1. Introduction

Over the last decade, the semiconductor industry has been exploring innovative lithographic techniques to extend Moore’s law beyond the limitations of the more direct extrapolations of optical projection lithography. As deep-UV immersion lithography at 193 nm wavelength is reaching its limits, extreme ultraviolet lithography (EUVL) has become the leading candidate for the fabrication of semiconductor chips at the 16 nm technological node and beyond. However, as industry prepares for high-volume manufacturing (HVM), the lack of practical, compact EUV metrology tools capable of assessing mask printability has become a major concern.

The transfer of defects from the lithographic mask to the printed wafer is of great concern in EUVL. The reduction of the critical dimension (CD) of printed features significantly increases the restrictions on the size and density of allowable defects on the mask. Furthermore, new additional technological challenges, inherent to the use of EUV light for printing, need to be implemented for defect mitigation. For example, the lack of pellicles in EUVL steppers increases the probability of particle deposition on the masks and in turn the probability of printing defects on the wafers [1]. Also, printable defects in the form of phase defects arising from particles or pits buried within the Mo/Si multilayer stack can affect lithographic printing [2]. Reduction of critical dimensions (CD) also imposes tighter tolerances on printed patterns, as printing errors on small lines can result in non-operational devices.

Actinic (at-wavelength) aerial image microscopes can render information on the printability of patterns and defects prior to wafer printing and are greatly needed for the implementation of EUVL at HVM [3]. Since these microscopes use illumination wavelengths within the reflectivity bandwidth of Mo/Si multilayers, they can obtain magnified images of the EUVL masks as seen by the wafer in an EUVL stepper. The mask images can be used to characterize defects and to quantify the printability of mask patterns.

In photolithography, several parameters are used to assess the printability quality of the mask patterns. Normalized image log-slope (NILS), measured as the derivative of the logarithm of the image’s intensity, assesses the steepness of the intensity slope along the imaged features. Line edge roughness (LER) is a statistical measure of the variation of a feature’s edge along the feature’s extent and is typically expressed as three times the standard deviation, 3σ_dev. When the LER is below 5–10% of the CD value for the specific line, it is considered acceptable, inducing no printing errors. The values of NILS and LER are inversely proportional and depend on several factors including the quality of the illumination, the uniformity of the light and the aberrations of the optical system; the quality of the multilayer coating and absorber pattern of the EUVL mask; and, if measured on a printed wafer, the response of the photoresist employed.

Current actinic microscopes are based on synchrotron radiation illumination [4,5], and although they have allowed great progress in the development of EUVL masks, they are not practical for HVM. The availability of compact EUV light sources with sufficient brightness for image acquisition is currently a main limitation to the implementation of these imaging systems for HVM. High repetition rate, table-top, EUV lasers are particularly well suited for full-field microscopy due to their high brightness and beam directionality. They also have high spectral purity which is advantageous when using wavelength-dependent optics such as diffractive zone plates. At 13.2-nm wavelength, a table-top microscope capable of rendering images of transmission samples with a spatial resolution better than 38 nm has been demonstrated at Colorado State University using table-top EUV laser illumination [6]. Using the same type of illumination source in combination with specially designed zone plate optics, a reflection microscope for actinic aerial imaging of EUVL masks with approximately 55 nm spatial resolution was developed [7]. However, in that first demonstration the quality of the images was insufficient for printability measurements, mostly due to illumination non-uniformities.

In this paper we demonstrate improved capabilities of a 13.2 nm EUV laser-based table-top actinic aerial image microscope for analyzing mask patterns and defect printability, based on NILS and LER measurements of mask images that have illumination non-uniformities below 10%. We also report on the analysis of the coherence of the illumination of the EUV laser-based actinic microscope. The combination of these results provides the information necessary to assess mask printability independently of resist response.

2. EUV laser-based aerial image measurement system

The table-top EUV aerial imaging microscope uses zone plate lenses and 13.2 nm wavelength laser illumination in a configuration similar to that described by Brizuela et al. [7]. The illumination is provided by a table-top, plasma-based, collisional EUV laser [8,9]. This laser has been recently operated at repetition rates of up to 2.5 Hz producing an average power of approximately 20 µW at 13.9 nm wavelength using a Ag target [10]. To obtain the results reported here the laser was operated at 1 Hz repetition rate to produce an average power of a several µW in the 13.2 nm line of nickel-like cadmium. The laser beam is generated by amplification of spontaneous emission in a transient population inversion produced by electron impact excitation of nickel-like cadmium ions [8,9]. This transient population inversion is generated by heating an 8-mm-wide cadmium slab target with a sequence of pulses from a chirped-pulse-amplification Ti:Sapphire laser system. Pre-pulses are focused into a 30 µm FWHM wide × 6.3 mm FWHM long line, to generate a cadmium plasma that is allowed to expand to reduce electron density gradients. Subsequently the plasma is rapidly heated by an intense laser pulse of 6 ps duration impinging at a grazing incidence angle of 23 degrees, generating the transient population inversion. The laser produces highly monochromatic (Δλ/λ<1x10⁻⁴) light pulses with a divergence of 9 ± 0.5 mrad FWHM parallel to the slab target and 10 ± 0.5 mrad in the perpendicular direction.

The optics and configuration of the microscope were selected to emulate the imaging conditions of a 0.25 NA, 4 × -demagnification lithographic stepper [3]. This ensures that the images obtained with the microscope are a magnified version of how the mask would print on the wafer. As shown in Fig. 1 , a Mo/Si multilayer-coated mirror positioned at a 42° angle of incidence collects the EUV beam onto a condenser zone plate that illuminates a 250 µm² region of the EUVL mask at an angle of 6°. The light reflected by the mask is collected by the objective zone plate which projects a magnified image onto an EUV-sensitive CCD. The zone plates were made on silicon nitride membranes using e-beam lithography [11]. The condenser, with a diameter of 5 mm and NA = 0.066, allows complete collection of the laser beam. The condenser was designed to have no central stop. This produces an illumination spot in the form of a disk and increases the condenser throughput by 12%. The objective is an off-axis zone plate with a NA of 0.0625, corresponding to the numerical aperture of the projection optics of the EUVL stepper as seen by the mask. With this zone plate, a spatial resolution of 55 nm half-pitch can be achieved on the mask, corresponding to a resolution of ~14 nm on the wafer, considering a 4 × -demagnification stepper. The off-axis design of the objective allows for near normal incidence imaging while providing a mechanism for separating the 0th order light from the image forming 1st order at the image plane. During image acquisition, the condenser was moved from shot to shot in the plane perpendicular to the propagation of the light using a two-axis piezoelectric stage. This improves the uniformity of the illumination and reduces coherent effects that might build up in the image such as diffraction along the edges of the features.

 

Fig. 1 Diagram of the microscope setup. The light from the laser is guided by a Mo/Si turning mirror onto a condenser zone plate which focuses the light onto the EUVL mask. The light reflected off the mask is collected by an off-axis zone plate which projects a magnified image onto an EUV-sensitive CCD.

Download Full Size | PDF

The optical elements and the sample are housed in a 70 × 45 × 40 cm³ vacuum chamber connected through a manifold to the output port of the EUV laser. Thin, 100 nm, zirconium filters are placed before the vacuum manifold to filter unwanted plasma light. The microscope can accommodate full size EUVL masks on position-controlled stages that allow the selection of the region of the mask to be imaged. Two samples were available for inspection. One consisted of a Mo/Si multilayer-coated wafer with an absorption test pattern consisting of elbow gratings with features as small as 80 nm. The second sample was a 6 × 6 in² Mo/Si multilayer-coated low thermal expansion glass EUVL mask that contained a series of dark-field and bright-field regions with a variety of test features and extended line patterns [12]. The coatings of both samples were centered at 13.5 nm wavelength.

Figure 2 shows the EUV image of a 180 nm half-pitch elbow absorption pattern in a bright field of a Mo/Si EUVL mask. The image was obtained with an exposure time of 90 seconds with the laser operating at a repetition rate of 1 Hz. The image shows a ~10% variation in intensity in the central 4 × 4 µm² region of the field of view. Speckle features arising from the Mo/Si mask roughness can be seen in the bright areas of the image [13]. EUV images for sample recognition were acquired with exposure times as short as 5 seconds. Longer exposure times of ~90 seconds were used to achieve the high image quality, such as that of Fig. 2, necessary for pattern printability analysis. In both cases, a magnification of 660 × was used, resulting in an equivalent pixel size of approximately 20 nm.

 

Fig. 2 EUV image of a 180 nm half-pitch absorption elbow pattern in a bright field obtained with an exposure of 90 seconds. The color bar indicates the number of counts per pixel.

Download Full Size | PDF

The EUV images are affected by the degree of coherence of the optical system. The degree of coherence, evaluated through the coherence parameter, σ, is influenced by the spatial coherence length of the illumination source and the numerical aperture of the objective optics. For a completely coherent optical system, σ equals zero, while σ = ∞ corresponds to completely incoherent illumination [14]. The coherence parameter can be determined by analyzing the modulation intensity from images of a grating pattern placed at different distances from the image plane (through-focus analysis) [15]. In an incoherent system high intensity modulation is obtained only at the position that fulfills the imaging condition. On the other hand, for a completely coherent system, images of the test pattern appear at the Talbot planes, located at a distance n·a²/λ from the focus, where a is the pitch of the grating, λ the illumination wavelength and n an integer. In this case, a plot of the image intensity modulation shows a series of peaks of the same intensity at different defocusing positions. If the system is partially coherent, the modulation of these secondary intensity peaks decreases as defocusing increases. The coherence parameter of the table-top actinic microscope was evaluated by analyzing the through-focus performance of a 200 nm half-pitch grating. As shown in Fig. 3 , the appearance of secondary maxima as far as 25 µm away from the focal plane reveals the system’s partial spatial coherence. Comparison with simulations using SPLAT [16] indicates that the coherence parameter for the microscope is σ ~0.25. This value is lower than the value of 0.5 of a 4 × EUVL stepper [3], but higher than that of current synchrotron-based systems that operate in a higher coherence regime (σ = 0.1–0.2) [15,17].

 

Fig. 3 Normalized modulation intensity of a bright-field EUV image of a 200 nm half-pitch grating at different positions from the focal plane. The intensity of the secondary maxima reveals a partial spatial coherence of approximately 0.25 as simulated using SPLAT.

Download Full Size | PDF

3. Evaluation of pattern printability of an EUVL mask

The table-top microscope was used to evaluate the printability of an EUVL mask containing different size features and patterns [12]. Figure 4 shows an image of horizontal and vertical 175 nm half-pitch 1:1 gratings located in one of the mask dark-field subfields. Both sets of lines are fully resolved and no discernable astigmatism is observed in the image. Although the resolution of the microscope was evaluated using absorber patterns in a bright field, the evaluation of pattern printability through the measurement of NILS and LER was done on bright lines in a dark field which present a better contrast due to reduction in flare [15].

 

Fig. 4 EUV image of 175 nm half-pitch gratings in a dark field.

Download Full Size | PDF

NILS and LER were measured for the EUV images of 175 nm and 225 nm half-pitch gratings shown in Figs. 5(a) and 5(d) respectively. To obtain these parameters, the images were normalized in intensity and for each an intensity threshold corresponding to the normalized intensity value for which the lines and spaces have a 1:1 ratio was obtained. These thresholds are 0.40 and 0.42 for the 175 nm and the 225 nm half-pitch gratings respectively. The averaged normalized intensity plots and intensity thresholds, for both images are shown in Fig. 5(b) and 5(e). NILS values were measured at the intensity threshold for each slope of the averaged cross-sections. LER for each line was measured by evaluating the magnification-corrected location of the occurrence of the intensity threshold for each pixel row of the images. The standard deviation of these values was subsequently calculated for each edge. The measurements were performed across the central 1 µm region of the gratings. Figures 5(c) and 5(f) show the resulting plots. In each case, the LER value is below 10% of CD (LER/CD < 0.1). The results of the NILS and LER analysis are summarized in Table 1 .

 

Fig. 5 LER analysis of a) 175 nm and d) 225 nm half-pitch gratings. Normalized intensity versus position plots are given in b) and c). e) and f) provide a graphical representation of the variation of the edge position along the features. The analysis shows that the 3σ variation is below 10% of the CD value.

Download Full Size | PDF

Table 1. NILS and LER measurements of two grating structures with indicated CD values

View Table

4. Summary

An actinic aerial image microscope based on a table-top EUV laser, capable of obtaining high quality images for the evaluation of EUVL mask pattern printability, was demonstrated. The system, designed to mimic the printing conditions of a 0.025 NA 4 × -demagnification EUVL stepper, can obtain images with a spatial resolution of approximately 55 nm within exposure times of 5 to 90 seconds. These results are comparable to current synchrotron-based tools. The uniformity and intensity of the illumination enable the measurement of key parameters of mask pattern printability such as NILS and LER. Extension to higher resolution and thus simulation of higher NA EUVL steppers necessary to reach smaller printing nodes can be done by increasing the NA of the off-axis objective zone plate. Using this approach, zone plates capable of simulating NA above 0.35 are currently available [17]. The results presented demonstrate the capabilities of the microscope for EUVL mask research and development and open the path to the realization of practical standalone EUVL metrology tools for HVM.