using the light from the another LED; wherein the other LED comprises: A pn junction having a metal oxide semiconductor (MOS) structure.
Citation Information
Patent Citations
Nitride semiconductor light-emitting transistor
JP2013168507A
High Speed Communication
US20140023376A1