Patent Application
20120241427
1. Field of the Invention
The present invention relates to the field of laser processing methods and systems, and specifically, to laser processing methods and systems for laser processing multi-material devices.
2. Description of the Related Art
Lasers can be used in the processing of microstructures in memory and integrated circuit devices. For example, laser pulses can be used to ablate conductive links or link portions in a memory device, such as DRAMs in order to substitute working redundant memory cells for defective memory cells during memory manufacture.
Recently, the use of new materials, such as aluminum, gold, and copper, coupled with the small geometry of these devices, have made the problem of link removal more difficult. Economics and device performance goals have driven the size for the DRAMs and logic devices to very small physical dimensions. Thus, it can be increasingly difficult to irradiate a target structure without damaging surrounding components such as the substrate and adjacent circuitry and links. Furthermore, as more links need to be processed for a given area of semiconductor circuitry, the time required to process a given die increases.
When a single laser pulse or burst of pulses is used to irradiate and sever each link designated for removal, the beam path of laser pulses may move relative to the substrate during the process of irradiation in an “on-the-fly” link blowing process. This relative movement may include moving the substrate and/or moving the beam, although substrate motion on an X-Y stage in conjunction with a vertically oriented and stationary beam is a currently common approach. In conventional laser processing systems, groups of arrayed microstructures are processed. The array may be links in a row, links in closely spaced rows, links in staggered rows and similar regularly spaced arrangements. The conventional processing is generally carried out with either an energy on demand system (e.g. pulse equalization) or an energy picking system (e.g. pulse picking). In the energy on demand system, an irradiation period is timed to coincide with a moving target and the processing rate is limited by a minimum period between energy on demand irradiation periods. In the energy picking system, the laser is pulsed in a continuously repeating sequence at a predetermined repetition rate (e.g. at a q-rate, pulse rate, or burst rate) and the arrayed microstructures in a group are moved synchronously with the repetition rate so that energy is available to process any microstructure in a particular group. The processing rate is limited by a period associated with the maximum repetition rate, and an acousto-optic device or other optical switching device blocks energy from reaching the substrate except when processing a selected synchronized target.
The conventional energy picking process is illustrated in
In a conventional processing system links and pulses are synchronized. T1 and the period of the laser pulse repetition rate (e.g. the pulse to pulse period of a q-switched laser controlled by trigger signals from the control computer 14 ) are made equal. With this method, a pulse is available to process every link. Pulses that are synchronized with links to be processed, such as links 200a, 200d, and 200f of
It will be appreciated that the time required to process a given set of links within a group of a row or a column of links is approximately the number of links times the time period T1, which in these systems equals the laser pulse repetition rate. If the laser used has a maximum pulse rate of 50 kHz, for example, completing the pass of the beam across the 11 links of
For further reference, the following co-pending U.S. applications and issued patents are assigned to the assignee of the present invention, describe many additional aspects of laser link blowing, and are hereby incorporated by reference in their entirety:
According to some aspects, a method of processing material of device elements by laser interaction is disclosed. The elements may be distributed at locations about a workpiece. The method may include generating a pulsed laser processing output along a laser beam axis, the output comprising a plurality of laser pulses triggered sequentially at times determined by a pulse repetition rate, generating a trajectory relative to locations of device elements designated to be laser processed, the trajectory comprising a motion profile of an optical system axis intercept point at the workpiece, driving relative motion of the intercept point and the workpiece along the trajectory; predicting the position of one or more designated device elements relative to the intercept point position on the trajectory at one or more laser pulse times, deflecting the laser beam axis relative to the optical system axis to sequentially offset focused laser spots from the intercept point within a predetermined deflection range based on the predicted position, wherein the predetermined deflection range defines a deflection field shape having at least four extended regions about a central deflection region, and irradiating the designated elements with pulses from the laser output at the offset laser spots.
Other methods of processing selected material within a deflection field by laser interaction where the material is distributed at locations about a workpiece are provided. One such method may include storing data representative of a selected processing field shape that comprises a portion of the deflection field, storing timing data representative of a laser pulse that can be emitted for processing material, storing data representative of one or more workpiece locations selected for processing, moving the workpiece relative to the deflection field, predicting the position of the one or more workpiece locations selected for processing on the moving workpiece within the deflection field at one or more laser pulse times based on the stored timing data, comparing the predicted position of the one or more workpiece locations with the selected field shape; and preventing laser interaction at any one of the one or more workpiece locations that is not within the selected field shape.
Another such method may include storing data representative of a selected processing field shape that comprises a portion of the deflection field, storing data representative of one or more workpiece locations selected for processing, moving the workpiece relative to the deflection field, comparing the position of the one or more workpiece locations in the deflection field with the selected field shape, and preventing laser interaction at any one of the one or more workpiece locations that is not within the selected field shape.
Systems and apparatus for processing material with laser interaction are also provided. In one embodiment, a laser processing apparatus for processing material of device elements by laser interaction where the elements are distributed at locations about a workpiece is provided. The system may include a laser source configured to generate a pulsed laser processing output along a laser beam axis, the output comprising a plurality of laser pulses triggered sequentially at times determined by a pulse repetition rate, a positioning system configured to carry the workpiece and generate a trajectory relative to locations of device elements designated to be laser processed, the trajectory comprising a motion profile of an optical system axis intercept point at the workpiece, the positioning system being further configured to drive relative motion of the intercept point and the workpiece along the trajectory. The system may also include a system controller configured to receive data corresponding to array elements designated for processing, and predict the position of one or more designated array elements relative to the intercept point position on the trajectory at one or more laser pulse times, and at least one beam deflector configured to deflect the laser beam axis relative to the optical system axis to sequentially offset focused laser spots from the intercept point within a predetermined deflection range based on the predicted position. The predetermined deflection range defines a deflection field shape having at least four extended regions about a central deflection region. Also provided is a processing lens configured to receive a deflected beam and to focus the deflected beam so as to irradiate the designated elements with pulses from the laser output at the offset laser spots.
In another embodiment, a system may include means for generating a pulsed laser processing output along a laser beam axis, the output comprising a plurality of laser pulses triggered sequentially at times determined by a pulse repetition rate, means for generating a trajectory relative to locations of device elements designated to be laser processed, said trajectory comprising a motion profile of an optical system axis intercept point at the workpiece, means for driving relative motion of the intercept point and the workpiece along the trajectory, means for predicting the position of one or more designated device elements relative to the intercept point position on the trajectory at one or more laser pulse times, means for deflecting the laser beam axis relative to the optical system axis to sequentially offset focused laser spots from the intercept point within a predetermined deflection range based on the predicted position, wherein the predetermined deflection range defines a deflection field shape having at least four extended regions about a central deflection region, and means for irradiating the designated elements with pulses from the laser output at the offset laser spots.
In another system for processing selected material by laser interaction where the material distributed at locations about a workpiece, the system may include a laser source configured to generate a pulsed laser processing output along a laser beam axis, one or more beam deflectors defining a deflection field, means for moving the workpiece relative to the deflection field; and a system controller. The system controller may be configured to store data representative of a selected processing field shape that comprises a portion of the deflection field, store timing data representative of a laser pulse that can be emitted for processing material, store data representative of one or more workpiece locations selected for processing, predict the position of the one or more workpiece locations selected for processing on the moving workpiece within the deflection field at one or more laser pulse times based on the stored timing data, compare the predicted position of the one or more workpiece locations with the selected field shape, and prevent laser interaction at any one of the one or more workpiece locations that is not within the selected field shape.
In another such system, the system may include a laser source configured to generate a pulsed laser processing output along a laser beam axis, one or more beam deflectors defining a deflection field, means for moving the workpiece relative to the deflection field, and a system controller. The system controller may be configured to store data representative of a selected processing field shape that comprises a portion of the deflection field, store data representative of one or more workpiece locations selected for processing, move the workpiece relative to the deflection field, compare the position of the one or more workpiece locations in the deflection field with the selected field shape, and prevent laser interaction at any one of the one or more workpiece locations that is not within the selected field shape.
Multi-axis inertialess beam positioning is used to access processing targets relative to the trajectory of a mechanical positioning system to sever conductive links at high rates. Various laser processing aspects using split and/or deflected beams are disclosed in US patent publication 20090095722. This document is incorporated herein by reference and forms part of this application. The present disclosure is primarily directed to rapid access with a single beam. In particular, the approach uses high speed positioning within a two dimensional random access field that moves along a trajectory relative to the wafer. Positioning laser spots within the field at a processing rate allows flexible access to links passing through the field along the trajectory with a throughput exceeding a conventional link pitch based processing rate. Elapsed time traditionally required for passing over unprocessed links can be reduced, a higher percentage of laser pulses are used for processing, and processing throughput can be increased.
Generally, the position of each blast in this scheme is determined by a combination of mechanical stage position (the nominal spot position along the trajectory) and a spot displacement. A stage carrying a target substrate moves along a processing trajectory, and periodic laser blasts are fired along the trajectory to process selected targets on the substrate. For each selected target, a control unit determines the exact time of a corresponding laser blast. The control unit also computes a spot displacement relative to an aligned field position for the blast using the target coordinates and stage coordinates that correspond to the blast time. An inertialess beam deflector deflects the laser beam axis according to the spot displacement and the laser is commanded to fire at the specified time, so that the laser spot is positioned on the target when blast is issued.
In this way, efficient processing is unencumbered by traditional assumptions about target locations such as regular target spacing, row allocation, and target orientation. Moreover, stage velocity can be selected over a continuous range of values to optimize throughput without the traditional constraints of matching a laser pulse rate to a uniform link pitch and the attendant trade-offs. The present approach allows for higher stage velocities and provides considerable flexibility so that arbitrary link placement can be handled as well as traditionally structured layouts.
As shown in
In at least one embodiment, detectors may be included in the system illustrated in
A system control architecture shown in
Many aspects of this invention are largely independent of laser material interactions and processing energy windows for various regimes of lasers and pulse types. These aspects relate primarily to improved beam positioning and throughput, however to the extent that positioning accuracy is improved or new types of lasers or new modes of operation are used, some aspects may be process related. In general, beam positioning aspects of this invention, using high-speed positioning within in a two dimensional field moving along a trajectory, can apply to many different types of laser processing.
Laser source (1) generates a laser processing output (3). In at least one embodiment, the processing output includes processing periods 3 as shown in
Output from the laser source is directed along a beam path to the input aperture of a first acousto optic beam deflector AOBD 7. As shown in
Acousto-optic beam deflectors may also be referred to as acousto-optic Bragg deflectors, acousto-optic deflectors (AOD), acousto-optic devices (AOD) or acousto-optic modulators (AOM). Any one of these terms applies to a Bragg regime deflector. AOBD and AOD are considered synonymous and generally refer to devices optimized for variable deflection. AOM usually refers to a Bragg cell that is optimized for high extinction and high efficiency as an amplitude modulator, however over small ranges with varied frequency input an AOM can provide variable beam deflection. The specific construction of the device in various configurations such as, off-axis designs, phased array, alternate materials etc. may be used as beam deflectors in this invention. Other types of acousto optic devices, for example variable filters, may also be considered as deflectors in some cases. It will be understood that any variable deflector operating in the Bragg regime is considered an AOBD for the purposes of this disclosure. Deflectors with similar or superior characteristics may be used in various aspects of this invention, for example deflectors that provide decreased access speed, increased time bandwidth product, improved efficiency, more addressable spots, or reduced beam distortion. Alternate deflectors may be improved AOBDs, electro-optic deflectors or any other type of high speed inertialess deflector.
It will be appreciated that each AOBD is designed for a specific wavelength and that the center frequency will correspond to a different deflection angle for different laser wavelengths. In the case of an optical system designed for different wavelengths, accommodation may be required for differences in deflection angle when the laser source wavelength is changed. In at least some embodiments shown in
It will be appreciated that AOBDs are driven by specialized RF drives (102, 103) that are capable of supplying multiple frequencies to the active deflector cell. Considerations for the RF driver include thermal stability, frequency range, stability and resolution, output power range stability and resolution, number of simultaneous frequencies, frequency switching time, modulation bandwidth, dynamic range, intermodulation, and signal to noise ratio. Drivers may be available as suitable versions from AOBD manufactures or custom as electronic modules.
In a preferred arrangement, four amplified DDS channels (A, B, C, and D in
For two axis deflection, the AOBD 1 may itself be a two axis device with multiple transducers on a single acousto-optic crystal or multiple AOBDs each with its own transducer or transducer array, such as AOBD 1 and AOBD 2 may be used to provide beam deflection in two axes as shown in
It is to be understood that each AOBD will generate a zero order, non-deflected beam in addition to the desired deflected beam. As a matter of routine, design, the zero order beams are fully attenuated for example with a knife edge. The spaced-apart layout provides access for separate knife-edges such as beam stops (9) and (12) or each deflection axis and prevents unwanted energy from the zero order of the first AOBD from propagating into the second AOBD. Other types of beam attenuators are possible, for example in polarization active AOBDs, polarizers maybe used to attenuate zero order energy. In addition to zero order beams, other undesired higher or lower diffraction order beams may be present and may be attenuated in a conventional manner.
Following first and second AOBDs, beam conditioning optics may be employed in the beam path, for example polarization control optics such as a Liquid Crystal Variable Retarder (17) which may be used to adjust polarization according to target type or link orientation as described in U.S. Pat. No. 6,181,728. The beam path may include relay optics (13) to modify the deflected output beam for entrance to the LCVR, for example to fit a well collimated beam into a limited active aperture. These relay optics may further image the pupil of the second AOBD to an intermediate image plane (15) and may provide further anamorphic optics in an anamorphic beam path arrangement.
Following the first and second AOBDs and beam conditioning relay optics, the image of the deflector pupil is expanded. A pre-expander relay (16) may reimage the deflector pupil, for example the intermediate image (15) of the deflector pupil described above to the input pupil of the system beam expander (19). As described in the 20090095722 publication, a beam expander, preferably a zoom beam expander is used to image the deflector pupil or an image of the deflector pupil to the entrance pupil of the processing objective (20). Position of the zoom beam expander can be used to adjust the deflector pupil image location at the objective pupil to improve telecentricity, and might be adjusted to different axial positions to improve telecentricity of either deflection axis. Beam expander optical groups, for example 3 groups as described in 20090095722, may be driven in linear motion precisely using Nanomotion HR2 piezo drives and MicroE Mercury 2 encoders. As the beam expansion is changed the beam diameter at the objective lens changes, and hence the spot size in the field changes accordingly.
This process will be explained with reference to
The number of focused spots that can be addressed in the range of the deflector over the field will be constant regardless of the beam expander setting. So, there is a direct trade-off between spot size and field size with small spots over a small field and larger spot over a larger field. In conjunction with the processing lens, methods according to U.S. Pat. No. 7,402,774 can be used to provide a range of field sizes and spot sizes without degradation of the spot over the field.
Preferably the processing lens (20) is a high numerical aperture objective lens of at least NA 0.7 to provide spots as small as 1.4 microns or 0.7 microns for the processing wavelengths 1064 nm and 532 nm respectively. The objective lens is preferably mounted on an air bearing (21) and translated axially according z height positioning commands as described in U.S. Pat. No. 6,483,071. Preferably the lens will have a working distance of 6 mm or more to avoid contamination from processing debris and to provide mechanical clearance. The lens may be achromatized to provide spot formation with broadband fiber laser sources or for imaging with auxiliary through the lens viewing equipment. Preferably, the lens will have a field of view of at least +−20 microns with the smallest spot setting and largest input beam. Preferably the field of view will be at least +−80 microns for the largest spot setting. Most preferably, the field of view will be +−80 for small spots and +−500 micron for large spots. Preferably, the field will be a flat field with a field curvature less that 10% of the spot depth of focus. Field flatness may be for example 0.1 micron over +−20 microns.
Generally, the field of view of the lens is circular and the deflection field shape is addressed within the lens field of view. The deflection field accessed can be selected as the entire lens field of view, or any portion of the lens field of view. This may be a circular truncation of a superscribed square deflection field, an inscribed shape such as an inscribed square or a partially truncated deflection field. The deflection field when using AOBD positioning is limited by the maximum number of spots available from each deflector. In some cases, for example with small spot sizes, the addressable field may be smaller than the lens field of view.
The wafer substrate (22) with links to be processed is mounted on a wafer chuck for processing. The spot formed by the objective impinges the surface of the wafer. The chuck is carried on a stage or mechanical positioning system (23) according to any of the well-known mechanical positioning configurations. One such configuration is the 2 axis fine stage supported by an air bearing that travels over a 2 dimensional portion of a wafer as found in GSI Group model M550. For this type of system, full wafer coverage is accomplished by stepping a beam delivery system in increments over the wafer and sequentially processing small areas of the wafer with fine stage motion. Alternately, full travel single axis stages in stacked or split arrangements or other configurations and various combinations including galvanometer positioning as know in the art can be used as the mechanical positioning system. Regardless of the particular mechanical positioning configuration, the mechanical positioner moves the substrate relative to a nominal laser beam axis to provide mechanical positioning of targets in a processing trajectory.
Mechanical positioning may also include auxiliary mirror based deflection to provide improved dynamic performance. This has been implemented in the form of galvanometer based field scanning and more recently using a two axis fast scan mirror for stabilization. Yet another approach to improve dynamic performance of mechanical positioning is the use of force cancellation technology, for example as described in U.S. Pat. No. 6,144,118. With force cancellation, mechanical system perturbations and resultant mechanical positioning errors are minimized.
Coordination of laser pulsing, selective pulse picking for blasting selected links, spot displacements to access positions in the deflection field and mechanical stage motion is generally achieved using a system controller (401). The controller is used to generate laser trigger timing signals, pulse picking commands, spot displacement commands and stage positioning commands.
Preferably, the controller generates trigger timing signals that fire laser pulses at a substantially constant repetition rate either continuously or for a minimum interval prior to blasting to provide uniform pulse energy. Conventionally, the trigger timing signals often correspond to link positions on a regular pitch at a particular stage velocity. However, in the present invention, trigger timing signals merely correspond to a position along the mechanical trajectory that will be defined as a virtual link position. The virtual link position represents a position along the trajectory that would be blasted without a commanded displacement. However, with a displacement command, the blast is deflected to the desired blast location at the real link with an offset from the virtual link location. With a constant PRF and a constant velocity along the trajectory, the virtual link locations can generally be regarded as conventional links aligned in along a row on a regular pitch with typical laser timing requirements.
Laser triggering may be initiated by a comparison of the current position of the laser beam axis relative to a target coordinate so that when the position of the laser beam and a virtual link position coincide, accounting for a known lag in the firing sequence, the laser is triggered and the blast is fired to process the target link at the displaced offset position. Alternately, blast times can be scheduled in advance to coincide with virtual link positions according to a planned trajectory and associated blast displacements.
Processing blasts are fired by gating the triggered laser pulses according to pulse picking commands with an optical device (such as AOM 5 of
The system controller (401) also controls blast displacement relative to the trajectory and provides offset commands and deflection signals to position blasts within the AOBD field. With the use of a deflection field, the controller may generate commands that result from a combination of both time and position processing domains. Displacement can be calculated based on set blast times, blast time can be set based on set displacements, for example if only a limited set of deflections is available, or both blast time and displacement can be set in combination. As a result of the flexibility of this approach, blasts may be fired without either regular target spacing or regular pulse spacing.
Stage positioning commands control the stage motion and position the targets with high precision along the trajectory. Position errors measured or characterized during the trajectory can be accommodated in different ways. For example, errors in either axis can be corrected with corresponding adjustments within the beam deflection field by the AOBDs. When the instant blast position is known to a high accuracy, this method of correction can be used in both constant and non-constant velocity processing. For errors in the direction of mechanical motion, small changes in the timing of scheduled blasts can also be used to correct blast position.
System operation is managed by a control program (400) that executes process steps and issues control signals. The program may require operator input or may run automatically to process single substrates or batches of substrates. The program may reside in a storage medium integrated with the system, may reside in a removable medium or may reside at a remote location for downloading of one or more steps to the system. The control program executes processing steps that result in laser processing of unrepaired memory devices to sever selected conductive links and thereby increase the yield of functional memory devices on one or more semiconductor substrates.
In at least one embodiment, processing occurs along a processing trajectory using a sequence of trajectory segments that position the virtual link positions, rather than the real link positions, relative to an aligned beam position. As shown in
Field access in the inertialess deflector field can include a general position offset that can be any combination of position either along or across the processing trajectory direction. With the capability to offset pulses along the processing direction, correction for measured position errors is an inherent feature. For scheduled blasts, adjustments to laser firing time are not strictly required. However, in some cases timing correction may be used to closely match current processing methods, or may be used in conjunction with inertialess access based error corrections.
In at least one embodiment, referring to
Generally, with current device layouts, links are formed in rows running thought the central axes of a die. Different local geometries may be used for example as shown in
Various optimizations known in the field of acousto-optic deflecting can be applied to the design and selection of the AOBDs used in various embodiments of the present invention. In at least one embodiment using a 1064 nm laser source, the AOBD selected is Crystal technologies model AODF 4090 1064 nm with a Te02 crystal, 90 MHz center frequency, 35 MHz bandwidth is used operating from 72.5 MHz to 107.5 MHz to generates 116 milliradian to 173.2 milliradian of beam deflection prior to beam expansion. For use at 532 nm AODF 4110 may be used. Preferably the 532 nm deflector is modified so that the beam entrance and exit are the same as for the 1064 nm version by adding a wedge so it fits easily into the path optical without major redesign and a common optical platform can be configured to operate at multiple wavelengths. Other vendors for AOBD devices include NEOS, Isomet and Sciner, and the devices may include alternate crystal materials and different constructions such as longitudinal mode, shear mode, and phased array devices among know AOBD device configurations.
Generally, an approach using spherical optics and round beams is preferred when a limited number of spots provide an adequate field of view and rapid access time is desired. For example, a 40 micron wide field including twenty-five 1.6 micron diameter spots maybe generated with the TeO2 device described above. For wider field systems, an anamorphic beam path can be used with an increased acoustic window dimension along the deflection axis. Generally this will increase in the number of spots that can be addressed, roughly proportional to the increased size of the in acoustic window and with a corresponding increase the access time required to fill the longer acoustic window of the AOBD. With TeO2, the shear mode acoustic velocity is 0.656 mm/μs, so an increase of 10 mm to the acoustic window would add about 15 microseconds to the access time. Increased access time will in effect reduce the maximum PRF. This effect is a result of the so called time-bandwidth product of the AOBD.
US Patent Publication 20090095722 among other art describes some aspects of AO design and optimization. Embodiments include the use of various AOBD types including on axis and off axis configurations. The AOBDs may be used to generate simultaneous spots, to generate rapid changes in spot shaping, to split a beam into various configurations having multiple spots along and across a row of links.
As discussed, a simple arrangement of stacked AOBDs can be used to provide two axis deflections. This configuration has the advantage of a short optical path length and a limited number of optical components. Disadvantages include beam spreading across the acoustic window of the second device due to the deflection range of the first upstream device. The deflection point is different for each axis which can affect telecentricity at the target surface. Compensation can be provided by adjusting the image location of each deflector with relay optics as described in the 20090095722 publication.
Preferably, deflectors are spaced apart with relay optics. In this arrangement, the window of the first AOBD is imaged on to the second AOBD. Advantages of this arrangement include the ability to pick-off the zero order beam from the first AOBD before the second AOBD, the elimination of beam spreading across the second deflector window and maintenance of a single deflection origin point and for telecentric spot imaging in the processing field.
In a preferred embodiment, from the laser output aperture to the processing field, a total of five relays are used. The laser output is imaged to the first AOBD with a first relay lens. Next the first AOBD is imaged to the second AOBD with a second relay which may be for example a pair of lenses spaced according to focal lengths (i.e. a 4 f relay) to achieve a 1× magnification. The second AOBD is imaged with a third relay, which also may be a spaced lens pair, to an intermediate image plane. An optional beam rotor may be located in the optical path of this relay. The intermediate AOBD image is imaged to the input of the zoom telescope relay with a fourth pre-expander relay that may be a spaced lens pair arranged with a magnification to fill the entrance pupil of the zoom beam expander relay. The LCVR aperture may be located in a collimated region of the optical path of the fourth relay. Finally, the zoom telescope relays the input pupil with variable magnification to the objective lens. Thus, the laser beam waist is imaged to AOBD 1, and AOBD 1 is imaged successively to AOBD 2, an intermediate image plane, the entrance pupil of the zoom beam expander and the objective lens in a manner that accommodates an optional beam rotator and a polarization controlling LCVR.
Conveniently, one turning mirror may be located at the intermediate image plane following the second AOBD (not shown) to provide field adjustment without translation. In this case the turning mirror is in the image of each deflector to provide alignment by way of a field angle offset without translating the pupil image.
In operation multiple relay deflection and imaging system may be characterized by the following typical performance parameters:
AOBD deflectors are diffraction based devices, and the deflection angle is linearly related to the ratio of the grating period in the Bragg cell to the wavelength of the processing beam. If the wavelength of the light entering the deflector is changed, the deflection angle exiting the deflector changes proportionally. As noted in publication 20090095722 and U.S. Pat. No. 7,466,466, diffraction effects can have undesirable effects that can affect the performance of a laser processing system.
Some lasers have very narrow emission spectra, which means very little spread in the deflected beam due to dispersion. However, some lasers, such as fiber lasers may have spectra that are more than an order of magnitude greater than rod based lasers, for example. When used in an AOBD, an increased spectral bandwidth in a laser source can result in undesirable spreading in the spot image and result in an out of round spot shape. In addition, chromatic focusing can further degrade imaged spot quality.
As described in US Patent Publication 20090095722, pre-dispersion gratings and prisms can be used to offset lateral effects of broad band laser sources. Preferably however, laser sources will have sufficiently narrow line widths to avoid spot shape and focus distortion. Advances in fiber lasers have resulted in fiber lasers with line widths narrowed for efficient conversion by way of frequency doubling, for example lasers described in US Patent Publication 20090016388. This type of fiber laser can be used to preserve the advantages of a fiber laser source including temporal pulse shaping capability, while at the same time providing minimal dispersion and defocus artifacts in an AOBD based system.
One aspect of AOBD optimization is the speed at which different position commands can be realized in a deflector according to the RF frequency applied to the AO crystal.
Another related aspect of AOBD optimization in random access positioning is the duration of an applied RF defection signal. Duration of the applied RF, using the optimized lead time, can be varied while deflection efficiency or other parameters are measured. In this way a minimum RF deflection period can be determined for any particular AOBD device. The minimum RF period in conjunction with the minimum lead time can be used to further optimize a laser processing sequence.
In a laser processing system the stage performance can be limited by many constraints such as maximum velocity, edge of travel, and thermal loading. Acceleration and the resulting g-force applied to moving substrates may be limited by coil current constraints or by dynamic considerations. Generally, for high-speed positioning, the stage is light weight and dynamically stiff to maintain high precision without substantial mechanical deflection. Relaxation of constraints can be achieved in part by considering aspects of precision machine design. For example applying forces along the center of gravity to avoid induced deflection and optimizing machine geometry to minimize Abbe errors. In general, while need for high speed positioning persists even with use of inertialess deflectors, the length of mechanical trajectory and therefore its duration may decrease significantly when several trajectory segments are “merged” together by processing their corresponding links in a single run.
Management of constrains and resulting stage performance can enjoy benefits of an inertialess deflection field. With deflectors and an objective with an appreciable field or view, the field of view can be used at the edge of the stage travel to access edge positions while the stage is offset from the edge. This may allow for modification in management of edge link groups, associated trajectory segments and motion parameters. For example, velocity can be arbitrarily slowed rather than incrementally slowed, especially near the stage edge, while maintaining a constant PRF. High velocity can be used on links that might otherwise be too close to the edge of field. In some cases the addressable field of the stage can be increased by the field of view of the objective. For example a 50 mm stage field with a 1 mm deflector field would be able to address a 51 mm square target area. Conversely, stage field can reduced while accessing the full field with the deflectors, for example a 49 mm stage field with a 1 mm deflector field could address links over a 50 mm square area.
Adjustment of the mechanical field and accessible field can have profound effects to enhance throughput. In one example, marginal links may barely miss fitting into a processing field. Considering the tiling of the entire wafer into rows and columns of processing sites, the ability to increase the processing field even if only buy 100 microns, may allow a row and/or column to be eliminated from the wafer processing cycle removing the associated overhead of one or more processing sites, which is significant. Additional mechanical margin around a stage positioning field can allow more aggressive high speed positioning.
Generally, system calibrations with be performed on a periodic basis with certain calibration supplied at the factory, at system installation, at system turn-on, at wafer loading, for each processing site or during a processing sequence. Longer calibration periods are generally desired and may be associated with systems having increased stability, performance and reliability.
Generally, system alignment will include conventional alignment techniques such as edge scanning of reflective alignment targets to achieve overall system positioning accuracies to 150 nanometers or less. A nominal AOBD field position such as the center frequency position can be used for the alignment routine. Of course, other positions may be used, for example field positions that are relatively low drift locations in the field. Multiple positions can also be used to add data redundancy or to include field calibration capability. As described in 20090095722, acousto-optic deflectors can be used in conjunction with target alignment scanning. For example, multiple points of an alignment feature edge within an inertialess deflection field can be sampled and averaged. Utilizing the extremely high bandwidth of AOBD, iterative edge scanning can be performed at high rates. Various combinations of stage motion and AO field scanning are possible.
Within the AOBD field, alignment targets may be L-shaped, square or other shapes can be scanned in both x and y axes without additional mechanical positioning steps. Alignment targets can be scanned on-the-fly during processing trajectory when they fall near link groups and can be traversed within the AO field as the field passes the alignment target.
In typical alignment scanning, first the alignment target is found to low precision with a pre-scan. Once the alignment target is located, high precision scanning over relatively short scan lengths is possible. With an appreciable AO deflection field, the pre-scanning process scan be achieved on-the-fly while the stage approaches the alignment target area, perhaps during a deceleration segment. On-the-fly pre-scanning can potentially eliminate associated overhead.
Mechanical positioning can be slowed or stopped for alignment with AO target scanning. This is especially attractive while scanning targets to determine focus characteristics in the z axis. While stationary, vibrations are reduced, thermal loading is minimal, and dynamic errors are eliminated. It is recognized that with high speed target scanning for focus, increased bandwidth in z positioning is attractive, for example, using axial piezo positioners to move the objective lens over a small range.
Routine field calibration may include calibration of static errors and slowly drifting errors by measuring fiducial positions in sufficient quantity spatially and temporally to determine correction values that can be applied to positioning commands in order to maintain positioning accuracy within a predetermined tolerance range during a processing operation. A typical tolerance range would be less than 10% of the size of target feature such as the width of a conductive link and less than half of the overall system accuracy. Preferably, the tolerance contributes only a minor fraction of the overall tolerance budget, for example 25 nanometers or less. Well-known techniques such as correction table generation and polynomial fitting can be applied. Recalibration periods can be determined with a combination of theoretical models and conventional system accuracy diagnostic routines. Calibration data may be generated during alignment scanning. For example, an AOBD field dimension may be calibrated by scanning multiple edges with a known separation or a single edge at different mechanical positions.
Acousto optic field scale may be determined theoretically based on a range of applied RF frequencies applied, may be measured in the beam path as a deflection angle or beam position, or in the processing field with field calibration features. Deflectors may be calibrated independently or preferably in combination in a 2 dimensional field.
Skew of a deflector relative to inertialess beam positioning coordinates can be adjusted by mechanical rotation of the deflector or rotation of one of more beam rotators. However, generally calibration of a 2 dimensional field will accommodate small residual skew errors resulting from mechanical mounting tolerances.
Generally, inherent linearity of AOBD deflections over small ranges of 10 to 100 spots across the field provides sufficient accuracy. However for improved accuracy, especially when a large number of spots across the field are used, linearity correction can be applied, for example using a correction table to transform real field positions to error corrected positions.
Compensation for variations of AOBD efficiency (AOBD efficiency is the ratio of pulse energy exiting the AOBD to the pulse energy entering the AOBD) by adjusting RF input power levels with field position is a well-known technique. Theoretical models can be used to predict efficiency performance versus angle and generate correction values; however each AOBD can have varying efficiency characteristics. As a result, efficiency characteristics, as shown in
However, AOBD efficiency versus angle also depends on the RF power level, so simple efficiency measurement at a static RF power level may be inadequate to accommodate this non-linear efficiency characteristic. Therefore, a more sophisticated correction scheme is needed. Dynamic measurements can be made by adjusting the RF level to match measured values to an efficiency target value over a range of selected deflection angles to generate an RF power versus deflection angle correction function for the efficiency target value. Alternatively, iterative measurements can be made across the deflection range for a nominal efficiency target value, starting with an initial RF correction function, determining residual efficiency errors versus angle based on efficiency measurements in subsequent steps, and generating an improved RF correction function using the residual error values. Other procedures may be used to accurately calibrate efficiency versus field angle such as generating an efficiency look-up table over the desired deflection and efficiency range. However, techniques that minimize data management overhead, such as determining sets of characteristic curves are preferred, especially when considering complexities of 2 axis deflection described below.
Modulating the RF power in an AOBD can be used to control optical attenuation. However since the efficiency curves change for different attenuations as shown in
For 2 axis AOBD deflection using a pair of deflectors, calibration is required in each deflection axis. The efficiency of the second AOBD is dependent both on its own deflection angle and the angle of the beam entering from the first deflector, so it needs calibration over the additional variable of input angle. The dependence of calibration at different attenuation values applied in either AOBD makes the task of simultaneous deflection and attenuation with an AOBD pair complex. Attenuation can be applied in the first AOBD, the second AOBD or both AOBDs and the ability to effectively provide calibrated attenuation across the two dimensional deflection field is an important consideration. In a preferred calibration routine, the first AOBD is calibrated in the dimensions of deflection angle and optical attenuation value, and the second AOBD is calibrated at a single efficiency target value versus the variable input angle and output deflection angle. Calibration of the second AOBD is not dependent on optical energy of the beam, so attenuation can be provided in the first AOBD without compromising either calibration of the second deflector or calibration over the 2D field. In this case each AOBD is calibrated over two variables and the data intensive burden of calibrating the second AOBD over three variables is avoided. Of course an additional AOM can be used to provide variable optical attenuation and further relax the calibration requirements of the AOBD deflectors.
Energy Calibration with Detector Configuration
In at least one embodiment, a detector 25 may be situated after the first AOBD and before the second AOBD 11 as shown in
In addition to providing beam deflection and attenuation, the AOBD can split the laser beam using 2 or more frequencies simultaneously in the acousto-optic crystal to deflect portions of the input to multiple angles. When beam splitting is used to generate multiple simultaneous spots, energy calibration is further complicated. Not only does the calibration need to account for two axis deflection and attenuation in multiple AOBDs, the calibration must also account for the balance or prescribed split of energy and the separation angle between split beams in at least one axis. When possible, single beam positioning is preferred, however aspects of beam splitting may be advantageous in certain circumstances to achieve high throughput rates.
One method used to measure pulse energy for the above calibration methods and other system routines includes use of an energy detector such as an in field integrating sphere and photodiode. This type of detector can measure single spot energy and the combined energy of multiple closely spaced spots. However, measuring individual spots from a group of multiple split spots is difficult when spots are closely spaced, for example spaced on the order of several to 10's of microns. In this case a pick-off at or near the spot image plane is required, which is difficult to achieve at this scale. However calibration for split-beam processing requires energy measurement of at least one, and preferably all split beams. Considering that efficiency calibration in AOBDs is dependent on the RF level applied, it is desirable to operate the AOBD at operating RF levels for direct energy measurement and calibration while splitting the beam.
In at least one embodiment, reflected energy is measured from various targets at the spot image plane in the processing field. By scanning split spots over a target such as an edge, independent energy measurement is possible even for closely spaced spots. However, at full processing RF levels, pulse energy can be high enough to damage the reflective targets. To remedy this and allow the AOBD to operate at full RF power for accurate calibration, an upstream attenuator can be used to reduce split pulse energy to an acceptable level where calibrations targets are not damaged. Since total energy of the split beams can be measured with the in field detector, absolute power measurement of each split beam is not strictly required. Relative measure of each spot's energy in conjunction with the total energy can be used to determine each spot's absolute energy. Generally the split ratio or energy balance is the primary calibration concern. This relaxes the requirement of the upstream attenuator so that a non-damaging energy range can be set for calibration with reflective targets without requiring a precise upstream attenuation adjustment.
Splitting the laser beam successively with a pair of AOBDs in the optical path generates an N×M array of spots. As illustrated by
In some cases, the objective lens may have residual field curvature and an annular field can be addressed. In this case it is preferred when splitting and directing the beam axis to two rows, to dispose the lens axis relative to the row position such that focus height of each row falls within the annular field and preferably on a focus common plane as shown in
Beam steering with AOBDs may be used to calibrate other pointing errors introduced in the optical system. For example, motion of zoom beam expander elements or other optical elements can generate repeatable pointing errors. Correction of repeatable pointing errors can be accommodated with pointing corrections applied with the AOBDs. In the zoom beam expander example, AOBD can be used with an appropriate correction look-up table to maintain pointing accuracy through the zoom range as spot sizes are changed.
Considering the complexity and subtlety of multi-axis AOBD calibration, there may be characteristic deflection field regions that can be more accurately and reliably calibrated and regions that are less accurately and less reliably calibrated. Analysis of field calibration fidelity can be used to identify preferred areas within a calibration domain. A laser processing sequence may be generated to use these preferred areas while avoiding other areas in the calibration domain. In effect, a sweet spot of field calibration is identified and exploited for increased processing performance. For example, characterization of AOBDs may identify angle ranges where efficiency has good linearity especially regarding variable RF power ranges used for attenuation. Even when performance is acceptable across the entire field, a selected portion of the field may be used for the convenience of limiting calibration requirements. A combination of trajectory planning and blast sequencing within the deflection field can be used to effectively avoid areas having lower performance or use only calibrated areas. The field portion or portions used should access all laterally offset blast locations and include sufficient length in the direction of motion to accommodate large scale pulse timing adjustments (e.g. link phase adjustment).
Sub-field shape may also accommodate shapes such as ring fields. For example when the objective lens has residual field curvature, an annular sub-field may be selected to limit processing to areas of best focus. Useable width of such an annulus may depend on spot size, for example a narrower annulus with smaller spots. Diameter of the sub-field annulus may vary with target distance. Other focus characteristics, such as irregular variations in focus or spot quality over the field of view may be used to determine sub-field shape selection.
As discussed in 20090095722, multiple frequencies can be used simultaneously for spot shaping. In a multi-axis AOBD system, shaping can take place in either axis to provide very rapid, pulse to pulse spot shape orientation. In a group of links having mixed orientation, this would allow spot shaping in concert with random access. Spot shaping can be extended to multiple spot dimensions, for example to rapidly form more square spots shapes or change the effective spot size in a sequence of pulses. These techniques might be applied for preheating, cleaning or other multiple pulse processing regimes.
One method of processing closely spaced links uses bursts of sub-pulses fitting within an envelope to allow standard constant motion substrate positioning while the burst is applied to a link. The length of the burst must be short enough to avoid so called pulse smearing effect whereby movement of the spot position during the burst exceeds a positional tolerance and compromises the energy window of the laser process. Aspects of U.S. Pat. No. 7,394,476 are directed to compensating for relative motion between a link and a burst of sub-pulses so that long burst periods can be used without adversely affecting the processing window.
With implementation of a fast inertialess two axis addressable field, further improvements in burst type processing are possible. Without reducing the processing rate, by processing multiple rows or other dense groups of links in a simultaneous trajectory, the velocity of the spot relative to links can be reduced. For example, if 4 rows are processed with a single spot, then the relative velocity of the link and spot can be reduced as much as 4 times. At slower relative velocities, longer bursts are possible without using link tracking techniques. For example, a 500 ns long burst may be the limit in high speed positioning systems that do not employ link tracking. However, when the relative velocity is reduced by a factor of 4, the burst length can be proportionally increased up to 2 μs. To the extent that the AOBD access time permits, longer bursts can be used without affecting throughput.
Application 20090095722 incorporated in its entirety, describes many aspects of link processing with AOBD scanning that may be used in the current invention. In one embodiment a scan axis is tilted in relation to wafer motion, for example tilted at a 45 degree angle. Among other benefits, tilted scanning can allow high speed access in multiple axes with a single inertialess scanner, spot shaping along a link, alignment with staggered link arrangements and control of telecentricity error. In other embodiments, an acousto optic device is thermally stabilized by driving with a near constant RF power.
Embodiments of the present invention using further aspects from published US patent application 20090095722 may include asynchronous processing; that is to say the product of link pitch times velocity may not correspond to the PRF. In at least one embodiment, for improved throughput, all links processed and unprocessed, will pass through the processing field at a rate exceeding the PRF with improved utilization of available pulses that are directed to links selected for processing. Processing may include mixed pitch layouts of links, for example moving along a trajectory a constant velocity and processing a variety of link pitches. Mixed phase is also possible, where groups of regularly spaced links may not be laid out on an overall regular pitch. Mechanical pitch phase adjustment from group to group can be accommodated with the inertialess deflectors. Channeled processing as described in US Patent Publication 20090095722, using a discreet set of deflections may be beneficial when a limited number of RF frequencies are available for rapid switching. In this case a preselected frequency corresponds to each discreet processing channel. These processing regimes among others, deviating from tradition equally spaced links in a single row, can be applied to various layouts of single or multiple rows with the benefit of inertialess positioning.
Two-axis AOBD positioning provides a convenient way to correct for either positional or temporal errors in a link blasting process. Positional errors, measured, calculated or estimated can be summed with two axis deflector position commands to correct the errors on a pulse by pulse basis. In addition, AOBD positioning can be used along the trajectory path to correct for temporal errors and delays, such as trigger timing adjustments. In much the way that convention laser processing systems correct position with temporal adjustments of the laser firing time, a blast firing error or adjustment can be accommodated with a corresponding position adjustment in the direction of travel.
Various error correcting aspects of AOBD positioning may, in some cases, allow higher dynamic positioning speeds where positional errors are increased and compensated. Also, since AOBD positioning with error correction can eliminate the need for pulse to pulse timing corrections, constant laser repetition is possible. Instabilities resulting from irregular pulse timing are therefore eliminated and stable laser pulse energy can be supplied, potentially at increased pulse rates where error adjustments are made in the AOBD positioning command.
Error correction can include predetermined errors that have been characterized and are applied by the controller to correct for known, planned or expected positioning occurring errors. Error correction may include estimated errors where a parametric model is used and based on process parameters an error is estimated for correction. Errors may also be measured directly in real time for correction.
Error limits may be used as input for trajectory optimization. For example a trajectory may be planned to keep errors within a range that can be corrected in the field of inertialess deflectors or within a specified tolerance band. Actively measured errors can be monitored and modifications to trajectory can be made when the measured error exceeds a predetermined level. For example, velocity may be slowed to maintain errors within a correctable range when a target error limit is approached or exceeded.
Aspects of beam rotation are generally described in Published US application 20090095722. The beam rotation can be used with single axis deflection to accommodate 2 dimensional field access in a polar coordinate fashion. In this case, as is well-known the output beam rotation angle is two times the beam rotator angle. When 2-axis deflection of a single beam is used, the system may be configured without a beam rotator and skew errors resulting for rotational misalignments of deflection axes can be calibrated out with a coordinate transformation. However, it may be desirable to include one or more beam rotators even when single beam two-axis deflection is used. This may also be used, for example, in conjunction with beam splitting. When beam splitting, the orientation of the plane of the split will be determined by the rotational orientation of the deflector along the beam axis. Of course, each deflector might be rotated directly, or a beam rotator might be used to align the deflecting and split axis with alignment feature or targets to be processed in the addressable field. With multiple deflectors, it is possible to use multiple beam rotators so that each deflector can be independently aligned. In practice, deflection axes can be relatively aligned to acceptable tolerance, for example so that field axes are orthogonal. In this case only a single rotator is used to adjust the orthogonal deflection field skew to mechanical beam positioning coordinates. The beam rotator can be any type known such as a Pechan prism or Dove prism, however, in a preferred arrangement; a K-mirror with three first surface mirrors is used. The K-mirror essential provides a large aperture hollow dove prism that can rotate one or more deflection axes without using large blocks of transmissive material. Advantageously, one or more reflective surfaces of the K-mirror can be adjusted to null out beam pointing and or beam offset errors. Such a K-mirror may be manually operated or may be motorized for automated adjustment or rotation. The K-mirror may be removable from the beam path and may be replaced with fixed path optics arranged to maintain axial beam length along the beam path.
Conventional processing systems such as the GSI Group M550 include a coarse stage movement for stepping the laser beam axis relative to the substrate from region to region. Stepping may be from a single device to a single device, from a part of a device to a different part of a device, or from a processing site that includes more than a single die to a different processing site. The coarse stage remains stationary during processing. While the coarse stage remains stationary, the fine stage positions the wafer relative to the beam axis according to a trajectory planned to process selected links in the local region of the wafer. When the trajectory is complete the coarse stage steps to a new region. The time penalty of repeated steps, lockdown of stepped optical components and alignment is offset by high-speed positioning of the wafer with the fine positioning stage.
Yet another conventional system uses a pair of long travel stages in a split stage architecture. One axis moves the optical axis while the other axis moves the wafer. A first axis is stepped to a location corresponding to one or more rows of links on the wafer. The orthogonal axis is then scanned at high velocity, generally along rows across the entire wafer and alignment may include may dice across the wafer. This provides for long stage motions at velocity, but heavy stages limit acceleration capabilities between link groups and at the edge of the wafer.
Other configurations are possible with various combinations and permutations of substrate and beam positioning to produce relative motion between target structure and the processing spot. Regardless of the configuration, generally coarse movement will be associated with relatively infrequent high inertia positioning. Coarse movement, especially considering acceleration and deceleration, can generate system perturbations. These perturbations may include for example mechanical vibration, center of gravity shifting, thermal loading, air turbulence, and electrical noise. In a step and settle regime, perturbations are allowed to attenuate over a settling period, and processing proceeds when a predetermined performance level in achieved. Various methods can be used to mitigate system perturbations as are known in the field of precision engineering. For example, force cancellation as disclosed by Cahill, et al. in U.S. Pat. No. 6,144,118 can be used as a means to mechanically counter acceleration forces. Moving mass can also be used to maintain balanced static loads on isolated support systems.
Some form of fine positioning is generally used for link processing to provide sufficient bandwidth for a high throughput system. As mentioned, a small travel fine stage can be used in conjunction with a large travel coarse stage. The fine stage may be for instance a 50 mm×50 mm travel moving magnet stage supported on a planar air bearing. In this case the coarse stage addresses the full wafer, which may be a 300 mm diameter wafer, in increments of 50 mm or less. With long travel linear stages covering the entire length of the wafer a fast steering mirror has been used to provide high bandwidth error correction.
Methods and systems of the present invention can be characterized as a superfine positioning providing access over a small field, generally smaller than a single die and larger that a single link, that can position laser blasts within the field on a blast by blast basis. In addition to throughput improvements, a superfine positioning system can correct dynamic errors, control relative beam to target velocity, and split a beam to multiple superfine positioned beams.
Conventionally, trajectory planning is largely independent of spot size and there is no deflection field to consider. However, when there is a deflection field and the dimension of the field can vary, as shown in
During a trajectory segment, links selected for processing enter and subsequently exit the deflection field. As the field moves relative to the substrate, links can be addressed and blasted at different positions in the deflection field from the point where a link enters the field to a point where the link exits the deflection field. The range of positions in the field where links can be blasted is in effect a spatial buffer that can include multiple addressable links at different positions when a laser pulse is available for blasting. Based on the size of the deflection field and the relative velocity between the substrate and the field, there is an associated time interval during which a link selected for processing dwells in the deflection field. A link can be blasted by any one of a number of different pulses in a pulse sequence that occurs over the interval. Therefore a deflection field of appreciable size can be considered as either a spatial buffer or a temporal buffer. During relative motion of the deflection field and the substrate, unprocessed links can accumulate in this buffer for processing with available pulses before exiting the deflection field. A maximum PRF of the laser source will limit the number of links that can accumulate in the buffer (not considering multiple simultaneous beams),
Various advantages of link buffering in a two axis deflection field can be used for trajectory planning. As a spatial buffer, leading or lagging links can be sequenced according to preferred trajectory scenarios. As a temporal buffer, link blasts can be advanced and delayed to provide improved laser utilization. In some cases, the buffer size may be exceeded and unprocessed links can processed during subsequent, partially overlapping passes. For example links from isolated dense groupings of links can be deferred and processed later in areas adjacent to relatively sparse processing areas.
A diagonal deflection field allows a single high speed deflector to process links spaced apart in different axes, for example Cartesian X and Y axes. Processing on the diagonal allows system operation without requiring different modes of operation for different axes as may be required when switching from an x offset to a y offset (e.g. modifying deflection orientation with a beam rotator or selecting from branched optical paths). Errors resulting from reconfiguration and subsequent requirement for recalibration are avoided. As show in
In conventional link processing systems, the laser processing rate is simply the substrate velocity divided by the link pitch. In terms of actual links processed, an effective link processing rate over a processing segment can be calculated by multiplying the conventional processing rate times number of links processed divided by number of links traversed. Generally, a fraction of links is processed and the resulting effective link processing rate is low compared the PRF.
With more efficient processing and higher relative motion velocities, the effective processing rate can be increased. One measure of link processing efficiency for a link group is the number of processed links (LP) divided by the total number of laser pulses (PTotal). The upper efficiency limit is 1 when LP=PTotal and all pulses are used to process links. Various embodiments disclosed provide for increased efficiency and therefore a higher link processing rate.
At a conventional processing velocity, throughput can be increased by simultaneous processing of multiple rows and shortening the overall trajectory by eliminating multiple passes over the rows. In the case where multiple links require processing at the same time, either the beam can be split to provide multiple processing spots or a preceding or subsequent laser blast can be used out of sequence with a spatial offset in the field along the direction of travel to blast the link. The blast selected might be the nearest available blast either preceding or following the nominal blast time, but other blasts can be used. To the extent that blasts are available, this can provide a doubling of throughput when 2 rows are processed simultaneously or a factor of N when N rows are processed simultaneously.
One aspect of random access inertialess positioning is the ability to perform laser processing at velocities different from conventional velocities and increase the effective processing rate. If the local density of links to be processed within the addressable field exceeds 1/N links per column, then there may not be enough available blast times. In this case, translation speed of the substrate can be slowed to provide more blast times until there are sufficient pulses available for complete processing. When the velocity is reduced, the random access field allows an arbitrary velocity to be used with correction to most if not all pulses. In a conventionally synchronized system a slowdown would be limited to an integer increment to maintain synchronous processing, e.g. ½ speed or ⅓ speed etc.
Not only can velocity be slowed for high local densities, but velocity can be raised for low local density. As disclosed in US Patent Publication 20090095722, various types of buffered processing such a channeled processing and asynchronous processing can be used to increase velocity. Within the limit of various constraints, such as maximum travel velocity and random access field size, velocity can be increased until the average blast density in time matches the process repetition frequency and all accessible blasts are used. This can apply to multiple rows as well as single row processing or randomly placed targets.
Another possibility to manage high link densities is to designate some links for processing in a subsequent pass. For example, if three rows are to be processed, rather than slowing velocity to process all links in a single pass, one row such as the middle row could be partially processed in a first pass and completed in a second pass. This technique may be especially useful when the spacing of a desired odd number of rows to be processed exceeds the random access field size. For the above example of three rows, rather than processing 1 row and 2 rows in separate passes, each pass can include essentially 1½ rows and average density can be managed to some extent when assigning a processing pass to links in the split row.
A number of different parameters can be used to calculate a processing trajectory velocity or starting values in iterative velocity optimizations. For example an average number of links the in field, an average link pitch, a constant sum of link velocities within the field, a rate of links entering the field, or a rate of links exiting the field may be used to calculate a processing velocity. Likewise a comparison of parameter values may be used, for example the difference between the numbers of links entering and exiting the field may trigger an increase or decrease in velocity to accommodate a respective depleting or accumulating number of links in the addressable field.
Other factors affecting a velocity or an acceleration value may be set based on predetermined parameter values, such as permissible levels of system perturbation.
In some cases, especially where travel velocity is determined by system constraints, the width of the field accessed relative to the direction of travel may be selected based on the velocity. For example a number of rows or width of the processing field accessed may be determined based on a desired effective processing rate at a predetermined velocity. Other factors affecting choice of width selected may be AOBD efficiency, orientation of links or rows, process window optimization or trajectory optimization.
In some cases the length of the field accessed relative to the direction of travel may be selected based on velocity and other factors. For example, a shorter length may be selected for use with reduced velocities or increased length may be used with increased velocities. Other factors may include AOBD efficiency, orientation of links or rows, process window optimization or trajectory optimization.
Pulse by pulse deflection can be used based on rapid position sampling and prediction of the optical system axis intercept point on the wafer at future pulse times. For example stage position encoders may be sampled at about a 3 MHz rate or about every 350 nanoseconds to provide dense position data that is used to accurately estimate the intercept point position at a planned pulse trigger time. For example, with laser pulse repetitions near 300 KHz, the fast sampling rate provides position data much faster than laser pulses are used for processing. Thus position estimates can be generated at and well above the laser repetition rate and up to the sampling rate, so accurate predicted positions are available for each pulse. An accurate predicted intercept point position can be used to generate corrected deflections relative to the intercept point for each pulse and may be generated, for example, in much less than the 3.3 microsecond time period between laser pulses for a 300 kHz laser.
The lead time afforded by predicting the intercept point for an upcoming pulse and rapidly generating corrected RF deflection signals generally accommodates the time required for AOBD acoustic wave set-up. Within each AOBD, there is a characteristic acoustic delay time for the RF generated acoustic wave to propagate through the acoustic crystal to fill the acoustic aperture used for beam deflection. So, the laser spot offset from the intercept point and the associated RF frequency and RF amplitude must be determined in advance of the laser pulse, which may be on the order of 10 microseconds. The delay depends on the acoustic crystal material properties (acoustic velocity) and the AOBD crystal geometry. When high repetition lasers are used such as lasers pulsed at greater than 100 KHz, the pulse repetition period may be less than the acoustic delay time. In one implementation of the invention, rapid sequential pulse transmission can be accommodated by stacking RF pulses in the AO crystal. For example, at about 300 KHz, three RF pulses may simultaneously propagate in the AO crystal and the RF generation may be several pulses ahead of the laser pulse. This aspect is illustrated and described with reference to
In these deflection systems, position prediction at future laser pulse times can ensure spot placement accuracy at high scanning speeds.
At a given time, a predictive processing sequence may be initiated as represented by 1503. The predictive processing may include predicting an X,Y coordinate of a position ahead. The predicted position is an accurate position based on a generated trajectory. The generated trajectory includes a motion profile of the optical system axis intercept point at a workpiece. The sequence may subsequently calculate relative deflection distances dX:dY along each axis for the link to blast based on the predicted nominally deflected position. These deflection distances may thus reflect the offset position of a deflected beam from the predicted intercept position. The offset position dX:dY may then be converted to frequencies Fx:Fy for the AOBDs to deflect the beam to based on the determined offsets. Subsequently, efficiency for beam transmission may be determined as represented by TRx and TRy to determine the appropriate RF energy to apply to the AOBD at the selected frequencies. Look-up tables or formulas may be used in order to determine RF frequency values and amplitudes corresponding to the amount of deflection desired and the desired pulse energy for blasting a link.
As represented by 1504, the predictive processing sequence may include a comparison of the offset position (dX:dY) with a field shape. At 1505, the system may determine whether a link blast should be executed with this pulse based on the comparison of (dX:dY) with the field shape. If the offset position lies outside of the deflection field for links under consideration for blasting, the system may determine that the laser pulse should not be used for link blasting. For example, the laser pulse may be left un-deflected and picked-off, attenuated or deflected to a dump position where no processing of links occurs. If the position is within the field shape, the sequence may continue to 1502 to initiate AOBD control for laser pulse 1502E. As illustrated in
The AOBD acoustic wave may require a predetermined amount of time to enter the AOBD deflection window. For example, this time is represented as a 5 μs propagation time to begin entering the AOBD deflection window as will be described in greater detail with reference to
The offset positions dX:dY may be compared with a particular field shape at decision block 1604. The particular deflection field shape, or parameters of the field shape, may be stored in a shape map as illustrated by block 1630. The method may load the coordinates of the deflection field from the shape map 1630 and compare the offset positions dX:dY with the loaded coordinates. If the offset positions are within the field shape, the method proceeds to block 1605 by initiating the deflection of the laser beam. The method may initiate the deflection by filling an AO window with an AOBD acoustic wave as will be described with reference to
If it is determined that the offset positions dX:dY are not within the field shape at decision block 1604, the method proceeds by determining whether the link to be processed is past the field shape at decision block 1608. The offset positions may be outside of a field shape in one of four possible positions. The offset position may be outside of the shape on either side, or laterally, with respect to the trajectory. The offset position may also be before or past the deflection field along the trajectory. The system may check whether the beam and corresponding shape is past the offset position of the link to be processed along the trajectory. If the beam and corresponding deflection field are past the offset position, the method may determine whether the link position to be processed should be deferred to a next processing pass at decision block 1609. If the link cannot be deferred to a next processing pass (for example, the system will not make additional passes in the vicinity of this link position), the method produces an error output. If the link can be deferred, the method determines whether all processing has been done at decision block 1610. The processing may be done when all links to be processed have been processed. If the processing is not done, the method may loop back to block 1602 to load one or more additional blast coordinates at block 1622. The blast coordinates may correspond to a link position to be blasted at a time corresponding to a future laser pulse as discussed above.
If it is determined that the offset position is not past the field shape at decision block 1604, the method may loop back to block 1603 where new offset positions dX:dY may be calculated.
In at least one embodiment a generally four arm star sub-field shape provides pulsed laser processing in AOBD deflection regions of a two-axis addressable deflection field. Within the sub-field shape, laser spot position and transmitted laser pulse energy can be rapidly and accurately corrected. Along each AOBD deflection axis, different drive frequencies result in corresponding deflection angles. Generally, AOBD transmission efficiency accounts for angular variation in diffraction efficiency, variation in applied RF power and RF driver to Bragg cell power transmission through the drive frequency range. The AOBD transmission efficiency is high near the center frequency and drops off substantially at extreme field positions. In 2 axis deflection for overall efficiency, transmission efficiency for each axis is essentially multiplied. The corners of a deflection field can have very poor efficiency that would be difficult to accurately correct. Multi-axis deflections with non-linear transmission efficiency in each deflection axis require generation of complex corrections, for example higher order polynomials or multi-variable polynomials that can require significant computation and attendant time. However some portions of the deflection range exhibit relatively flat efficiency response. The flat areas, generally near the center of each deflector range, can be used to simplify 2-axis efficiency correction by deflecting one axis only within a narrow flat portion of the deflection range as the orthogonal axis is deflected through a larger range. The narrow field in one axis in combination with the wider deflection range in the other axis can be applied in different orientations along each orthogonal axis. When superimposed, two narrow axial fields may combine to generate a star shape with four branches aligned along each AOBD deflection axis.
Another consideration in field shape is scan lens distortion that requires correction to avoid spot positioning errors. Lens distortion can increase non-linearly with field position with highest distortion at the corner of the field. It is desirable to limit the magnitude and complexity of distortion correction needed to provide accurate spot positioning. The star field shape provides reduced distortion magnitude by eliminating the extreme corners and with narrow deflection near each deflection axis, relatively simple distortion correction can be used avoiding more complex multi-axis corrections that might for example result in higher residual positioning error. It will be appreciated that there may be coupling from the focus axis (z) into x and y axes. As such, focus contributions as well as other measured or characterized position error sources can be accommodated with beam deflection.
Simplified beam deflection and energy corrections over reduced field areas as discussed above can be applied more rapidly than complex corrections over larger areas. When the corrections can be determined and applied in less than the pulse to pulse period of the laser, correction can be applied on a pulse by pulse basis.
One advantage of simpler corrections is expected higher stability over time and lower sensitivity to perturbations. For example, by limiting or eliminating high order terms, non-linear drift contributions are reduced and corrected deflection performance is stabilized over time.
The star shape provides symmetry for substrate positioning in 2 full axes while limiting the area of the field and limiting the requirements for various calibrations and corrections, such as limiting use of high order polynomial terms, cross terms, slow calculations (e.g. non-integer) and overly sensitive or unstable corrections. The shaped field dimensions may accommodate offsets in the direction of substrate motion for tracking links in each column of an array of links as well as 2 axis offsets for applying laser spot position corrections based on measured or predicted link positions at the blast time.
In at least one embodiment, a portion of the star field shape provides a wide cross-axis field, generally along array columns transverse to substrate motion and a narrow on-axis field generally along array rows and in line with substrate motion. This embodiment may access many rows simultaneously with available laser pulses while addressing only one or a few columns, for example with array scanning and relatively uniform positioning velocity.
In at least one embodiment, the star field shape provides a narrow cross-axis field, generally along array columns transverse to substrate motion and a wide on-axis field generally along array rows and in line with substrate motion. This embodiment may be preferred when the array is limited to only a few rows (e.g. <=2 rows) and many columns are simultaneously accessed, for example to accommodate dynamic errors, velocity optimization or beam splitting.
One suitable star subfield shape comprises a cruciform shape, which may be characterized as a plus sign with straight arms, such as shown in
An array of links for processing can be sequenced for high speed processing within a desired field shape while maintaining deflections within a cruciform or truncated compass rose. For a cruciform shape, a raster along each of sequential columns or a random sequence along each column can be used. Generally a column by column processing is advantageous within the narrow field, however if multiple columns fit within the field shape, links may be processed in different columns out of turn in the column sequence.
With the compass rose shape, larger combined x-y offsets can be used near the center, with smaller combined x-y offsets used progressively toward the extremes of each axis. This will be described in greater detail with reference to
For example, as illustrated in
The sequence may effectively be configured as a raster scan modified with the line to line increment occurring at the middle of the line rather than the end of the line. Since the AOBDs are being used to provide random access, there is effectively no retrace required. As explained above, when the array is centered along the cross axis field, symmetric pairs of links can be transposed to modify the sequence, for example by transposing 2 and 5 of
In at least one embodiment, when the relative velocity is the conventional link pitch based velocity (e.g. single row processing, fixed q-rate) is divided by the number or rows, the narrowest useable field dimension for regular link arrays will depend on the virtual link pitch and the array row pitch. The row pitch is the link to link pitch in a row and the virtual link pitch is the row pitch divided by the number of rows when all links are targeted (i.e. not considering velocity optimization to provide increased blast utilization). To process adjacent columns in a row, the minimum deflection is the row pitch less the virtual link pitch (motion between pulses) assuming laser timing adjustment is used to correct for on axis phase and position errors along the rows. If the deflector is used to correct for phase and position errors (e.g. using a fixed laser repetition rate), then the minimum deflection requires an additional row pitch, so the minimum deflection is twice the row pitch less the virtual pitch. The field may be increased to accommodate a mechanical position error band, limit pulse dropout and provide margin for velocity optimization.
For a predetermined scan pattern with adequate margins for mechanical error correction, the scan pattern may fit the field shape such that every available blast is positioned within the field shape. In other embodiments, one or more designated link positions may be compared with the predicted position of the field shape at the next blast opportunity. A blast sequence to initiate pulse picking and processing of a designated link is initiated if one or more compared link positions are within the predicted field shape position. A link that falls outside the shape position remains in the queue for processing with a subsequent blast. When a designated link is passed over, unprocessed, this may flag a system fault or flag the link for processing at a later time. Essentially this comparison based routine operates on a FIFO basis to process links as soon as possible and may be preferred with optimized positioning velocities and when processing in acceleration or deceleration. When multiple designated links are available for processing at the next blast opportunity when using the compass rose shape, links at outer rows where the field is narrow may be given preference. The actual field shape used may be determined experimentally, analytically, theoretically or by approximation and may be stored digitally, for example as a graphical shape table. While constant pulse rates are preferred, the technique is flexible and may be used effectively when uniform pulse energy is available at non-constant rates.
The following table lists some exemplary parameters solely for illustrative purposes and is not to be considered limiting.
It will be evident that the nominal deflections of this table fit the compass rose shape with a 0.5 micron deflection range at the extremities, a 1.5 micron intermediate range and a 2.5 micron range toward the center. In practice, margins would be added to accommodate at least the expected position error band, for example +−1 micron in X and +−2 microns in Y. This margin is reflected as field dimension in the table.
The four arm star sub-field shape may take on any number of possible configurations.
|dY|≦B−A|dX| for |dY|≦|dX|
|dX|≦B−A|dY| for |dY|<|dX|, and
|dY|≦C, and
|dX|≦C Eq. (1)
wherein |dY| represents the absolute value of deflection in a Y axis, |dX| represents the absolute value of deflection in an X axis, B represents a distance from a center of the compass rose shape to the point of a star shape corresponding to a full tapered arm along the X or Y axis, and C represents the distance from a center of the compass rose shape to an edge of the field in the X or Y axis. As illustrated in
|dY*dX|≦D
2, and
|dY|≦C, and
|dX|≦C Eq. (2)
wherein D represents a predetermined deflection field value, and C represents the distance from a center of the compass rose shape to an edge of the X or Y axis as discussed above. As illustrated in
|dX|≦C, and
|dY|≦E,
OR
|dX|≦E, and
|dY≦C| Eq. (3)
wherein C represents a distance from a center of the field to an edge of the field along an X or Y axis, and E represents a distance from a center of the field shape to a start of an arm of the field shape along the X or Y axis.
|dX|, and |dY|≦F,
OR
|dX|≦C, and
|dY|≦E,
OR
|dX|≦E, and
|dY|≦C Eq. (4)
wherein F represents a distance from a center of the field to an edge of a predetermined square having a perimeter of 8 F, E represents a distance from a center of the field to an edge of an arm of the field along an axis corresponding to a width direction of the arm, and C represents a distance from a center of the field to an edge of the field along an X or Y axis. The field shape illustrated in
One of skill in the art will recognize that field shapes other than those illustrated in
Aspects of certain embodiments may be practiced in a single path optical system where all beams are incident on the same set of optical components. In a single path system multiple beams may be offset from an optical path axis propagating with non-collinear beam axes but generally each beam propagates in the same direction in the same sequence near the optical path axis through common optical elements. The non-collinear beams are generally centered with respect to the entrance pupil of the laser processing lens so that beam positioning at each target position in the field of view is telecentric. As shown in
U.S. Pat. No. 6,951,995, U.S. Publication 2002/0167581, and U.S. Pat. No. 6,483,071 disclose systems for beam positioning alignment, splitting, and the like as well as various material processing components, systems, and methods that can be used in conjunction with the inventions disclosed herein. Each of these documents is incorporated by reference herein and foams part of this disclosure.
This application is a Continuation-In-Part of U.S. application Ser. No. 12/976,539, entitled Link Processing with High Speed Beam Deflection, filed on Dec. 22, 2010, which application claims priority to U.S. Provisional Application 61/291,282, filed on Dec. 30, 2009. This application also claims priority to U.S. Provisional Application 61/446,943, filed on Feb. 25, 2011.
| Number | Date | Country | |
|---|---|---|---|
| 61291282 | Dec 2009 | US | |
| 61446943 | Feb 2011 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 12976539 | Dec 2010 | US |
| Child | 13404930 | US |
