Method and apparatus for evaluating the thickness of thin films.

Technical Field

0001The subject invention relates to an apparatus for evaluating the thickness or optical constants of thin films. The subject invention is particularly useful in measuring films that are less than 100Å thick.

Background of the Invention

0002There is considerable interest in developing systems for accurately measuring the thickness of thin films and the optical constants of those thin films. The need is particularly acute in the semiconductor manufacturing industry where the thickness of thin film oxide layers on silicon substrates are measured. To be useful, the measurement devices must be able to determine the thickness of films with a high degree of accuracy and be able to make the measurements within very localized areas (ie. on the micron scale).

0003One such device that achieves these goals has been recently introduced by the assignee herein under the trademark Optiprobe. The approach used to obtain the measurements in this device is described in detail in commonly owned U.S. Patent No. 4,999,014, issued March 12, 1991, to Gold et. al. and incorporated herein by reference.

0004As described in that patent, a high power, spherical, microscope objective is used to tightly focus a probe beam substantially normal to the surface of the sample in a manner to create a spread of angles of incidence. An array of discrete photodetector elements is provided to measure the intensity of individual rays within the reflected probe beam. These individual rays correspond to discrete angles of incidence with respect to the surface of the sample. Using the angularly dependent intensity measurements, the processor can evaluate the parameters of the thin film. Parameters such as layer thickness and index of refraction can be determined using the Fresnel equations.

0005Some of the fundamental concepts that were disclosed in Patent No. 4,999,014 were extended to develop a high resolution ellipsometer which is described in commonly owned U.S. Patent No. 5,042,951, issued August 27, 1991, to Gold et. al. and incorporated herein by reference. In that device, the change in polarization state of the probe beam, caused by the reflection off the surface of the sample, was analyzed so that the conventional ellipsometric parameters could be used to evaluate the sample. As in the earlier cited patent, the detector arrangement in the ellipsometer included an array of discrete detector elements so that angularly dependent intensity measurements could be made.

0006Both of the devices as described above provide information about thin film layers. However, and as described in both patents, additional accuracy can be obtained by measuring not only the angularly dependent intensity of the rays with an array of discrete elements, but in addition, by measuring the full power of the reflected probe beam. The advantage to measuring the full power of the reflected probe beam is that it provides a much better signal to noise ratio. It should be noted that a measurement of the full power of the reflected probe beam could not be used to determine thin film thickness alone, since this signal is not linear, but varies sinusoidally with thickness. However, if the approximate thickness can be derived using the angular dependent intensity measurements, the actual thickness can be more accurately determined using the additional full power measurement.

0007In practice, using both angular dependent intensity detection and full probe beam power detection, accurate measurement of the thickness and optical constants of thin films down to 100Å has been achieved. Below 100Å, the interference effects of the film are so small that the signal to noise ratio associated with the discrete element photodetector makes analysis by angular dependent intensity measurements difficult. In addition, the sensitivity of the standard full power measurement in the region below 100Å is also low.

0008Therefore, it is an object of the subject invention to provide an approach for measuring the thickness of films less than 100Å thick.

0009It is another object of the subject invention to provide an apparatus for determining the optical constants of films less than 100Å thick.

0010It is a further object of the subject invention to provide a device for evaluating parameters of a thin film with micron scale resolution.

0011It is a further object of the subject invention to provide a method and apparatus which can be used in conjunction with existing systems to further refine the measurement of the thickness and optical constants of thin films.

Summary of the Invention

0012In accordance with these and other objects, a method and apparatus is disclosed for evaluating parameters of thin films. The apparatus includes a means for generating a probe beam of radiation. A high power, spherical, microscope objective is used to tightly focus the beam substantially normal to the surface of the sample to create a spread of angles of incidence. The probe beam is preferably focused to a spot less than a micron in diameter. A means is provided for retarding the phase of one polarization state of the beam with respect to the other. In the illustrated embodiment, where the probe beam is initially linearly polarized, a quarter-wave plate is used to retard the phase of one polarization state of the reflected probe beam by ninety degrees.

0013The reflected beam is then passed through a linear polarizer to create interference between the two polarization state components. The power of the beam is measured by a photodetector along two orthogonal axes. In the preferred embodiment, a quad cell is used having four radially disposed quadrants. Each quadrant will receive all the rays having different angles of incidence and generate an output signal of total power that effectively integrates the angular intensity measurements. As will be described in greater detail below, because of the geometry of the optical components, the power in each quadrant will be the same except for one term which will be positive for one set of diametrically opposed quadrants and negative for the remaining quadrants. By subtracting the sum of the outputs of the diametrically opposed quadrants, a signal can be obtained which is linearly proportional to the thickness of the thin film.

0014Since the signal in each quadrant of the detector derives its magnitude from the power in an entire quadrant of the beam, the signal to noise ratio will be better than that achieved using only small discrete detectors that measure intensity only at specific points within the beam. Further, since the measurement includes information over a range of angles of incidence, the signal is highly sensitive in the region of thin film thickness below 100Å. This result can be compared to the low level of sensitivity available in a standard full power measurement.

0015Although the sensitivity of this method, similar to the full power measurement, varies sinusoidally with thickness, it has been found that the subject measurement will also be useful above 100Å. More particularly, in film thickness regions between zero and 5000Å, the sensitivity of this method is greatest in those regions where the full power measurement is a minimum. Accordingly, the subject method can be advantageously implemented with the angular dependent measurements described above to further refine the analysis in thickness regions where the total power measurement is less sensitive.

0016It should be understood that since this approach functions to integrate the angular intensity measurements, information available by measuring individual rays will be lost. Thus, a detailed analysis of the parameters of very thin films would be difficult based on this measurement alone. However, the measurements derived from this approach can be compared to known samples to give a highly accurate assessment of the specific parameters of very thin films. In most cases, the unknown parameter to be evaluated will be the thickness of the thin film. If the thickness is known, the system can be used to determine another optical constant such as the index of refraction of the thin film.

0017Further objects and advantages of the subject invention will become apparent from the following detailed description taken in conjunction with the drawings in which:

Brief Description of the Drawings

0018Figure 1 is a block diagram of the apparatus of the subject invention.

0019Figure 2 is a schematic diagram of the surface of a quad cell photodetector used in the subject invention.

0020Figure 3 is a graph illustrating a comparison of the variation in the amplitude of the output signal with respect to oxide layer thickness for a detector that measures total reflectivlty and one that measures the signals in accordance with the subject invention (quad-cell difference signal).

0021Figure 4 is an enlarged version of the graph of Figure 3 illustrating the signal in the region of oxide thickness below 500Å.

0022Figure 5 is a block diagram of an alternate embodiment configured to minimize errors due to asymmetries in the probe beam.

Detailed Description of the Preferred Embodiments

0023Turning to Figure 1, an apparatus 10 is illustrated for performing the method of the subject invention. The apparatus 10 is designed to evaluate the thickness of a thin film layer 12 on a sample 14. In particular, the apparatus is capable of measuring the thickness of a thin film oxide layer on a silicon substrate.

0024Apparatus 10 includes a laser 20 for generating a probe beam 22 of radiation. One suitable laser source is a solid state laser diode which emits a linearly polarized beam. Such a laser is available from Toshiba, model TLD 9211 having a 3 milliwatt power output at 670nm.

0025Probe beam 22 is turned towards the sample 14 with a 50/50 beam splitter 24. The probe beam is focused onto the surface of the sample with a lens 26. In the preferred embodiment, lens 26 is defined by a spherical, microscope objective with a high numerical aperture on the order of .90 NA. The high numerical aperture functions to create a large spread of angles of incidence with respect to the sample surface. The spot size is on the order of one micron in diameter.

0026A fraction of the probe beam power also passes through splitter 24 and falls on an incident power detector 30. As discussed in the above cited patents, the incident power detector 30 is provided to monitor fluctuations in the output power of the probe laser. As will be discussed below, the incident power detector can be modified for this invention to minimize measurement errors which arise due to asymmetries of the beam.

0027Light reflected from the surface of the sample passes up through splitter 24 towards photodetector 40. Prior to reaching detector 40, the beam 22 is passed through a quarter-wave plate 42 for retarding the phase of one of the polarization states of the beam by 90 degrees. It should be noted that the quarter-wave plate could be located in the beam path prior to the probe beam striking the sample so that the system would operate with circularly polarized light. The latter approach might have some advantages in reducing the aberrations created by lens 26. In addition, while a phase retardation of 90 degrees will maximize the desired signal, other intermediate levels of retardation would be possible.

0028The beam is then passed through a linear polarizer 44 which functions to cause the two polarization states of the beam to interfere with each other. In order to maximize the desired signal, the axis of the polarizer should be oriented at an angle of 45 degrees with respect to the fast and slow axes of the quarter-wave plate 42.

0029In accordance with the subject invention, detector 40 is configured to generate independent signals from regions along two orthogonal axes. In the preferred embodiment, this goal is achieved by using a quad cell photodetector. As illustrated in Figure 2, the detector surface includes four radially disposed quadrants 52, 54, 56 and 58. Each quadrant will generate an output signal proportional to the magnitude of the power of probe beam striking the quadrant. This signal represents an integration of the intensities of all the rays having different angles of incidence with respect to the sample surface. While this integration approach results in the loss of some information content as compared to an analysis of individual rays, the composite approach does provide significantly greater sensitivity through enhanced signal to noise performance.

0030The probe beam 22 should be centered on the detector so that each quadrant intercepts one quarter of the probe beam. The probe beam should underfill the detector.

0031As noted above, the geometry of the optical elements results in creating a signal response that can be processed to form a result that varies linearly with layer thickness. This response can be understood from the following analysis which is described with reference to linearly polarized light. As noted above, the system can operate with circularly polarized light as well.

0032In the following analysis, it is assumed that the incident light has the form E_i(r,φ)ê_x where r is the radial distance from the center of the beam and φ is the angle relative to the x-axis. After passing through the objective lens 26, reflecting off of the sample surface, and returning through the lens, the electric field will be spatially dependent and have the form <maths id="math0001"><math display="inline"><mrow><msub><mrow><mtext>E</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msub><mrow><mtext>(r,φ) = E</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><msub><mrow><mtext>(r,φ){(R</mtext></mrow><mrow><mtext>p</mtext></mrow></msub><msub><mrow><mtext>cos²φR</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><msub><mrow><mtext>sin²φ)ê</mtext></mrow><mrow><mtext>x</mtext></mrow></msub><msub><mrow><mtext> + (R</mtext></mrow><mrow><mtext>p</mtext></mrow></msub><msub><mrow><mtext> - R</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><msub><mrow><mtext>)sinφcosφê</mtext></mrow><mrow><mtext>y</mtext></mrow></msub><mtext>}   (1)</mtext></mrow></math><img file="EP0549166A2_D0001.tif" /></maths> where ê_x and ê_y are the unit vectors along the x and y axes, R_p is the complex p-wave amplitude reflection coefficient, and R_s is the complex s-wave amplitude reflection coefficient. The radial position within the beam, r, is related to the angle of incidence upon the sample, Θ, through the expression r=dsinΘ, where d is the focal length of the objective lens. R_p and R_s are functions of Θ and are independent of φ.

0033If the beam 22 is passed through the quarter-wave retardation plate 42 oriented such that the x-component is retarded 90° relative to the y-component, and then is passed through the linear polarizer 44 oriented at angle α to the x-axis, the electric field will be of the form <maths id="math0002"><math display="inline"><mrow><msub><mrow><mtext>E</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msub><mrow><mtext>(r,φ,α) = E</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><mtext>[</mtext><mtext mathvariant="italic">i</mtext><msub><mrow><mtext>cosα(R</mtext></mrow><mrow><mtext>p</mtext></mrow></msub><msub><mrow><mtext>cos²φR</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><msub><mrow><mtext>sin²φ)ê</mtext></mrow><mrow><mtext>x</mtext></mrow></msub><msub><mrow><mtext> + sinα(R</mtext></mrow><mrow><mtext>p</mtext></mrow></msub><msub><mrow><mtext>-R</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><msub><mrow><mtext>)sinφcosφê</mtext></mrow><mrow><mtext>y</mtext></mrow></msub><mtext>]   (2)</mtext></mrow></math><img file="EP0549166A2_D0002.tif" /></maths> The light intensity is equal to the square of the field magnitude, I_R=|E_R|². I_R can be expressed in terms of the ellipsometric parameters ψ and δ through the following relationships:<maths id="math0003"><img file="EP0549166A2_D0003.tif" /></maths> and<maths id="math0004"><img file="EP0549166A2_D0004.tif" /></maths> Carrying out the expansion of |E_R|², we get <maths id="math0005"><math display="inline"><mrow><msub><mrow><mtext>I</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msub><mrow><mtext>(r,φ,α) = I</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><msub><mrow><mtext>(r,φ)|R</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><mtext>|²{[tan²ψcos⁴φ+sin⁴φ]cos²</mtext></mrow></math><img file="EP0549166A2_D0005.tif" /></maths> α <maths id="math0006"><math display="inline"><mrow><mtext>+ [tan²ψ+1]sin²φcos²φsin²α + 2tanψcosδcos²φsin²φ(cos²α-sin²α) + 2tanψsinδcosφsinφcosαsinα}   (5)</mtext></mrow></math><img file="EP0549166A2_D0006.tif" /></maths> For very thin films, tan ψ is independent of thickness and δ is linearly proportional to the thickness. Hence, the last term in the above expression is of greatest interest. As noted above, this term is maximized when the analyzer angle, α, is 45°.

0034The total power impinging on each quadrant of the photodetector is equal to the integral of I_R over the quadrant. If the incident laser beam, I_i, is circularly symmetric, the power in each quadrant is given by<maths id="math0007"><img file="EP0549166A2_D0007.tif" /></maths> The sign of the second integral is positive for quadrants 54 and 58 and negative for quadrants 52 and 56. If the signals from quadrants 52 and 56 are summed and subtracted from the sum of the signals from quadrants 54 and 58, the remainder will have a simple form that is linearly proportional to film thickness for thin films:<maths id="math0008"><img file="EP0549166A2_D0008.tif" /></maths> For very thin films or small δ then<maths id="math0009"><img file="EP0549166A2_D0009.tif" /></maths>    The graphs in Figures 3 and 4 illustrate the variation of the output signal S with thickness for a silicon oxide (SiO₂) film on a silicon (Si) substrate. The incident laser beam was assumed to have a Gaussian intensity profile and the numerical aperture of the objective lens was 0.9. For comparison, the sum of the four quadrants is also shown which corresponds to the total reflectivity measurement described in the above cited patents.

0035These curves demonstrate the superior sensitivity of the quad-cell difference signal for films thinner than 100Å. They also show that the sum and difference signal techniques are complementary. More specifically, when the difference signal is at a peak or valley, and thus where its thickness sensitivity is at a minimum, the total reflectivity signal has a steep slope. Conversely, when the total reflectivity signal is at a peak or valley, the difference signal has a steep slope. Thus, by using a combination of the two techniques, it is possible to take full advantage of the signal-to-noise benefits of large-area detection for all thicknesses from 0Å to at least 5000Å.

0036To carry out the analysis, the output signals from the detector 40 are supplied to a processor 60. Processor 60 functions to find the difference between the sums of the signals from diametrically opposed quadrants 52, 56 and 54, 58. The difference value can be compared with a calibration table of values for known thicknesses. As can be seen from Figure 4, there is a significant variation in signal in the region below 500Å so that accurate measurements can be made. In the case where thickness of the film is known, the difference signal could be used to determine the other optical constants of the film, specifically the index of refraction and the extinction coefficient.

0037In the broadest sense, the quad-cell difference signal provides a single, very accurate measurement of the ellipsometric parameter δ (when measuring very thin films) which is related to the optical thickness (nt) in thin films and the extinction coefficient of the substrate and can be used to evaluate either parameter. For thicker films, the information provided by the quad-cell difference signal is a complicated function of ψ and δ, but the point remains that the difference signal can be used to refine the calculation of any unknown quantity (thickness, index of refraction or extinction coefficient) given the others.

0038It is envisioned that the subject approach will be incorporated into the devices described in the above cited patents. This incorporation can be very straightforward since the devices described above were already equipped with most of the elements necessary to carry out the subject invention. Significantly, the prior devices both included a full power detector (element 340 in both patents). This standard full power detector can be replaced with a quad cell detector. As can be appreciated, if a full power measurement is desired, all four quadrants can be summed. In contrast, if a difference signal is desired in accordance with the subject approach, the processor can subtract the sums of the diametrically opposed quadrants. The prior devices would be further modified by the addition of the appropriate retardation and polarizing elements.

0039It is envisioned that for films having thicknesses greater than 100Å, the initial evaluation would be based on the angularly dependent intensity measurements. Refinement of the measurement would be made using either the full power measurement or the subject difference measurement. The choice will be based on the sensitivity of the signal in the particular thickness region being measured.

0040The mathematical analysis set forth above assumed that the probe beam was circularly symmetrical or independent of φ. If it is not, the leading terms in equation (5) will not cancel upon subtraction. This residual term introduces an offset in thin-film measurements and thereby complicates the interpretation of the results. Furthermore, the extra term can degrade the signal-to-noise of the device since it increases the noise level without contributing useful information about the film thickness. Various signal processing techniques can be used to minimize such errors.

0041One approach to minimizing such errors would be to rotate the linear polarizer 44 and record the signals at α=0° and α=90°. From the expression for I_R, equation (5), it can be seen that <maths id="math0010"><math display="inline"><mrow><msub><mrow><mtext>I</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msub><mrow><mtext>(r,φ,α=45°) = ½{I</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msub><mrow><mtext>(r,φ,α=0°) + I</mtext></mrow><mrow><mtext>R</mtext></mrow></msub><msub><mrow><mtext>(r,φ,α=90°)} + sinδtanψcosφsinφI</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><msub><mrow><mtext>(r,φ)|R</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><mtext>|²   (10)</mtext></mrow></math><img file="EP0549166A2_D0010.tif" /></maths> By subtracting 0° and 90° signals from the 45° signal, the term proportional to small δ can be isolated. Since the subtraction is done for each individual quadrant, circular asymmetry will not effect the result. As indicated in Figure 1, rotation of polarizer 44 can be placed under the control of the processor 60.

0042A second approach is illustrated in Figure 5. In this approach, two identical quad-cell detectors 70 and 72 are used simultaneously to measure the right and left-handed circular components of the reflected beam. This can be accomplished most directly by replacing the final linear polarizer 44 with a polarizing beam splitter 74. After passing through the polarizing beam splitter 74, two beams are created. One beam will be the right-handed component and the other will be the left-handed component. The power hitting the two quad cells is identical except for a sign change in the term linear in small δ. By this arrangement, a subtraction of the signal generated by one quadrant in one detector from the signal generated by the corresponding quadrant in the other detector will produce a result that is linear for thin films. The signal can be maximized by performing a quadrant by quadrant subtraction and summing the result.

0043While either of the above two techniques will guarantee the removal of terms insensitive to small δ, the remaining linear term will still be weighted by any asymmetry of the beam through its dependence on I_i(r,φ). For the most part, the r and φ dependence of the incident light can be calculated by measuring the signals from known samples and comparing them to theory. However, to keep track of dynamic changes in the incident beam, it is necessary to monitor the incident intensity with a separate photodetector.

0044In the devices described in the above cited patents, a single photodetector measures the total intensity of the incident beam to remove the effects of the variation in the intensity of the beam generated by the laser. This concept can be extended by modifying the incident beam detector to a quad cell detector (shown as 30 in Figure 1). By using a quad cell detector, small changes in the symmetry of the incident beam can also be measured. By individually normalizing the signals from the quadrants of detector 40 with the incident power measured in the corresponding quadrants of detector 30, the effects of symmetry changes can be greatly reduced.

0045In summary, there has been disclosed a method and apparatus for evaluating a parameter of a thin film. The apparatus includes a linearly polarized probe laser beam which is tightly focused on the sample surface to create a spread of angles of incidence. The reflected probe beam is passed through a quarter-wave plate and a linear polarizer before impinging on a quad cell photodetector. The output signals from the photodetector represent an integration of the intensity of individual rays having various angles of incidence. By taking the difference between the sums of the output signals of diametrically opposed quadrants, a value can be obtained which varies linearly with film thickness for very thin films. The subject device can be used in conjunction with other prior devices to enhance sensitivity for thicker films.

0046While the subject invention has been described with reference to the preferred embodiments, various changes and modifications could be made therein, by one skilled in the art, without varying from the scope and spirit of the subject invention as defined by the appended claims.