Ultra-compact LED projector

Abstract

A monolithic display / projector comprising a single die having a mechanically isolated array of LED pillars is disclosed. Each pillar has a height greater than its width, and the spacing between pillars is less than the height of the pillars. The die comprises an LED display portion bonded to a silicon substrate addressing portion, with each pixel having one metal contact. When projected onto a human retina, the resolution of the display is preferably about the same as the resolution of the human retina, so that the image projected onto the retina can be indistinguishable from the real world. The display can be packaged into a contact lens with focusing optics embedded into the contact lens. To electrically contact the N-type semiconductor layer, the pillars are surrounded by a reflective cathode metal grid, so that the cathode current is coupled via the vertical sides of the N-type layer. The metal grid mechanically connects the isolated LED pillars and optically isolates each LED pillar. The active layer can emit blue light, and a wavelength conversion layer can be used to generate red and green light.

Description

[0001] This application is a divisional application of the invention patent application with an international application date of October 17, 2018, which entered the Chinese national stage on April 13, 2020, with Chinese national application number 201880066775.5 and invention name “Ultra-Dense LED Projector”. Technical Field

[0002] The present disclosure relates to light emitting diode (LED) displays, and in particular to monolithic ultra-dense LED displays such as for contact lenses. Background Art

[0003] Conventional direct-emitting LED displays use discrete LEDs that emit red, green, and blue light, arranged in an addressable array of composite pixels. Because separate LED dies are used, these displays have a relatively large pixel pitch. These displays typically have a resolution of up to 500 pixels per inch (composite white pixels / inch), and the pitch from one color pixel to the next is about 25 microns.

[0004] In another approach, LEDs emitting red, green, and blue light are combined on a single die. However, the practical minimum pixel pitch achievable with conventional monolithic LED display technology is about 5-10 μm, with a pixel size of a few microns. Since the area of ​​each pixel is a few square microns, such LED displays can be referred to as microdisplays. Very small displays may require a die size of 1 mm or less. Such LED displays constructed using conventional technology are typically limited in resolution or composite white pixel count.

[0005] Therefore, better methods are needed to form ultra-dense (and therefore correspondingly higher-resolution) LED displays. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Embodiments of the present disclosure have other advantages and features that will become more apparent from the following detailed description and appended claims when taken in conjunction with the examples in the accompanying drawings, in which:

[0007] Figure 1A Shown is a top view of a front panel for a femtoprojector display, along with a zoomed-in view of the hexagonal LED array within the front panel.

[0008] Figure 1B A schematic diagram showing some of the circuitry on the backplane of a femtoprojector display.

[0009] Figure 2 yes Figure 1A A cross-sectional view of three pixels in a display section.

[0010] Figure 3 is a cross-sectional view of the layers in an LED wafer before etching the LED wafer to form semiconductor pillars / pixels.

[0011] Figure 4 Pictured Figure 3 A portion of an LED wafer after being etched to form trenches surrounding hexagonal pillars, where each pillar acts as an addressable pixel, and after a dielectric layer is formed over the surface.

[0012] Figure 5 The diagram shows the N-type layer after further trench etching. Figure 4 structure.

[0013] Figure 6 The diagram shows that the sidewalls of the P-type layer and the active layer are coated with a transparent dielectric material to prevent electrical shorting of the PN junction, and then a reflective N-metal is deposited in the trench to electrically contact the vertical sidewalls of the N-type layer in the pillar.

[0014] Figure 7A The top surface of the LED wafer is shown after planarization using CMP, where the reflective metal above the P-type layer and the N-type metal contacts are exposed and coplanar.

[0015] Figure 7B yes Figure 7A Top view of the structure.

[0016] Figure 8 The diagram illustrates the formation of a metal contact bump over the anode (P) contact.

[0017] Figure 9 The bonding of the LED die to the backplane substrate wafer is illustrated, wherein an electrical connection is made between the anode metal contact on the P-type layer in each column and a bump on the backplane substrate wafer to selectively apply an anode voltage to each LED pixel, and wherein an electrical connection is made between the N-metal contact (cathode common to all columns) and a bump on the backplane substrate wafer to apply a reference voltage to each LED pixel.

[0018] Figure 10 The bonding of the LED die to the backplane substrate wafer is illustrated again, with more details on possible dimensions and backplane substrate construction.

[0019] Figure 11A The bonded structure is illustrated after the sapphire growth substrate is removed from the LED die and after the exposed surface of the LED die is planarized to expose the N-type layer in each pillar.

[0020] Figure 11BAn alternative embodiment for Cu-Cu bonding is illustrated, where glass (SiO2) is formed between copper electrodes, and the top surfaces of the copper and SiO2 are coplanar.

[0021] Figure 12 Illustrated is the formation of a distributed Bragg reflector (DBR) and the deposition of red and green photon energy down-converter materials over a blue pump LED to form red, green and blue addressable pixels with a width of less than 1 um.

[0022] Figure 13 Figure 2 is a top view of a hexagonal metal pad on the top surface of the backplane substrate. The pad can be used for Cu-Cu bonding to the LED die, or can support a gold bump for bonding to a corresponding gold bump on the LED die.

[0023] Figure 14 The diagram illustrates the display sections and how the pixel resolution of the central portion is higher than that of the outer portions to substantially correspond to the varying resolution of the human retina.

[0024] Figure 15 is a cross-sectional view of another type of femtodisplay in which no down-conversion material is used and all peak wavelengths of the LED pixels are the same.

[0025] Figure 16 is a cross-sectional view of another type of femtodisplay in which all LEDs emit IR light and upconversion materials are used to create red, green, and blue emitting pixels.

[0026] Figures 17A to 17N Shown for manufacturing Figure 1A to Figure 1B The process of the femto display.

[0027] Figure 18 A cross-sectional view of adjacent LED columns of a front plate for a femtoprojector display is shown.

[0028] Figure 19 A cross-sectional view of an eye-mounted display incorporating a femtoprojector in a contact lens is shown. DETAILED DESCRIPTION

[0029] The drawings and the following description relate to the preferred embodiments by way of illustration only. It should be noted that from the following discussion, alternative embodiments of the structures and methods disclosed herein will readily be identified as viable alternatives that can be employed without departing from the principles of the claimed invention.

[0030] A "femtoprojector" is a small projector that projects an image from a display contained inside a contact lens onto the user's retina. The display and associated optical system are small enough to fit inside a contact lens. In order to meet this size requirement while still achieving reasonable resolution, the pixel size in the image source is typically much smaller than that of image sources for other applications. For example, conventional LED direct emission displays use discrete red, green, and blue emitting LEDs with a resolution of up to 500 pixels per inch (composite white pixels / inch) and a spacing of approximately 25um (micrometers) from one color pixel to the adjacent color pixel. In contrast, displays for femtoprojectors preferably have a spacing of less than 1um in the emission area. 2 The pixel size is 2 μm and the pixel pitch is 2 μm or less, so that when the displayed image is projected onto the retina, the image resolution can also be commensurate with the density of photoreceptors in the retina.

[0031] Ultra-dense LED arrays suitable for monolithic "femto-LED displays" (or femto-projectors) are disclosed, preferably with a resolution that substantially matches or exceeds that of the human retina. The design preferably allows for pixel sizes less than 1 μm in an emitting area with a pixel pitch less than 2 μm. 2 , so that when the displayed image is projected onto the retina, the retinal image has a resolution similar to that of the real image being enhanced or replaced. In one embodiment, the spacing between pixels is less than 0.2 μm, and the preferred pixel diameter is about 0.5 μm. If each pixel is square, its area is about 250x10 -15 m 2 (or 250 square femtometers).

[0032] In one approach, a higher density of LED pixels can be created by forming a single LED die and etching the die to form a dense array of individually addressable LED pixels. Such LED pixel segments can be as small as a few microns or less. If the LED die is GaN-based and the LED emits blue light, segments of red and green pixels can be formed by depositing photon energy down-conversion phosphors on top of the blue LEDs. Alternatively, if the LED die emits IR, red, green, and blue pixels can be created by depositing energy up-conversion materials on top of the IR LEDs.

[0033] One possible use for this monolithic, ultra-compact LED display is to embed it into a contact lens, so that the displayed image overlays the wearer's view of the real world. The LED display can be less than 1 mm wide, making it invisible to objects viewing the wearer and substantially blocking light from the real world that enters the wearer's pupil. The display projects directly onto the wearer's retina.

[0034] Another possible use for monolithic ultra-dense LED displays is in eye protection (e.g., glasses or goggles) to create an immersive visual experience or an image overlaid on the wearer's real-world view (e.g., augmented, mixed, or artificial reality applications). The LEDs in such an ultra-dense display can be distinguished from conventional LEDs by their aspect ratio of height (or thickness) to width. Conventional LEDs are flat, with an aspect ratio (height:width) ranging from less than 1:100 to 1:2. The ultra-dense LEDs disclosed herein are taller than they are wide, with an aspect ratio (height:width) greater than 1:1, and preferably greater than 5:1. These LEDs can be visualized as an array of LEDs in the shape of a tightly packed "soda can," with aspect ratios ranging from similar to a 12-ounce soda can (2:1) to similar to 5 soda cans stacked on top of each other (10:1).

[0035] In one embodiment, one surface of the LED pixel (opposite the growth substrate) is a P-type layer, and the growth substrate is removed to expose the N-type layer underneath. Each pixel is formed into a thin column of semiconductor material. The top and bottom surfaces of the column are less than 1 μm apart. 2 The small area and small spacing between adjacent LEDs (e.g., less than 0.5um) make it difficult to efficiently electrically contact the N-type layer "below" the P-type layer and conduct current. A preferred embodiment provides a filler between the pillars. In one embodiment, the filler is a reflective N-metal (e.g., Al, Ag, Ni / Ag, Ti / Al) that vertically surrounds and fills the space between the LED pillars to electrically contact the N-type layer along its vertical sidewalls. As a result, there is a relatively large cathode contact area to achieve low resistance and good current handling capability. The sidewall contact is also used to provide mechanical support for individual LED pixels and optically isolate them. The reflective N-metal surrounding each LED pixel and any reflective metal covering the N-metal prevents optical crosstalk between pixels and improves efficiency. The reflective metal performs the combined functions of electrically conducting current, optically isolating pixels, reflecting backlight into the pixels, and mechanically supporting pixels. In alternative embodiments, the filler between the pillars may include other metals or materials in addition to the reflective N-metal.

[0036] In embodiments utilizing photon energy down-conversion, in one approach, for red and green pixels, a blue emitting LED is covered with a wavelength converting material (eg, phosphor or quantum dots) to convert from blue to the desired pixel color.

[0037] In embodiments where photon energy upconversion is desired, in one approach, the IR LEDs are covered with a photon energy upconversion material to convert from IR wavelengths to the appropriate pixel color of red, green, or blue.

[0038] Reflective aluminum can surround the color-converting portion of the pixel, where the aluminum contacts the top surface of the N-metal. The N-metal and aluminum form a grid around the pixel. The optical isolation between pixels is useful for maintaining high-resolution color images, while the high aluminum reflectivity maintains high efficiency.

[0039] Preferably, for use in imaging directly on the retina (eg, imaging using a contact lens display), the pixels may form a honeycomb pattern (ie, a hexagonal array).

[0040] A distributed Bragg reflector (DBR) may be formed on top of the display to enhance directivity.

[0041] The resolution of the retina varies substantially from a peak at the fovea to lower resolutions toward the periphery. This presents a significant challenge for the aspirations of contact lens-based retinal projection displays, where the resolution and shape of the LED pixels vary depending on location. The "soda can" LED array design described in this disclosure is advantageous because the display is monolithic, with a pixel size and shape that can be photolithographically defined to match the variable resolution of the human eye. By varying the pixel resolution to match the varying resolution of the retina, the die area can be used most efficiently by not providing a higher resolution than the human eye can perceive.

[0042] To address the pixels in a monolithic LED display, the LED display is mounted on a backplane substrate that contains addressing circuitry.

[0043] After forming reflective P-metal contacts on the P-type surfaces of the pillars, and after filling the spaces between the pillars with N-metal to contact the N-type layers of the pillars, the top surface of the LED wafer is flattened, such as using chemical mechanical polishing (CMP), to form planar metal anode and cathode contacts. The LED wafer is then singulated, and the die is bonded to a backplane substrate wafer containing the addressing circuitry. The anode and cathode contacts of the LED die are on the same plane, and the array of metal contacts on the backplane substrate wafer is also planar to achieve reliable bonding. The diameter of the LED die can be less than 1 mm. Copper-copper bonding is preferably used to bond the LED die to the backplane substrate wafer.

[0044] The growth substrate is then removed from the bonded LED die, exposing the light emitting surface of the pixel.

[0045] In one embodiment, the color conversion layer is formed as a separate die layer and individually bonded over the light emitting surface of the LED die.The backplane substrate is then singulated.

[0046] The wavelength conversion layer is preferably thin but blocks blue pump light so as not to sacrifice the purity of the color emitted from the conversion layer. One technique is to form a polymer matrix that deterministically arranges the quantum dot crystals to achieve good conversion and reduce absorption of the converted light. Equally important for the fabrication of the color conversion layer is the use of a reflective metal (e.g., aluminum) to separate the individual pixels so that photons converted in the layer are reflected back from the sides and emitted only from the top. Thus, photons from, for example, a green pixel are not emitted laterally into the red pixel region.

[0047] A monolithic display can be embedded in a contact lens or as part of glasses or goggles that project images onto the wearer's retina. The display can also be added to a smartphone to provide projector functionality.

[0048] On the other hand, gallium nitride (GaN) is a common material system used to manufacture LEDs. In a typical manufacturing process, a GaN layer is deposited on a substrate such as sapphire. From the substrate upward, the GaN region can include a buffer region, an n-doped region, an active region (e.g., InGaN quantum wells), and a p-doped region. Because the first few microns of GaN directly adjacent to the substrate are typically of poor quality, the distance from the substrate to the active region (i.e., the buffer region and the n-doped region) can be particularly thick (e.g., 5 μm or more). This is sometimes referred to as the buffer region. Therefore, the GaN region is grown thicker so that the material farther from the substrate has the desired quality. If the pixel area is small (e.g., in a femto projector), the resulting LED structure can be tall and thin (e.g., a 5 μm tall column with a 1 μm wide active region). This is undesirable because the light generated by the active region reflects many times from the sidewalls before leaving the LED column, resulting in higher optical losses. Structures with high aspect ratios can also cause manufacturing difficulties. Deeper, narrow trenches are more difficult to etch and filling the trenches with metal and / or other materials is also more difficult.

[0049] As a result, reducing the thickness of this GaN region is beneficial. In one approach, a wafer with a GaN region on a substrate is bonded to a backplane wafer containing the LED driver circuitry. The substrate is separated from the GaN region, exposing the GaN buffer layer, which is first made thinner by chemical mechanical polishing and then by etching. Thinning removes the buffer layer and reduces the height of the LED column, which in turn reduces light loss. Furthermore, if trenches are etched to separate the GaN region into individual LED columns, thinning the GaN region reduces the aspect ratio of the trench, which makes the etching process easier.

[0050] Other embodiments are described.

[0051] In many embodiments, a femtoprojector display includes a front plate and a back plate. Figure 1Ais a diagram of the front panel containing the LED array. Figure 1B is a schematic diagram of a backplane containing addressing and driver circuitry.

[0052] Figure 1A A top view of a front plate 10 for a femtoprojector display is shown, along with a magnified view of the hexagonal LED array within the front plate. In one application, an ultra-dense LED display using this front plate 10 can be used as a projector in a contact lens to project an image directly onto the wearer's retina. The LED array within this front plate 10 is shown as having a hexagonal shape, but other shapes are possible. The die containing the front plate can have a rectangular shape. The dimensions described below are also examples.

[0053] The front plate 10 includes a central pixel region 12, a dead space region 14, and an n-annular region 16. Region 16 is a terminal region for electrically connecting one contact of all LEDs in the array to a common electrical contact on the back plate. The diameter of the front plate assembly shown can be approximately 0.7 mm, and the diameter of the pixel region 12 can be approximately 0.5 mm. The width of each pixel is less than 2 μm, and preferably is approximately 0.6 μm. In one embodiment, the display contains more than 400,000 pixels, with variable pixel sizes ranging from a minimum of 0.6 μm to a maximum of 2 μm.

[0054] Figure 1A Also shown is an enlarged view of a portion of the pixel area 12, which shows a red pixel (R) 18, a green pixel (G) 20, and a blue pixel (B) 22. In one embodiment, the LED is a GaN-based LED, and the active layer of the LED within the pixel outputs blue light. The red and green pixels are formed using phosphors, quantum dots, or other color conversion mechanisms to down-convert the blue pump light to a longer wavelength. The gap 24 between the pixels is less than 0.5 μm, and preferably about 0.2 μm, to increase the density, fill factor, and resolution of the display. The space 25 between the pixels is filled with a reflective metal (e.g., aluminum).

[0055] The die can also be rectangular even though the display portion 10 is hexagonal.The die can also contain various silicon circuitry for processing image signals, powering the device, addressing pixels, and the like.

[0056] Figure 1B The back plate 46 is shown with Figure 1A A schematic diagram of certain circuits used in conjunction with the femtoprojector front panel 10. Figure 1B One possible addressing technique used to address a particular pixel by applying a voltage to its associated contact on the backplate 46 is schematically illustrated.The width of the die is approximately 0.5-1 mm.

[0057] The image signal can be transmitted to the backplane 46 using wireless or other means. In one embodiment, a radio frequency signal (e.g., approximately 13 MHz) is received by an antenna and processed by a receiver / processor 110. Power for the backplane 46 can be received by the antenna via resonant inductive coupling and converted to the appropriate voltage and polarity by a power converter 112. The power signal and the image signal can be at different frequencies, making it possible to separate the signals. The power converter 112 and the receiver / data processor 110 can be integrated into the backplane chip 46, or can be integrated into a separate power / data chip, wherein the data receiver / processor 110 and the power converter 112 are electrically connected to the display backplane 46 via conductors. The small size allows the femtoprojector display to be packaged in a contact lens. The image signal can include an addressing signal decoded by a column decoder 114 and a row decoder 116. Traces 118 in the device layer of the backplane 46 form an array of pixel locations. The control voltage on the selected column and row lines causes the transistors to turn on to conduct current to the selected pixels. Color brightness can be controlled by pulse width modulation, amplitude modulation, or other means. Low power CMOS switches can be used to address the pixels.The relative brightness of the red, green, and blue pixels in a single full-color pixel determines the perceived color for that composite pixel.

[0058] In an example where the display is incorporated into a contact lens, the power converter 112 and the receiver / processor 110 can be separated from the backplate 46 in separate chips, and both chips can be packaged separately in the contact lens. The power / data chip is positioned away from the pupil to avoid obstructing vision. Thin wires connect the metal pads on the backplate 46 to the metal pads on the power / data chip. A thin wire loop antenna is also connected to the pads on the power / data chip and packaged in the contact lens.

[0059] Figure 2 It is captured by red pixel 18, green pixel 20 and blue pixel 22 Figure 1A As shown in the figure, the pixels are much taller than they are wide. The height of the semiconductor LED portion of the pixel can be approximately 4 μm, while the width can be approximately 0.6 μm. The height (i.e., thickness) of the wavelength conversion material can be less than 2 μm. The GaN semiconductor regions are optically isolated from each other and electrically connected to each other via reflective metal N-contacts (reflective N-metal 54) that fill the spaces between the pixels.

[0060] Red pixel 18 includes a thin P-type layer 26, an active layer 28, a relatively thick N-type layer 30, a distributed Bragg reflector (DBR) 32 that passes blue light but reflects red light, and a red down-converter 34, such as a phosphor or quantum dots. Overlaying red down-converter 34 may be a reflective DBR 35 that reflects blue light but passes red light.

[0061] Green pixel 20 is identical to red pixel 18, except that DBR 36 reflects green light, and a green down-converter 38 covers the blue pump LED. Covering green down-converter 38 may be a reflective DBR 37 that reflects blue light but passes green light.

[0062] The blue pixel 22 is identical to the red pixel 18, except that no DBR or wavelength converter is required. A transparent dielectric material 40 and a protective layer 42 can be formed over the blue pump LED to maintain planarity with the red and green pixels. If the blue pump light is not the desired blue display wavelength, for example, when using deep blue light <430 nm, a suitable DBR and down-converter material can be used to generate the desired blue display wavelength (preferably in the range of 455 nm to 470 nm).

[0063] A protective transparent oxide layer (not shown) may be formed over top of the display.

[0064] Reflective P-metal electrodes 44 (anode electrodes) are formed on P-type layer 26 and are in electrical contact with associated metal pads on backplane substrate 46. They also improve optical efficiency by reflecting light toward the desired output face. Substrate 46 may comprise silicon and include addressing circuitry.

[0065] After forming the LED semiconductor layers, they are masked and etched (e.g., by RIE) to form hexagonal trenches around each pixel area. These trenches form pillars of the semiconductor layer in a honeycomb pattern. The trenches are essentially vertical, but may have a slight inward angle due to the nature of RIE etching.

[0066] A transparent dielectric material 50, such as oxide or nitride, is formed around the bottom portion of each LED sidewall to insulate the sides of the P-type layer 26 and active layer 28 in the PN active junction region. This can be accomplished using a mask and etching step. The sidewalls of the N-type layer 30 are exposed.

[0067] Reflective N-metal 54 (cathode electrode) is then deposited in the trenches between the hexagonal pixels to electrically contact the larger vertical sidewall areas of the N-type layer 30. Preferably, more than 80% of the pillar height is in electrical contact with the N-metal 54. The N-metal 54 can comprise nickel, silver, gold, aluminum, titanium, alloys thereof, or other reflective metals to achieve at least 80% reflectivity and can include multiple metal layers. Reflective and low-resistance metal-semiconductor contacts are crucial for the metal in direct contact and within 100 nm of the N-type layer 30. Further away from the N-type layer 30 are other metal layers (e.g., copper) selected for mechanical strength, thermal conductivity, and electrical conductivity. Sidelight that penetrates the dielectric material 50 is also reflected back by the N-metal 54. As a result, virtually all light generated by each blue pump LED is emitted from the top surface (through the top of the N-type layer 30), and there is virtually no crosstalk between pixels. The N-metal 54 also serves to mechanically support the pillars and dissipate heat.

[0068] The blue light then passes through the DBR 32 / 36 to be converted by the red or green down converter material 34 / 38. The DBR 32 / 36 allows the blue light to pass but reflects back the red or green light, so the efficiency is very high.

[0069] Aluminum 25 is deposited over the N-metal 54 in the trenches between the hexagonal pixels to surround the DBR 32 / 36, down converter material 34 / 38, and transparent dielectric material 40. The aluminum 25 provides high reflectivity to limit crosstalk.

[0070] N metal 54 extends to Figure 1A N ring region 16 surrounds display 10. N ring region 16 is bonded to a cathode electrode on backplane substrate 46. In another embodiment, aluminum 25 terminates along the perimeter of display 10 in one or more electrodes for contacting the cathode on substrate 46.

[0071] A significant amount of heat is generated in the LED pixels and in the down-converting layers above them. The combination of N-metal 54 and aluminum 25 provides an excellent thermal conductor to remove heat from the large area of ​​LEDs in the pixel array and down-converting array and spread the heat across the width of the die.

[0072] In a preferred design, individual RGB pixels are optically fully isolated with no substantial optical path between pixels, thereby preventing degradation of the display resolution and maintaining a large color gamut. The device preferably has an optical isolation between adjacent pixels greater than 1000:1 and a color gamut defined by the blue, red, and green primaries (with light exclusion from adjacent pixels greater than 1000:1).

[0073] Figure 3-Figure 12 Some example manufacturing techniques are illustrated. Figure 3 The various layers in the LED wafer that form the pump LEDs are shown. These layers are epitaxially grown on a sapphire substrate 58 which is subsequently removed after the LED die are bonded to the silicon backplane substrate 46.

[0074] An AlGaN buffer layer 60 is grown on the sapphire substrate 58, followed by the undoped GaN layer 62, the N-type layer 30, the active layer 28, and the P-type layer 26. A reflective metal (e.g., Ni / Au / Al, ITO / Al, Ni / Ag, Ag) anode contact 63 is formed on the P-type layer 26 to make electrical contact therewith.

[0075] A metal anode contact 63 is deposited on the planar epitaxial wafer and annealed to form a low resistivity reflective anode contact. Alternatively, the metal anode contact 63 is deposited in the form of a pixel contact using a lift-off technique.

[0076] exist Figure 4 In the process, the various LED semiconductor layers and the metal anode contact 63 are masked and etched to form grooves around each hexagonal column. Each column will be a primary color pixel about 0.5-1um wide, with a gap of about 0.6um or less between pixels. The etching depth is lower than the PN junction. The dielectric material 50 is formed on the metal anode contact 63, the P-type layer 26, and insulates the PN junction in the active layer 28. The preferred transparent dielectric material 50 is SiO2, SiN or Al2O3. These materials are selected because they have optical transparency, high electrical breakdown voltage and the availability of pinhole-free deposition technology. For Al2O3, atomic layer deposition (ALD) can be used to form a thin aluminum layer, and then the Al layer is wet oxidized. The dielectric sidewall thickness is preferably less than 0.05μm.

[0077] exist Figure 5 In the embodiment of the present invention, the gaps between pixels are etched to extend through the transparent dielectric material 50 in the bottom of the etched trench, down to below the N-type layer 30, and optionally down to the substrate 58. The sides of the N-type layer 30 are exposed. The N-type layer 30 of each pillar is isolated from each other. There is no continuous N-type layer connecting adjacent pillars.

[0078] like Figure 6 As shown, a suitable reflective N-metal 54 (e.g., Al, Ti / Al) is deposited in the trenches and on the pillars. The N-metal 54 electrically contacts the sides of the N-type layer 30, but is insulated from the anode contact 63, the P-type layer 26, and the active layer 28 by the dielectric layer 50. Because the N-metal 54 contacts a large area, it has good current distribution, a constant cathode voltage between pixels, and low contact resistance.

[0079] like Figure 7AAs shown, a chemical mechanical polishing (CMP) process is performed to planarize the top surface and expose the anode contact 63, the reflective N-metal 54 (cathode electrode), and the dielectric 50 on the same plane.

[0080] Figure 7B yes Figure 7A , and shows the pixel region after CMP, where the top surface including the anode contact 63, the reflective N-metal 54 (cathode electrode) 54 and the dielectric material 50 is planar.

[0081] exist Figure 8 In FIG. 5 , metal bump 66 is formed on anode contact 63, and at the same time, metal bump 67 is formed in N ring region 16 which is electrically connected to N-metal 54. Note that the LED column below N ring region 16 is shorted by metal bump 67 and does not emit light by design. The combination of anode contact 63 and metal bump 66 forms Figure 2 The P-metal electrode 44 in the.

[0082] Preferred bump materials include indium, tin, AuSn, gold, and copper. The planarity of the "soda can" LED device after CMP and the small display area make the soda can LED compatible with Cu-Cu bonding to Cu pads on a silicon CMOS backplane wafer. Cu-Cu bonding requires planarity of less than a few nanometers across the array, which is achieved by the soda can LED device design.

[0083] The LED wafer is then singulated and the LED die are bonded to a silicon backplane wafer as described below.

[0084] exist Figure 9 In, as Figure 2 As shown, metal bumps 66 and 67 on the LED die are bonded to corresponding metal bumps 68 and 69 on the silicon backplane substrate 46. Ultrasonic bonding (typically using gold bumps) or thermocompression In-In bonding or Cu-Cu bonding may be used.

[0085] The mating bump metals can include Au-Au, AuSn-Au, In-In, Sn-Sn, and most preferably Cu-Cu. Cu-Cu is not used in conventional LED array manufacturing due to the extremely high flatness required on both the backplane and device sides (typically <2nm RMS flatness over the Cu-Cu contact area of ​​the entire array). The structural design of the soda can LED device meets the requirements of Cu-Cu bonding so that the surface anode and cathode contacts are coplanar (preferably within 2nm) by CMP. The cathode contact is shown in Figure 9 on the left side and is Figure 1AThe preferred design of the N ring region 16 shown in FIG is formed on the same surface as the anode contact along the edge of the die and is electrically connected to the N-metal 54. CMP can be performed after the metallization step to planarize the anode and cathode bumps (e.g., Cu). The distance between each LED anode and cathode bump is less than a few microns (e.g., less than one micron), so that Cu recessing during CMP can be minimized to less than a few nanometers (which can be less than 1 mm) over the entire array.

[0086] Substrate 46 comprises silicon 70 with an insulating oxide layer 72. MOSFET drivers, addressing circuitry, and various conductors are located in device layer 74. Other types of substrates can be used. For example, oxide layer 72 can be omitted. Gold bumps 66 and 68 can be deposited on hexagonal metal pads to match the shape of the pillars. As previously mentioned, Cu-Cu bonding can also be used.

[0087] The remaining drawings are not within the field of view of the N annular region 16 .

[0088] Figure 10 FIG2 is a simplified view of LED die 76, showing sapphire substrate 58, semiconductor layer 78, and gold bumps 66. Backplane substrate 46 is shown without an oxide layer. Conductive traces are represented by Cu metal layer 80. Copper contacts 82 electrically contact the conductors, and oxide layer 84 is located between copper contacts 82. Gold bumps 68 are formed over copper contacts 82. In some embodiments, the oxide layer may extend to the top surface of gold bumps 68.

[0089] To excite the pixels, Figure 1A N ring region 16 (electrically contacting all N-type layers 30 of the pillar), gold bump 68 on substrate 46 that is electrically contacting the P-type layer of one of the pillars, and N-metal 54 on substrate 46 ( Figure 2 ) is electrically contacted to another gold bump. For example, using row and column lines, N-metal 54 can be connected to ground and gold bump 68 is connected to addressing circuitry, where the intersection of the column and row lines illuminates the pixel at that location. A transistor can be at each pixel location in device layer 74 to apply a voltage / current to the corresponding gold bump 68. This is called an active matrix. In one embodiment, CMOS circuitry is used to address the pixels, and a digitally controllable current source applies a variable current to the pixel to control its brightness. Conventional addressing techniques can be used.

[0090] Figure 11A The figure shows the process of removing the sapphire substrate 58 ( Figure 9). In one embodiment, after removing the sapphire substrate 58, the trench is etched down to the substrate, leaving no semiconductor connections between the LED columns. In another embodiment, after removing the substrate 58 to isolate the individual LED columns, the AlGaN buffer layer 60 ( Figure 3 ) and an undoped GaN layer 62. In all cases, the sides of the LEDs that were previously connected to the sapphire substrate 58 are planar. The backplane substrate 46 provides mechanical support.

[0091] Figure 11B Pictured Figure 11A A variation of an embodiment in which the bonding is Cu-Cu bonding. Glass 85 (SiO2) is formed around each LED pixel, such as by deposition and CMP down to expose the copper anode contact 63 and cathode contact. Glass 85 is not formed over the reflective anode contact 63. A similar layer of glass 86 is formed over the substrate 46. If plating is to be used, a copper seed layer is formed in the openings of glass 85 and 86. Copper 87 and 88 are then plated or otherwise deposited to fill the opening areas of glass 85 and 86. The LED wafer surface and the substrate 46 surface are then polished using CMP so that both surfaces are planar (flatness within 2 nm). Copper 87 / 88 form electrodes for the substrate 46 and the LED pixels. If the copper is not recessed enough during CMP planarization, a short wet etch of copper 87 / 88 is used to slightly recess copper 87 / 88.

[0092] The LED wafer is then cut and the die is bonded to the substrate 46 wafer as described below. The LED die is aligned with the substrate 46 wafer and the structure is pressed together (e.g., at 100 psi) under heat (e.g., 200°C) so that the glass 85 / 86 surfaces are bonded. As the heat increases (e.g., to 300°C), the thickness of the copper 87 / 88 expands greater than the thickness of the glass 85 / 86, and the opposing copper electrodes contact each other to form a low-resistivity contact. The copper atoms diffuse to create a good mechanical bond. The temperature expansion coefficient of copper is about 16 ppm / C, while the temperature expansion coefficient of glass is about 0.6 ppm / C. The bonding process may take 30 minutes.

[0093] In this configuration, bonding first occurs between the SiO2 on the silicon backplane side and the SiO2 on the LED side at a low temperature of approximately 200°C under thermal compression. The bonded wafers are then heated to above 200°C, allowing the copper on both the silicon backplane and the LED device side to expand sufficiently to contact each other and form a thermocompression bond. In this case, compression is solely due to the different coefficients of thermal expansion between the SiO2 and copper.

[0094] The remaining figures assume the use of Figure 11Abonding technique, but any bonding method can be used.

[0095] exist Figure 12 In the example, DBR 89 is formed above the blue pump LED column for red pixels 90. DBR 89 consists of many layers that collectively pass blue light but reflect red light. This way, the LED does not absorb red light. Red down-converter material 92 covers DBR 89. The down-converter material can be phosphor or quantum dots.

[0096] Similarly, a DBR 96 is formed over the blue pump LED column for the green pixel 98. The DBR 96 passes blue light but reflects green light. A green down-converter material 100 covers the DBR 96. The down-converter material can be a phosphor or quantum dots.

[0097] Above the blue pixel 102 is a transparent dielectric 104 so the top surface is planar.

[0098] Aluminum 25 is deposited in the hexagonal gaps between adjacent pillars of down-converter material. The aluminum 25 surrounding the pillars forms a reflective containment for the converted light to prevent optical crosstalk between different color pixels within the color conversion layer. The aluminum 25 also helps mechanically support the down-converter material and determine its shape. Figure 12 The top view is similar to Figure 1A The enlarged part of .

[0099] The thickness of the down conversion material should be thin to maximize efficiency and minimize the height of the entire device. For quantum dots, this thickness should be on the order of 1.5um. The latest color conversion layer thicknesses are 10-50um, so special design considerations must be taken to convert more than 99% of the pump blue light to the desired color. It is desirable to convert more than 99% of the pump blue light to the desired color to maintain the color purity of the converted color pixels and thus maintain a good color gamut for the entire RGB display. One technique is to form a polymer matrix in which the quantum dot crystals are deterministically arranged to achieve optimal conversion and maximum solid density. For quantum dots with a diameter of 6nm, a quantum dot density of >10 21 / cm 3 , to achieve over 99% conversion within a thickness of 1.5um.

[0100] A DBR 106 can be deposited on top of the display to reflect the incident blue pump light to reduce leakage of the pump light through the color converted pixels. The DBR 106 can also be designed to partially collimate the pixel emission. Figure 12In the case shown, blue light directly from the LED is used for the blue pixel, and the reflective DBR 106 is removed from the area above the blue pixel. In one embodiment, the area above the blue pixel is replaced by a transparent dielectric 109, which can be planarized, for example, by CMP.

[0101] The wavelength conversion layer above the N-type layer 30 can be formed as a separate wafer or die and then bonded on top of the LED die or LED wafer.

[0102] Figure 13 1 is a top view of hexagonal metal pads 107 on backplane substrate 46 supporting gold bumps 68 ( FIG. 11 ), which correspond to gold bumps 66 on the LED die. Metal pads 107 may alternatively be copper for Cu-Cu bonding to the LED die.

[0103] Figure 14 The diagram illustrates how the pixel density (resolution) is reduced from the central portion 108 of the display 10. The display 10 projects its image directly onto the retina. The retina has a fovea, which contains the densest concentration of cones. Outside the fovea, the density of cones is lower. Therefore, the display 10 does not need to have a high resolution outside its central region 108. This improves the efficiency of the display 10 because the resolution is no higher than the resolution outside the foveal area. The resolution of the display 10 in the central region 108 should be indistinguishable from the resolution of the real world as perceived by the fovea. The resolution of the display outside the central region can be smoothly reduced, consistent with the reduction in resolution of the retina away from the fovea. Therefore, over the entire displayed image (including the portion of the image in the high-resolution fovea and the area away from the fovea as the resolution decreases), the displayed image can be indistinguishable from reality.

[0104] In the foveal area, 20 / 20 resolution on the retina requires a spacing between white (RGB composite) pixels on the retina of approximately 6 μm. If the display is enclosed in a contact lens, the display is approximately 25 mm from the retina. If the optics in front of the display have a magnification of approximately 3X and the white composite pixels are composed of RGGB pixels, the spacing between white pixels on the display must be less than 2 μm (6 μm / 3), and the spacing between individual color pixels must be further reduced by a factor of 2, so must be less than 1 μm. It is desirable to keep the size of the display less than 1 mm so as not to substantially block real-world light from entering the pupil. The minimum pupil diameter is typically 2 mm, which means that for a minimum pupil size of 2 mm in diameter, a circular obstacle with a diameter of 1 mm will block approximately 25% of real-world light. In order to image a relatively large field of view (FOV) on the retina, the foveal pixel spacing in an LED display should be less than 1 μm to display an image on the retina that is indistinguishable from reality. The spacing is therefore less than the height of each pixel (column).

[0105] The size of the LED pixels can be varied across the display to achieve a projected image that matches the resolution of the retina, with the highest resolution at the center of the fovea.

[0106] Figure 15 The backplane substrate 46 is schematically illustrated as being used to provide a plurality of electrodes to the pixel through the associated gold bumps 68 ( Figure 12) is one possible addressing technique for addressing the pixels by applying a voltage. The width of the die can be 0.5-1 mm. The image signal can be transmitted to the backplane substrate 46 using RF or other means. In one embodiment, a radio frequency signal (e.g., approximately 13 MHz) is received by an antenna and processed by a receiver / processor 110. The power converter 112 and the receiver / data processor 110 can be integrated into the display backplane or integrated into a separate power / data chip with a data receiver / processor 110 and a power converter 112 electrically connected to the display backplane by wires. The antenna can receive power via resonant inductive coupling and can be converted to the appropriate voltage and polarity by the power converter 112. The power signal and the image signal will be at different frequencies so that the signals can be separated on the substrate 46 using appropriate coupling circuits on the substrate 46. This allows the monolithic display die to be packaged in a contact lens or eyeglass lens. The image signal includes an addressing signal decoded by a column decoder 114 and a row decoder 116. Traces 118 in the device layer of the substrate 46 "intersect" at each pixel location. Control voltages on selected column and row lines turn on transistors, conducting current to the selected pixel. Color brightness can be controlled by PWM or the amplitude of the current. Low-power CMOS switches can be used to address the pixels. The relative brightness of the red, green, and blue pixels in a single full-color pixel determines the perceived color for that composite pixel.

[0107] In an example where the display is incorporated into a contact lens or eyeglass lens, the power converter 112 and receiver / processor 110 are separated from the backplane substrate 46 in separate chips, and the two chips are separately packaged in the contact lens or eyeglass lens. The power / data chip is located away from the pupil to avoid obstructing vision. Thin wires connect metal pads on the backplane substrate 46 to metal pads on the power / data chip embedded in the lens. A thin wire loop antenna is also connected to the pads on the power / data chip and is packaged in the lens.

[0108] All manufacturing processes are preferably performed on a wafer scale. As described above, the LED wafer is diced and individual or grouped LED dies are attached to the backplane substrate 46 wafer. If the color conversion layer is formed as a separate die, the color conversion die is attached to the LED die above the backplane substrate 46 wafer, which is then singulated to form individual display dies (i.e., substrate die bonded to LED pixel dies) with edges less than 1 mm.

[0109] In the embodiments disclosed above, all LEDs emit blue light, and down-converter materials are used to create red and green light. Figure 15The diagram illustrates a technique for making a monochrome display. Therefore, no wavelength conversion material is required. All LEDs are based on GaN to generate a single color (e.g., blue or green), or the LEDs are based on AlInGaP to generate a single color (e.g., red or amber). Four green LEDs 120 are schematically illustrated. The reflective N-metal 54 that electrically contacts the N-type layer and surrounds the hexagonal columns can be the same as previously described. The gold bumps 68 that contact the P-type layer can be the same as previously described. The DBR layer 128 can provide collimation. The transparent oxide layer 130 protects the device.

[0110] Figure 16 Another type of LED pixel is shown in which all LEDs 134 emit infrared pump light (e.g., 975 nm wavelength) and convert the light to red, green, or blue using red upconverter material 136, green upconverter material 138, and blue upconverter material 140. A first DBR layer 142 passes the IR pump light and reflects visible wavelengths. A second DBR layer 144 passes visible light and reflects IR light back to improve efficiency. The growth substrate is GaAs.

[0111] The individual pillars may also be resonant cavity LEDs or vertical cavity surface emitting lasers.

[0112] Although the devices described herein are full-color displays using columns of LED pixels, similar techniques can be used to form arrays of vertical transistors, photodetectors, and other semiconductor devices.

[0113] Preferably, the fabrication process is performed on a wafer scale. Figures 17A to 17N Shows the use of Figure 1A to Figure 1B Another process to manufacture the femtoprojector display is to use a front panel and back panel. Figure 17A is a legend showing the cross-hatching pattern used in this series of figures. Metals and other conductive materials are represented by cross-hatching patterns. TCOs are transparent conductive oxides (e.g., indium tin oxide, aluminum zinc oxide, indium zinc oxide, indium cadmium oxide, and carbon nanotube layers). GaN is a gallium nitride region suitable for making the active region of an LED. GaN is a direct bandgap III-V semiconductor material well suited for making LEDs. Figure 17A In CMOS, dielectrics and silicon dioxide (SiO2) are materials that can provide electrical isolation, and the dielectric stack is used to provide wavelength selectivity.

[0114] Figure 17BA front plate wafer 300 and a separated back plate wafer 350 are shown. The front plate wafer 300 includes a GaN region 310 epitaxially grown on a substrate 302, which will be referred to as the front plate substrate. Starting from the substrate, the GaN region 310 includes a buffer region (not labeled), an n-doped region 312, an active region 314 (indicated by a doped line), and a p-doped region 316. Typically, the GaN region 310 can have a total thickness of approximately 4-6 μm. If an InGaN multiple quantum well structure is used, the active region 314 is typically very thin (e.g., 0.1 μm or less). The p-doped region 316 is also thin (perhaps 0.2 μm, typically less than 0.5 μm). The rest is the n-doped region 312 and a relatively thick buffer region. This is because a certain thickness is required to allow the GaN growth to reach sufficient quality. The GaN directly adjacent to the substrate 302 is a buffer region of poor quality.

[0115] The P-contact metal 320 provides electrical contact to the P-doped region 316. In some designs, the P-contact metal 320 also acts as a reflector for light generated by the active region 314. Examples of substrate 302 include sapphire. Other examples include silicon and silicon carbide. Figure 17B The frontplane wafer 300 in FIG is not patterned. That is, the GaN region 310 has not yet been patterned into individual LEDs. This significantly reduces the alignment requirements when attaching the frontplane wafer 300 to the backplane wafer 350.

[0116] The backplane wafer 350 includes LED driver circuitry on a substrate 352 . Figure 17B The actual LED driver circuit is not shown, but copper pads 362, 366 are shown for making electrical contact between the driver circuit and the LED contacts. Copper pad 366 provides an electrical connection from the n-contacts of all LEDs to a common cathode on the backplane. Copper pad 362 provides a connection from the p-contact metal 320 of each LED to the addressable driver circuitry for that LED. The filler 368 between the copper pads 362, 366 can be SiO2 or SiN which acts as an etch stop in subsequent processing steps. x Typically, the backplane wafer 350 is a processed CMOS on a silicon wafer. Figure 17B Some alignment marks 390 are also shown.

[0117] Attach the front panel wafer 300 to the back panel wafer 350 to obtain Figure 17Cstructure. In the example shown, attachment is performed by non-welding surface bonding (e.g., by surface diffusion) between the p-contact metal 320 and the copper pads 362, 366. A conductive bonding agent 330 is deposited on the p-contact metal 320 of the front plate wafer. Examples of bonding agents include aluminum, indium tin oxide, aluminum-doped zinc oxide, and aluminum with a surface coating of silicon, germanium, or titanium to prevent aluminum oxidation. The bonding agent 330 is bonded to the copper pads 362, 366. This mechanically attaches the two wafers to each other. It also provides an electrical connection between the p-contact metal 320 and the copper pads 362, 366. This bonding step requires only rough alignment because the GaN region 310 has not yet been patterned into individual pixels. The bonding agent 330 can be selected to be compatible with further silicon wafer processing, and the bonding process itself occurs at a temperature and pressure that does not affect the already processed CMOS structures on the backplane.

[0118] In an alternative approach, both metal layers can be coated with a bonding agent. In this example, both the p-contact metal 320 and the copper pads 362, 366 can be coated with a bonding agent and then bonded together.

[0119] In an alternative approach, the frontplane wafer 300 is coated with a solder such as Sn or In and attached to the backplane wafer 350 using solder bonding. In this approach, the copper pads 362, 366 may also be coated with solder.

[0120] After bonding the two wafers 300, 350 together, the front plate substrate 302 is removed. Laser lift-off can be used to remove the sapphire substrate 302. Chemical methods can also be used to remove the sapphire substrate 302. This exposes the buffer region of GaN 310. Removing the sapphire substrate 302 can generate significant vibrations, so from a mechanical stability perspective, it is helpful to perform this step before patterning the GaN region 310.

[0121] like Figure 17D As shown, a dielectric filler 335 such as silicon dioxide is deposited to planarize the surface. Other examples of materials include SiN, benzocyclobutene (BCB), and spin-on-glass.

[0122] The GaN region 310 is thinned to obtain Figure 17E Chemical mechanical polishing and / or blanket dry etching can be used to reduce the thickness of GaN region 310. Thinning removes most or all of the GaN buffer region.

[0123] like Figure 17FAs shown, the GaN region is patterned into individual LED columns 342R, G, B, which form the LEDs for the red, green, and blue pixels of the display. In one method, a deep etch is performed with the GaN region 310 and the conductive bonding layer 330. Etching in the SiN x The CMOS process stops at the surface of the SiO2 region 368. This also exposes the pad 366 for the common cathode.

[0124] By etching trenches between the pillars, Figure 17E The GaN region 310 is patterned as Figure 17F The individual LED columns 342 are arranged in a pattern. These grooves can be 0.2-0.3 μm wide. If the GaN is not thinned, the gaps between the columns will be 6 μm high and 0.2-0.3 μm wide, with an aspect ratio in the range of 20-30. It is difficult to etch narrow grooves with such a high aspect ratio. It is also difficult to further process the narrow grooves (for example, coating or filling them). Thinning the GaN region reduces the aspect ratio of the grooves, which makes etching and other processes easier.

[0125] Thinning the GaN region also reduces the aspect ratio of the LED columns 342, thereby improving their optical performance. Figure 17 is not drawn to scale. Instead, the figures are drawn to illustrate the sequence of process steps and the spatial relationship between the various material layers. For a height-to-width aspect ratio in the range of 4-12, the unthinned LED columns 342 can be 4-6um high and 0.5-1um wide. Thinning the GaN region can remove 2-4um of material, reducing the LED columns to a height of no more than 2um and reducing the aspect ratio by a factor of 2 to 3.

[0126] Note that when the GaN pixels are patterned, certain areas of the bonding agent 330 are etched away. The conductive bonding agent 330 is selected so that it does not coat (eg, sputter as an etch byproduct) the sidewalls of the GaN LED pillars 342 to prevent shorting of the LEDs.

[0127] exist Figure 17G In one embodiment, the sidewalls of the LED pillar 342 are passivated. In one approach, atomic layer deposition is used to deposit a passivation material 344 over the entire structure, including on the top and sidewalls of the LED pillar 342. Example passivation materials include Al2O3, TiO2, SiO2, SiN X , HfO X and NbO X A directional etch is used to etch the passivation material 344. This removes the passivation material from the level that includes the top of the individual LED columns, but leaves the passivation material on the sidewalls of the LED columns. This method does not require any photolithography or fine alignment. The passivation material 344 electrically isolates the sides of the LED columns 342 to prevent short circuits of the LEDs.

[0128] exist Figures 17H-17I The electrical connection between the LED column 342 and the common cathode 366 is formed in Figure 17H In the process, a thin reflective metal liner (metal 346) such as aluminum or ruthenium is deposited, followed by a filler metal (metal 347) such as copper, aluminum or gold. It is then planarized via chemical mechanical polishing to expose the GaN surface. Figure 17I In the embodiment of the present invention, a thin layer of transparent conductive oxide 348 (eg, ITO) is deposited as a current spreading layer that provides electrical connection to each pixel.

[0129] This completes the LED structure in the femtoprojector. Figure 17I Each LED's individual p-contact goes from p-contact metal 320 through conductive bonding agent 330 to copper pad 362 to the addressable driver circuit. All LEDs' n-contacts go through transparent conductive oxide 348 to metal 346 / 347 to copper pad 366 to the common cathode.

[0130] Note that this process uses only one photolithography step. Figure 17F In FIG, high resolution photolithography is used to align the LED pillars 342 with their corresponding driver pads 362. However, the sidewalls ( Figure 17G ) insulates and forms an electrical network to the top contacts of the individual LED columns ( Figures 17H-17I ) is completed without any photolithography steps.

[0131] Figure 18 A cross-sectional view of two adjacent LED columns 342 is shown. For illustration purposes, the specific dimensions are drawn to scale. Figure 18 , but other dimensions can be used to construct LED displays. In this example, the LED column 342 includes 2.75um n-doped GaN 312, a 0.05um InGaN multi-quantum well active region 314, and 0.2um p-doped GaN 316. Light is generated in the active region 314, so the optical path from the active region to the LED exit at the far end of the n-doped region 312 is 2.75um long. If the width of the LED column is 1um, the aspect ratio of the optical path is 2.75:1. Before reaching the exit, the light may reflect from the sidewalls multiple times. Each reflection introduces some light loss.

[0132] As described above, the sides of the LED column 342 are coated with a dielectric 344 (0.02um), a metal 346 (0.05um) that serves as a bottom reflector 346, and a metal filler 347 (0.16um) that provides structural support and can also provide an electrical connection to the common cathode pad. To manufacture these structures, 3um high LED columns are used, and a 3um high and 0.3um wide trench is first etched between the LED columns. This is a trench with a height-to-width aspect ratio of 10:1. Because the trench electrically isolates the p-doped GaN 316 from adjacent pixels, the depth of the trench is 3um. The trench also extends through the bottom metal 320 and the bonding agent 330, so the total trench depth can be greater than 3um. The sidewalls are then conformally coated with a dielectric 344 and a reflector 346. When the trench is filled with metal 347, this narrows the trench to 0.16um (but still 3um high).

[0133] If the GaN region is not thinned, the LED column 342 will be taller (e.g., 6 μm). At this height, the optical path from the active region 314 to the exit has an aspect ratio of 5.75:1, doubling the number of reflections before leaving the LED. Additionally, the trench to be etched will have an aspect ratio of 20:1, and the metal fill 347 will occupy a space with an aspect ratio of almost 40:1.

[0134] The dimensions given above are examples only. Typical ranges are as follows. For the gap between pillars: the entire gap width is 0.2-1um, the dielectric 344 is 0.01-0.05um, and the reflector 346 is 0.03-0.10um. For the LED pillars: n-GaN 312 is 1-5um, MQW 314 is 0.05±0.025um, and p-GaN 316 is 0.2±0.1um. In GaN ( Figure 18 ) as follows: <50 nm for the p-contact metal 320 (ITO, Ag or NiAu) and <1 um for the bonding agent 330 .

[0135] Return to Figure 17, 17A to 17I The steps in FIG. 3 are described in the context of attaching the LED frontplane wafer 300 to the silicon CMOS backplane wafer 350. This can also be done at the die level. For example, the LED frontplane wafer can be diced and individual or groups of LED frontplane dies 300 attached to corresponding backplane dies 350. The backplane die can also be in wafer form or diced into individual or groups of dies.

[0136] Figure 17J-17N The additional step of adding a color conversion material, which in this example is a quantum dot material, is shown. Figure 17JIn the embodiment of the present invention, a multilayer dielectric stack 373 is deposited on the top surface. The dielectric stack 373 provides wavelength selectivity. For example, if the LED produces blue light, the dielectric stack 373 can be designed to transmit blue light and reflect red and green light. A thick oxide (e.g., SiO2 or SiN X ) is deposited on the dielectric stack 373. Figure 17K In FIG, dielectric stack 373 and thick oxide 374 are patterned into columns 372R, G, B aligned with LED columns 342R, G, B. Figure 17L In the , a thin reflective metal liner 375 such as aluminum or ruthenium is deposited, followed by a fill metal 376 such as copper or aluminum. Figure 17M In the process, it is planarized by chemical mechanical polishing to expose the oxide 374. Figure 17N In the embodiment of the present invention, sacrificial layer 374 is removed, for example using a wet or dry chemical etch. This leaves a space 377 into which the color conversion material can be deposited. A typical height of space 377 is in the range of 1-3 μm. Different materials (e.g., quantum dot material 377R for color conversion from blue to red, quantum dot material 377G for color conversion from blue to green, and material 377B for blue pixels) can be deposited in different spaces 377. Another alternative is light scattering particles (e.g., TiO2) for blue pixels.

[0137] Alternatively, the color conversion layer can be formed as a separate die and then attached to the LED die over the backplane. It is then singulated to form individual femtoprojector displays, typically less than 1 mm in width.

[0138] One possible use for this monolithic, ultra-dense LED display is to embed the display into a contact lens so that the displayed image overlays (or replaces) the wearer's view of the real world. Figure 19 A cross-sectional view of an eye-mounted display incorporating a femtoprojector 500 in a contact lens 550 is shown.

[0139] Figure 19 An embodiment using a preferred scleral contact lens is shown. Scleral contact lenses are preferred because they are designed not to move on the cornea, but contact lenses do not have to be scleral. The aqueous humor of the eye is located between the cornea 574 and the lens 576 of the eye. The vitreous humor fills most of the eye (including the volume between the lens 576 and the retina 578). The iris 584 limits the eye's aperture.

[0140] The contact lens 550 is preferably less than 2 mm thick, and the femtoprojector 500 is preferably assembled in a volume of 2 mm by 2 mm by 2 mm or less. The contact lens 550 is comfortable to wear and maintains eye health by allowing oxygen to reach the cornea 574. The femtoprojector 500 includes an image source (backplate 512 / frontplate 514) and an optical system 530. The image source includes the backplate 512 and the frontplate 514, examples of which have been described above. In this example, the backplate 512 is a CMOS application-specific integrated circuit (ASIC) and the frontplate 514 includes a GaN LED array. The backplate electronics 512 receives data packets from a source external to the eye-mounted display. The backplate ASIC 512 converts the data packets into a drive current to drive the frontplate GaN LED array 514, which produces light that is projected by the optical system 530 onto the user's retina 578.

[0141] The light emitter array 514 can have a non-uniform resolution. For example, the central area of ​​the array can be imaged onto the fovea centralis, and therefore the central pixels have a higher resolution (i.e., the spacing between pixels is smaller) than the pixels on the periphery of the array. The spacing of the front plate 514 and the back plate 512 can be matched, in which case the area of ​​each pixel driver is smaller in the center of the back plate than in the periphery. Alternatively, the back plate 512 can have a uniform spacing, wherein the front plate 514 still has a variable spacing. In one approach, a wiring layer is bridged between the uniform spacing back plate 512 and the variable spacing front plate 514. By using different wiring layers, the same back plate can be used with different front plates.

[0142] Eye-mounted femtoprojector displays can use a 200x200 array of color pixels. The display can be monochrome or color. A three-color display with three LEDs per color pixel can have a total of at least 120,000 LEDs.

[0143] Another possible use for monolithic ultra-dense LED displays is in eyewear (e.g., glasses or goggles) to create an immersive visual experience or overlay an image on the wearer's view of the real world (e.g., in augmented, mixed, or artificial reality applications).

[0144] Although the detailed description contains many details, these details should not be interpreted as limiting the scope of the present invention, but are merely used to illustrate different examples. It should be understood that the scope of this disclosure includes other embodiments not discussed in detail above. Without departing from the spirit and scope defined by the appended claims, various other modifications, changes and variations that are obvious to those skilled in the art can be made to the arrangement, operation and details of the methods and apparatus disclosed herein. Therefore, the scope of the present invention should be determined by the appended claims and their legal equivalents.

Claims

1. A method for manufacturing a femtoprojector display, the method comprising: attaching a frontplane die to a backplane die, the backplane die including an array of LED driver circuits on a backplane substrate, and the frontplane die including a gallium nitride region grown on the frontplane substrate; removing the front plate substrate; thinning the gallium nitride region; as well as patterning the gallium nitride region into individual LED pillars, the individual LED pillars having active areas no wider than 2 μm, wherein the gallium nitride region is thinned such that the height of the individual LED pillars measured from the active area to the top of the gallium nitride region is reduced to less than 3 μm, Wherein patterning the gallium nitride region into individual LED columns comprises: photolithographically masking regions of the gallium nitride area to define the individual LED columns; as well as etching trenches into unmasked areas of the gallium nitride region; Without any photolithography steps: passivating the sidewalls of the individual LED columns; and Electrical contact is made to the top contacts of the individual LED columns. 2 . The method of claim 1 , wherein the growth of the gallium nitride region on the front plate substrate results in a buffer region in the gallium nitride region adjacent to the front plate substrate, and the buffer region is removed by thinning the gallium nitride region. 3 . The method of claim 1 , wherein thinning the gallium nitride region reduces the ratio of the height to the width to less than 5:

1. The method according to claim 1 , wherein the gallium nitride region is thinned such that a thickness of the gallium nitride region is reduced by at least 3 μm.

5. The method of claim 1 , wherein attaching the frontplane die to the backplane die comprises: The metal layer on the frontplane die is bonded to the metal layer on the backplane die.

6. The method according to claim 5, wherein the bonding uses a bonding agent selected from the group consisting of: aluminum; indium tin oxide; aluminum-doped zinc oxide; aluminum with a surface coating of silicon, germanium or titanium; Sn solder and In solder. 7 . The method of claim 5 , wherein the backplane substrate is a silicon substrate, the LED driver circuit is a CMOS circuit, and the metal layer on the backplane substrate is a copper layer.

8. The method of claim 1 , wherein removing the front plate substrate comprises: The front panel substrate is removed using a lift-off process.

9. The method of claim 1 , wherein patterning the gallium nitride region into individual LED columns comprises: Trenches are dry etched into the gallium nitride region to pattern the individual LED columns.

10. The method of claim 9, wherein the width of the trench separating adjacent individual LED columns is no greater than 0.5 um.

11. The method of claim 9, wherein the trenches separating adjacent individual LED columns have a height-to-width aspect ratio of at least 4:

1.

12. The method according to claim 1, further comprising: After removing the frontplane substrate, the frontplane die is planarized before thinning the gallium nitride region.

13. The method of claim 1 , wherein passivating the sidewalls of the individual LED columns comprises: depositing a passivation material on the sidewalls and tops of the individual LED columns using atomic layer deposition; as well as The deposited passivation material is etched without photolithography to remove the passivation material from the top of the individual LED column but to retain passivation material on the sidewalls of the individual LED column.

14. The method according to claim 1, further comprising: A multi-layer dielectric stack is deposited on the top surface of the individual LED columns, the multi-layer dielectric stack providing wavelength selectivity.

15. The method according to claim 14, further comprising: depositing a sacrificial layer on the multilayer dielectric stack; patterning the multilayer dielectric stack and the sacrificial layer into individual pillars aligned with the individual LED pillars; depositing a filler material in the spaces between the individual pillars of the sacrificial layer; removing the sacrificial layer to create a void defined by the fill material; as well as At least some of the voids are filled with a wavelength conversion material that converts light generated by the individual LED rods to different wavelengths, wherein the multilayer dielectric stack passes light from the individual LED rods and reflects light converted to different wavelengths. The method of claim 15 , wherein the wavelength converting material comprises quantum dots.

Images