Patent Application
20110197429
Current technology can produce a wide array of sensor and actuator devices made from different materials. The ability to integrate these devices into complete systems and products is valued by industry because of the potential range of applications. The standard tool kit of fabrication consists mainly of bulk and surface silicon micromachining, laser micromachining and other lithographic techniques. However, industrial trends indicate that future generations of MEMS, sensors, and actuators will be integrated with other electronic and optical components onto a substrate to yield powerful and complex systems. Therefore, the development of large-scale, heterogeneous, and parallel integration techniques to assemble MEMS, electrical and optical devices onto a single substrate is critically important for the realization of low cost, high density, and high performance systems.
As the market for low cost and high density integrated circuits increases, new assembly techniques must be developed. Methods must be found to enable the assembly of devices onto dissimilar substrates. To improve performance and reduce the cost of assembly, compound semiconductor devices must often be integrated monolithically in order to activate circuitry that is embedded within the substrate. A primary interest is the integration of devices with CMOS technology in order to increase the number of on-wafer functions and ultimately reduce the cost, size, and weight of systems. However, combining different materials comes with inherent difficulties. Chief among these are mismatches of lattice properties and thermal expansion coefficients. For example, there are very large variations of thermal expansion coefficients between silicon and III-V compounds used for optoelectronic devices.
Current integration strategies often rely on ‘pick and place’ serial assembly techniques. These techniques encounter immediate and insurmountable speed and cost constraints especially in applications that require the assembly of a large number of components with high precision. Also, surface forces must be carefully controlled in order to prevent the unwanted adhesion of devices with each other and the tools of the assembly process. Because of these limitations, new, low-cost, and parallel assembly techniques are being investigated and introduced to industry.
As the dimensions of micro-electrical, micro-optical and micromechanical devices and systems decrease, there is a need for techniques that simplify the effective parallel process of assembly. Several approaches have been proposed for such assembly. They include selective area growth, flip chip bonding, epitaxial lift off, electrostatic alignment, and fluidic self-assembly. However, all of these approaches come with drawbacks and other technological issues that limit their application.
Selective area growth has been investigated as a potential method; its limitation comes from lattice and thermal mismatches.
Another method is epitaxial lift off. An epitaxial layer is released from its growth substrate; the layer, which is typically supported by a polymer membrane, is then bonded onto a host substrate by van der Waals forces. The devices can be processed either before or after the transfer of the layer depending on the requirements of the process. The technique suffers from various disadvantages, including the handling of potentially extremely thin epitaxial layers, which can be difficult, and any pre-processed devices that need to be aligned onto existing circuitry, which is time consuming.
Yet another approach is wafer bonding that is used to transfer a primary layer onto a secondary wafer. The primary layer and secondary wafer are then bonded together and processed into discrete devices. The methods major drawback is thermal expansion coefficient mismatch when the layer and wafer are comprised of different materials.
A common approach to bulk parallel processing is to align devices without individual manipulation. Techniques that follow that paradigm are vector potential parts manipulation, DNA and electrophoresis-assisted assembly, and fluidic self-assembly.
Vector potential parts manipulation allows for the alignment of devices by using electrostatics to direct and place units.
The DNA and electrophoresis-assisted assembly uses two sets of a DNA-like polymer film. One film is formed onto the individual parts and another complimentary film is deposited in substrate recesses on the wafer where the parts are to be placed. The parts will only adhere into those locations with matching DNA patterns.
In the fluidic self-assembly approach, carefully etched devices are placed onto a substrate with matching recesses. Separate devices are aligned and placed in the substrate recesses without active individual manipulation. However, the method requires the devices to be formed with beveled edges, which is costly to achieve, and the process is statistical therefore does not yield 100% accurate assembly.
Other methods to assemble devices utilize magnetic fields. The Magnetically Assisted Statistical Assembly (MASA), which was created at the Massachusetts Institute of Technology (Cambridge, Mass.), uses magnetic layers deposited onto devices and into substrate recesses. These magnetic layers caused the devices, which are slurried over the substrate, to adhere in the substrate recesses. The Magnetic Field Assisted Assembly (MFAA), which was created at the New Jersey Institute of Technology (Newark, N.J.), uses an external magnetic field to position the devices in substrate recesses. A third method being pursued at the Institute of Microelectronics (Singapore) uses an array of permanent magnets, placed under a substrate, which then drives devices that have been coated with a magnetic layer into substrate recesses by vibration.
These techniques are not free of important and limiting issues. While MFAA is deterministic, MASA and the method that employs the array of permanent magnets are statistical and do not guarantee 100% yield. Additionally, the methods suffer issues with respect to frustration and cross-interference as devices compete to reach a substrate recesses and the actual placement of devices into substrate recesses (for example, devices may enter substrate recesses in various angles and orientations which may be impossible to correct without additional and costly assembly steps).
The objective of the present invention is to provide a method for the parallel (simultaneous) processing of various devices. The method is capable of both manipulating individual devices and processing a large number of devices simultaneously, deterministically, and non-statistically. The method eliminates the issues with frustration and competition between devices and recesses and offers the ability to correct errors thereby achieving 100% yield. It is a technique for assembly that does not place special geometric or material constraints upon the devices or the substrate recesses. The only special preparations required are a layer of magnetic material deposited on the surface of the devices and a bonding agent applied in the substrate recesses. Also, because it is a room-temperature process, materials with different lattice and thermodynamic properties can be integrated with out damaging the devices and substrates.
In addition to the ability to assemble devices, the proposed method can be used to manipulate the material properties of a substrate by assembling onto it components with passive functionality. For example, components that manage stress and strain and that alter electrical and optical properties in semiconductors. The proposed method can be used to form composite materials by assembling different layers of materials onto a substrate.
In reference to
The upper chamber 105 of the enclosure 100 contains the array of electromagnets 110 along with a cooling mechanism, such as a heat sink or a heat bath, and various other leads connected to an external control unit that programs each and every electromagnet of the array 110. The array of electromagnets 110 produces localized magnetic fields (depicted by 205, see
The shield 115 separates the upper 105 and lower 135 chambers of the enclosure 100. Its purpose is to protect the devices 120 from damage that may be caused by direct physical contact with the array of electromagnets 110 and to provide a smooth surface across which the devices 120 are moved by electromagnet array 110. Optionally, the shield 115 may be deformed either by mechanical or electromagnetic mechanisms to facilitate the process of assembly.
The enclosure 100 contains the assembly chamber 135 where the devices 120 are injected, positioned, and assembled.
The top surface of the devices 120 is to be coated by a layer of soft magnetic material 125; the thickness of the layer 125 is to be such that it allows the devices 120 to be securely suspended against gravity by the force of magnetic attraction generated by the array of electromagnets 110. The action of the electromagnet array 110 causes the devices 120 to be held against the underside of the shield 115 and to be moved across that surface. Also, the devices 120 may be rotated into any required orientation by manipulating the field of the array of electromagnets 110 (depicted by 320, see
The devices 120 are introduced into the assembly chamber 135 through the injection ports 160 and are suspended against the underside of the shield 115 by the force of magnetic attraction produced by the array of electromagnets 110. The electromagnet array 110 produces a localized magnetic field 205 that surrounds each device 120 on the shield surface. This field 205 is generated at certain rates, along certain paths, and draws the devices 120 from the injection ports 160 to desired locations immediately above matched recesses 145 on the template 140 (also 310, 315, see
Once a device 120 is moved above its desired location 145 on the template 140, it is disengaged from the shield 115 by weakening the local magnetic field 205 below a certain threshold value. Each device 120 requires a minimum strength of the local magnetic field 205 to keep it suspended against the shield 115. If the local magnetic field 205 around the device 120 is weakened below that threshold, the device 120 falls from rest in the absence of an atmosphere under the action of gravity 130 from the shield 115 to the template 140 without deflection.
The template 140 is a magnetically passive construction upon which the devices 120 are temporarily placed in desired locations 145. The orientation of the template 140 can be altered, and the distance between the shield 115 and the template 140 can be varied (as depicted by
The interface 155 between the injection ports 160 and the assembly chamber 135 is a partition that maintains the integrity of the vacuum within the enclosure 100.
Injection ports 160 around the perimeter of the enclosure 100 connect the assembly chamber 135 to bins (not depicted in the drawings) that contain the devices 300 prior to assembly (see
To complete the assembly process, the template 140 may be fully or partially populated by the devices 150 (see
If a device 120 is not placed in its desired location within a certain allowable tolerance, then an error-handling process is activated. The template 140 is raised up to the shield 115 and the array of electromagnets 110 activates above those devices 120 that were not properly placed. Once the devices 120 are reattached to the shield 115, they may be re-positioned until they are placed within tolerance. Also, devices 120 that were not properly placed may be omitted entirely from the integration step by a mechanical (or electrical) manipulation of the template 140 prior to the substrate 400 being put into contact with the template 140.
The process of assembly along with error-handling is automated by the external control unit. The rate at which the devices 120 are injected into the assembly chamber 135, the rate at which they are placed onto the template 140, the initial injection locations, the final desired positions, and the paths taken, along with such parameters as the device weights and dimensions, are aspects of the process that are monitored and manipulated by the external control unit. The simultaneous processing ability of the assembly and its high (100%) yield capability are achieved by controlling these parameters via real-time sensing and/or feedback.
