Semiconductor device manufacturing: process – Coating of substrate containing semiconductor region or of... – Insulative material deposited upon semiconductive substrate
Reexamination Certificate
2001-06-04
2002-12-24
Ho, Hoai (Department: 2818)
Semiconductor device manufacturing: process
Coating of substrate containing semiconductor region or of...
Insulative material deposited upon semiconductive substrate
Reexamination Certificate
active
06498113
ABSTRACT:
FIELD OF THE INVENTION
The invention relates to a method for the production of a high quality free-standing layer of Gallium Nitride or similar material by heteroepitaxial deposition and subsequent removal from a transparent substrate.
BACKGROUND
Gallium Nitride (GaN) has been recognized as having great potential as a technological material. For example, GaN is used in the manufacture of blue light emitting diodes, semiconductor lasers, and other opto-electronic devices, as well as in the fabrication of high-temperature electronics devices.
One of the greatest challenges for the large-scale production of GaN-based devices is the lack of a suitable native GaN substrate. GaN is not found in nature; it cannot be melted and pulled from a boule like silicon, gallium arsenide, sapphire, etc., because at reasonable pressures its theoretical melting temperature exceeds its dissociation temperature. However, the fabrication of very high crystal quality, thin layers of GaN, and its related alloys, for use in electronic devices, requires that they be deposited homoepitaxially onto an existing GaN surface. Such high quality device layers cannot be directly grown heteroepitaxially, for reasons that are outside the scope of this invention.
The techniques currently in use for the fabrication of high quality GaN and related layers involve the heteroepitaxial deposition of a GaN device layer onto a suitable but non-ideal substrate. Currently such substrates include (but are not limited to) materials such as sapphire, silicon, silicon carbide, gallium arsenide, lithium gallate, lithium aluminate, and lithium aluminum gallate. All heteroepitaxial substrates present challenges to the high-quality deposition of GaN, in the form of lattice and thermal mismatch. Lattice mismatch is caused by the difference in interatomic spacing of atoms in dissimilar crystals. Thermal mismatch is caused by differences in the coefficient of thermal expansion (CTE) between joined dissimilar materials, as the temperature is raised or lowered.
For the purpose of clarity, heteroepitaxial growth is defined herein as a process whereby the atomic lattices of two dissimilar materials are intimately joined together by atomic bonds across their common interface. When the cross-linking bonds are made in a regular and orderly array displaying long-range order, the interface is said to be coherent. When the cross-linking bonds are broken, bent, twisted, or otherwise distorted such that there is no long-range order, the interface is said to have lost coherency. Coherent interfaces are much stronger than incoherent interfaces, due to the greater number of cross-linking bonds between the materials. The loss of coherency may be partial; if only a percentage of cross-linking bonds are broken or distorted in an interface, the interface is partially coherent. The percentage (by area) of broken or distorted bonds represents the level of incoherency or loss of coherency for that interface.
The most commonly used heteroepitaxial substrate for GaN deposition is sapphire (Al
2
O
3
), which has both a large thermal mismatch and a large lattice mismatch compared to GaN. In addition, the sapphire substrate is not electrically conductive, and has poor thermal conductivity, limiting its heat sinking capabilities, further reducing device performance and complicating device processing. For reasons unrelated to the scope of this invention, sapphire otherwise possesses superior properties as a hetero-substrate. However, the large lattice mismatch results in films that have very high defect densities, specifically in the form of dislocations, which are especially undesirable from a device fabrication point of view. (The formation of dislocations at regular intervals along the interface does not affect its coherency, as defined for the purposes of this application, for the dislocations themselves exhibit a type of long-range order in their distribution.) As with other epitaxial crystal growth processes, it is necessary to grow a buffer layer of GaN on the sapphire surface prior to the formation of device-quality layers. The buffer layer will vary, depending on device tolerance to dislocations, whether or not special growth techniques (such as growth through a mask pattern, use of low temperature buffer layers, etc.) are employed, as well as other factors. Typically, this GaN buffer thickness is less than one micron to tens of microns thick. Defect densities, however, predominantly in the form of dislocations, remain high (~10
10
cm
−2
) resulting in diminished device quality. In addition to the conventional buffer layer, a low temperature GaN buffer layer is nearly always used. This layer is the first layer deposited on the sapphire. The buffer layer is initially amorphous and typically is 30-50 nm thick; it is recrystallized at the growth temperature.
Besides dislocations and lattice mismatch problems, thermal mismatch is also a consideration. Typically the GaN is deposited onto sapphire at a temperature of between 1000-1100° C.; as the sample cools to room temperature, the difference in thermal expansion (contraction) rates gives rise to high levels of stress at the interface between the two materials. Sapphire has a higher coefficient of thermal expansion (CTE) than does GaN. As the sapphire substrate and GaN layer cool, the mismatch at the interface puts the GaN under compression and the sapphire under tension. Up to a point, the amount of stress is directly related to the thickness of the deposited GaN, such that the thicker the film, the greater the stress. Above a film thickness of approximately 10 microns, the stress levels exceed the fracture limits of the GaN, and cracking and peeling of the film may result. Cracks in this layer are much less desirable than high dislocation densities, and should be avoided because of the risk of their catastrophic propagation into the device layer during subsequent processing steps.
One method to prevent such thermal stress-related problems involves separating the sapphire substrate from the deposited film. This may be done by physically removing the substrate (lapping and polishing), or by focusing a very high-intensity light source (such as from a laser) from the substrate side of the sample. The light source emits photons having an emission energy that is not absorbed by the sapphire. This second technique utilizes the difference in absorption between the two materials: GaN has a room temperature electron bandgap of approximately 3.45 eV, whereas sapphire has a bandgap of 9.9 eV. Photons with an energy greater than approximately 3.45 eV and less than 9.9 eV (corresponding to vacuum wavelengths less than 359 nm but greater than 125 nm) are able to pass through the back side of a sapphire wafer, where they are absorbed in various amounts, depending on energy, by the GaN at the interface. Once absorbed, the photons are converted to heat, which locally disrupts the Ga—N bonds. If the incident radiation is intense enough, large-scale local disruption results in a complete loss of coherency between the lattice of the sapphire substrate and the GaN. At lower radiation levels, the loss of coherency may only be partial and incomplete, resulting in a film that is still attached to the sapphire substrate, but is no longer completely bonded to it.
Both aforementioned techniques have limitations. A free-standing film must be sufficiently thick to have the required mechanical strength necessary for subsequent device processing. Typically, this requires a minimum thickness on the order of 50-100 microns. Deposition of a crack-free film with this thickness onto sapphire is feasible if done carefully, however thermal stresses will cause severe bowing in the wafer as it cools to room temperature. Conventional lapping and polishing processes are not effective at removing a concave substrate; alternatively, use of a laser to remove the GaN from the sapphire can create unstable localized regions of stress in the partially-removed film, leading to layer fracture during the lift-off process.
Referring to the drawings, F
Miller David J.
Solomon Glenn S.
CBL Technologies, Inc.
Ho Hoai
Hoang Quoc
Isenberg Joshua D.
JDI Patent
LandOfFree
Free standing substrates by laser-induced decoherency and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Free standing substrates by laser-induced decoherency and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Free standing substrates by laser-induced decoherency and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2994773