Patent Application
20050058172
The present invention relates to an excimer or molecular fluorine laser system.
Semiconductor manufacturers are currently using deep ultraviolet (DUV) lithography tools based on KrF-excimer laser systems, operating at wavelengths around 248 nm, as well as ArF-excimer laser systems, which operate at around 193 nm. Vacuum UV (VUV) tools are based on F2-laser systems operating at around 157 nm. These relatively short wavelengths are advantageous for photolithography applications because the critical dimension, which represents the smallest resolvable feature size that can be produced photolithographically, is proportional to the wavelength used to produce that feature. The use of smaller wavelengths can provide for the manufacture of smaller and faster microprocessors, as well as larger capacity DRAMs, in a smaller package. In addition to having smaller wavelengths, such lasers have a relatively high photon energy (i.e., 7.9 eV) which is readily absorbed by high band gap materials such as quartz, synthetic quartz (SiO2), Teflon (PTFE), and silicone, among others. This absorption leads to excimer and molecular fluorine lasers having even greater potential in a wide variety of materials processing applications. Excimer and molecular fluorine lasers having higher energy, stability, and efficiency are being developed as lithographic exposure tools for producing very small structures as chip manufacturing proceeds into the 0.18 micron regime and beyond. The desire for such submicron features comes with a price, however, as there is a need for improved processing equipment capable of consistently and reliably generating such features. Further, as excimer laser systems are the next generation to be used for micro-lithography applications, the demand of semiconductor manufacturers for powers of 40 W or more to support throughput requirements leads to further complexity and expense.
In laser systems used for photolithography applications, for example, it would be desirable to move toward higher repetition rates, increased energy stability and dose control, increased system uptime, narrower output emission bandwidths, improved wavelength and bandwidth accuracy, and improved compatibility with stepper/scanner imaging systems. It also would be desirable to provide lithography light sources that deliver high spectral purity and extreme power, but that also deliver a low cost chip production. Requirements of semiconductor manufacturers for higher power and tighter bandwidth can place excessive, and often competing, demands on current single-chamber-based light sources.
Excimer and molecular fluorine lasers typically utilize a fast avalanche gas discharge to excite the laser gas. A discharge voltage on the order of 20-40 kV typically is delivered to a peaking capacitor of a solid state pulser module, which then delivers a breakdown voltage to the main discharge electrodes that is sufficient to properly excite the gas. In order to optimize the efficiency of this energy transfer, the impedance of the discharge circuit is matched with the relatively low impedance of the gas. The typical voltage pulse applied to the main electrodes shows a rise time of about 20 ns, with a pulse duration of about 50 ns. The solid state pulser module transforms the charging voltage to the required voltage range, and compresses the pulse to obtain the necessary fast rise time at the electrodes.
In a standard excimer laser gas discharge using a circuit such as shown in
FIGS. 8(a)-(c) show schematic views of several electrode arrangements that can be used in accordance with various embodiments of the present invention.
Systems and methods in accordance with embodiments of the present invention can overcome deficiencies in existing excimer and molecular fluorine laser systems by changing the way in which voltage is applied to the discharge electrodes. An improved discharge can help to optimize the temporal shape of the discharge pulse, such as to provide a pulse that is longer, has a lower peak value, and/or is substantially more uniform at the peak.
The stretching of the overall pulse can be a result of the superimposition of gain in the gas medium as a result of the separated discharges. A delay between discharges on the order of about 20 ns, for example, can provide for some “overlapping” of the gain produced by each pulse such that a lengthened gain period is introduced into the gas medium. The temporal separation, or delay, between discharges should be greater than 0 sec in order to stretch the output pulse and lower the output peak power, but less than the temporal duration of one of the discharges, in order to ensure some temporal overlap of the separate pulses. The temporal overlap is necessary to ensure that a single, stretched pulse is generated instead of a plurality of relatively short pulses. The overlap also can be selected such that the overlap of the pulses creates a more uniform peak intensity, as will be described with respect to
Further, a proper combination of capacitor and inductor values used to set the relative timing of the discharges can help to reduce acoustic waves and shock waves in the laser chamber, especially for repetition rates above 4 kHz. Any discharge in the laser gas can produce acoustic and/or shock waves that travel in both directions along the resonator axis. As the discharges are separated temporally and spatially, these waves are less focused and can interfere destructively as they propagate in opposing directions, such that an effective damping of the shock and sound waves occurs. Further, separating the discharge such that each separate discharge has a lower energy dissipation can cause the shock and/or sound waves to have a lower initial intensity, such that less instability is introduced into the laser chamber.
Generating multiple discharge pulses in the resonator chamber, each of which can be controlled in timing, allows for an optimization of the temporal shape of the resulting output laser pulse. Applying the energy to the segmented electrodes rather than to a standard electrode pair also can improve the stability of the gas discharge. This higher stability can be specifically attractive to obtain higher repetition rates. Separated electrode segments can be used with laser systems of any appropriate wavelength, and can be used for applications such as microlithography where an optimization of the laser pulse length is desirable. The use of electrode segments also can be beneficial for high repetition rate lasers and high power level lasers.
A long pulse with low peak power can be desirable for applications such as microlithography where damage to downstream optics can be avoided and/or lessened, and the compaction effect in fused silica can be reduced. A longer pulse also can facilitate line narrowing used for microlithography lasers, and can significantly reduce the amount of amplified spontaneous emission (ASE). Performance at the high repetition rate can be improved by the enhanced stability of the discharge. The use of multiple electrode segments also can affect the spatial distribution of the laser discharge, as the multiple discharges can enable modulation of the effective gain length. A reduction in the effective gain length can be used to further reduce the ASE level of the laser, as the ASE level emitted by an excimer laser typically is relatively high prior to formation of the lasing pulse. The ASE level can drop significantly in the presence of the laser pulse. In order to reduce the ASE during the start of the laser pulse, the effective gain length and the inversion can be reduced during the initial phase of the pulse.
Pre-Ionization
One way to improve the stability of a discharge in an excimer or molecular fluorine laser system 700 is to utilize some means of laser gas pre-ionization 702. Means of pre-ionization are discussed, for example, in U.S. patent application Ser. No. 10/776,137, entitled “EXCIMER OR MOLECULAR FLUORINE LASER WITH SEVERAL DISCHARGE CHAMBERS,” filed Feb. 11, 2004, hereby incorporated herein by reference. Pre-ionization can create a sufficient amount of free electrons and ions in the laser gas to provide for a substantially homogeneous avalanche discharge. In commercial excimer lasers, for example, a corona discharge or spark-UV-pre-ionization can be used to pre-ionize the gas. Placement of the pre-ionizing elements homogeneously to the main discharge can lead to a stable, homogeneous gas breakdown. The pre-ionization can be used to determine the timing of at least one of the discharges in a segmented electrode system, as well as the breakdown voltage. In one embodiment the pre-ionization pulse is applied prior to the arrival of a discharge voltage. The corona discharge can be driven by the same voltage pulse applied to the main electrodes.
Pre-ionization also can be obtained using UV-laser radiation, such as radiation of a wavelength on the order of about 193 nm. Pre-ionization can be applied to each separated discharge, or can be applied only to one of the separated discharges as shown in
An electronic control module can be used to control the timing of a trigger ionization of gases between at least one pair of segmented electrodes. By controlling this trigger ionization, the precise timing of the actual discharge(s) can be more finely controlled. Each ionization control can include, for example, a high voltage (HV) power supply or high voltage pulse generator in electrical communication with an ionization element or electrode 702. There can be a single ionization pulse generator, or one pulse generator for each ionization element. Other ionization configurations are possible, such as separate ionization circuits in series with a high frequency transformer, multiple such circuits in series, or a single such circuit, in order to obtain the appropriate voltage. Proper ionization of the gas can produce a sufficient level of electrons, ions, and charged particles to start an avalanche gas discharge in the entire volume of a discharge gap.
Ensuring sufficient ionization can provide for a “fine” control over the timing of a discharge. Firing an ionization pulse after the electrodes or electrode segments are charged can ensure that the actual discharge occurs with a controlled timing or delay. Even if the charging of the electrodes can vary on the order of about 10 ns, the trigger ionization can be fired after this period of potential variation in order to more accurately control the timing of each discharge. Since the timing of the ionization pulse can be controlled to within about 1 ns, the timing of the discharge then also can be controlled to within 1 ns even if the charging of the main electrodes varies by 10 ns. The ionization can be used to adjust the delay between electrode segments. In an exemplary approach, the ionization can be obtained using a corona discharge component that provides sufficient ionization after arrival of the main voltage pulse. The design and configuration of a corona rod used for trigger ionization in accordance with various embodiments of the present invention can utilize any of a number of corona rod configurations that are presently used in conventional pre-ionization approaches, such as described in U.S. patent application Ser. No. 10/696,979, filed Oct. 30, 2003, entitled “MASTER OSCILLATOR—POWER AMPLIFIER EXCIMER LASER SYSTEM,” hereby incorporated herein by reference. The result of this ionization is a precise timing of the gas breakdown close to the point where the peaking capacitors are charged to a maximum voltage.
The circuitry for the trigger ionization can be separated from the circuitry for the main discharge pulse, such that the timing of the ionization can be controlled independently. The discharges can be synchronized to a higher accuracy than in existing systems, provided that the trigger ionization pulse timing is more precisely controlled than the timing of the main voltage pulse. An advantage of such an approach lies in the fact that requirements on the timing of the main discharge voltage pulse can be greatly reduced. The switching of the ionization can require a fairly low amount of power, such as on the order of tens of Watts or less, such that a fast pulsed source of high voltage can be used without multiple stages of compression and the associated delay uncertainty. Such a circuit can have sufficiently low inductivity and stray capacity, however, in order not to produce displacement current through the corona rod as the voltage on the main discharge electrode rises.
In one embodiment an effective preionization energy can be obtained for a discharge of at least one of the segmented electrodes using a “plasma cathode” arrangement as known in the art, such as the arrangement shown in
Beam Dimension
For various applications it is necessary to have a relatively high power laser beam, while it is desirable to have the energy density and/or power density of the beam relatively low in order to minimize the damage to the optics within, and external to, the laser. One way to obtain these levels is to optimize the geometric arrangement of the electrode segments. The segmented electrodes described above can be arranged in any of a number of different configurations, such as those shown in FIGS. 8(a)-8(c), with the arrangement in x, y, and z directions being dependent upon a number of factors. For example, the x- and z-directions can be determined by the gas exchange requirements of the discharge volume between laser pulses. High repetition rate lasers of 4 kHz and higher can have a narrow x- and/or z-dimension in order to obtain the proper gas clearing at a reasonable gas velocity, such as a velocity on the order of about 30-50 m/s. In one arrangement 800 such as that shown in the top view of
The anode segments also can be at an angle with respect to each other and/or with respect to the resonator axis. For example, the arrangement 802 of
For applications where the gain length of the laser is of less importance than the total gain, this approach can allow a beam to be extracted with increased x-dimension and, thus, reduced energy density. The discharge width can remain small for each of the segments, and the gas clearing requirements can still be met. The use of segment electrodes also allows for different spacing between segments. The electrode spacing can be used to optimize the breakdown voltage of the gas and the performance of the laser. The technique also can be used to generate separated output pulses.
Separation of the Pulses
In order to generate decoupled output pulses from a common high voltage source, a solid state pulser module can be used, such as is described in U.S. patent application Ser. No. 10/699,763, entitled “EXCIMER OR MOLECULAR FLUORINE LASER SYSTEM WITH PRECISION TIMING,” filed Nov. 3, 2003, which is hereby incorporated herein by reference. As shown in
A reset current can be applied to the inductor of each final compression stage 910, 912 in order to provide accurate sub-nanosecond timing control between the voltage outputs for each voltage pulse path, or channel, as driven by the common pulser. The basic approach to introducing variable timing delays between branches or channels of a circuit is described in U.S. Pat. No. 6,005,880, entitled “PRECISION VARIABLE DELAY USING SATURABLE INDUCTORS,” incorporated herein by reference above. Using such an approach with a common pulser system, a reset current component for each channel can apply a separate reset current to the final inductor of each final compressor stage, which can function as a tuning component for the main discharge pulse of each channel. The reset current applied can be determined using a computer or processing component in combination with a mechanism for monitoring the timing of the discharges.
In one embodiment, a reset current supplied to one or both of the final compressor stages can be used to adjust the delay of the corresponding circuit loop. Controlling the individual delay to the final compressor stage for each channel of the system can provide a control of the delay of the output pulse from each final compressor stage.
∫V(t)·dt=N·A·ΔB
where V(t) is the applied voltage, N is the number of winding turns, A is the magnetic core cross sectional area, and ΔB is the magnetic flux density swing. For a square voltage pulse the saturation time is given by
In an exemplary setup, the magnetic core of a final compressor has 4 turns and an operating voltage of 30 kV, whereby the saturation time is on the order of about 100 ns. The number of turns of the inductor of the final compression stage can be selected to adjust the saturation time as needed. For example, the use of five windings instead of four windings can lead to a suitable shift in the delay of about 25 ns.
The saturation flux density Bsat can be reached faster if the core is not completely reset before the pulse to—Bsat. Reasons for variation in the magnetization between pulses can include fluctuations in the reset current, variations in the time between pulses in the burst mode, and magnetization by “reflected” pulses. For each switching cycle, the core can be driven through the magnetization curve, where the pulser current drives the magnetic material into positive saturation and the reset current drives the core back to a defined point on the magnetization curve.
The exact position to which the core is reset on the magnetization curve can be a function of the reset current. With higher magnetization, the magnet will take longer to saturate, such that the forward current will encounter a longer delay. The influence of the reset current on the pulser delay has been found to be several nanoseconds per ampere of reset current. This makes feasible a modulation of the resulting pulser hold-off delay by fine adjustment of the reset current. The reset current can be used to adjust the nominal delay difference between the two discharges. For a stable laser operation in certain embodiments, the difference in the delay of the two discharges is critical and must be stable within 1 ns.
The timing between the multiple discharges can be further controlled by modulating the saturation time for at least one of the inductor cores in the compressor stage(s) of the common pulser module. In order to modulate the saturation time, an additional voltage can be superimposed onto the operating voltage. Once the start condition of the core is reached by applying a reset current, the additional voltage can be applied before the operating voltage. The additional voltage then can begin to pull magnetic flux from the core in order to drive the core towards the saturation point. Using a relatively low voltage on the order of about 10V can lead to a longer saturation time, such as on the order of about 300 μs. Application of this additional voltage for a controlled time prior to the application of the main voltage pulse can allow the core to be set to virtually any point on the B-H curve. For example, the saturation time can be reduced by about 20 ns by applying an additional voltage pulse of 10V for approximately 60 μs.
If the timing shift between the output pulses applied to the various electrode segments becomes sufficiently large with respect to the saturation time, the final discharge pulse can compete for energy. It therefore can be desirable to utilize additional decoupling of the circuits, which can be achieved in at least one embodiment by splitting the winding of the last common compressor core 1102 of the system 1100 as shown in
MOPA Systems
Excimer lasers used in lithography often should be line-narrowed, and should work with high repetition rates above 1 kHz and energy levels between 5-15 mJ. The length of a single laser pulse can be of great importance for such lasers, especially for wavelengths below 193 nm. The short pulses can have a high peak power, which can severely damage the optics of the laser or of a scanner, stepper, etc. MOPA systems can be used to address this problem, as MOPA technology can separate the bandwidth and power generators of a laser system, as well as to control each gas discharge chamber separately, such that both the required bandwidth and pulse energy parameters can be optimized. Using a master oscillator (MO), for example, an extremely tight spectrum can be generated for high-numerical-aperture lenses at low pulse energy. A power amplifier (PA), for example, can be used to intensify the light, in order to deliver the power levels necessary for the high throughput desired by the chip manufacturers. The MOPA concept can be used with any appropriate laser, such as KrF, ArF, and F2-based lasers.
Components of one MOPA laser system are discussed generally in U.S. patent application Ser. No. 09/923,770, filed Aug. 6, 2001, hereby incorporated herein by reference, which discloses a molecular fluorine (F2) laser system including a master oscillator (or seed oscillator) and power amplifier. The master oscillator comprises a laser tube including multiple electrodes therein, which are connected to a discharge circuit. The laser tube is part of an optical resonator for generating a laser beam including a first line of multiple characteristic emission lines around 157 nm. The laser tube can be filled with a gas mixture including molecular fluorine and a buffer gas. The gas mixture can be at a pressure below that which results in the generation of a laser emission, including the first line around 157 nm having a natural line width of less than 0.5 pm, without an additional line-narrowing optical component for narrowing the first line. The power amplifier increases the power of the beam emitted by the seed oscillator to a desired power for applications processing. A power amplifier (PA) typically includes a discharge chamber filled with a laser gas, such as a gas including molecular fluorine, and a buffer gas. Electrodes positioned in the discharge chamber are connected to a discharge circuit, such as an electrical delay circuit, for energizing the molecular fluorine in the chamber. The discharge of the PA can be timed to be at, or near, a maximum in discharge current when a pulse from the master oscillator (MO) reaches the amplifier discharge chamber. Various line-narrowing optics can be used, such as may include one or more tuned or tuneable etalons.
In a MOPA, the oscillator can produce narrow-band pulses with low energies, such as on the order of about 1-2 mJ, and the amplifier can amplify the pulses to pulse energies on the order of about 10-15 mJ. MOPA arrangements can be used with XeCl excimer lasers, for example, which can be used for thin film transistor (TFT) annealing, where annealing energies of above 1 J can be necessary and the stability σ can be under 1%. One potential problem with existing MOPA configurations is that the optics in the amplifier (and any subsequent scanner/stepper) can be damaged as the pulse exiting the amplifier of a typical MOPA system is short but intense, having a relatively high energy level. It can be difficult to obtain effective pulse stretching in a standard MOPA system, as it can be difficult to produce the necessary high repetition rates (>4 kHz).
In various MOPA systems, it can be advantageous to utilize pulse stretching as discussed herein for the power amplifier chamber. Other systems can utilize pulse stretching with the master oscillator chamber, or with both chambers. Another possible configuration for a MOPA system 1200 in accordance with one embodiment of the present invention is shown in
It should be recognized that a number of variations of the above-identified embodiments will be obvious to one of ordinary skill in the art in view of the foregoing description. Accordingly, the invention is not to be limited by those specific embodiments and methods of the present invention shown and described herein. Rather, the scope of the invention is to be defined by the following claims and their equivalents.
The present application claims benefit from U.S. Provisional Patent Application Ser. No. 60/502,073, filed Sep. 11, 2003, which is incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 60502073 | Sep 2003 | US |
