High-resolution Junction Photo-voltage Mapping of Sheet Resistance and Leakage Current Variations with ms-timescale Annealing Methods

Transcription

1 High-resolution Junction Photo-voltage Mapping of Sheet Resistance and Leakage Current Variations with ms-timescale Annealing Methods Vladimir Faifer a, Michael Current b, N. Ohno c, Jeffri Halim a, Jason Lu a a Frontier Semiconductor,199 River Oaks Pkwy, San Jose, CA USA b Current Scientific, 179 Comstock Way, San Jose, CA 9514 USA c Frontier Scientific Japan, Noborito 679-4, Kawasaki, JAPAN A non-contact method for measurement of sheet resistance and leakage current (RsL) for ultra-shallow junction characterization is described where high-density (30,000 pixel) sheet resistance maps provide detailed evaluation of local and global variations of dopant activation arising from implant and anneal conditions. Various examples of RTP-spike, flash and laser anneals and new low-dose implant results are discussed. Introduction The needs of advanced ultra-shallow junction (USJ) technology, which uses lowenergy ( 00 ev/dopant) implants and millisecond anneals, has encouraged the development of non-contact junction monitoring metrology. Junction photo-voltage (JPV) measurements of carrier spreading and recombination in p-n junctions provide a non-contact method for evaluation of sheet resistance (Rs) and leakage current (Io) [1,]. The range of Rs measurements span implant doses from threshold adjust ( ions/cm ) to source-drain doping ( ions/cm ), including accurate evaluations of shallowextension ( 10 nm) junctions in halo/well profiles [3]. Leakage current densities, including the effects of carrier recombination and trap-assisted tunneling, can be obtained over a dynamic range of 5 orders of magnitude [3,4]. The analysis of JPV measurements is based on rigorous solutions of a onedimensional Poisson equation governing the absorbtion of light and creation of carrier pairs and by integrating the three-dimensional continuity equations for carrier spreading over vertical spatial coordinates perpendicular to the wafer surface [1]. Excellent correlation of the contact-less RsL technique with contact four-point probe (4PP) measurements was demonstrated for deep (>50 nm) p-n junctions with low-leakage currents []. In the case of USJs formed in heavily-doped regions, e.g., halo or well, no correlation between RsL and 4PP measurements was observed due to probe penetration damage and high p-n junction leakage current [,4]. The analysis methods of RsL measurements has been independently verified and analyzed in detail by analytical modeling and numerical simulation [5]. The advantages of the RsL technique for monitoring of USJs formed with low-energy implants and millisecond anneal and by lowtemperature CVD have been demonstrated [6]. In this work, we will explore the highresolution RsL capabilities for monitoring of micro and global non-uniformities related to implant and annealing process.

2 Measurement Background The basis of the RsL measurement is to use photo excitation of carriers in the p-n junction and wafer substrate and to monitor, in a spatially resolved manner, the JPV signals inside and outside the illumination area, when absorbtion of modulated light flux, Φ (t)= Φ 0 (1-cos(πft)), creates electron hole pairs in the semiconductor material. Two electrodes, a circular transparent electrode (1) with diameter r 0 at the center of the probe and second round arc conducting electrode () subtending an angleβ and coaxial with the first electrode a small distance away, are used to measure JPV voltages V 1 and V (Fig. 1) [1]. (a) Modulated Light Beam V 1 V 1 Junction P+ h Depletion Substrate N e Spreading (b) 1 r 1 r 0? r Figure 1. (a) Photo-excitation and carrier drift with a modulated light source and two capacitor electrodes for monitoring the induced junction photo-voltage in a spatially resolved manner; (b) electrode configuration. The JPV voltages, V 1 and V, under low-level light excitation are: q (1 R) Φ 0RS V1 = η kr k [ 1 I ( kr ) K ( )] 0 (1) (1 R) Φ 0βRS V = qη I1( kr0 )[ r1 K1( kr1 ) r K1( kr )] () πk r 0 k = R G i πfr C (3) S p n + S p n where I 0, I 1, K 0 and K 1 are modified Bessel functions, Φ 0 the incident photon flux density modulated at frequency f, R the reflectivity, η the quantum efficiency, R S, G p-n, C p-n the p-n junction sheet resistance, capacitance and conductance [1,5].

3 By measuring the JPV signals at the two electrodes with different light modulation frequencies, calibrated by with reference to JPV measurements on a wafer with a deep p-n junction with known sheet resistance and low-leakage, the sheet resistance, R s, conductance G p-n and capacitance of the p-n junction, C p-n can be simultaneously determined using measured voltages V 1 and V and Eqs. (1) and (). The measurement sequence is to first analyze V 1 and V under high light modulation frequencies, where R s xc p-n, in Eq. 3 is the dominant term in the solution. This allows for a direct determination of the USJ sheet resistance, R s, independent of junction depth or leakage current effects. The analysis is then repeated for a lower light modulation frequency, where G p-n xr s, is a more important factor. Since R s is already determined at this point, G p-n (and J RsL ) can be determined directly. For high resolution measurements (>1,000 pixels/wafer), a focused laser beam with a diameter < 0.1 mm is used. In this case, the spatial resolution is mainly determined by the diameter of light beam. Collecting Rs data over a test wafer junction, with pixel counts from 1,000 to up to 70,000/ wafer, provides detailed evaluation of local and global dopant activation uniformity for various implant and anneal conditions. Sub-millimeter spatial resolution with these techniques has been demonstrated with scanned laser annealed test structures. High Density Rs Mapping Process effects which have been clearly identified with RsL mapping include: implant dose variations due to insufficient overlap of beam scan paths, variations in local heating during spike-rtp ( 1050 C/ 1 s) and ms-anneals ( 1350 C/ 1 ms) related to heater lamp signatures, laser scan paths (Fig. ) and pulse-to-pulse laser power variations (Fig. 3). Figure. A 973-point Rs map and 11-point Rs line scan along a 45 o diameter across a 300 mm wafer implanted with 0. kev B and annealed with a scanned CW-laser. Laser beam scanning traces resulted in 3% Rs variations over the wafer.

4 Figure point Rs (left) and Io (right) maps for a 0.5 kev B implant annealed with a step-and-pulse laser on a 00 mm wafer. Regions of high Rs and Io are indicated with the darker (blue) colors. Laser pulse-to-pulse power variations resulted in a global 5% variation in Rs around an average of 996 Ohm/square and Io values from <10-7 to x10-5 A/cm [7]. High-resolution (>1,000 pixels/wafer) are essential for local evaluation of densely patterned local annealing conditions which can be achieved by scanned laser annealing (Fig. 4). The use of a high-resolution RsL probe with a suitably designed laser scan pattern can yield process data on dopant activation for hundreds of separate anneal conditions on a single wafer. Such capabilities are vital to achieve rapid process learning for process proto-typing and analysis of in-line problems. Figure 4. High-resolution (30,000 pts/ 300 mm wafer) Rs map of a locally-patterned, scanned-laser-annealed wafer for rapid proto-typing of multiple process conditions.

5 Flash annealers, which heat the entire wafer surface to >1300 C with sub-ms pulses of light from various arc-lamp sources, rely on a highly-uniform distribution of light power over the wafer surface (just as scanned and pulsed laser annealers require tightly controlled laser beam energies and optics). With high-resolution RsL mapping, local variations in dopant activation can be evaluated with mm-scale resolution over 300 mmd areas. In the example shown in Fig. 5, a flash annealer that employed a bank of linear arc lamps resulted in a variety of dopant activation variations; local high (bright) and low (dark) sheet resistance near the wafer edges and a subtle but clearly discernable stripping pattern over the wafer which mirrors the lamp array. Figure 5. High-resolution (30,000 pts/ 300 mm wafer) Rs maps of a flash-annealed wafer (bottom map after 90 o rotation of wafer shown on the top) with a global Rs uniformity of 10.6% and an average of 3,468 Ohms/square. The stripping pattern across the wafer reflects the array of linear arc lamps used in this tool.

6 RTP spike annealers, which anneal wafers at 1050 C for peak temperature exposures of 1 s, often show Rs patterns that reflect the combined effects of multiple halogen lamps controlled in rings centered on a rotating wafer. High-resolution Rs mapping provides data sufficient to allow tuning of individual lamp-array rings. Figure 6. High-resolution (30,000 pts/ 300 mm wafer) Rs maps of wafers annealed in two different RTP-spike annealers using circular arrays of lamps and wafer rotation during annealing. Rs uniformities are 4.8% (top) and.56% (bottom). Although the focus of most high-performance logic processes is optimization of USJ with doping levels of >10 0 dopants/cm 3 and junction depths <15 nm, new areas interest are also developing for high-breakdown voltage circuits for automotive controllers and deep-well optical sensors which call for deep junctions and much lower doping levels than USJs. In this regime of deep and lightly-doped junctions, RsL analysis, with light sources chosen with deep enough penetration to excite carriers throughout the junction and depletion region, provides a unique probe for dopant activation beyond the reach of contact probes and conventional optical analysis. A submm pixel resolution Rs map for a 90 kev B + implanted low-dose (4x10 11 B/cm ) junction (Fig. 7) shows fine details of the local over-doping (dark bands) resulting from the overlap of beam paths in an x-y scanned beam over the wafer.

7 Figure 7. A high-resolution (30,000 pts/150 mmd, 0.6 mm /pixel) Rs map of a 90 kev B +, 4x10 11 B/cm implanted wafer with an x-y scanned ion beam. The average Rs is 57,671 Ohms/square with a uniformity of.4% Summary High resolution ( 30,000 pixels per wafer) Rs maps derived from RsL analysis of JPV signals, coupled with spatial analysis of junction leakage current variations, provides detailed insight into across-wafer process variations arising from implant and annealing with sub-mm resolution. References 1. V.N. Faifer, M.I. Current, D.K. Schroder, Appl. Phys. Lett (006).. V.N. Faifer, M.I. Current, T.M.H. Wong, V.V. Souchkov, J. Vac. Sci. Technol. B4, (006). 3. V.N. Faifer, D.K. Schroder, M.I. Current, T. Clarysse, P.J. Timins, T. Zangerle, W. Vandervorst, T.M.H. Wong, A. Moussa, W. Lerch, S. Paul, D. Bolze, J. Halim, J. Vac. Sci. Technol. B5(5) (007). 4. V.N. Faifer, M.I. Current, D.K. Schroder, T. Clarysse, and W. Vandervorst, in Analytical and Diagnostic Techniques for Semiconductor Materials, Devices, and Processes 7, ECS Transactions 11, (007). 5. T. Clarysse, A. Moussa, T. Zangerle, F. Schaus, W. Vandervorst, V.N. Faifer, M.I. Current, J. Vac. Sci. Technol. B6, (008). 6. T. Clarysse, A. Moussa, F. Leys, R. Loo, W. Vandervorst, M.C. Benjamin, R.J. Hillard, V.N. Faifer, M.I. Current, R. Lin, D.H. Petersen, Mater. Res. Soc. Symp. Proc C05-07 (006). 7. M.I. Current, V.N. Faifer, T.M.H. Wong, T. Nguyen, A. Koo, Proc. in Ion Implantation Technology-006, AIP Proc. CP (006).