Reference-Free Total Reflection X-ray Fluorescence Analysis of

Publication Date (Web): September 20, 2007 ... and Si(Li) detectors are used to monitor the exciting radiant power respectively the fluorescence radia...
0 downloads 0 Views 875KB Size
Download PDF
Anal. Chem. 2007, 79, 7873-7882

Reference-Free Total Reflection X-ray Fluorescence Analysis of Semiconductor Surfaces with Synchrotron Radiation Burkhard Beckhoff,* Rolf Fliegauf, Michael Kolbe, Matthias Mu1 ller, Jan Weser, and Gerhard Ulm

Physikalisch-Technische Bundesanstalt, Abbestrasse 2-12, 10587 Berlin, Germany

Total reflection X-ray fluorescence (TXRF) analysis is a well-established method to monitor lowest level contamination on semiconductor surfaces. Even light elements on a wafer surface can be excited effectively when using high-flux synchrotron radiation in the soft X-ray range. To meet current industrial requirements in nondestructive semiconductor analysis, the Physikalisch-Technische Bundesanstalt (PTB) operates dedicated instrumentation for analyzing light element contamination on wafer pieces as well as on 200- and 300-mm silicon wafer surfaces. This instrumentation is also suited for grazing incidence X-ray fluorescence analysis and conventional energydispersive X-ray fluorescence analysis of buried and surface nanolayered structures, respectively. The most prominent features are a high-vacuum load-lock combined with an equipment front end module and a UHV irradiation chamber with an electrostatic chuck mounted on an eight-axis manipulator. Here, the entire surface of a 200or a 300-mm wafer can be scanned by monochromatized radiation provided by the plane grating monochromator beamline for undulator radiation in the PTB laboratory at the electron storage ring BESSY II. This beamline provides high spectral purity and high photon flux in the range of 0.078-1.86 keV. In addition, absolutely calibrated photodiodes and Si(Li) detectors are used to monitor the exciting radiant power respectively the fluorescence radiation. Furthermore, the footprint of the excitation radiation at the wafer surface is well-known due to beam profile recordings by a CCD during special operation conditions at BESSY II that allow for drastically reduced electron beam currents. Thus, all the requirements of completely reference-free quantitation of TXRF analysis are fulfilled and are to be presented in the present work. The perspectives to arrange for reference-free quantitation using X-ray tube-based, table-top TXRF analysis are also addressed. Metallic and nonmetallic contamination on wafers can cause considerably reduced yields of structured semiconductors. Typical cleanliness requirements of basic semiconductor materials, such as silicon wafers, previously ranged between 109 and 1011 atoms/ cm2. These levels have been steadily decreasing for more than a decade. New and more differentiated requirements are being * Corresponding author. E-mail: [email protected]. Fax: +49 30 63925082. 10.1021/ac071236p CCC: $37.00 Published on Web 09/20/2007

© 2007 American Chemical Society

considered for lighter elements and compounds. Through continuing research and development,1-10 total reflection X-ray fluorescence analysis (TXRF) has become a well-accepted inspection method for monitoring wafer cleanliness. Whole wafer surface control is accessible either by scanning or by sweeping TXRF techniques, providing information on the local contamination, or by vapor-phase decomposition (VPD) preparation of the native silicon oxide, yielding some kind of average value of the integral wafer surface contamination. Collecting the entire dissolved surface contamination by VPD from the wafer and depositing it on a single spot leads to improved lower limits of detection (LLD or LOD), however, inhomogeneously dried VPD droplets9 may result in severe problems of TXRF quantitation due to both varying thickness and lateral extension profiles. Experimental stations dedicated to the TXRF analysis on silicon wafer surfaces using synchrotron radiation mainly in the hard X-ray range were commissioned at both the Stanford Synchrotron Radiation Laboratory2 and the European Synchrotron Radiation Facility3 for the analysis of transition metals and selected lighter elements. Complementarily, the Physikalisch-Technische Bundesanstalt (PTB) has laid emphasis on the methodological development of TXRF wafer analysis in the soft X-ray range, allowing for the efficient excitation of all light elements. For this purpose, special TXRF instrumentation fulfilling industrial requirements for wafer handling has been commissioned. Furthermore, calibrated detectors have been used, enabling reference-free quantitation of contamination. The present paper addresses the following topics: a detailed description of the dedicated TXRF instrumentation (1) Klockenka¨mper, R. Total-Reflection X-Ray Fluorescence Analysis; John Wiley and Sons Inc.: New York, 1997 (2) Pianetta, P.; Takaura, N.; Brennan, S.; Tompkins, W.; Laderman, S. S.; Fischer-Colbrie, A.; Shimazaki, A.; Miyazaki, K.; Madden, M.; Wherry, D. C.; Kortright, J. B. Rev. Sci. Instrum. 1995, 66, 1293. (3) Comin, F.; Navizet, M.; Mangiagalli, P.; Apostolo, G. Nucl. Instrum. Methods, B 1999, 150, 538. (4) Streli, C.; Wobrauschek, P.; Kregsamer, P.; Pepponi, G.; Pianetta, P.; Pahlke, S.; Fabry, L. Spectrochim. Acta, Part B 2001, 56, 2085. (5) Mori, Y.; Uemura, K.; Iizuka, Y. Anal. Chem. 2002, 74, 1104. (6) Sakurai, K.; Eba, H.; Inoue, K.; Yagi, N. Anal. Chem. 2002, 74, 4532. (7) Pahlke, S. Spectrochim. Acta, Part B 2003, 58, 2025. (8) Streli, C.; Wobrauschek, P.; Fabry, L.; Pahlke, S.; Comin, F.; Barett, R.; Pianetta, P.; Lu ¨ ning, K.; Beckhoff, B. Total-Reflection X-Ray Fluorescence (TXRF) Wafer Analysis. In Handbook of Practical X-Ray Fluorescence Analysis; Beckhoff, B., Kanngiesser, B., Langhoff, N., Wedell, R., Wolff, H., Eds.; Springer: Berlin, 2006; pp 498-553. (9) Hellin, D.; De Gendt, S.; Valckx, N.; Mertens, P. W.; Vinckier, C. Spectrochim. Acta, Part B 2006, 61, 496. (10) Klockenka¨mper, R. Spectrochim. Acta, Part B 2006, 61, 1082.

Analytical Chemistry, Vol. 79, No. 20, October 15, 2007 7873

built by PTB, detection sensitivities11-14 achieved so far by PTB and various cooperation partners at the electron storage ring BESSY II, the systematic optimization of detection limits of selected elements such as Al, the speciation of contamination, and recent developments in reference-free quantitation in TXRF analysis including its error budget as well as perspectives of this quantitation mode in table-top instruments. Instrumentation. Reference-free quantitation in TXRF analysis requires specially designed instrumentation involving high standards in wafer positioning as well as very well-known X-ray excitation and detection characteristics. For the purpose of semiconductor surface analysis, PTB built XRF and TXRF instrumentation14-19 that can handle 15-75 mm as well as 200and 300-mm silicon wafers. The TXRF and XRF instrumentation, which is suitable for handling 15-75-mm Si or SiC wafers, is described elsewhere.12,15-19 This instrumentation is located in the focal plane of the PTB’s plane grating monochromator (PGM) beamline20,21 for undulator radiation within the PTB laboratory22 at BESSY II and, additionally, can be positioned below a local minienvironment to avoid unintentional cross-contamination of the wafer samples. The instrumentation of interest in this paper enables reference-free TXRF quantitation of contamination on relevant industrial wafer sizes and also utilizes the excitation conditions at the PGM beamline: 300-mm Si wafers, as well as 200-mm wafers, are transported directly from their shipping cassettes, i.e., front opening unified pod (FOUP) or standard mechanical interface (SMIF) types, via a prealigner into a highvacuum load-lock by an adapted commercial equipment front end module (EFEM). After the pump down, a vacuum robot located inside the load-lock takes the wafer and places it inside the analysis chamber on an electrostatic chuck (ESC) mounted on an eightaxis manipulator. Figure 1 shows a schematic top view of this arrangement, and Figure 2 shows an overview of the facility located at the PGM beamline. The analysis chamber is oriented with respect to the focal plane of the PGM beamline, and the Si(Li) detector stands perpendicular to the incoming beam. On the right side, the load-lock is attached to the EFEM under a class 100 cleanroom. To take advantage of the linear polarization of the (11) Streli, C.; Wobrauschek, P.; Beckhoff, B.; Ulm, G.; Fabry, L.; Pahlke, S. X-Ray Spectrom. 2001, 30, 24. (12) Beckhoff, B.; Fliegauf, R.; Ulm, G.; Pepponi, G.; Streli, C.; Wobrauschek, P.; Fabry, L.; Pahlke, S. Spectrochim. Acta, Part B 2001, 56, 2073. (13) Streli, C.; Pepponi, G.; Wobrauschek, P.; Beckhoff, B.; Ulm, G.; Pahlke, S.; Fabry, L.; Ehmann, Th.; Kanngiesser, B.; Malzer, W.; Jark, W. Spectrochim. Acta, Part B 2003, 58, 2113. (14) Beckhoff, B.; Fliegauf, R.; Ulm, G.; Weser, J.; Pepponi, G.; Streli, C.; Wobrauschek, P.; Ehmann, T.; Fabry, L.; Mantler, C.; Pahlke, S.; Kanngiesser, B.; Malzer, W. Proc. Electrochem. Soc. 2003, 2003-03, 120. (15) Beckhoff, B.; Ulm, G. Adv. X-Ray Anal. 2001, 44, 349. (16) Kawahara, N.; Shoji, T.; Yamada, T.; Kataoka, Y.; Beckhoff, B.; Ulm, G.; Mantler, M. Adv. X-Ray Anal. 2002, 45, 511. (17) Kolbe, M.; Beckhoff, B.; Krumrey, M.; Ulm, G. Spectrochim. Acta, Part B 2005, 60, 505. (18) Mu ¨ ller, M.; Beckhoff, B.; Ulm, G.; Kanngiesser, B. Phys. Rev. A 2006, 74, 012702. (19) Zarkadas, C.; Karydas, A.; Mu ¨ ller, M.; Beckhoff, B. Spectrochim. Acta, Part B 2006, 61, 189. (20) Senf, F.; Flechsig, U.; Eggenstein, F.; Gudat, W.; Klein, R.; Rabus, H.; Ulm, G. J. Synchrotron Radiat. 1998, 5, 780. (21) Scholze, F.; Beckhoff, B.; Brandt, G.; Fliegauf, R.; Klein, R.; Meyer, B.; Rost, D.; Schmitz, D.; Veldkamp, M.; Weser, J.; Ulm, G.; Louis, E.; Yakshin, A. E.; Oestreich, S.; Bijkerk, F. Proc. SPIE 2000, 4146, 72. (22) Beckhoff, B.; Klein, R.; Krumrey, M.; Scholze, F.; Thornagel, R.; Ulm, G. Nucl. Instrum. Methods, A 2000, 444, 480.

7874 Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

Figure 1. Schematic view of the arrangement for 200- and 300mm Si wafers.

Figure 2. Picture of the entire instrumentation for 200- and 300mm Si wafers at the PGM beamline.

Figure 3. Schematic view of the principle arrangement and the available degrees of freedom.

exciting radiation, the ESC is moved into a vertical orientation during the TXRF measurements. Figure 3 shows a schematic view of the principle arrangement. The whole surface of a 200-mm or a 300-mm wafer can be scanned. To extend the capability of the system from the TXRF into the XRF regime, the angle of incidence can be increased from grazing incidence up to 45°, allowing for the analysis of thin multielemental, multilayered structures. The main components of the instrumentation are described below.

A. PGM Beamline. The PGM monochromator beamline20,21 for undulator radiation provides high photon flux of high spectral purity in the soft X-ray range. At a typical stored electron current of 200 mA, the photon flux at the PGM ranges from 2.4 × 1012 at 180 eV to 2.4 × 109/s at 1800 eV using the U49 undulator and an exit slit size of 40 µm. Depending on both the exit slit size and the trigonometric ratio of the incident to the exit angle at the 1200 l/mm grating, the PGM resolving power E/∆E ranges from about 1000 to 9000. Typical settings of the beamline result in a focus size of about 70 µm (vertical) × 140 µm (horizontal) in the PGM focal plane. B. TXRF Instrumentation for 200- and 300-mm Silicon Wafers. (1) Vacuum System. To achieve the vacuum conditions required by the PGM beamline and to prevent cross-contamination on the silicon wafer samples, the analysis chamber is designed as an ultrahigh-vacuum system. It is based on a DN500 COF flange on which the sample manipulator is mounted and that is covered by a 650-mm-high cylindrical chamber. The system can be baked out up to 100 °C, reaching a base pressure below 5 × 10-9 mbar after cooling down to room temperature, which increases to a typical working pressure of 1 × 10-8 mbar while the vacuum stepper motors are driven. The vacuum is generated by three turbomolecular pumps with an overall pump speed of 860 L/s. As roughing pump, a common dry scroll pump is used, which also serves the turbomolecular pump of the load-lock. The loadlock is made of aluminum and, because of the higher pressure tolerated here, mainly sealed by O-rings. Its base pressure is ∼2 × 10-7 mbar, but for the sample exchange a pressure of only below 1 × 10-6 mbar is required, which can be achieved in ∼30 min. During venting and purging, the turbopump is isolated by a gate valve and runs at operating speed. A second scroll pump is used to pump down below 2 × 10-1 mbar when it becomes possible to open the gate valve. This procedure prevents vibrations induced by the turbopump during ramping up or down and ensures short pump-down times. (2) Wafer Handling System. To minimize the risk of unintentional cross-contaminations of the wafer samples, an adapted commercial EFEM is employed for the wafer handling. The EFEM consists of a loadport for 200-mm wafers (SMIF) and one for 300-mm wafers (FOUP), a robot that can handle both wafer sizes, a prealigner, and an enclosure with an integrated filter fan unit to ensure class 1 cleanroom conditions. This arrangement allows for handling wafer samples without any manual interaction if the wafers are delivered in SMIF or FOUP boxes. After loading a load port with a SMIF or FOUP box to be opened, the robot takes one wafer out of the respective box and moves it to the prealigner. The prealigner centers the wafer and moves it to a well-defined position with respect to the notch. To realize fast sample exchange, the vacuum robot has two arms and can handle two wafers of each size simultaneously. If the load-lock is vented and the rectangular gate valve is open, the vacuum robot can take the wafer from the prealigner with one arm and put a wafer already investigated back to the prealigner with its other arm. After moving both arms back into the load-lock, closing the gate valve, and pumping down to 1 × 10-6 mbar, the system is ready to exchange the wafer sample in the analysis chamber. (3) Cleanroom. The sensitive area including the EFEM, where wafers are handled, is completely covered by a mobile class

100 cleanroom. This cleanroom has a size of 2 × 2 m2. A class 10 workbench is mounted on one side of this cleanroom that allows for single-wafer manipulations by means of vacuum pincers for 200- and 300-mm wafers. The entry to this area is accessible via a cleanroom lock, also of class 100, having a size of 1 × 2 m2. The complete arrangement can be moved together with the analysis chamber a few meters away without switching off the cleanroom or venting the analysis chamber. This feature allows for relatively fast instrumental changes in the PGM experimental station as required by weekly alternating routine and research activities ranging from TXRF, grazing incidence, or conventional XRF to X-ray reflectometry involving a Kirkpatrick-Baez refocusing mirror arrangement for laterally resolved mask characterizations in the extreme ultraviolet range. (4) Electrostatic Chuck. The only known way to clamp a wafer to a very flat reference plane inside the UHV TXRF analysis chamber and to move the whole assembly into a vertical orientation without touching any point of the wafer surface is to use an ESC.23 The bipolar ESC employed was developed by the Fraunhofer Institut fu¨r Angewandte Optik und Feinmechanik (IOF). Not only does it generate enough force to clamp a 300-mm wafer vertically at relatively low voltages (typically 300 V), it also includes a stepper motor-driven lift pin mechanism to remove the wafer from the surface while being oriented horizonatally. The flatness of the chuck surface is ∼2.2 µm, and its parallelism to the mounting surface of the vacuum goniometer is below 30 µm. The ESC is also designed to withstand a bake-out temperature of 100 °C. (5) Eight-Axis Manipulator. The adjustment of the angle of incidence of the exciting radiation with respect to the wafer surface and the lateral mapping of the whole wafer surface are ensured by a UHV manipulator offering 8 degrees of freedom, as shown in Figures 3 and 4. Four of them are needed for the basic alignment and additional options when dealing with patterned structures on wafers. The other four axes are crucial for the TXRF experiments. The incident angle RY is realized by a one-circle goniometer operating outside the vacuum, which is coupled to the goniometer head inside by means of a heavy steel rod. The mapping of the samples is ensured by the rotation RZ of the ESC by a small one-circle goniometer inside and by the linear position TY of the one-circle goniometer outside the vacuum (cf. Figure 4). To move the ESC between the horizontal position for sample transfer and the vertical position for the TXRF measurements, the rotation TILT is available. All axes, inside or outside the vacuum, are stepper motor driven. Table 1 provides detailed information on the respective parameters of the manipulator axes. (6) Support Frame. To realize the geometric conditions for the TXRF measurements as well as to control and verify them easily, the analysis chamber and the load-lock are mounted on a common support frame with a two-axis stage. This enables the analysis chamber together with the load-lock to move in both the horizontal and vertical directions by means of several stepper motors. Because of the size and the weight of the EFEM, it is not mechanically coupled with the load-lock but rather with the support frame. Consequently, the stage must be in a well-defined position to allow for the wafer transfer between the prealigner (23) Kalkowski, G.; Risse, S.; Guyenot, V. Microelectron. Eng. 2002, 61-62, 357.

Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

7875

Figure 4. Engineering detail drawing of the analysis chamber with the 8-axis manipulator and the ESC. RY, TY, RZ: the main axes for measurements. The synchrotron radiation propagates through the center of the X normal to the paper plane. Table 1. Characteristics of the Stepper Motor-Driven Axes of the Sample Manipulator, the Support Frame, and the Detector Stage of PTB’s Industrial Purpose TXRF Instrumentation axis

range

resolution

uncertainty

speed

TY RY RZ TILT

Sample Manipulator Main Axes 180 mm 0.1 µm 10 µm (45° 0.0001° 0.001° (200° 0.005° 0.01° -3° to 92° 0.0003° 0.1°

5 mm/s 10 °/s 20 °/s 0.3 °/s

J-TX J-TZ J-RX J-RZ

Sample Manipulator Alignment Axes (12 mm 2.5 µm 0.1 mm (12 mm 2.5 µm 0.1 mm (15° 0.0025° 0.1° (15° 0.0025° 0.1°

0.5 mm/s 0.5 mm/s 0.5 °/s 0.5 °/s

TX TY

(15 mm (15 mm

Support Frame 0.025 µm 10 µm 0.025 µm 10 µm

5 mm/s 0.25 mm/s

TX TY

(15 mm (25 mm

Detector Stage 0.25 µm 10 µm 0.25 µm 10 µm

1 mm/s 1 mm/s

and the atmospheric robot. Normally the whole arrangement stands on adjustable machine feet, but to make room for other experimental setups, it can easily be put on wheels and moved several meters away without having to demount anything. (7) Detection Systems for Fluorescence Radiation. So far, all TXRF measurements have been performed with a commercial Si(Li) detector having an ultrathin entrance window without any supporting grid. But one of the design goals of the instrumentation was to provide the capability to use various other detection systems. For example, the three-axis detector stage, on which the 7876

Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

Si(Li) is mounted, can handle a weight of up to 100 kg. Thus, it is even possible to mount and carefully align the LHe-LN-cryostat of a superconducting tunnel junction detector24,25 offering a higher energy resolution than conventional semiconductor detectors such as Si(Li) or silicon drift detector26 systems. For further demands, such as the use of wavelength-dispersive detection systems involving a spherical grating-CCD camera assembly, the detector stage can easily be removed, and there is a DN160 CF flange in the chamber, now reduced to DN40 CF. (8) Options. For several tasks, especially the alignment and measurement of the intensity of both incoming and reflected beams, calibrated photodiodes are mounted on a feed-through 270 mm behind the focal plane. On the upper side of the analysis chamber, there is a DN160 CF flange to mount a five-axis manipulator carrying a polycapillary optic. The optic and manipulator have been successfully tested at a different setup.24 In order to allow for the future use of a refocusing mirror optics, the beam entrance flange is a DN200 CF flange. The analysis chamber is not the terminal point of the beamline, thereby allowing other experimental setups to be mounted behind it and operated alternately. If the free space conditions around the PGM experimental station change at sometime or if it is necessary to move the facility to another PTB beamline, it can also be turned 180°. Determination and Improvement of TXRF Detection Limits. At PTB’s monochromator beamline for undulator radiation, the absolute detection limit values of TXRF were determined for low-Z elements11-14 such as C, N, Na, Mg, and Al to be in the 300 fg-1.3 pg range employing several sets of wafer samples that were contaminated intentionally by light elements. The detection limits deduced for VPD samples, based on the assumption that the dried droplets were collected from a 200-mm wafer surface, ranged from 2 × 107 to ∼108 atoms/cm2 for Na, Mg, and Al, thus fulfilling the respective requirements of the technological roadmap. In a beam geometry where the detection channel is oriented in line with the polarization vector of the excitation radiation, lowest detection limits were achieved. In general, this arrangement requires a vertical positioning of the wafer surface to be investigated when employing synchrotron radiation in the TXRF excitation channel. TXRF detection limits that are derived from oneelemental contamination, thus resulting in general in a domination of the corresponding spectral count rates by the fluorescence line of this element, represent the ideal case of the repartitioning of the spectral distribution detected in favor of the fluorescence line of interest. Hence, when dealing with real multielemental contamination at the maximum count rate capacity of the detection system employed, the corresponding detection limits can be higher by a factor of 2 or 3 depending on the specific spectral distribution caused by the corresponding fluorescence radiation. Instead of conventional peak fitting functions, a deconvolution technique for XRF spectra15 that uses experimental detector response functions,27,28 including an optional modeling of these response functions, was developed. In the very soft X-ray range, (24) Beckhoff, B.; Fliegauf, R.; Ulm, G. Spectrochim. Acta, Part B 2003, 58, 615. (25) Bechstein, S.; Beckhoff, B.; Fliegauf, R.; Weser, J.; Ulm, G. Spectrochim. Acta, Part B 2004, 59, 215. (26) Eggert, T.; Boslau, O.; Kemmer, J.; Pahlke, A.; Wiest, F. Nucl. Instrum. Methods, A 2006, 568, 1. (27) Scholze, F.; Procop, M. X-Ray Spectrom. 2001, 30, 69. (28) Krumrey, M.; Gerlach, M.; Scholze, F.; Ulm, G. Nucl. Instrum. Methods, A 2006, 568, 364.

Figure 5. TXRF investigation of a 100-µL droplet containing 500 pg of Na dried on a 200-mm Si wafer. Part a of the figure shows the lateral distribution of Na, deducted from the Na count rate recorded during 5-s periods, using a 150-µm-wide excitation beam. The maximum Na count rate is indicated in gray. The TXRF spectrum in part b was recorded at the respective maximum Na position during 100 s real time. The dotted line indicates the assumed background contribution caused by bremsstrahlung events.

it could be shown that only small amounts of continuous scattering and bremsstrahlung background15 exist when exciting the wafer samples with radiation of high spectral purity.20,21 However, the continuous contribution resulting from the resonant Raman scattering effect18,19,29 has to be included when exciting energetically close below the absorption edge of a main constituent of the wafer material. The commissioning of PTB’s industrial purpose TXRF arrangement involved 200-mm silicon wafers that were intentionally contaminated with 100-µL droplets containing metal and light element contamination in the picogram range, which were prepared by the Central Analytical Laboratories of Wacker Siltronic. In a study,14 the element Na was excited with monochromatic radiation of 1206-eV photon energy at an incident angle of 1° to optimize its detection limit. Part a of Figure 5 shows a rapid lateral scan across a droplet containing 500 pg of Na. The TXRF spectrum at the maximum of this lateral scan was deconvoluted by means of detector response functions, leaving the dotted line in part b of Figure 5 as the assumed bremsstrahlung background, both of which are used to calculate a LLD value for Na of 40 fg with respect to a typical accumulation time of 1000 s. As an alternative to this spectral decomposition method, a TXRF spectrum is recorded at an off-droplet position, which is a reliable estimate for the spectral background. Then by taking into account the ratio of the respective Na deposition present in the lateral maximum position to the total Na deposition in the droplet, a LLD value of 34 fg can be found with respect to the same accumulation time, thus being well in line with the approach of spectral decomposition. The most prominent advantage of exciting ultratrace contamination on a Si wafer surface at photon energies below the energy of the Si-K absorption edge consists in avoiding Si-KR,β fluorescence originating from the wafer matrix material. Otherwise, at excitation energies above the Si-K absorption edge, SiKR,β fluorescence radiations can clearly dominate the spectral distribution of the fluorescence radiation emitted, thus deteriorating drastically the detection limits of all elements other than Si due to limited count rate capacities of the energy-dispersive detection system employed to detect the fluorescence radiation. At excitation energies below the Si-K absorption edge, however, (29) Baur, K.; Kerner, J.; Brennan, S.; Singh, A.; Pianetta, P. J. Appl. Phys. 2000, 88, 4642.

Figure 6. Lateral distribution of Al within a 100-µL droplet containing 500 pg of Al by means of a 70-µm broad (fwhm) beam having a photon energy of 1742 eV (ψin ) 0.9°).

resonant Raman scattering (RRS) at the Si wafer occurs, which limits the detection limits of elements having their fluorescence line energies close to the excitation energy. In order to investigate this effect quantitatively, a systematic study on its impact on the detection limit of Al was performed in the present work, varying both the energetic and the angular characteristics of the incident radiation. As X-ray Raman scattering is a resonant effect that decreases with decreasing incident energy, improved detection limits for Al are to be expected likewise due to the energy dependence of the photoelectric cross section of Al. Regarding the angle of incidence of the excitation radiation with respect to the wafer surface, the situation is somewhat complicated: with increasing incident angle still below the critical angle of total reflection, the intensity of the X-ray standing wave (XSW) field at the wafer surface increases, thus inducing higher fluorescence radiation of the contaminants. The RRS background, originating from the top 3 nm of the Si wafer bulk material, increases equally below the fluorescence lines of the contamination. At incident angles above the critical angle of total reflection, the penetration depth of the incident radiation increases drastically to several hundred nanometers, resulting in a likewise enhanced RRS background. For the investigation of the impact of RRS background on the LLD values of Al, a 100-µL droplet sample containing 500 pg of Al was deposited on a 200-mm silicon wafer. Figure 6 shows the lateral distribution of Al within this droplet by means of a 70-µm Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

7877

Figure 7. TXRF spectra recorded during 100 s real time at an off-droplet position depicted in part a and at the maximum position of the lateral distribution of Al depicted in part b at an excitation energy of 1742 eV (ψin ) 0.9°). The lower spectrum in part b consists of the assumed bremsstrahlung background and, above ∼1000 eV, the difference between the spectra recorded at the maximum and off-droplet positions, revealing clearly the Al fluorescence line. Table 2. Lower Limits of Detection (LLD) of Aluminum on Si Wafer Surfaces for Different Photon Energies and Different Angles of Incidencea LLD (∆LLD) (in fg) in dependence of the angle of incidence photon energy/eV

0.7°

0.8°

0.9°

1.0°

1622 1682 1742 1802

210 (11) 324 (16) 427 (21) 567 (28)

148 (8) 224 (11) 286 (14) 525 (26)

86 (4) 147 (7) 376 (19) 447 (22)

138 (7) 247 (12) 299 (15) 359 (18)

Figure 8. Rotatable wheel containing appropriate absorption filters to modify the spectral distribution detected.

a The values of the respective relative uncertainties ∆ LLD are indicated in parentheses.

broad beam having a photon energy of 1742 eV at an incident angle of 0.9°. Far away from this lateral distribution, part a of Figure 7 shows a TXRF spectrum dominated by the RRS distribution. At a given measurement time, this RRS reference measurement can contribute to the deconvolution of the TXRF spectrum recorded at the maximum position of the Al distribution as shown in part b of Figure 7. Table 2 presents the detection limits (LLD) of aluminum on Si wafer surfaces for different photon energies and different angles of incidence with respect to an accumulation time of 1000 s. The optimal photon energy for the detection of aluminum is slightly larger than the K-absorption edge of aluminum (1560 eV). Here, the maximum of the RRS spectral distribution is quite close to the Al-KR,β fluorescence lines, but the absolute value of the RRS contribution is rather small because the photon energy is far away from the resonance region of RRS. The optimal angle of incidence is slightly below the critical angle of total reflection as the effective excitation intensity, which is the product of the intensity of the excitation radiation and the angular-dependent XSW field value, increases with increasing angle of incidence. Therefore, the Al fluorescence and the background are growing in parallel whereas the LLD value is decreasing. Above the critical angle, the penetration depth increases rapidly and considerably more Si atoms can contribute to the scattering of photons by RRS. Hence, the background increases strongly, and the lower limit of detection increases as well. For specific groups of energetically adjacent elements, the detection limits can be also improved by appropriate modifications of the spectral distribution detected. Figure 8 shows a rotatable 7878 Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

Figure 9. Comparison of filtered and unfiltered TXRF spectra of the same droplet sample, 58 pg of Ni of which is being excited. Both spectra were recorded during the same accumulation time as shown by the identical zero lines caused by an electronic pulser.

wheel on which several thin self-standing filters are mounted. Each of these filters has a different transmittance allowing for a specific spectral modification of the TXRF spectra recorded. Figure 9 shows how the use of such an optional absorption filter in the TXRF detection channel modifies the spectral distribution detected, thus improving the detection limits (LLD) by a factor of ∼3-5. However, the incident flux needs to be increased by a factor of 10-30 in order to achieve these improved LLD values. Speciation of Contamination by TXRF Combined with Near-Edge X-ray Absorption Fine Structure (NEXAFS). Due to the energetic tunability of synchrotron radiation, NEXAFS30-36 (30) Sto ¨hr, J. NEXAFS Spectroscopy; Springer Series in Surface Sciences 25; Springer-Verlag: Berlin, 1992. (31) Drube, W.; Treusch, R.; Sham, T. K.; Bzowski, A.; Soldatov, A. V. Phys. Rev. B 1998, 58, 6871. (32) Soldatov, A. V.; Kasrai, M.; Bancroft, G. M. Solid State Commun. 2000, 115, 687.

Figure 10. TXRF-NEXAFS spectra of 500 pg of two different compounds of light elements recorded at the N-, C-, and O-K absorption edges. The respective fluorescence count rates are normalized to the incident radiant power and compared to the off-droplet background (solid gray line and dotted gray line scaled to the droplet count rates) for C and O. Relevant resonance structures are indicated by vertical dotted lines.

investigations can be combined with TXRF analysis to contribute to the speciation of contamination by light or organic compounds at extremely low levels. As these contaminants are gaining an increasing interest in the quality control of wafer surfaces, various NEXAFS experiments14,37-40 of droplet depositions of low-Z compounds, nano- and microparticles, and nanolayers were performed at the PGM beamline that offers, apart from the required tunability (33) Rehr, J. J.; Albers, R. C. Rev. Mod. Phys. 2000, 72, 621. (34) de Groot, F. Chem. Rev. 2001, 101, 1779. (35) de Groot, F. Coord. Chem. Rev. 2005, 249, 31. (36) Rehr, J. J. Radiat. Phys. Chem. 2006, 75, 1547. (37) Pepponi, G.; Beckhoff, B.; Ehmann, T.; Ulm, G.; Streli, C.; Fabry, L.; Pahlke, S.; Wobrauschek, P. Spectrochim. Acta, Part B 2003, 58, 2245. (38) To ¨ro¨k, S.; Osan, J.; Beckhoff, B.; Ulm, G. Powder Diffr. J. 2004, 19, 81. (39) Osan, J.; To ¨ro¨k, S.; Beckhoff, B.; Ulm, G.; Hwang, H.; Ro, C.-U.; Abete, C.; Fuoco, R. Atmos. Environm. 2006, 40, 4691. (40) Hoffmann, P.; Baake, O.; Beckhoff, B.; Ensinger, W.; Fainer, N.; Klein, A.; Kosinova, M.; Pollakowski, B.; Trunova, V.; Ulm, G.; Weser, J. Nucl. Instrum. Methods, A 2007, 575, 78.

of the excitation radiation, a relatively high resolving power while ensuring both a sufficient photon flux and a high spectral purity. To simulate minute surface contamination, low-Z compounds were diluted and deposited as droplets on wafer surfaces. In TXRFNEXAFS experiments14,37 the K absorption edges of C, N, O were probed. Figure 10 shows the corresponding TXRF-NEXAFS spectra recorded from droplets containing 12 ng of N deposited on a 200-mm silicon wafer. The reproducibility of these measurements at the N edge was good, whereas background contamination at the C-K and O-K edges can interfere considerably. Furthermore, the spectral shape of the NEXAFS structure may depend on the adsorbate orientation with respect to the wafer surface. In the soft X-ray range, analytical and speciation information of transition metals can be recorded preferably at the Liii,ii absorption edges. The TXRF-NEXAFS spectrum of another droplet deposition in Figure 11 demonstrates a limit of detection for the Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

7879

Table 3. Relative Uncertainty of the Analytical Results Obtained by Reference-Free TXRF Analysis at PTBa

Figure 11. TXRF-NEXAFS spectrum of 36 pg of Ni deposited on a wafer surface showing resonances associated with the Liii and Lii absorption edges. The nonresonant variation of the mass absorption coefficients at the Liii and Lii edges is represented by the gray dotted line.

speciation of only a few picograms of Ni or its compounds. With respect to the reliability of the theoretical modeling of X-ray absorption spectra, single-electron NEXAFS or XANES calculations of the density of states using multiple scattering33,36 work well for the K edges in general. However, at the Liii,ii absorption edges of metals, the agreement of such calculations with experiments is rather poor as one does not observe the density of states in such absorption processes due to the strong overlap of the core wave function with the valence wave function according to de Groot.35 Reference-Free Quantitation in TXRF Analysis. Knowing the relevant instrumental and atomic fundamental parameters,41-50 the mass deposition mi/FI of the element i with unit area FI can be calculated by combining basic quantitation51 in X-ray fluorescence analysis with the modulation1 of the incident radiant power by the XSW intensity:

{

mi -1 ) ln 1 FI µtot,i

}

Pi Ωdet 1 1 P0,Wsurfτi,E0Q 4π sin ψin µtot,i

where E0 is the photon energy of the incident (excitation) radiation, P0 ) S0/σdiode,E0 is the radiant power of the incident radiation, S0 is the signal of the photodiode measuring the incident (41) Bambynek, W.; Crasemann, B.; Fink, R. W.; Freund, H.-U.; Mark, H.; Swift, C. D.; Price, R. E.; Venugopala, Rao, P. Rev. Mod. Phys. 1972, 44, 716. (42) Krause, M. O.; Nestor, C. W.; Sparks, C. J.; Ricci, E. Oak Ridge National Laboratory (ORNL) report no. 5399, 1978. (43) Saloman, E. B.; Hubbell, J. H.; Scofield, J. H. At. Data Nucl. Data Tables 1988, 38, 1. (44) Henke, B. L.; Gullikson, E. M.; Davis, J. C. At. Data Nucl. Data Tables 1993, 54, 181. (45) Chantler, C. T. J. Phys. Chem. Ref. Data 2000, 29, 597. (46) Elam, W. T.; Ravel, B. D.; Sieber, J. R. Radiat. Phys. Chem. 2002, 63, 121. (47) Ebel, H.; Svagera, R.; Ebel, M. F.; Shaltout, A.; Hubbell, J. H. X-Ray Spectrom. 2003, 32, 442. (48) Brunetti, A.; Sanchez del Rio, M.; Golosio, B.; Simionovici, A.; Somogyi, A. Spectrochim. Acta, Part B 2004, 59, 1725. (49) Shaltout, A.; Ebel, H.; Svagera, R. X-Ray Spectrom. 2006, 35, 52. (50) Zschornack, G. H. Handbook of X-Ray Data; Springer: New York, 2007 (51) Mantler, M.; Willis, J. P.; Lachance, G. R.; Vrebos, B. A. R.; Mauser, K.-E.; Kawahara, N.; Rousseau, R. M.; Brouwer, P. N. Quantitative Analysis. In Handbook of Practical X-Ray Fluorescence Analysis; Beckhoff, B., Kanngiesser, B., Langhoff, N., Wedell, R., Wolff, H., Eds.; Springer: Berlin, 2006; pp 309410.

7880

Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

parameter

estimate x

S0 σdiode,E0 IWsurf ψin

120 nA/100 mA 0.264 A/W 2.0 15.7 mrad [22.7 and 43.6 mrad] 20000/(40 s) 0.75 0.05 sr b b b

Ri det,Ei Ωdet τi,E0 µtot,i,E Q total:

standard measurement uncertainty u(x)

contribution to the relative uncertainty of mi/FI /10-2

12 pA/100 mA 0.0026 A/W 0.1 0.25 mrad

0.01 1.0 5.0 0.7

140/(40 s) 0.01 0.002 sr b b b

0.7 1.5 4.0 7.0 [10.0 and 13.0] 7.0 [10.0 and 13.0] 9.0 [12.5 and 15.5 ] 15.0 [20.0 and 25.0]

a For the determination of the mass deposition of an element on the wafer by TXRF, the following contributions to the relative uncertainty (k ) 1) are given for an excitation energy of 1622 eV (1060 and 520 eV, respectively). b The relative uncertainties of the fundamental atomic parameters are estimated for the groups of excited elements according to ORNL data.42,50

radiation, σdiode,E0 is the spectral responsitivity of the photodiode, IWsurf is the relative intensity (as calculated by the software package IMD52) of the XSW at the wafer surface, P0,Wsurf ) P0IWsurf, ψin is the angle of incidence with respect to the wafer surface, Ei is the photon energy of the fluorescence line l of the element i, Ri is the detected count rate of the fluorescence line l of the element i, det,Ei is the detection efficiency of the Si(Li) detector at the photon energy Ei, Pi ) Ri/det,i, τi,E0 is the photoelectric cross section of the element i at the photon energy E0, µi,E is the absorption cross section of the element i at the photon energy E, µtot,i ) µi,E0/sin ψin + µi,Ei/sin ψout, ψout is the angle of observation, which equals 90° in a typical TXRF beam geometry, Ωdet is the effective solid angle of detection, ωXi is the fluorescence yield of the absorption edge Xi (of the element i), gl,Xi is the transition probability of the fluorescence line l belonging to the absorption edge Xi, and jXi is the jump ratio at the absorption edge Xi, Q ) wXigl,Xi(jXi - 1)/jXi. The concept applied in the above quantitation approach is the following: Ultratrace contamination on top of the wafer surface and in the first few nanometers, in general less than 5 nm, is excited by the effective excitation radiant power P0,Wsurf at the wafer surface that is reduced following the usual exponential decrease involving µi,E0/sin ψin during penetration into the wafer. The absorption of the element-specific fluorescence radiation induced below the wafer surface is taken into account with the exponential decrease involving µi,Ei/sin ψout. In Table 3, the contributions to the relative uncertainty of the analytical results are given. The contributions associated with the relative uncertainties of fundamental atomic data can be reduced by dedicated experiments15,18,19,53 aiming at more reliable determinations of selected fundamental parameters of interest. In contrast to the detection geometry in conventional XRF analysis, where the source region of induced fluorescence radiation is somewhat restricted and the distance of this region to the detector is rather fixed, the effective solid angle of detection in TXRF beam geometry depends on several parameters (angle of (52) Windt, D. Comput. Phys. 1998, 12, 360. (53) Scholze, F.; Beckhoff, B.; Kolbe, M.; Krumrey, M.; Mu ¨ ller, M.; Ulm, G. Microchim. Acta 2006, 155, 275.

Figure 12. Beam spot profile in the focal plane of the PGM beamline for an exit slit of 70 µm.

Figure 14. Illustration of the solid angle of detection employing the extended beam spot on the wafer surface.

Figure 13. Beam spot profile projected on the wafer surface for an incident angle of 0.9°.

incidence, beam profile, detector collimator, distance between the wafer and the detector if used for active count rate adapting), requiring a more detailed description. To adapt the recorded fluorescence count rate to the maximum count rate capacity of a given fluorescence detector, the distance between the wafer sample and the detector is normally reduced to a minimum. Consequentially, due to the comparable dimensions of the effective detector area and the irradiated area on the wafer, the simplification of assuming a point source is no longer valid. For example, at a grazing incident angle of 0.5°, a horizontal beam width of just 100 µm is extended to an almost 11.5-mm-long illuminated region on the wafer surface, a rather large value compared to a conventional detector diameter of ∼5 mm. When dealing with typical distances of 5-50 mm between the wafer surface and the detector entrance, the determination of the solid angle of detection becomes crucial for a reference-free quantitation of mass depositions in TXRF analysis. Figure 12 shows the beam spot in the focal plane of the PGM beamline at typical operation parameters. The beam spot was recorded with a CCD system directly positioned in the PGM focal plane during special operation conditions of BESSY II allowing for drastically reduced stored electron beam currents. Here, the fwhm value is ∼135 µm in the horizontal direction and ∼80 µm in the vertical direction, the latter selected by a corresponding exit slit size scaling. Using a grazing incident angle of 0.9°, the projected beam on the surface of the sample has an increased horizontal fwhm of 8.6 mm as shown in Figure 13. In the vertical direction, the approximation by a point source is still valid, but in the horizontal direction, this no longer holds. Hence, the horizontal intensity distribution has to be taken into account in order to determine the effective solid angle of detection.

Figure 14 shows the extended beam spot on the surface of the sample. The determination of the solid angle is divided into characteristic regions. From each point in the region marked as B, the solid angle can be described in the usual manner by the area of the detector and the distance between sample surface and detector entrance plane. In the region marked as A, the detector area cannot be described as a circle anymore. The area decreases with increasing distance from the center point at the sample surface directly opposite to the detector center. In addition, fluorescence radiation originating from a point outside the regions A and B is not detectable at all. The following formulas show the determination of the effective solid angle Ω:

dΩ(x) )

[

π‚h‚(f/2)/ab for region A πr2/R2 for region B

]

In region A, the detection area changes from a circle to an ellipse,

Ω)

∫

∞

-∞

I(x) dΩ(x) dx

with r being the radius of the detector area, R being the distance from the sample, a and b being position-dependent distances between sample and detector, and f and h describing the illuminated ellipse on the detector. Figure 15 shows the solid angle function dΩ, which has to be convoluted with the intensity distribution. The effective solid angle depends on the intensity distribution on the sample surface, the dimension of the detector, and the distance between sample and detector. Thus, for example, the solid angle of detection for a detector with a radius of 3.1 mm and a distance to sample of 13.5 mm changes from the 0.1656 sr as obtained in the usual point source model to an effective solid angle of 0.0854 sr in the model including source, wafer, and detector characteristics as described above. Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

7881

Figure 15. Solid angle acceptance function (lateral contribution) to be convoluted with the intensity distribution on the wafer surface.

Figure 16. TXRF spectrum recorded from the surface of a 200mm Si wafer and deconvoluted with experimentally obtained detector response functions with respect to C, N, O, Fe, Cu, Na, and Al fluorescence radiation, bremsstrahlung background, RRS background and Rayleigh scattering (Scatt.). The excitation energy was 1622 eV and the angle of incidence 0.9°.

Figure 16 shows a TXRF spectrum recorded from the surface of a 200-mm Si wafer and deconvoluted with experimentally obtained detector response functions with respect to C, O, Fe, Cu, Na, and Al fluorescence radiation, taking various background contributions into account. The fluorescence count rates deduced from this spectrum were then converted into mass depositions according to the equation for mi/FI introduced at the beginning of this section. Ordered with increasing fluorescence line energy some selected results of this measurement are 1600 pg/cm2 of Fe, 114 pg/cm2 of Cu, 5 pg of Na/cm2, and 190 pg/cm2 of Al. Perspectives for Reference-Free Quantitation in TableTop Instrumentation. The synchrotron radiation-based TXRF instrumentation presented in the current work differs from conventional table-top TXRF instruments. However, the increasing interest of manufacturers to extend the applicability of their instruments results in the need to better understand the characteristics of relevant instrumental components such as X-ray tubes, X-ray optics, and X-ray detectors. The emission characteristics of X-ray tubes can be calculated54,55 or measured directly by means of appropriate pinhole detector arrangements. The reliability of the calculations depends on the knowledge of X-ray production and absorption cross sections. The measurement of angular(54) Ebel, H. X-Ray Spectrom. 1999, 28, 255. (55) Ebel, H. X-Ray Spectrom. 2003, 32, 46. (56) Graessle, D. E.; Soufli, R.; Aquila, A. L.; Gullikson, E. M.; Blake, R. L.; Burek, A. J. Proc. SPIE 2004, 5165, 469.

7882

Analytical Chemistry, Vol. 79, No. 20, October 15, 2007

dependent spectral distributions of X-ray tube spectra is energetically restricted by the effective thickness of the pinhole’s land and, in addition, depends quantitatively on the a priori knowledge of the pinhole diameter and position as well as of the efficiency of the detector employed. The band-pass performance of monochromatizing X-ray optics such as multilayers employed in TXRF can be also calculated and measured. The reliability of the calculations depends on the knowledge of the optical constants44 and of the specific multilayer structure. At wavelength experiments of the angular, viz. energy-dependent reflectivity of a multilayer can be performed both with laboratory or synchrotron radiationbased instruments, the latter offering smaller divergences and tunable radiation.56 The physically based description of the efficiency and response behavior of X-ray detectors26-28 allows for good estimates of the real detector characteristics if enough knowledge is available on the specific quality and structure of a given detector type. In addition, X-ray detectors can be absolutely calibrated with respect to their efficiency and response behavior by monochromatic radiation of known and appropriate radiant power. When arranging X-ray source, optics, and detectors with respect to the sample site in table-top instruments, special care has to be taken with respect to relevant distances and angles as well as to a very stable mechanics to allow for a completely reference-free TXRF quantitation. The development and investments needed to produce such table-top TXRF instruments will depend on the flexibility and quantitative reliability expected in future applications, e.g., when dealing with frequently differing novel materials for which not enough appropriate reference standards exist. CONCLUSION In the present work, we have shown that a completely reference-free quantitation in total reflection X-ray fluorescence analysis can be performed when instrumental parameters are wellknown and controlled by appropriate instrumentation. Apart from the instrumental parameters, the relative uncertainty of the analytical results depends on the knowledge of the atomic fundamental parameters involved, which can be improved in dedicated experiments. Reference-free total reflection X-ray fluorescence analysis employing synchrotron radiation can substantially contribute to the contamination analysis and speciation on semiconductor surfaces. The continuing development of analytical methods and related instrumentation by PTB is dedicated to high end investigations in the R&D of semiconductor samples related to industrial applications such as the assessment of wafer cleaning procedures by off-line reference measurements. The perspectives for using the methods and technology discussed here include the transfer to novel table-top instrumentation. ACKNOWLEDGMENT The authors thank Siltronic AG for Si wafer samples and SiCrystal AG for SiC wafer samples. For the design and setup ofthe TXRF arrangements of PTB at BESSY II, we also thank our colleague M. Bock.

Received for review June 12, 2007. Accepted August 7, 2007. AC071236P