Principles of Light Sources for Lithography

It has been proven that the design lenses which images near the diffraction are limited only over narrow bandwidths in wavelength. This is a consequence of the phenomenon of dispersion where index of refraction changes with the wavelength of the light.

spectrum of color by prism

Figure 1: White light is broken into a spectrum of color by prism

Glass has an optical property by which white light is changed by a prism into a spectrum of separated colors. Moreover, it can even vary the bandwidth of light produced by a line-narrowed laser. Thus, it is necessary for lens designer to include corrections for the varying optical property of glass materials over the bandwidth of light. But there are limits to range of wavelengths over which lenses can be color corrected i.e. maintaining practical size and reasonable costs.

Consequently, high resolution stepper lenses image only over a fairly narrow range of wavelengths. Moreover, intense light sources are needed for cost effective stepper throughput, so exposer systems for optical lithography have operated and will continue to operate at wavelengths for which there are intense sources of illuminations over narrow bandwidths. This has led to a short list of possible light sources for use in lithography. But there are some benefits on the restrictions to particular standard wavelengths-

  • Additional technology that depends upon the wavelength is required.
  • Manufacturers need to develop resists that will perform optimally at the specific wavelengths used.
  • Pellicles need to be optimized for the specific wavelengths used.
  • Limited number of wavelengths at which lithography is practiced enables R&D resources to be well focused.

Mercury-arc lamps and excimer lenses have been the sources of actinic light for nearly all projection in photolithography (actinic light: Radiation that can induce chemical reactions in photoresists is term actinic). The lines of the Mercury spectrum and excimer lasing wavelengths used in lithography are listed in table 1.

Class of light source

Specific type

Wavelength

Location in the electromagnetic spectrum

Mercury-arc

g-line

436 nm

visible

h-line

405 nm

visible

i-line

365 nm

mid-ultraviolet

DUV

240 -255 nm

deep-ultraviolet

Excimer lasers

KrF

248 nm

deep-ultraviolet (DUV)

ArF

193 nm

deep-ultraviolet (DUV)

F2

157 nm

vacuum-ultraviolet (VUV)

Table 1: Light Sources used in Photolithography

The mercury-arc lamp has three intense spectral lines in the blue and ultraviolet portions of the electromagnetic spectrums (i.e. mentioned in table 1) along with some continuum emission in between these spectral lines.

g-line: Basically, mercury g-line is blue light (λ = 436 nm) in the visible part of spectrum. The first commercial available wafer stepper, the GCA DSW4800, was operated at the mercury g-line, as did the first steppers of Nikon, Canon, ASML, and TRE, which are built based upon the same concept with GCADSW4800. (TRE was an early supplier of wafer stepper which later changed their name to ASET and discontinued in the early 1990s)

Stepper lenses designed to meet the extreme resolution and field size requirements of microlithography are only over a very narrow range of wavelengths. For mercury-arc lamp based systems, this has been over the bandwidths of the arc lamps, i.e. 4-6 nm. These bandwidths are much larger than the natural bandwidths of mercury atomic emissions because of the collision (pressure) and thermal (Doppler) broadening that can be considerable in a high-pressure arc lamp operating at temperatures approaching 2000°C.

Some systems are built by Ultratech images over broader range of wavelengths (390-450 nm), but these have resolutions limited to 0.75 µm or larger. This use of multiple wavelengths has significant advantage in terms of reducing standing-wave effects, which Utlratech steppers utilized effectively for lithography at ≥1.0 µm feature size. Unfortunately, the reticle quality of broadband Ultratech lens design, which prints the reticle at 1:1 ratio, has largely prevented the use of Ultratech steppers for critical applications in deep sub-micron lithography. Aside from Ultratech lenses, there are few more lenses imaging at both the g- and h-lines but with limited acceptance.

h-line: While the Mercury h-line was used on a few wafer steppers, most of the stepper manufacturers made a transition to i-line lithography in the late 1980s as the need to print submicron features, while maintaining depth-of-focus ≥1.0 µm, arise.

i-line: Mercury i-line dominated as the leading-edge light source for lithography until the advent of deep-UV lithography in the mid-1990s.

DUV: There is a strong band of DUV emission (λ= 240 – 255 nm) from mercury-xenon arc lamps and these were used on early DUV exposure tools, such as the Micrascan I and Micrascan II from SVGL. Most of the DUV systems today use excimer lasers as light sources with higher resolution versions of Micrascan platforms and a bandwidth requirement of less than 1.0 pm.

Mercury-arc Lamp Structure and Operation

In Figure 2 a fused silica bulb is filled through a tip with a small quantity of mercury and argon or xenon, which, once filled up, is sealed. Operation is initiated by applying a high-frequency high voltage (> 10 kV) across the two electrodes and ionize the insert gas. The resulting discharge evaporates the mercury, and the mercury begins to contribute to the discharge. The plasma, being electronically conducting, cannot support such high voltage, so the voltage drops. A steady lamp output is maintained by operating the lamp at constant current at relatively low DC voltages, i.e. 50V-150V, where high voltage is needed to ignite the plasma. Condensation of mercury on the cooler walls of the bulb near the electrodes is inhibited by reflective coatings. Pressure inside the bulb can exceed 30 atm during operation and catastrophic failure is always a concern. The electrodes are made of refractory metals such as tungsten in order to withstand the internal temperatures that can be as high as 2000°C. Thorium coatings are used to reduce the electrode work functions and provide electrons to the plasma more easily.

Mercury-arc Lamp Structure

Figure 2: Mercury-arc Lamp Structure

During the operation, electrode material is gradually deposited onto the insides of the bulb, which reduces light output. Lamps are usually replaced after several hundred hours of operation in order to maintain high stepper throughput. The probability of a lamp explosion also increases with time and the prevention of such explosion is another reason why lamps are replaced at the end of their rated lives. The two most common mechanism of catastrophic failure are degradation of glass-metal seals or fracture of the glass bulb.

Less than 1% electrical power supplied to mercury-arc lamp is converted to actinic light. The rest of energy is heat that is removed by the air exhaust and cooling water, the air exhaust also serves as a safety function. Improved stability of light output has also been obtained by monitoring and controlling the external lamp temperature which will exceed 700°C. As a practical matter, the lamp base temperature typically between 150°C and 200°C, is more easily measured than the bulb temperature. Temperature control is obtained by adjusting the air exhaust.

Excimer Lasers Light Source

Another light source applied to lithography is the excimer laser. Excimer lasers are much larger and more complicated than arc lamps. Because of their size, installing excimer laser machines inside the cleanroom will occupy floor space, which is expensive. Thus, excimer lasers are usually placed outside of the cleanroom (mentioned in below Figure 3). The intrinsic directionality of the laser light enables the lasers to be placed up to 25 meters away from the stepper and the light will delivered by a series of mirrors and lenses without a significant loss of energy.

Configuration of Excimer Laser Light Sources

Figure 3: Configuration of Excimer Laser Light Sources – Lasers is placed far from the stepper.

Excimer Laser’s Subsystems

Excimer laser consist of several subsystems (shown in figure 4). A high repetition rate is desirable for these pulsed light systems and excimer laser suppliers have improved the available rates from 200 Hz to 2 kHz and now to 4 kHz. The higher rates allow for high doses in short times without requiring high peak light intensities and this reduces damage to optical elements. Measurements for stepper self-metrology take place no faster than permitted by the excimer laser frequency, so stepper set-up time is reduced with high repetition rate lasers

Schematic of an excimer laser

Figure 4: Schematic of an excimer laser – key subcomponents

Faster exchange rates place significantly greater requirements on fans and motors with attendant concerns for reliability degradation. In principle, glass damage can be reduced by “stretching” the pulses from their current length of 25-30 nsecs, reducing the peak energy as well as the pulse bandwidth.

An example of starched pulse is shown in figure 5.

Excimer laser power vs. time for a normal and starched pulse

Figure 5: Excimer laser power vs. time for a normal and starched pulse from a Cymer ArF excimer laser.

Because laser pulse intensity does not evolve symmetrically in time, temporal pulse length needs definition. The most commonly used definition is the integral-squared pulse width, which is the most relevant definition of pulse duration equation to the issue of glass damage.

τ=[∫I(t)dt ]²/∫I2 (t)dt

The requirements for the laser gases increased the cost of installation of excimer laser steppers relative to costs for arc lamps (listed in table 2).

ARC Lamps

KrF Excimer Lasers

Availability 99.5% 97%
Cost for consumables $40,000 / Year $100,000 / Year
Installation Costs $1000 – $10,000 / Stepper $100,000 / Stepper

Table 2: Cost of Operation of Lithography Light Sources

Moreover, fluorine gases have a couple of safety requirements that needs to be addressed by an innovative solid source for fluorine. Fluorine leads to the etching of silicon dioxide. The windows of the excimer laser are typically made of calcium fluoride. But laser gases must be pure because impurities will cause degradation of laser performance. Since fluorine is a primary excimer laser gas, special materials must be used for handling this very chemically reactive gas.

Beam Delivery systems

Beam delivery systems involve another expense associated with excimer laser steppers that does not exist on arc lamp systems where the illumination source is built into the stepper. Main purpose of the beam delivery system is to provide an optical path from the excimer laser, which usually located outside of the cleanroom to the wafer stepper. Basically beam delivery system is a set of tubes contains lenses and mirrors through which the DUV light traverses the distance between laser and stepper (mentioned in figure 5). Performance of imaging optics required alignment of the input beam to within 1 deg, positioned to within 150 µm, while laser may be separated from the stepper by as much as 25 meter. This place tight constraints on tolerate levels of vibration throughput an extended part of the fabrication. Maintaining the 150 µm placement of a beam delivered from 10 meters away requires 3-arc sec of angular control. A robust way to ensure good alignment of the laser beam is to employ active beam sensing and control of the mirrors in the beam delivery unit.

Beam Delivery System Structure

It contains optics to address the natural beam divergence of an excimer laser. Laser beam of cross section 1 cm x 2 cm diverge to 5 mrad and 2 mrad, respectively. At the distance of 25 meter beam spreads to a cross section in excess of 13= cm x 6 cm without optics to refocus the beam. It also requires mirrors to bend the beam around obstructions between the laser – stepper and lenses require antireflection coatings to maintain the light intensity at the actinic wavelength and also at an optical wavelength that usually 632.8 nm (HeNe laser). It also enclosed and often purged with nitrogen to avoid photochemically deposited contaminants on the optical elements. Lenses and mirrors require coatings to be resistant to damage by high intensity light. The gas inside the lasers requires periodic replacements since small amounts of impurities in the gas reduce lasing efficiency and can deposit coatings on windows.

There is a direct relationship between the maximum scan speed and the number of pluses required to achieve a specified dose –

Ws= Vmn/f

ws is the slit width

Vm is maximum wafer stage scan speed

n is the minimum number of pulses required to achieve specified dose i.e. also related to dose control

f is laser repetition rate

The standard variation of the dose σD is related to the pulse to pulse variation σp-p

σD= σp-p/√n

If σp-p is large then n cannot be too small or this will be an inadequate dose control.

KrF and ArF Excimer Lasers

In KrF excimer lasers, excited dimers are created by placing a strong electric field across a gas mixture containing Kr, F2and Ne. Early excimer lasers required voltages > 20 kV and many components failed from high-voltage breakdown. A high voltage is used to produce and electrical discharge which in turn and drives the reactions which ultimately result in lasing (reactions are in table 3).

F2 + e- –> F- + F Negative Fluorine Production
Kr + e- –> Kr* + e-Kr* + e- –> Kr+ + 2e- Two-step positive krypton production
Kr+ + F- + Ne –> KrF* + Ne Excimer formation
KrF*–> Kr + F + hv Spontaneous emission
KrF* + hv –> Kr + F + 2hv Stimulated emission
F + F + Ne –> F2 + Ne Recombination

Table 3: Chemical Reactions in KrF and ArF Excimer Lasers

Reliability has improved through the laser designs that require lower voltages in the range of 12 – 15 kV to produce the electrical components. Excimer lasers produce light in pulses at rates up to several kilohertz. This has been a transition from thyratron-based discharging electronics to solid-state electronics and this has also contributed to improvements in laser reliability. This improvement in excimer light sources has played a critical role in bringing DUV lithography to production worthiness. To appreciate the degree of reliability required for use in manufacturing consider that modern KrF and ArF excimer lasers are capable of pulsing at 4 kHz with a duty factor of only 10%.

The unnarrowed fluorescence spectrum for the KrF emission is shown in figure 6. There is sufficient gain over only part of this spectrum for lasing about 400 pm. Free running ArF have lasers similarly broad bandwidths with full width half-maxima of about 450 pm. These line widths are much too large for use with all-refractive optics and require narrowing for catadioptric lenses as well.

KrF* fluorescence spectrum

Figure 6: KrF* fluorescence spectrum

KrF lasers have a natural bandwidth of approximately 300 pm which is too wide for high resolution wafers steppers. All-refractive optics require bandwidth of < 1.0 pm and even catadioptric, moderate-NA systems require bandwidth < 100 pm. The requirements of catadioptric lenses are met easily while the line narrowing demanded by all refractive lenses has presented a challenge to laser designers. For refractive systems there is a need to be able to vary the center wavelength in a controlled manner and this wavelength is tunable over a range of up to 400 pm.

Prisms are sufficient to narrow the bandwidth for application to catadioptric systems. The different options are demonstrated in figure 7.

Different types of bandwidth narrowing optics

Figure 7: Different types of bandwidth narrowing optics

Etalons are based upon the transmission properties of light through a transparent, parallel, plane plate. If the reflectance from an individual surface of the plate is R, then the transmitted intensity through the plate It is normalized to the incident light intensity Ii

It / Ii = 1 / (1 + F Sin2C δ/2)

Where

δ = 4 π / λ nt cosθ

And

F = 4R / (1 – R)2

In these above equations –

t is the thickness of the parallel plate

θ is the angle of incidence

n is the refractive index of the plate

λ is the wavelength of the light

Resulting light is shown in figure 8

Transmitted light intensity vs. δ through an etalon

Figure 8: Transmitted light intensity vs. δ through an etalon

Because of their high resolving power, etalons are used to measure wavelength and bandwidth with the intensity reduced sufficiently to avoid damage. All refractive optics requires very good control of the wavelength. Variations of the wavelength center-line can cause shifts in focus distortion and other aberrations.

For ArF lasers, there is a carbon spectral line at 193.0905 nm that is used to establish absolute wavelength calibration. F2 lasers emit over two fairly distinct peaks, as shown in figure 10, and laser optics only need to eliminate one of the peaks. It is expected that the 157 – nm lithography, if ever practiced, will use this remaining peak perhaps with some narrowing.

Spectrum of an unnarrowed F2 laser

Figure 9: Spectrum of an unnarrowed F2 laser – The curve labeled “b” is 8x that is curve “a”

MOPA

Cymer introduced a two-chamber laser system known as Master Oscillator Power Amplifier (MOPA) which produces light with less spatial coherence then typical injection locking lasers. In MOPA first chamber, the master oscillator is used to create a laser pulse with a narrow bandwidth. This pulse is then injected to second chamber, the power amplifier which amplifies the signals. There is an inactive discharge in the amplifier chamber that must be synchronized with the pulse generated in the master oscillator. Cymer’s MOPA differs from traditional injection locking by not having resonator optics as part of amplifier chamber and this allows for reduced levels of spatial coherence. Lambda-Physik has also introduced a two-chamber excimer laser system for lithography applications. Gigaphoton has introduced a laser with resonator optics in the power amplifier stage, which appears to produce light with spatial coherence that is low enough for lithographic applications.

MOPA Configuration

Figure 10: MOPA Configuration

Table 4 litsts the objective narrow bandwidth and high output of these dual chamber lasers. The bandwidth, whether measured in terms of full-width-half-maximum (FWHM) or E95%, which is the bandwidth containing 95% of the light energy, is cut in half on the dual chamber lasers relative to single chambers lasers. Output power, on the other hand, is doubled.

Cymer Nanolith 7000 (Single Chamber) Cymer XL-100(MOPA) Lambda – Physik Gigaphoton GT40A (MOPA)
Pulse energy 5 mJ 10 mJ > 12.5 mJ 15 mJ
Power 20W 40W > 50W 60W
Spectral bandwidth (FWHM) ≤ 0.5 pm ≤ 0.25 pm ≤ 0.25 pm ≤ 0.18 pm
Spectral bandwidth (E95%) ≤ 1.3 pm ≤ 0.65 pm ≤ 0.55 pm ≤ 0.15 pm

Table 4: Objective Bandwidth and Power Output of Each Chamber Lasers

Note:
If you are interested in our services, please visit our site.

Subscribe to our newsletter to receive our new articles directly in your mail box.

If you liked this article, please give it a quick review in StumbleUpon, Facebook or Pinterest.

Leave a Reply

Your email address will not be published. Required fields are marked *

* Please arrange the below number in increasing order

You may use these HTML tags and attributes: <a href="" title="" rel=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

CommentLuv badge