Improving angle control in sub-65nm implantation
11/01/2006
A new angle control technology has been developed to automatically measure beam angles before implant in both the horizontal and vertical planes and to correct for any deviation from the desired implant angle. A symmetric parallelizing lens that corrects angles without bending the beam enables calibration of the horizontal angle mask to crystal planes with one or two wafers. This paper describes the measurement and calibration methods of this angle control system.
 |
In situ methods to measure beam angles in both the horizontal and vertical planes are required for accurate control of the implant angle. It is necessary that these methods be calibrated to the surface or crystal planes of the wafer to achieve the required angle control.
Implant angle control is increasingly important with each new device node. Some devices have demonstrated a sensitivity of threshold voltage of ~100 mV/deg. for implant angle and require implant angles to be held within ±0.2° for process control. There are many sources of angle variation in single wafer implanters. Mechanical orientation can usually be controlled to high precision, but accurate control of the implant angle requires knowledge of the actual beam angle relative to the surface or crystal planes of the wafer.
Each succeeding device node has placed more stringent requirements on angle control, and implant angle control has become a necessary component of implanter process control. Early medium-current implanters used raster-scanned beams to span wafers up to 150mm dia. with scan angles of up to ±2° across the wafer. Channeling effects on profile depth were tolerated or managed by using nonchanneling crystal orientations. While these methods provided adequate angle control for devices at that time, it became clear that 200mm wafers would need better angle control. Parallel beam implanters were introduced to avoid the scan angle variation [1-3] and offered implant angle control of <1°, which was satisfactory for devices until recently.
Device structures that can lead to asymmetry in source/drain extensions of transistors due to shadowing of gates or narrowly spaced photoresist patterns are placing even tighter limits on angle control. Devices in which Vt varies by up to 100mV/deg. require angle control of 0.2° [4]. Such tight control requires in situ measurement of the implant angle in horizontal and vertical planes and is smaller than the present tolerance on crystal-cut error.
Sources of variability
All ribbon beam or ribbon-like beam implanters today use a magnetic or electrostatic lens to image ions to infinity after they diverge in one plane from a waist near the focal point of the lens [1-3, 5]. The focal lengths of these lenses are in the range of 600-1000mm. Magnetic lenses achieve parallelism by bending the beam in a wedge-shaped dipole magnet through an angle ranging from 40°-70° for the central ray. Electrostatic lenses allow the central ray to continue in a straight line, while the scanned rays have the scan angle canceled by the vector sum of velocities in the acceleration or deceleration gap between the shaped electrodes, which are symmetric about the central ray.
For some simple effects, the focal properties of these lenses can be approximated by the thin lens equation. If the object point of the lens or position of the scan vertex moves to one side of the focal point of the lens, the image is still focused to infinity (parallel rays), but at an angle θ, where Δx is the lateral displacement and F is the focal length,
 |
Also, if the implanter focusing elements make the apparent vertex move backward or forward along the beam by Δz, the beam will be converging or diverging at the edge of a 150mm radius wafer by
 |
For small errors of 5mm in position and F = 800mm, these amount to 0.36° and 0.07° angles, respectively.
Electrostatic lenses are designed for a fixed ratio of lens voltage to extraction voltage; for example, the Optima MD lens, a 1% error in the ratio results in a 0.05° error at the edge of the 300mm wafer. A magnetic lens is designed for a fixed bend radius or ratio of (ME/Q2)1/2/B. For a 50° bend angle, a 1% error in the field B due to hysteresis, for example, can result in an error of just under 0.5° in the average angle. The field must be set to achieve parallelism across the wafer, however, and any residual error in the average angle must be corrected by tilting the wafer about a vertical axis [6]. Charge exchange of ions that have partially completed the bend on the corrector magnet may strike the wafer at the wrong angle, which can also result in dose nonuniformity [7].
Deceleration/acceleration of a beam with an angle error magnifies/demagnifies the angle by a factor M,
 |
where Ein is the energy before decel/accel and Eout is the energy after. Other effects, such as beam blowup and portions of divergent beams interacting with apertures, also affect the final effective angle at the wafer. Vertical angle errors may also occur due to source/extraction alignment or by vertical deflectors or focusing aberrations [8].
Horizontal angle measurement and calibration
If a recipe has a limit on maximum implant angle, beam-tuning software can be used to automatically measure and correct angles to satisfy the recipe limit. If angles are diverging or converging as one moves from the center toward the left
ight edges of the wafer, corrections can be made by adjusting the voltage of a parallelizing lens (P-lens) (Fig. 1) to make the lens stronger or weaker. The P-lens voltage can be adjusted independently of the beam energy. Average angle errors are corrected by steering the pencil beam before the scanner and P-lens horizontally by two electric elements of the middle quadrupole. For a given angle error, the required change in the P-lens voltage and steerer is computed in a model-based, closed-loop control algorithm that accounts for the effects of acceleration or deceleration; angles are usually corrected with one iteration.
 Figure 1. The path of ion rays from scanner to wafer with the P-lens voltage on and off. |
Angle measurement devices can be calibrated to the crystal planes of the wafer to provide a proper reference for channeled implants and to confirm that a 0° implant angle is actually normal to the surface of the wafer. Axially symmetric parallelizing P-lenses, for example, enable a method for horizontal angle calibration. During normal operation, the P-lens is set to a voltage that cancels the scan angle to produce parallel rays. The horizontal beam angles of the scanned beam are measured using a moving profiler behind a mask with vertical slots at seven points across the wafer [9]. If the P-lens voltage is turned off, and with zero voltage on the accel column, the scanned beam travels at the extraction energy of 45keV in straight lines from the scanner to the wafer, as shown in Fig 1. In that case, the implant angle ranges linearly ~±4° across the wafer, and the beam angle is easily predicted from the geometry. When a wafer is implanted in this mode, at some point on the wafer, the ions will have the angle that is parallel to the crystal planes, and a measurement using a Therma Wave Inc. Therma-probe tool (TW), or sheet resistance, will reach a minimum.
As shown in Fig. 2, for a portion of the wafer, beam angles measured across the wafer form a straight line plot of angle vs. position while the TW values form an approximately parabolic shape with a minimum at the position of maximum channeling.
The corresponding position on the angle plot is +0.05° for this wafer. If the wafer had no crystal-cut error, one would calibrate the mask to read 0° at that position. However, to compensate for crystal-cut error, one typically uses two wafers implanted with a 180° difference in twist.
Vertical angle measurement
The vertical beam angle monitor (VBA) is mounted on the arm that supports the electrostatic chuck such that there is a fixed offset between the angle of the surface of the VBA and the surface of the wafer of about 30°. The VBA is a faraday with a thick slotted mask that limits the current transmitted to the collector as the slots in the mask are tilted relative to the beam angle. This will produce a current as a function of the tilt angle of the VBA that peaks when the slots are aligned with the beam and that allows accurate determination of the "center of mass vertical angle" of the beam.
The fixed offset between the VBA and the wafer surface allows one to set the chuck tilt relative to the actual beam angle to the value specified in the recipe. The vertical beam angle is measured at the start of a batch by tilting the chuck so that the VBA is in line with the ion beam. The chuck and VBA are tilted ±5° about this point to collect the data; at these tilt angles, the wafer is safely above the ion beam. The chuck then moves to the implant position and scans through the beam at the desired tilt angle relative to the measured beam angle, while the VBA remains fixed to the tilt mechanism and is safely out of the beam.
Vertical beam angle calibration
As with horizontal angles, it is preferred to calibrate the vertical beam angle measurement to the crystal planes of a wafer as it enables the verification that a 0° tilt is indeed perpendicular to the wafer. In this case, implants are done at a range of tilt angles with wafers from a single boule. The VBA is used to measure the beam angle, then implants are performed at tilts ranging ±1.5° and the average TW data are fit with a second order trendline to determine the minimum accurately.
 Figure 3. Calibration of VBA (or beam-tilt angle) measurement to crystal planes using 370keV P++, 5E13 at near 0°/0° orientation. |
The data in Fig. 3 show an offset of 0.54° between the average angle of the uncalibrated VBA and the crystal planes. Results are not shown, but these implants are also run with 180° twist to compensate for crystal-cut error. The angle offset is then zeroed so that subsequent angle measurements are referenced to the ideal crystal plane orientation normal to the surface of the wafer. Implants are then performed with constant focal length scanning, ensuring the wafer has the same beam focus and angle content at all positions regardless of tilt.
 Figure 4. SIMS profiles of channeled implants of 450keV P++, 5E13 at 0°/0° in {100} wafers at five points compared to implants at 0.2° and 0.5° tilts. |
An implant of 450 keV P++ was made at a tilt/twist of 0°/0° and the implanted profile was measured on center and at a radius of 140mm at four points: 3, 6, 9, and 12 o’clock. As shown in Fig. 4, the profiles match well with each other, indicating all points of the wafer were implanted with ions at the same angle. Profiles from wafers implanted at tilts of 0.2° and 0.5° show the sensitivity of this implant to small angle errors.
Conclusion
Precise implant angle control becomes a determining factor for device performance in the sub-65nm era; some devices demand a threshold voltage of ~100mV/deg for implant angle and require implant angles to be held within ±0.2° for process control. As demonstrated in this article, in situ methods that measure beam angles in both the horizontal and vertical planes (and the calibration of these methods to the surface or crystal planes of the wafer) prior to implant enable angles to be corrected and controlled within 0.2° at most conditions, ensuring proper implant process control.
References
1. D.W. Berrian, et al., Nucl. Instr. and Meth. B37/38, pp. 500-503, 1989.
2. A.M. Ray, J.P. Dykstra, R.B. Simonton, Nucl. Instr. and Meth. B74, pp. 401-404, 1993.
3. T. Kawai, et al., Nucl. Instr. and Meth. B96, pp. 470-473, 1995.
4. A. Agarwal, private communication, 2006.
5. N.R. White, M. Sieradzki, A. Renau, IEEE Proc. of 11th Intl. Conf. on Ion Implantation Tech., Austin, TX, pp. 396-399, 1996.
6. J.C. Olson, A. Renau, IEEE Proc. of Intl. Conf. on Ion Implantation Tech., Alpbach, Austria, pp. 670-673, 2000.
7. M.R. LaFontaine, et al., IEEE Proc. of Intl. Conf. on Ion Implantation Tech., Alpbach, Austria, pp. 246-250, 2000.
8. D.C. Sing, M.J. Rendon, Nucl. Instr. and Meth. B237, pp. 318-323, 2005.
9. R.D. Rathmell, et al., IEEE Proc. of Intl. Conf. on Ion Implantation Tech., Kyoto, Japan, pp. 392-395, 1998.
R.D. Rathmell received his BS in physics at Purdue U. in 1966 and his PhD in nuclear physics at the U. of Wisconsin in 1971. He is a beamline scientist at Axcelis Technologies Inc., 108 Cherry Hill Drive, Beverly, MA 01915; ph 978/787-4000, [email protected].
B. Vanderberg received his BS in mathematics and physics at the U. of Strasbourg, France, in 1984 and his PhD in industrial electrotechnology at the Royal Institute of Technology, Stockholm, Sweden, in 1994. He is a staff scientist at Axcelis Technologies Inc.
A.M. Ray received his BSME at the U. of Texas in 1980 and is manager of systems technology at Axcelis Technologies Inc.