The invention relates to an optical arrangement for homogenizing a laser pulse from a pulse-laser light source, more particularly a laser, preferably an excimer laser, in particular for a projection exposure unit.

PRIOR ART

Pulse-laser light sources, e.g. excimer lasers, for UV

lithography have a repetition rate of approximately 1000 to 4000 pulses per second. Each individual pulse has a pulse length of approximately 20 to 30 ns . There are significant modulations of the laser output power over time within each pulse as a function of the gas, the state of the laser, more particularly the optical components, and as a function of the resonator length.

In this context, it was found in practice that there is a significant disadvantage in that there is significant power peaking (peak power) due to the functioning of a pulse-laser light source and this has a very negative effect on the optical materials, in particular glassy materials. Particularly in the case of quartz glass and an operational wavelength of 193 nm, nonlinear optical effects are generated as a result of the temporal power division and these effects damage the material over an envisaged service life of the product. Results of this also include transmission losses and an uncontrolled increase in the refractive index.

It is also known that the intensity distribution is generally not homogeneous across a coherent light field. This particularly holds true for the radiation emitted by an excimer laser. When illuminating a surface with a coherent, inhomogeneous laser- light bundle, interferences are generated, which can be perceived as spatially differing luminance and which moreover also vary in terms of the interference from different

observation directions due to the phase relations changing in the process. This interference, which can be perceived as sparkling, is referred to as "speckle" in the art. Optical arrangements have been developed that avoid, or at least reduce, the occurrence of speckle. For this purpose, the coherence of the light bundle must be destroyed, so to speak, so that the light bundle can no longer interfere with itself and thereby generate speckle. This is usually achieved by virtue of a light bundle being split up and brought back together again on paths of different length, wherein the path length difference should be of the order of the coherence length of the light bundle.

EP 0 785 473 A2 has disclosed a device of the type mentioned at the outset, by means of which the light coming from a pulse- laser light source is subdivided into a plurality of beamlets, which pass through loop paths of different length. This widens the bundle or causes a subdivision into a plurality of beamlets arranged next to one another, the coherence of which is reduced or removed. These beamlets are introduced arranged next to one another into an illumination apparatus.

However, a disadvantage thereof is that this produces a

plurality of optical axes next to one another, and the

illumination apparatus must be adapted accordingly. Moreover, the alignment of the aforementioned device is fixed.

DE 195 01 521 CI describes an arrangement for reducing

interference in a coherent light bundle by reducing the temporal coherence. Therein, the use of micro-structured phase plates is proposed, through which the laser beam bundle passes for

reducing the temporal coherence, with, in a particular refinement of the invention, such a phase plate with a phase- changing surface structure having a reflecting design. If the laser light passes through the micro-structured phase plate, the coherence of the laser light is lifted.

US 7,369,597 B2 has disclosed a device of the type mentioned at the outset, by means of which the light coming from a pulse- laser light source is divided into two beamlets by a beam splitter, with one beamlet passing through a fixed loop path. These beamlets are introduced superposed into an illumination apparatus .

SUMMARY OF THE INVENTION

The present invention is based on the object of developing a device with the aid of which damage to components lying in the beam path of the pulse-laser light source is avoided and

undesired speckles are reduced, with the lowest possible losses in the efficiency of the pulse-laser light source and the greatest possible flexibility.

According to the invention, this object is achieved by an optical arrangement for homogenizing an at least partly coherent light field in a light pulse from a pulsed light source, more particularly a laser, preferably an excimer laser, as per Claim 1. For this purpose, provision is made for at least one optical loop, with a beam-splitter apparatus being provided in the optical loop. The beam-splitter apparatus separates the light field into two beamlets, polarized perpendicular to each other, with a first and a second polarization direction, the first beamlets with the first polarization direction reaching a surface to be illuminated without passing through the loop and the second beamlets with the second polarization direction passing through the optical loop. A polarization rotator is provided in the optical loop and rotates the second polarization direction of the second beamlets by an angle, which can be defined in advance, and so at least part of the second beamlets again pass through the optical loop with the first polarization direction. The beam-splitter apparatus decouples the other part of the second beamlets with the second polarization direction from the loop and the latter reach the surface to be illuminated superposed, offset in time, with respect to the first beamlets.

Advantageously, the second beamlets pass through the optical loop a number of times.

This stretches the light pulse in time and the power of the light pulse thus is reduced.

In one embodiment, the angle is changed during every loop such that a predetermined intensity of the second beamlets is decoupled by the beam-splitter apparatus and reach the surface to be illuminated superposed, offset in time, with respect to the first beamlets. This allows the pulse shape and width to be influenced in a targeted fashion.

It is particularly advantageous for the optical loop to have such a length that the difference between the optical path lengths of the paths of the beamlets is greater than the

temporal coherence in the light field. This reduces the

coherence of the laser light emerging from the arrangement and thus the occurrence of speckle is avoided or at least reduced.

In a further embodiment, reflecting components, embodied as mirrors, form the optical loop. In a further advantageous embodiment, reflecting components, embodied as prisms, form the optical loop. This reduces the transmission loss in the loop.

If the second beamlets are coupled into the prisms at Brewster's angle, the transmission losses can thus be reduced further.

In one embodiment, the beam-splitter apparatus is designed as a polarizing beam splitter.

The superposition, offset in time, of the beamlets and the path length difference advantageously reduce the coherence of the light field at the surface to be illuminated.

The multiple superposition, offset in time, of the beamlets particularly advantageously reduces the peak power of the light field at the surface to be illuminated.

In a further advantageous embodiment, the polarization rotator is designed as a Pockels cell. This allows very rapid change of the polarization direction.

The polarization rotator (14, 24) advantageously has a nonlinear optical crystal, selected from the following group of crystals: beta barium borate (BBO) , potassium dihydrogen phosphate (KDP) , deuterated potassium dihydrogen phosphate (DKDP) or lithium triborate (L1B ₃ O ₅ , LBO) . The advantage of this is that a broad wavelength spectrum of the light can be influenced in its polarization, with losses that are as low as possible.

A further embodiment has between the light source and the beam ^¬ splitter apparatus a polarization-adjusting element for

adjusting the polarization of the light field. The advantage of this is that the polarization direction of the light entering the optical arrangement can be preset in a targeted fashion.

Further advantageous refinements and developments emerge from the remaining dependent claims and from the exemplary

embodiments described in principle hereinbelow with the aid of the drawing.

BRIEF DESCRIPTION OF THE DRAWINGS

Herein :

Figure 1 shows a refinement of the invention with beam splitter and polarization rotator with mirrors in the deflected path;

Figure 2 shows a refinement of the invention with beam splitter and polarization rotator with deflection prisms in the deflected path;

Figure 3 shows an illustration of the principle of a deflection prism; and

Figure 4 shows a diagram to illustrate the angular rotation of a polarization rotator per loop;

Figure 5 the principle of the time-offset superposition

DETAILED DESCRIPTION

According to the embodiment according to Figure 1, a beam bundle 10 from a pulse-laser light source, e.g. from an excimer laser, is incident on a first beam-splitter apparatus 15 in the form of a well-known polarization beam splitter, for example a

polarizing beam-splitter cube. A polarizing beam-splitter cube splits unpolarized light into two mutually orthogonally

polarized beamlets. One beamlet is transmitted, the other is reflected. The transmitted beamlet 10b is polarized parallel to the plane of incidence of the beam-splitter cube (p- polarization) and the reflected beamlet 10a is polarized

perpendicular to the plane of incidence (s-polarization) . Other elements, which split polarized radiation into two beamlets with mutually perpendicularly oriented polarization directions, are also feasible.

If the light is linearly polarized and the beam bundle 10 is incident on the beam-splitter apparatus 15 for example with its polarization directions at 45° (hence, this is at 45° to the plane of the drawing in Figure 1 and indicated as a mixture of both polarization directions by means of the double-headed arrows and crossed colons) , 50% of the entire beam is passed, as a linearly polarized beamlet 10b, unhindered and without loss through the beam-splitter apparatus 15 toward a surface 12 to be illuminated, the beamlet having a polarization direction of 0° after the beam splitter apparatus 15.

In the example, the polarization direction is then parallel to the plane of the drawing (p-polarized) , as indicated by the two parallel double-headed arrows. The other 50% of the beam are reflected and travel as beamlet 10a over an optical loop

(referred to as loop path below) , which is formed by beam- deflecting components, between which the beamlet 10a can pass over a certain path. In the following text, this is referred to as loop path in an abbreviated fashion. In the present

embodiment, four mirrors lla-d, as beam-deflecting elements, form a loop path by each deflecting the beamlet 10a by 90° and transmitting it to the next respective mirror. Downstream of the beam-splitter apparatus 15, beamlet 10a is linearly polarized with a polarization direction of 90°, that is to say

perpendicular to the plane of the drawing (s-polarized) in the example. This is indicated by the crossed colons. The length of the loop path in the arrangement shown above, from the beam ^¬ splitter apparatus 15, via the mirrors lla-d and back to the beam-slitter arrangement 15, is dimensioned such that the pulse can be completely taken in.

Hence, in the case of a 20 ns pulse, for example, the loop path would need to have a length of 6 m (20-10 ^-9 s * 3-10 ⁸ m/s = 6 m) , divided between the four paths between the mirrors lla-d.

On its rear side, the beam-splitter apparatus 15 likewise acts as a beam splitter. When the s-polarized beamlet 10a arrives at the beam-splitter apparatus 15 without changing its

polarization, it is reflected at the rear side of the beam- slitter apparatus 15. Thus, without any further measures, the beamlet 10a, now with s-polarization, would leave the loop path in the direction of the surface 12 to be illuminated, offset in time to the beamlet 10b. In the process, after passing through the loop path, the beamlet 10a has travelled a longer path than the beamlet 10b. If the path difference between the two beamlets is greater than the temporal coherence length of the laser pulse in the incident light bundle 10, this would already mean a reduction in the coherence and the peak power as a result of the "smearing" of the laser pulse. However, such a design still is very inflexible because there are no further options for

influencing the loop time, the pulse shape and path difference in order to reduce coherence.

According to the invention, in the embodiment according to

Figure 1, a polarization rotator 14 with a control unit 13 has been introduced into the beam path in the arm of the loop path in front of the beam-splitter apparatus 15, which polarization rotator can rotate the polarization direction of a light beam by an angle a in a targeted fashion. By way of example, this can be a Pockels cell made of a nonlinear optical crystal, as known from WO 2005/085955 A2, for example. A Pockels cell rotates the polarization direction of radiation passing through the cell when an electrical voltage is applied by the control unit 13, the angle of rotation being proportional to the applied voltage. In one embodiment, the nonlinear optical crystal is transparent to light in a wavelength range below 200 nm. Crystals

transparent below this wavelength are particularly suitable for use in microlithography, in which work for generating very fine semiconductor structures is performed with illumination light at wavelengths under 200 nm, more particularly at 193 nm or 157 nm. In one development, the nonlinear optical crystal basically consists of beta barium borate (BBO) , potassium dihydrogen phosphate (KDP) , deuterated potassium dihydrogen phosphate

(DKDP) or lithium triborate (L1B ₃ O ₅ , LBO) . Crystals made of these materials are also transparent at wavelengths of less than

200 nm. KDP and DKDP have a transmission range between

approximately 190 nm and approximately 1500 nm. In LBO, the transmission range reaches from approximately 160 nm to

approximately 2600 nm. Thus, applications in the visible or infrared spectrum are also possible. The Pockels cell must have a quick switching time, because the polarization direction must, for example, be switched after one loop by a pulse. This can lie in the range of a few ns; in the current example < 20 ns .

Before the looping beamlet 10a reaches the polarization rotor 14, the polarization rotor 14 is set such that it rotates incident linearly polarized light by an angle a .

If the angle a = 90° were set during the first pass S i by the polarization rotator 14 and if the angle a set on the polarization rotator 14 were immediately reset to zero after the pulse passed therethrough, the coupled-in light remains in the illustrated arrangement without intensity being decoupled because the beamlet 10a now is p-polarized as a result of the 90° rotation and can pass the beam-splitter apparatus 15 without hindrance .

The beamlet 10a passes through the loop path one more time. This is indicated by the two double-headed arrows in parentheses. After another pass, or after a plurality of up to n passes S2 to Sn-i, the polarization rotator 14 is finally switched again before the beamlet 10a arrives such that said polarization rotator rotates the polarization direction by 90°. As a result, the beamlet 10a is s-polarized again in the final loop S _n , reflected by the beam-splitter apparatus 15 and thereby

decoupled. This is indicated by the crossed colons in braces. Hence, the s-polarized beamlet 10a can leave the loop path in the direction of the surface 12 to be illuminated after the beamlet 10b, displaced in time by the n-fold time compared to the beamlet 10b. This can achieve an almost arbitrary delay of the beamlet 10a as a multiple of the loop time in the loop path formed by the mirrors lla-d. Even if the pulse is intended to be stretched e.g. from 20 ns by fifty times to 1 ys, this has no influence on the installation space of the proposed arrangement, but rather only the number of loopings is increased. Hence the loop time of the pulse can be controlled.

If the angle a set by the polarization rotator 14 deviates from 90°, the beamlet 10a has an s-polarized and a p-polarized component in respect of the beam-splitter apparatus 15. The component of the beamlet 10a s-polarized relative to the

splitter plane of the beam-splitter apparatus 15 is decoupled. The component of the beamlet 10a p-polarized relative to the splitter plane of the beam-splitter apparatus 15 is passed through the beam-splitter apparatus 15 and once again passes through the loop path. At the next pass through the polarization rotator 14, the polarization is again rotated by an angle CC _n , which however need not necessarily correspond to the rotational angle a from the first loop. When the beam-splitter apparatus 15 is reached again, the s-component is in turn decoupled and the p-component is transmitted. Selecting the angle CC _n can set what intensity is intended to be decoupled from the device according to the invention. Thus, for example, it is possible in each case to decouple the same intensity proportion of the beamlet 10a as s-polarized light if the angle cc _n is selected to be the same in each case. Since the overall intensity of the looping light becomes ever smaller, the decoupled light intensity also becomes ever smaller in the process. Skillful selection of the angles CC _n affords the possibility of obtaining intensity of the decoupled light that is the same in each case. By way of example, if the described procedure is started not with linearly polarized light at 45°, but with an s-polarized light bundle 10, 100% of the light bundle is coupled into the loop path as beamlet 10a.

Figure 4 shows how the angle CC _n of the linear polarization must be changed from loop to loop if the pulse should be stretched by a factor of 10 and the intensity profile of the stretched pulse over time should remain constant. What is plotted is the angle a, against the number n of loops in the loop path.

In the last (n-th) loop, the orientation of the linear

polarization is finally rotated by 90° (~ 1.57 rad) such that then the entire remaining light is decoupled from the beam ^¬ splitter apparatus 15.

The profile of the angle CC _n to be set varies from loop to loop. In the example, the values are based on the condition that a tenth of the overall intensity is decoupled in each loop, i.e. the orientation of the linear polarization must in each case be rotated such that the square of the decoupled components

corresponds to 10% of the initial intensity. The number of mirrors (lla-d) should merely be understood as being exemplary. Any other number of mirrors and spatial

arrangement thereof, which can form an optical loop path, is possible. In the example, the mirrors are embodied as planar mirrors. Other mirror shapes, for example concave mirrors with a parabolically curved surface, are also feasible, as are

combinations of different shapes.

In order to be able to set the polarization direction of the beam bundle 10 incident on the beam-splitter apparatus 15 in a targeted fashion, a polarization-adjusting element 17 can be provided upstream of the beam-splitter apparatus 15 in the beam path of the beam bundle 10. This polarization-adjusting element 17 can for example be a polarization rotator analogous to the polarization rotator 14, which rotates an already present polarization of the beam bundle 10, or a polarizer, which for example generates linearly polarized light at a certain angle from unpolarized light.

In the process, the polarization can be set by a static element, for example a polarization filter, or by a variable element, for example a Pockels cell. Any other suitable element for

influencing the polarization and polarization direction is likewise feasible.

The principle of the time-offset superposition in Figures 5a-c is explained in more detail below.

Figure 5a shows, much simplified, the profile of the power of a laser pulse, which for example in Figure 1 is incident on the beam-splitter apparatus 15 as an s-polarized light bundle 10. The pulse is completely coupled into the loop path and a quarter of the power is decoupled after every loop in the present example by appropriate selection of the angle cc _n on the

polarization rotator 14. Figure 5b illustrates how the four decoupled pulses are distributed, in a successively overlapping fashion, over time. The pulses reach the surface 12 to be illuminated as a pulse that is composed of the four superposed pulses from Figure 5b. This is illustrated in Figure 5c in an exemplary fashion. The resultant pulse is not only stretched in time, but the power is reduced overall. Additionally, the coherence of the laser pulse is reduced if the optical path length of the loop path is greater than the coherence length of the light of the laser pulse.

Hence, the pulse of the pulse-laser light source is smoothed. At the same time, this reduces the peak power of the pulse and hence damage to optical elements is avoided. In the beam

splitter 15, the split is brought about as described previously such that different radiation components pass through the loop a different number of times and are combined together after being decoupled, with changes in the wavefronts of the individual radiation components resulting due to the multiple splitting and the optical path length difference, and a reduction in the coherence in the laser light emerging from the arrangement resulting therefrom. This avoids the occurrence of speckles, or at least reduces it.

A further embodiment of the invention in Figures 2 and 3 is described below.

The mirrors lla-d from the exemplary embodiment according to Figure 1 lead to undesirable transmission losses in the

deflected beamlet because losses occur at each reflection on a mirror due to a reflectivity that can at most lie at

approximately 95% - 98%.

Figure 2 shows an embodiment in which the beam deflection is formed by prisms 21a and 21b. A polarized light bundle 20 is incident on a beam-splitter apparatus 25. There, light bundle 20 is split into two beamlets 20a and 20b. After the beam-splitter apparatus 25, the transmitted beamlet 20b is p-polarized and the reflected beamlet 20a is s-polarized. Beamlet 20a is incident on the entry surface of the prism 21b, ideally at Brewster's angle on the boundary between the material of the prism 21a, 21b and the adjacent atmosphere, for example air or a vacuum. If the incident light 20a is completely p-polarized, the losses even equal zero in the ideal case. Beamlet 20a is reflected in the prism 21b such that it is incident on the second prism 21a after emerging from the prism 21b. There, beamlet 20a is reflected again and it thereafter passes through a polarization rotator 24 with a control apparatus 23 and from there it is incident again on the beam-splitter apparatus 25. In respect of the

functionality thereof, reference is made to the analogous description of the embodiment according to Figure 1, with the only difference being that now the prisms 21a, 21b are used as a deflection apparatus for forming the deflection path rather than the deflection mirrors lla-d.

Thus, the light is deflected in the prism 21a, 21b by using total reflection and leaves the prism 21a, 21b again at

Brewster's angle. In theory, the transmission loss in such a beam deflection equals zero. However, due to the finite surface quality, the loss in each deflection prism will not achieve the theoretically possible value, but it does lies in the range of ≤ 0.5%. Provision can again also be made in this case for a polarization-adjusting element 27 for adjusting the polarization direction of the light bundle 20 between light source and the beam-splitter apparatus 25.

An advantageous refinement of a deflection prism 21 will be explained in more detail with the aid of Figure 3. The prism 21 has four side surfaces, with the side surfaces including four angles a, b, b ^x and c. Angles b and b ^x are the same in the example, and so an incident beamlet 20a± and an emerging beamlet 20a _e run parallel to one another. Beamlet 20a± is incident at an angle Θ _Β to the surface of the prism 21 and is refracted into the prism 21 at an angle Θ _Β' · After two total-internal reflections, the beamlet 20a _e again emerges from the prism 21 and then runs on parallel and in the opposite direction to the incident beamlet 20ai in the deflection path.

In the example, the refractive index of the prism material is n _P = 1.5 and that of the surrounding atmosphere is n _L = 1. The angles are then calculated as follows: a = 180° - 2·Θ _Β

b = 45° + 0.5· (Θ _Β + ΘΒ

c = 90° + (ΘΒ - ΘΒ

Θ _Β = arctan (η) : Brewster's angle

Θ _Β' = arcsin (1 ^/ n · sin (Θ _Β ) )

With n = n _P / n _L

This results in: Θ _Β = 56.31°, Θ _Β « = 33.69°, a = 67.38°, b = 90° and c = 112.62°.

It goes with out saying that other refinements and number of prisms are also feasible for deflecting beams, which prisms deflect light beams with losses that are as low as possible and in this fashion form a loop path. Other forms of optical

elements for beam deflection operating on the principle of total reflection at boundaries are also suitable for an embodiment of a deflection path according to the invention.