WO1994016310A1

WO1994016310A1 - Zeeman ellipsometer

Info

Publication number: WO1994016310A1
Application number: PCT/NL1993/000283
Authority: WO
Inventors: Klaas Hemmes; Maurits Matthijs Wind; Rudolf Lepoole; Philip Ernst Habing
Original assignee: Technische Universiteit Delft
Priority date: 1992-12-31
Filing date: 1993-12-30
Publication date: 1994-07-21
Also published as: AU5844194A; NL194893C; NL9202303A; NL194893B

Abstract

Ellipsometer comprising at least a Zeeman laser (Z) to generate two beams (g1, g2) which are slightly shifted in frequency and (after transmission through a birefringent crystal (quarter-wave plate)) are both polarized linearly but perpendicular to one another, a non-polarizing beam splitter (N) downstream of which a working beam (g'm2) mofidied by a sample (S) interferes with a reference beam (g1), a unit (W) for separating two orthogonal (p- and s-) components of the interference beam thus composed and providing two a.c. voltages (V1, V2) corresponding thereto, polarization means (G1, G2) being provided to adjust the polarization directions of the working beam (g2) and/or the reference beam (g1) in order to effect the desired interference of the two beams.

Description

Zeeman ellipsometer

The present invention relates to an ellipsometer comprising at least a light source which during operation provides at least one light beam, a non-polarizing beam splitter for causing a work¬ ing beam modified by a sample, and a reference beam to interfere, a unit for separating two orthogonal components of the (interference) beam thus composed and providing two a.c. voltages corresponding thereto. An ellipsometer of this type is disclosed by H.F. Hazebroek, W.M. Visser, "Automated laser-interferometric ellipso etry and precision reflectometry", J.Phys.E: Sci. Instrum., Vol. 16, 1983, pp. 654-661. To give a better understanding of the present inven¬ tion, the ellipsometer disclosed by the above-mentioned article is first described globally with reference to Figure 1. A light source L, preferably consisting of an He-Ne laser, emits a beam of mono¬ chromatic, non-polarized light g, having a wavelength of, for example, 632.8 nm. The light beam g first passes a beam splitter B which taps off a part g of the beam g for the benefit of signal processing purposes. Said beam g is detected by a photodiode D3 which converts the calibration beam g into an electrical signal S3. The signal S3 is supplied to suitable electronic signal pro¬ cessing means (not shown) . The beam g passes a polarizer, for example a Glan-Thompson prism G, which polarizes the beam g linear- ly in such a way that the electric field of g is then at an angle of 45° with the normal on the plane of the working set-up shown in Figure 1.

The light beam then impinges on a non-polarizing beam split¬ ter N. The non-polarizing beam splitter N reflects at least sub- stantially half of the incident beam as the working beam g_m in the direction of a sample S. The remaining part g_r of the incident beam passes the non-polarizing beam splitter N and serves as the refer¬ ence beam.

The working beam g_m is reflected by the surface of the sample S to be analyzed. The beam reflected by the sample S is autocol- limated very accurately by a mirror M which reflects the working beam g_m towards the sample S. The working beam g_m is thus reflected twice by the surface of the sample S, whereupon a modified working beam g' is produced which is shifted, in amplitude and phase, with respect to the original working beam g_m, the amplitude and the phase shift being a function of the properties of the surface of the sample S. The reference beam g_r is reflected by a cubic reflector C which is coupled to an (electromagnetic) drive unit (not shown) in order to be moved back and forth with a constant speed. As a result, the reference beam g_r is shifted in frequency, namely as a result of a Doppler shift. The reflected beam g' , which propagates along a different trajectory than the incident reference beam g_r, impinges on the rear side of the non-polarizing beam splitter N. At that point at the rear side of the non-polarizing beam splitter N the modified working beam g' likewise comes in, as a result of which interference arises between the modified working beam g' and the Doppler-shifted reference beam g' . The interference beam, schematically designated by g' + g' , is split into two orthogonal polarization modes by a Wollaston prism w corresponding to the p- and s-directions, as conventionally defined with respect to the specimen. Each of these two orthogonal polarization modes is inter- cepted by a photodiode D1 and D2, respectively. The photodiodes D1 , D2 convert the intensities of the two orthogonal polarization modes into corresponding a.c. voltage signals V1 , V2. The sinusoidal a.c. voltage signals V1 , V2 both have a frequency which is equal to the Doppler shift caused by the moving cubic reflector C. The frequency coupled to the Doppler shift is equal to twice the speed of the (electromagnetic) drive unit (not shown) of the cubic reflector C, divided by the wavelength of the light used. By means of the known device it is possible, for example, to achieve a Doppler shift of 175 Hz. This is also the measuring frequency of the known ellipso- meter. The electrical signals V1 , V2 are supplied to a suitable electronic measurement processing system (not shown) . The amplitude ratio of the electrical signals V1 , V2 and the phase difference between these two electrical signals V1 , V2 provide the desired ellipsometric information concerning the sample S in terms of the angles ψ and Δ.

By placing the mirror M at the position of the sample S and causing the working beam to be reflected along the same path, the device can be calibrated in a simple manner. By means of the known_^device of Figure 1, the two orthogonal polarization modes can be made visible directly. The known device was used for measuring the optical properties of equilibrium sys¬ tems, for example metal surfaces, thin layers or interfaces between liquid and gas phase. Likewise, slowly changing processes on the surface, such as corrosion or the growth of an electrochemical film, can also be studied accurately with the device according to Figure 1. Owing to the limited measurement frequency which is link¬ ed to the maximum speed at which the cubic reflector C can be driven, it is however not possible to study very rapidly changing processes on the surface of a sample S. Moreover, considerable averaging of various successive measurement results is necessary to eliminate fluctuations. Furthermore, the algorithm used to obtain ellipsometric data from the two sinusoidal signals V1 , V2 is too time-intensive to provide for (quasi) real-time measurements. It is not possible to carry out continuous measurements, because the linear mirror movement has to reverse direction sooner or later.

The use of analog processing electronics to carry out real¬ time data analysis is possible in principle, but is made rather difficult owing to the not entirely constant Doppler-shift differ¬ ence frequency and the non-continuity of the mirror motion (causing Doppler shift) .

One object of the present invention is to eliminate (entire¬ ly) fluctuations in the signal frequency. A further object of the invention is to increase the measure¬ ment accuracy.

Another object of the invention is to provide an ellipsometer by means of which very rapidly changing processes can be measured, the option existing of presenting the measured ellipsometric para- meters Δ and ψ in real time as analog signals.

Finally it is an object of the present invention to be able to carry out continuous measurements.

To this end, an ellipsometer of the type mentioned in the preamble is characterized in that the light source is a Zeeman laser to generate two beams which are slightly shifted in frequency and which are both polarized linearly but perpendicular to one an¬ other, and in that polarization means are provided for adjusting the polarization directions of the working beam and/or the refer- ence beam in order to effect the desired interference between the modified working beam and the reference beam.

With an ellipsometer of this type, all the advantages of the known system according to Figure 1 are retained, while it is also possible to achieve the above-mentioned objects.

It should be noted in this context that an ellipsometer pro¬ vided with a Zeeman laser to generate two beams which are slightly shifted in frequency and which are both polarized linearly but per¬ pendicular to one another, polarization means being provided for adjusting the polarization directions of the working beam and/or the reference beam, is disclosed per se by EP-A-200.978. In this conventional ellipsometer, the reference beam and the working beam modified by the sample surface both impinge, at different angles, directly onto the surface of a photodiode, where interference would then have to occur. Beams crossing at an arbitrary angle are not, however, able to interfere effectively with one another, so that this conventional ellipsometer probably cannot function properly. Moreover, data are lacking concerning the signal processing of the signal originating from the photodiode. In a general sense, any suggestion is lacking in this publication how the use of a Zeeman laser can be combined with the above-mentioned prior art in order to retain the advantages of the ellipsometer according to the prior art.

The invention further makes provision for an ellipsometer of the above-mentioned type, which is provided with a mirror, for example a cubic reflector, for receiving the reference beam pro¬ duced by transmission of part of the beam generated by the light source through the non-polarizing beam splitter, and for reflecting it to the rear side of the non-polarizing beam splitter, character- ized in that, in the optical path between the rear side of the non- polarizing beam splitter and the cubic reflector, a polarization unit is positioned which, for example, rotates by 45°.

In a preferred embodiment, the ellipsometer is provided with a polarizing beam splitter for separating the two beams originating from the Zeeman laser, the one beam serving as the working beam and the other as the reference beam.

In a subsequent embodiment, the ellipsometer is characterized in that, in the path of the working beam between the polarizing beam splitter and the sample, a polarization element is positioned which rotates by an angle of, for example, 45°, and in the path of the reference beam between the polarizing beam splitter and the rear side of the non-polarizing beam splitter another polarization element is positioned which rotates the polarization direction by an angle of, for example, 45°, while the working beam impinges on the surface of the sample directly after transmission through the polarization element.

In a further preferred embodiment, the ellipsometer is designed in such a way that the beam transmitted by the polarizing beam splitter impinges on the front side of the non-polarizing beam splitter and the part thereof which is reflected by the non-polar¬ izing beam splitter provides the working beam and the part thereof which is transmitted by the non-polarizing beam splitter is blocked or scattered.

The present invention will be explained below in more detail with reference to a few drawings which explain the principle of the invention in more detail and which are not intended to limit the scope of the present invention. In the drawings: Figure 2 shows a set-up to explain the basic principle of the invention, by means of which optical parameters of the working beam, not necessarily Δ and ψ, can be measured;

Figures 3 and 4 show preferred embodiments of ellipsometers. Prior to the description of the figures 2, 3 and 4, a few general guidelines are first given for interferometric ellipso¬ meters which employ a Zeeman laser. By employing a Zeeman laser, a fixed instead of a moving cubic reflector C can be used. Use of a cubic reflector C is not necessary; in principle it is also poss¬ ible to use a different type of mirror. It is assumed that the correct ellipsometric information can be derived from the electri¬ cal a.c. voltages V1 , V2 (Figure 1). It is very important that the measurement signal and the reference signal are selected carefully and that they can interfere with one another, because otherwise information from the sample S will be lost. An ellipsometer accord- ing to the invention therefore has to satisfy the following three criteria: a) a polarization unit or a polarizing beam splitter in the path of the working beam between the sample S and the Wollaston prism W should be avoided. Such a polarization unit, namely, selectively transmits components of a light beam with the cor¬ rect polarization vector, as a result of which such a projec¬ tion of the polarization vector will in general lead to infor- mation loss concerning the sample S; b) a different frequency dependence of the working beam and the reference beam should be achieved, as only then is it possible to achieve a non-trivial phase difference between the two a.c. voltages V1 , V2. If both beams have the same frequency depen- dence, it can be readily demonstrated that no phase difference arises between the signals V₁ and V- and only the amplitude ratio of the a.c. voltages V1 , V2 comprises information con¬ cerning the sample S; c) the two frequencies of a Zeeman laser used can only be separ- ated optically by employing the different polarization states.

If the light beam passes a polarization unit between a Zeeman laser and a first beam splitter, both waves would be given the same polarization state and could no longer be separated from one another. Therefore, no polarization unit should be posi- tioned between the Zeeman laser used and the first beam split¬ ter.

A first ellipsometer which satisfies the above-mentioned criteria a), b) and c) can be designed based on the device accord¬ ing to Figure 2. The device according to Figure 2 differs in three points from the device according to Figure 1 : instead of a mono¬ chromatic laser L, a Zeeman laser Z is used, a Glan-Thompson prism G is not used, and the cubic reflector C takes up a fixed position. The Zeeman laser Z emits two waves g g₂ which, for example, have a frequency difference of approximately 1.8 MHz. Such a Zeeman laser is known and requires no further explanation here. The waves g₁, g₂ encounter a non-polarizing beam splitter N which separates them into two at least almost equal parts. A first part is produced by reflection on the surface of the non-polarizing beam splitter N and provides working beams g_ml, g_m2 which are directed at the sample S. In the same way as the modified working beam g' in Figure 1 is effected, two modified working beams g'_m-_|. g'_m2 ^are P^r°duced in the configuration according to Figure 2, which are reflected onto the non-polarizing beam splitter N. Another part of the beams g-, g₂ generated by the Zeeman laser Z is transmitted by the non-polarizing beam splitter N and forms two reference beams g - , g_r2. The fixed cubic reflector C reflects the reference beams g_r1, g_r2 to the rear face of the non- polarizing beam splitter N. The remaining components in Figure 2 are identical to those in Figure 1. The configuration according to Figure 2 cannot function as such, because the modified working beams g'_m1, g'_m2 do not interfere correctly, at the rear side of the non-polarizing beam splitter N, with the reference beams g_rl, g_r2. Correct functioning requires the working beam g'_mι, <?'_m2' respec¬ tively, to interfere with the reference beam g_r2 and g_rl, respect¬ ively. In order to bring this about, additional optical elements should be placed in the configuration according to Figure 2, i.e. polarization units or, for example, quarter-wave plates. In prin- ciple, there are five possibilities of doing so, which are indi¬ cated in Figure 2 by a, b, c, d, e. When placing optical elements of this type, the above-mentioned three criteria a, b, c should be satisfied. This leads to the following criteria for the configur¬ ation according to Figure 2: d) a non-polarizing beam splitter N is inevitable, because a pola¬ rizing beam splitter is not permitted, owing to criterion a; e) polarization units are not permitted at the positions a (owing to criterion c), b (owing to criterion a) or e (owing to cri¬ terion a) ; f) a polarization unit or quarter-wave plate is required at posi¬ tion c or d (owing to criterion b) .

In Figure 2, the working beams g_m1, g_m2 impinge on the sample S at an angle of, for example, approximately 45°, just as in the known device according to Figure 1. This is not strictly neces- sary, however, although it is indeed preferable. The normal of the sample S may form an angle between 0 and 90° with the incident working beams g_m1, g_m2 just as in the configuration according to Figure 1. The most suitable choice to achieve a construction in which the sample S is perpendicular to the measuring waves g_m1, g_m2, is to set up, at position d in the system according to Figure 2, a quarter-wave plate with an orientation of 45°. Such a quarter-wave plate in combination with the cubic reflector C rotates the polar¬ ization of the reference beam by an angle of 90°. The following can be demonstrated for a system of this type: the output signals V1 , V2 of the photodiodes D1 , D2, which are obtained by means of this set-up, are valid signals from which optical information can be derived; - strictly speaking, this set-up cannot be called an ellipso¬ meter, because a somewhat different combination of r and r_s is measured; nevertheless, this combination may still be suitable for some applications; the different frequency dependence of the s- and p-waves in the reference beams g_r1, g_r2 and the working beams g_ml, g_m2 make the system sensitive to vibrations of the optical elements used;

- owing to this different frequency dependence it is also more difficult to carry out a reference measurement than with the set-up according to Figure 1. If the sample S forms an angle of at least almost 45° with the incident working beams g_m1, g_m2, this has the advantage that the frequency dependence of the p- and s-polarizations show strong resemblance, although they are then not yet identical. The sensi¬ tivity to vibrations of the optical elements is considerably reduced. The orientation of 45° has the result that, if the laser Z is mounted in its normal position, i.e. that one of the polariz¬ ations is perpendicular to the plane of the drawing of Figure 2, the working beams g_m1, g_m2 reflected by the sample S are directed downwards from the optical plane. If this is to be avoided, the laser beams themselves or the laser Z should be rotated by 45°. The latter is simpler and can be achieved by placing, at point d in Figure 2, a Glan-Thompson polarization unit set to 45°. The follow¬ ing can be demonstrated for such a set-up: the set-up does not, strictly speaking, provide ellipsometric information;

- up to the fourth order in the reflection amplitudes, the cor¬ rect ellipsometric reaction is observed.

The drawbacks of the devices based on Figure 2 are over¬ come by means of the ellipsometers according to the Figures 3 and 4.

In the ellipsometer according to Figure 3, a Zeeman laser Z generates two beams g.,, g₂. A first beam splitter B1 diverts a part of the beams g-, g₂ for signal processing purposes. The diverted beams, after transmission through a polarizing element (G₃), are intercepted by a photodiode D3 and converted into a sinusoidal a.c. voltage V3 having a frequency equal to the fre¬ quency difference between g₁ and g₂. The use of a beam splitter B1 of this type for signal processing purposes is known per se and is not explained here in more detail. Its use is not essential for the action of the ellipsometer itself.

The remaining part of the beams g₁, g₂ passes the beam splitter B1 and reaches a second polarizing beam splitter B2. This second polarizing beam splitter B2 splits the two Zeeman beams g₁, g₂, because this polarizing beam splitter B2 is capable of differ¬ entiating between the different polarization directions of the two Zeeman beams g.,, g₂. The beam g₂ is used, in the present example, as a working beam and encounters the sample S, after the polariz- ation direction has been rotated with the aid of, for example, a Glan-Thompson prism G1 by an angle of 45°. The working beam g' ₂ reflected by the sample surface S then contains the optical infor¬ mation with regard to the surface of the sample S.

The Zeeman component g₁ propagates from the polarizing beam splitter B2 via a different path and first encounters a mirror M. After reflection on the mirror surface, said Zeeman component g._|, which in this case is used as a reference beam, encounters the rear side of a non-polarizing beam splitter N, after the polariz¬ ation direction has been rotated by a polarization unit, for example a Glan-Thompson prism G2, by an angle of 45°. The modified working beam g' ₂ encounters the front face of the non-polarizing beam splitter N, and after transmission through the non-polarizing beam splitter N the modified working beam g' ₂ interferes with the reference beam g₁. The composite beam is then split into two ortho- gonal polarizations (for example p and s) with the aid of, for example, a Wollaston prism W. The two orthogonal polarizations are again each converted, by means of a photodiode D1 , D2, into elec¬ trical a.c. voltages V1 , V2. The two electrical a.c. voltages V1 , V2 contain the same optical information as can be achieved with the set-up according to Figure 1 , albeit that the frequency of these a.c. voltages V1 , V2 is now equal to the frequency difference of the original Zeeman beams g.,, g₂, which is, for example, in the order of magnitude of 1 MHz. Therefore the measuring frequency is much higher than in the known device according to Figure 1. It can be seen in Figure 3 that the working beam g₂ arrives at an angle of 45° at the sample S. This considerably simplifies the interference of the working beam g'₂, modified by the surface of the sample S, with the reference beam g₁ on the rear side of the non-polarizing beam splitter N. Should it be desired to have the working beam g₂ arrive at the sample S at a different angle, additional mirrors or suitably disposed glass fibre cables should be employed to arrange for interference to take place between the modified working beam g' ₂ and the reference beam g^ after combination in the non-polariz¬ ing beam splitter N.

Various changes in the set-up of Figure 3 can be applied without moving outside the scope of the invention. Thus it is not strictly necessary for the polarization state of both the working beam g₂ and the reference beam g_, both to be rotated by a polariz¬ ation unit, for example the Glan-Thompson prisms G1, G2 shown, by an angle of 45°. It is also possible for one of the two measuring beams, that is either the working beam g₂ or the reference beam g₁ not to undergo polarization rotation and for only the other one to undergo a polarization rotation of 90°. It is only essential that the polarizations of the beams g₁, g₂ separated downstream of the beam splitter B2 match each other in such a way at the rear face of the non-polarizing beam splitter N, that interference can occur. Furthermore it is not absolutely necessary for the reference beam g-_j to be directed in the correct manner onto the rear face of the non-polarizing beam splitter N with the aid of one mirror M. If desired, it is also possible to use a plurality of mirrors, or alignment of the reference beam g₁ can take place with the aid of suitably chosen polarization-retaining glass fibre cables. An advantage of the set-up according to Figure 3 may be that the working beam g₂ reflects only once on the surface of the sample S. As a result, weakly reflecting samples can also be studied.

A preferred embodiment of an ellipsometer based on a Zeeman laser Z is shown in Figure 4. The same reference numerals refer to the same components as in the previous figures, and the description of these will not be repeated here. The Zeeman beam g₂ transmitted by the polarizing beam splitter B2 now impinges on the front side of a non-polarizing beam splitter N after passing through a polarization unit G1 which rotates the polarization direction by, for example, 45° and may be, for example, a Glan- Thompson prism. That part of the Zeeman beam g₂, which is reflected by the front face of the non-polarizing beam splitter N, impinges on the surface of the sample S as the working beam g_m2. After auto- collimation of this working beam g_m2 by a mirror M and repeated reflection on the surface of the sample S, a modified working beam g'_m2 is produced which propagates along the same path as the work- ing beam g_m2, albeit in the opposite direction. This double reflec¬ tion on the specimen is in general a major advantage of the ellip¬ someter according to Figure 1 , which is retained in the construc¬ tion of Figure 4. A doubling of the effect due to the specimen arises, and consequently the relative measuring error is reduced. The polarizing beam splitter B2 taps off the Zeeman beam g₁ and directs this onto, for example, three successive mirrors M1 , M2, M3, all this in such a way that the tapped-off beam g- can impinge, as the reference beam, on the rear face of the non-polar¬ izing beam splitter N. At a suitably chosen point between the polarizing beam splitter B2 and the rear face of the non-polarizing beam splitter N, the polarization state of the reference beam g₁ is rotated by an angle of 45° with the aid of a polarization unit G2, for example a Glan-Thompson prism. In Figure 4, the Glan-Thompson prism G2 is drawn between the mirror M3 and the rear face of the non-polarizing beam splitter N, but that is not strictly necessary. Any other position on the path along which the reference beam g₁ propagates is likewise possible. Just as in the set-up according to Figure 3, the reference beam g₁ interferes with the modified work¬ ing beam g'_m2 after combination on the rear face of the non-polar- izing beam splitter N. The composite reference beam is then split again with the aid of a Wollaston prism W into the two orthogonal polarizations (p and s). The electrical signals V1 , V2 contain the same optical information as those in Figure 3, apart from the double reflection in Figure 4, in analogy to the ellipsometer according to Figure 1.

The set-up according to Figure 4 has important resemblan¬ ces to the set-up according to Figure 1. An important difference is not only the use of a Zeeman laser Z, but also that the beam g₂, which impinges on the front face of the non-polarizing beam split¬ ter N, is only used for the benefit of the measuring objectives. Obviously, the non-polarizing beam splitter N transmits part of the incident Zeeman beam g₂, just as the non-polarizing beam splitter N in the set-up according to Figure 1 transmits part of the incident beam g to provide a reference beam g_r. In the set-up according to Figure 4, however, that part of the Zeeman beam g₂ which is trans¬ mitted by the non-polarizing beam splitter N is blocked or scat¬ tered, because this transmitted part no longer has to serve as a reference beam. Instead of three mirrors M1 , M2, M3, it is also possible to use, if desired, more or fewer mirrors. Moreover, the mirrors M1 , M2, M3 may be replaced by a suitably arranged polariz¬ ation-retaining glass fibre cable. It is further possible, if desired, to place additional polarization units in the paths of the beams g₂ and g₁ (while observing the criteria a, b, c) .

The set-up according to Figure 4 has advantages over that according to Figure 3. All the components of the ellipsometer, except the mirror M (and obviously the sample S) can be arranged in one compact housing, in which there is an opening for the working beam g_m2 and the returning modified working beam g'_m2. Because, in the set-up according to Figure 3, the incident working beam g₂ and the modified working beam g' ₂ travel along different optical paths, this is not as simple with the set-up according to Figure 3. More¬ over, the alignment of the sample S with respect to the ellipso- meter is much simpler. Using the set-up according to Figure 4, it is possible, in a simple manner, by changing the orientation of the sample S and the position of the mirror M, to adjust, if required, the incidence angle of the working beam _^ on the surface of the sample S. This can be effected, for example, by means of an accu- rate Θ-2Θ rotation apparatus as used in some ellipsometers and in x-ray diffraction equipment, the specimen rotating by an angle θ and the detector or, in the case of the set-up of Figure 4, the mirror M simultaneously rotating by an angle of 2Θ.

By virtue of the real-time option of the set-up of Figure 4, it is thus possible to make a rapid scan of the specimen as a function of the incidence angle. As stated earlier, this is not simple with the set-up according to Figure 3, and in the case of an incidence angle other than 45°, additional optical elements, for example mirrors or glass fibre cables, should be used.

The present invention is not limited to the embodiments as shown in the accompanying figures. If desired, the optical paths of the various working and reference beams may be varied with the aid of mirrors, glass fibres and/or polarization units.

Claims

1. Ellipsometer comprising at least a light source (Z) which during operation provides at least one light beam, a non- polarizing beam splitter (N) for causing a working beam (g'_ml. g'_m2' g'₂) modified by a sample (S), and a reference beam (g_r-_|. g_r ; g-,) to interfere, a unit (W) for separating two orthogonal components of the (interference) beam thus composed and providing two a.c. voltages (V1 , V2) corresponding thereto, characterized in that the light source is a Zeeman laser (Z) to generate two beams (g.,, g₂) which are slightly shifted in frequency and which are both polar¬ ized linearly but perpendicular to one another, and in that polar¬ ization means are provided for adjusting the polarization direc¬ tions of the working beam (g_ml, g_m2; g₂; g_m2) and/or the reference beam (g_r1. g_r2; g^ in order to effect the desired interference between the modified working beam and the reference beam.

2. Ellipsometer according to Claim 1 , provided with a mir¬ ror (C), for example a cubic reflector, for receiving the reference beam (g_r1, g_r2), produced by transmission of part of the beam (g_,, g₂) generated by the light source (Z) through the non-polarizing beam splitter (N), and for reflecting it to the rear side of the non-polarizing beam splitter (N), characterized in that, in the optical path between the rear side of the non-polarizing beam splitter (N) and the cubic reflector (C), a polarization unit is positioned (at point d) which, for example, rotates by 45°.

3. Ellipsometer according to Claim 1 , characterized in that a polarizing beam splitter (B2) is provided for separating the two beams (g , g₂) originating from the Zeeman laser (Z), the one beam (g₂) serving as the working beam and the other (g.,) as the reference beam.

4. Ellipsometer according to Claim 3, characterized in that, in the path of the working beam (g₂) between the polarizing beam splitter (B2) and the sample (S), a polarization element (G1 ) is positioned which rotates by an angle of, for example, 45°, and in the path of the reference beam (g.,) between the polarizing beam splitter (B2) and the rear side of the non-polarizing beam splitter (N) another polarization element (G2) is positioned which rotates the polarization direction by an angle of, for example, 45°, while the working beam (g₂) impinges on the surface of the sample (S) directly after transmission through the polarization element (G1 ) .

5. Ellipsometer according to Claim 3, characterized in that this is designed in such a way that the beam (g₂) transmitted by the polarizing beam splitter (B2) impinges on the front side of the non-polarizing beam splitter (N) and the part thereof which is reflected by the non-polarizing beam splitter provides the working beam (g_m2) and the part thereof which is transmitted by the non- polarizing beam splitter (N) is blocked or scattered.

*****