- Email: [email protected]
Image based in situ electron-beam drift detection by silicon photodiodes in scanning-electron microscopy and an electron-beam lithography system
Microelectronic Engineering 103 (2013) 137–143
Contents lists available at SciVerse ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Image based in situ electron-beam drift detection by silicon photodiodes in scanning-electron microscopy and an electron-beam lithography system Yi-Hung Kuo a, Cheng-Ju Wu a, Fu-Tsun Kuo a, Jia-Yush Yen a,⇑, Yung-Yaw Chen b a b
Department of Mechanical Engineering, National Taiwan University, Taipei 106, Taiwan Department of Electrical Engineering, National Taiwan University, Taipei 106, Taiwan
a r t i c l e
i n f o
Article history: Received 19 December 2011 Received in revised form 28 April 2012 Accepted 17 August 2012 Available online 20 September 2012 Keywords: Electron-beam lithography Electron-beam drift Back scattered electron Electron detectors
a b s t r a c t A silicon-photodiode detector can be used to sense the position of the electron beam in a scanningelectron microscope. In order to validate the implementation of a electron beam drift detector, a silicon photodiode was constructed with a low profile and small working distance. The performance in detecting the drift of the electron beam over time was analyzed. It was also shown that a back scattered-electron image can be created with electron scanning, which allows the development of highly sensitive in situ beam position feedback in the electron-beam direct-write lithography system. Ó 2012 Elsevier B.V. All rights reserved.
1. Introduction An electron-beam lithography system uses an electron beam as the source, and is not subject to the diffraction limitation present in focusing in optical lithography. This greatly improves the resolution and makes it possible to write photoresist patterns on the wafers without using masks. The next generation of lithography systems might employ multiple parallel direct writing systems [1], which would avoid the insufficient direct-writing throughput of the traditional electron-beam system. The use of a multipleelectron-sensor array would enable the position of each electron to be monitored and the drift of the electron beams to be detected, which is due to electron charges in the system and the thermal effect in the process of the electron-beam exposure during long-time operation interfering with the electron beams and the electron fields of the electromagnetic lens. The beam drift reduces the resolution and accuracy in the alignment of electron beams. Therefore, if the electron beams cannot be calibrated frequently it is impossible to achieve high throughputs and yields. In an early attempt, Ogasawara et al. [2] used a semi-blocked Faraday cup (1995) to detect the beam drift by detecting if the e-beam current was earthed. Later, Ando et al. [3] investigated the beam drift by observing the contamination layers on a second shaping aperture. This effort was an off-line process. Goodberlet et al. [4] started to detect the drift of electron beams using interferometer platforms, but this approach cannot detect the drift of each electron beam ⇑ Corresponding author. E-mail address: [email protected] (J.-Y. Yen). 0167-9317/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.mee.2012.08.012
in a multiple-electron-beam system. For real-time beam position control, Hastings et al. [5] devised a fiducial grid on the substrate to achieve online e-beam positioning signal feedback. This approach would require very expensive preprocessing. A silicon photodiode [6] can be used to detect back scattered electrons in a single-electron-beam system. The properties of high gain, small sensitive area, and rapid response of such a photodiode mean that it has potential as a secondary-electron detector [7]. The small-diameter central hole provides a high electron collection capacity because most of the back scattered electrons are close to this hole. Moreover, the silicon photodiode can be easily integrated into an electron-beam lithography system without requiring a large operating voltage (less than 50 V), preventing stray effects on the paths of the electron beams. The silicon photodiode has the potential to replace other secondary-electron detectors such as MCP and E–T detectors [8]. This paper explores the use of the photodiode for e-beam position/drift detection. The back scattered electrons are highly dependent to the incident beam. Using a quadrature detector to detect the deflection of the back scattered beam provide a measure to the incident beam location [9]. This approach, however, gives very low signal to noise ratio. Even with all the efforts of time domain data regression one still cannot surpass resolutions that are in the micrometer range. This is a resolution too low to meet the need for current day electron beam lithography. To achieve the required resolution, this paper proposes to use the image processing technique. Although the time domain signal is very noisy, the bulk image from the space domain is very consistent and is high sensitive to the location of the beam. With the advancement of multiple
138
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
e-beam system, it is possible to dedicate some of the e-beam sources for positioning. Notice that the concern is on the beam drift which is a macro property of the source relative to the target. The characteristics of a few typical beams will be able to describe the behavior of the neighboring areas. We have implemented the concept on a SEM e-beam writer. The experimental results show that the analyses yield results with sub-nanometer resolution. 2. Electron-beam imaging for beam drift detection This paper explores the use of the photodiode for e-beam position detection. The back scattered electrons are very sensitive to the position of the incident beam. Using the photodiode to detect the distribution of the back scattered electrons can be used as a sensitive measure of the electron beam position. Traditional Moiré patterning for IC wafer alignment achieves sub-micrometer accuracy which is insufficient for direct write ebeam lithography applications. Typically, one desires at least nanometer resolution for achieving lithography half pitch in the range of ten nanometers. An intuitive approach was to use a quadrature detector to measure the distribution of the back scattered electrons. This distribution caused imbalance in the detector currents and thus gave the electron beam deflection. And as described in the introduction, the back scattered electron signal was very noisy and yielded very poor measurement results. The resolution was in tens of micrometer range. Our attempts to deduce statistical properties from the time sequence had limited effect due to the inherent random shifting characteristic of the rebounding electrons and the drifting property of the beam. During our study, we noticed that the instantaneous electron image obtained from the back scattered electron is itself consistent with very rich positioning information. It would be possible to deduce position information from the electron image. Furthermore, this measurement on the relative position between the electron source and the substrate is the crucial information for the e-beam lithography process and can be used for direct servo feedback. All other positioning measurements such as the laser interferometry are non-collocated and require extreme precision and servo rigidity to achieve the same accuracy. This section then explains the procedures to derive the positioning information from the back scattering electron images.
sample and chuck. For the drift measurement study, an image was obtained every 2.5 min for the duration of 1 h. 2.2. Image-procession algorithm The image processing procedure follows the phase correlation approach [10] in which two images taken from the same position at different time instances are compared and the difference between the two frames was measured directly from their phases. First, the 2-D Fourier transform was applied to two frames:
Ga ¼ Ffg a g
The cross-power spectrum was then calculated by taking the complex conjugate of the second result, and normalizing this product as follows:
U¼
Beam drift can be detected by comparing the frames in a series of scanning-electron images using a convenient positioning mark, in this case a preprocessed silicon tip as shown in Fig. 1. The silicon tip had a pyramidal shape, as shown in the secondary electron images in Fig. 2. Before drift measurements, the sample with photoresist patterns was placed in the vacuum chuck for at least 5 hours so as to eliminate thermal interactions between the
Ga Gb jGa Gb j
ð2Þ
Finally, the inverse Fourier transform applied to seek the peak point corresponding to the difference in the spatial domain:
/ ¼ F 1 ½U ðDx; DyÞ ¼ maxf/g
ð3Þ
x;y
Fig. 3(a) and 1(b) show two sets of e-beam images of a preprocessed positioning reference mark on the substrate. This leads to the beam drift from the detection of the reference mark notable of the sample, while assumed being stationary. The offset between the images computed are then shown in Fig. 4. But beam drift and the difference between the two frames are in the opposite direction as shown in Fig. 4. For the x15000 magnification, there are 1280 horizontal pixels, the corresponding distance of 7935 nm, so a pixel is equivalent to 6.199 nm. 2.3. Responsivity For characterizing the magnification of the photodiode, we define the responsivity of a silicon photodiode as:
R¼ 2.1. Drift detection
ð1Þ
Gb ¼ Ffg b g
Iph Io ðE=eÞ
ð4Þ
where Iph is the measured photodiode current, Io is the incident electron-beam current, E is the electron landing energy, and e is the unit charge of the electron. 3. Experiment setup This section describes the measurement setup for the study. 3.1. Detector configuration
4 µm
Cr
0.1 µ m
2.6 µ m Si Fig. 1. Profile of the silicon tip.
A silicon photodiode is placed in a scanning-electron microscope (SEM). Four pico-ammeters were used to detect the quadrant currents. The central hole of the silicon photodiode had a diameter of 0.5 mm. The each quadrant sensing area was 11mm2. The responsivity, beam drift, and electron image could be obtained by computation. The entire system based on the silicon photodiode is shown in Fig. 5. A screen with a 60 V bias voltage was positioned directly in front of the silicon photodiode, as shown in Fig. 6(a). Notice that a negative bias was applied on the screen. Instead of the common use of positively biased screen to help trapping more secondary electrons for enhancing the images, we have used a negatively biased screen to filter out the secondary electrons. This was done in an effort to improve the positioning
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
139
Fig. 2. Secondary-electron image of the silicon tip.
Fig. 3. (a) Original frame. (b) Translated frame. The difference between these two frames were (3, 2) pixels.
The detector circuit was mounted in the linear motion feedthrough as shown in Fig. 7. This allows the photodiode detector move to the position just in front of the object lens by the distance of 1 mm, as shown in Fig. 8. The working distance between the photodiode detector and the sample was the normal working distance minus 3 mm. Fig. 9 shows the four quadrant photodiode detectors in the experiment. The back scattered electron forms a distribution over these quadrants. When the difference of the quadrature currents are used the accuracy and the displacement scope would be limited by the sizes of the back scattered electrons on the detectors. The smaller spots would result in a more-linear displacement scope. The image processing algorithm proposed in this paper, however, takes only the sum of all the detected currents for back scattered electron imaging. The position where the incident electron beam strikes the sample can then be analyzed from this image. Fig. 4. The resulting beam drift detected.
characteristics of the back scattered electron signals. Different voltages were tested; and as discussed before, the quad photodiode sensor signals were always too noisy and random to be used directly for positioning. The detector circuit shown in Fig. 6(b) and (c) was combined with the silicon photodiode and the printed circuit board by welding and wire bonding. The printed circuit board has 1 mm thickness, double-sided silver plated. The screen was meshed by 0.3 mmØ stainless wire and installed by 1 mm height above the photodiode detector.
3.2. Measurement procedure The pico-ammeters can convert the four-quadrant electron currents to the four channel analog voltage signals. By the operational amplifiers circuit for analog signal processing, the combined signal is set at ± 5 V range that is acceptable to the auxiliary input of the built-in imaging current board of the SEM. The back scatteredelectron image can be obtained by switching to the auxiliary input on the SEM for scanning. Notice that the strength of the back scattered electron is related to the atomic number. Collecting the four-quadrant currents with electron-beam scanning allows the composition image of the sample to be obtained. When scanning
140
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
Fig. 5. Architecture of the system hardware.
Fig. 7. The linear motion feed-through.
where A, B, C, and D are the currents in the four quadrants shown in Fig. 9. If necessary, the edge of the sample can be emphasized by subtracting the upper- and lower-quadrant currents or the left- and right-quadrant currents:
ðA þ BÞ ðC þ DÞ ðA þ DÞ ðB þ CÞ
Fig. 6. (a) The silicon photodiode detector. (b) The front view of the detector circuit. (c) The back view of the detector circuit.
the sample, the four quadrant currents can be added together to form an image according to
AþBþCþD
ð5Þ
ð6Þ
The experiments in this paper were carried out on an SEM; however, the underlining concept is developed for multiple ebeam direct write systems. The designs of these systems always involve tens of thousands of electron beams operating in parallel and on the same wafer. It is possible to dedicate a small number of the beams in convenient locations for servo feedback purpose. These beams can be outside of the circuit area or where the beams do not affect the circuit layout. We have checked with the semiconductor manufacturer to confirm that pre-patterned wafer for manufacturing is feasible. The authors are also developing the technique to integrate the sensors into the MEMS process for the electron optics. Such imaging can be taken parallel to the e-beam writing process and thus allow in situ feedback for the servo system. Notice that the proposed method provides relative position between the electron source and the wafer for correcting the effect
141
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
Fig. 8. The photodiode detector was positioned in front of the object lens of the SEM.
Fig. 9. The error-measurement theory of four-quadrant photodiodes [11].
of beam drifting over time. Until now there has been very little report on in situ detection of beam drift. The proposed method was not intended to provide information for scanning e-beam positioning. In the following subsections we explain the experimental setup first because the positioning calculation depends on the setup.
-4
2
x 10
1nA
1.9
500pA
4.1. Responsivity The responsivity of the photodiode was first tested to provide an assessment to the magnitude of the detected current. The detectors were exposed to the back scattered-electron from the interaction between the incident beam and the sample. The measurements were carried out for electron landing energy varying from 5 K to 20KeV. The working distance was fixed to 3 mm and the probe current was selected to be 500 pA and 1 nA. During these tests, the output current from the photodiode ranged from 0.3 to 4 nA. The resultant responsivity of the setup for probe currents of 500 pA and 1 nA are shown in Fig. 10. These results are in consistent with other measurements such as the results from [7]. The dark current, measured with the electron source off were 0.2 pA. This is enough to eliminate any possible measurement bias due to the beam blanking. 4.2. Electron-beam drift The beam drift experiments were carried out for probe current set at 500 pA and 1 nA. The settings are widely used in
RESPONSIVILITY (nA / W)
1.8
4. Experimental results
1.7 1.6 1.5 1.4 1.3 1.2 1.1 0
5
10 15 ELECTRON ENERGY (KeV)
20
25
Fig. 10. Responsivity of the photodiode for a working distance of 2 mm.
electron-beam exposures. The acceleration voltages were 5 and 10 kV. The working distance was defined as the distance between the silicon-photodiode detector and the sample. For the measurements performed in this study, the working distance was fixed to 3 mm. Data from the silicon-photodiode detector were plotted in Fig. 11. Similar procedures could be carried out for secondary
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
electron images, and those from the secondary-electron detector that was built into the SEM were plotted in Fig. 12. From these figures, one sees that the two figures are very similar. Both show larger amount of drift for higher acceleration voltages and for larger probe currents. We believe this is a clear hint on the thermal build up, higher voltages and currents resulted in more residual energy in the substrate and thus larger thermal deformation. From this point of view, it is also reasonable to believe that thermal effect on the substrate is the dominating effect on the misalignment between the electron source and the substrate. It is worth noting that the experiment was conducted in a scanning electron microscope whose electron gun and the substrate stages both carry very large thermal inertia. In a multiple beam system, the electron source and optics would be very small and placed very close to the substrate. The thermal effect could be further enhanced. The quantitative measures of the results are computed in Tables 1 and 2 for the images from the photo detector and from the secondary-electron detector, respectively. And the results for the two detectors are compared in Table 3. From Table 1, the back scattered electron images indicates more than 4 nm/min of drifts in both the X and the Y component of the motion axes. The higher acceleration voltage and the high probe current both resulted in larger drift speed. The secondary-electron detector in Table 2 provides the similar measurement. Table 3 listed the differences in the measurements. The secondary-electron detector gives a smaller drift measurement in general. The drift velocity ranges from 6.3 nm/min for 5 keV acceleration voltage and 500 pA probe current to 8.4 nm/min at 10 keV acceleration voltage and 1000 pA probe current. For e-beam lithography targeting 15 nm or less half pitch, this would be a very serious source of error for compensation.
0
-100
RELATIVE DRIFT (nm)
142
-200
-300 Y 5kV 500pA Y 10kV 500pA Y 5kV 1nA
-400
Y 10kV 1nA X 5kV 500pA
-500
X 10kV 500pA X 5kV 1nA X 10kV 1nA
-600
0
10
20
30 TIME (min)
40
50
60
Fig. 12. Electron-beam drift over time relative to the tip (for the secondary-electron detector).
Table 1 Beam drift parameters (for the silicon-photodiode detector). Probe current (pA)
500
Acceleration voltage (kV)
5
10
1000 5
10
X slope (nm/min) Y slope (nm/min)
4.699 4.262
5.445 5.119
5.916 5.382
6.157 5.682
Table 2 Beam drift parameters (for the secondary-electron detector).
4.3. Electron-beam images Comparing the scanning results obtained for the back scatteredelectron images with those for the secondary-electron detector at x100000 magnification revealed that the images sensed by the silicon photodiode showed much less noticeable Z-contrasts as shown in Fig. 13(a) and (b). The difference in contrast basically resulted from the limited input range provided by the auxiliary input of the SEM. But there are techniques available to further enhance the image contrast. Subtracting the left and right quadrant signals clearly revealed the edges of the sample Fig. 13(c), and these were
0
Probe current (pA)
500
Acceleration voltage (kV)
5
10
1000 5
10
X slope (nm/min) Y slope (nm/min)
4.371 4.188
5.964 5.355
5.964 5.758
6.460 5.943
Table 3 Comparison of parameters for the silicon-photodiode detector relative to the secondary-electron detector. Probe current (pA)
500
Acceleration coltage (kV)
5
10
1000 5
10
X slope (nm/min) Y slope (nm/min)
7.51% 1.76%
–8.70% –4.41%
–0.80% –6.53%
–4.70% –4.39%
RELATIVE DRIFT (nm)
-100
even more distinct when subtracting the upper and lower quadrant signals Fig. 13(d). Different subtraction also leads to enhancement in different portion of the image. Future study will look into whether the images from different subtraction can provide different meaning to the measurements.
-200
-300 Y 5kV 500pA Y 10kV 500pA Y 5kV 1nA
-400
5. Conclusions
Y 10kV 1nA X 5kV 500pA
-500
X 10kV 500pA X 5kV 1nA X 10kV 1nA
-600
0
10
20
30 TIME (min)
40
50
60
Fig. 11. Electron-beam drift over time relative to the tip (for the silicon-photodiode detector).
The usefulness of a silicon-photodiode detector in detecting electron drift has been demonstrated by comparing the images obtained by a silicon photodiode and formed from secondary electrons. Due to its low profile and manufacturing convenience, a silicon-photodiode detector is an ideal candidate for a multi-beam sensor array. The image technique, although more computational intensive, is necessary for achieving nanometer measurement resolution under the current setup.
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
143
Fig. 13. Comparison of images of Au particles obtained with a secondary-electron detector operating at a 10-kV acceleration voltage (a), and a silicon-photodiode detector operated at 10 kV with A + B + C + D quadrant processing (b), (A + B) – (C + D) quadrant processing (c), and (A + D)–(B + C) quadrant processing (d).
Acknowledgements This study was financed by the National Science Council of Taiwan under project No. NSC98-2622-E-002-003-CC1. References [1] M.J. van Bruggen, B. van Someren, P. Kruit, J. Vac. Sci. Technol. B 23 (6) (Nov/ Dec 2005). [2] M. Ogasawara, K. Ohtoshi, K. Sugihara, Jpn. J. Appl. Phys. 34 (1995) 6655–6657. [3] A. Ando, H. Sunaoshi, S. Sato, S. Magoshi, K. Hattori, M. Suenaga, H. Wada, H. Housai, S. Hashimoto, K. Sugihara, Jpn. J. Appl. Phys. 35 (1996) 6426–6428.
[4] J.G. Goodberlet, J.T. Hastings, H.I. Smith, J. Vac. Sci.Technol. B: Microelectron. Nanometer Struct. 19 (6) (Nov. 2001) 2499–2503. [5] J.T. Hastings, F. Zhang, H.I. Smith, J. Vac. Sci. Technol. B 21 (6) (Nov/Dec 2003). [6] International Radiation Detector, Inc. Available: http://www.ird-inc.com/ axuv.html. [7] C.S. Silver, J.P. Spallas, L.P. Muray, J. Vac. Sci. Technol. B: Microelectron. Nanometer Struct. 24 (6) (Nov. 2006) 2951–2955. [8] Y.-H. Kuo, C.-J. Wu, J.-Y. Yen, S.-Y. Chen, K.-Y. Tsai, Y.-Y. Chen, Nucl. Instrum. Methods Phys. Res. A 645 (1) (July 2011) 84–89. [9] Yi.-Hung Kuo, Cheng-Ju Wu, Jia.-Yush Yen, Sheng-Yung Chen, Kuen-Yu Tsai, Yung-Yaw Chen, Nucl. Instrum. Methods in Phys. Res. A 645 (2011) 84–89. [10] S. Ertürk, IEEE Trans. Consumer Electron. 49 (4) (Nov. 2003) 1320–1325. [11] L.P. Muray, C.S. Silver, J.P. Spallas, J. Vac. Sci. Technol. B: Microelectron. Nanometer Struct. 24 (6) (Nov. 2006) 2945–2950.
Contents lists available at SciVerse ScienceDirect
Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee
Image based in situ electron-beam drift detection by silicon photodiodes in scanning-electron microscopy and an electron-beam lithography system Yi-Hung Kuo a, Cheng-Ju Wu a, Fu-Tsun Kuo a, Jia-Yush Yen a,⇑, Yung-Yaw Chen b a b
Department of Mechanical Engineering, National Taiwan University, Taipei 106, Taiwan Department of Electrical Engineering, National Taiwan University, Taipei 106, Taiwan
a r t i c l e
i n f o
Article history: Received 19 December 2011 Received in revised form 28 April 2012 Accepted 17 August 2012 Available online 20 September 2012 Keywords: Electron-beam lithography Electron-beam drift Back scattered electron Electron detectors
a b s t r a c t A silicon-photodiode detector can be used to sense the position of the electron beam in a scanningelectron microscope. In order to validate the implementation of a electron beam drift detector, a silicon photodiode was constructed with a low profile and small working distance. The performance in detecting the drift of the electron beam over time was analyzed. It was also shown that a back scattered-electron image can be created with electron scanning, which allows the development of highly sensitive in situ beam position feedback in the electron-beam direct-write lithography system. Ó 2012 Elsevier B.V. All rights reserved.
1. Introduction An electron-beam lithography system uses an electron beam as the source, and is not subject to the diffraction limitation present in focusing in optical lithography. This greatly improves the resolution and makes it possible to write photoresist patterns on the wafers without using masks. The next generation of lithography systems might employ multiple parallel direct writing systems [1], which would avoid the insufficient direct-writing throughput of the traditional electron-beam system. The use of a multipleelectron-sensor array would enable the position of each electron to be monitored and the drift of the electron beams to be detected, which is due to electron charges in the system and the thermal effect in the process of the electron-beam exposure during long-time operation interfering with the electron beams and the electron fields of the electromagnetic lens. The beam drift reduces the resolution and accuracy in the alignment of electron beams. Therefore, if the electron beams cannot be calibrated frequently it is impossible to achieve high throughputs and yields. In an early attempt, Ogasawara et al. [2] used a semi-blocked Faraday cup (1995) to detect the beam drift by detecting if the e-beam current was earthed. Later, Ando et al. [3] investigated the beam drift by observing the contamination layers on a second shaping aperture. This effort was an off-line process. Goodberlet et al. [4] started to detect the drift of electron beams using interferometer platforms, but this approach cannot detect the drift of each electron beam ⇑ Corresponding author. E-mail address: [email protected] (J.-Y. Yen). 0167-9317/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.mee.2012.08.012
in a multiple-electron-beam system. For real-time beam position control, Hastings et al. [5] devised a fiducial grid on the substrate to achieve online e-beam positioning signal feedback. This approach would require very expensive preprocessing. A silicon photodiode [6] can be used to detect back scattered electrons in a single-electron-beam system. The properties of high gain, small sensitive area, and rapid response of such a photodiode mean that it has potential as a secondary-electron detector [7]. The small-diameter central hole provides a high electron collection capacity because most of the back scattered electrons are close to this hole. Moreover, the silicon photodiode can be easily integrated into an electron-beam lithography system without requiring a large operating voltage (less than 50 V), preventing stray effects on the paths of the electron beams. The silicon photodiode has the potential to replace other secondary-electron detectors such as MCP and E–T detectors [8]. This paper explores the use of the photodiode for e-beam position/drift detection. The back scattered electrons are highly dependent to the incident beam. Using a quadrature detector to detect the deflection of the back scattered beam provide a measure to the incident beam location [9]. This approach, however, gives very low signal to noise ratio. Even with all the efforts of time domain data regression one still cannot surpass resolutions that are in the micrometer range. This is a resolution too low to meet the need for current day electron beam lithography. To achieve the required resolution, this paper proposes to use the image processing technique. Although the time domain signal is very noisy, the bulk image from the space domain is very consistent and is high sensitive to the location of the beam. With the advancement of multiple
138
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
e-beam system, it is possible to dedicate some of the e-beam sources for positioning. Notice that the concern is on the beam drift which is a macro property of the source relative to the target. The characteristics of a few typical beams will be able to describe the behavior of the neighboring areas. We have implemented the concept on a SEM e-beam writer. The experimental results show that the analyses yield results with sub-nanometer resolution. 2. Electron-beam imaging for beam drift detection This paper explores the use of the photodiode for e-beam position detection. The back scattered electrons are very sensitive to the position of the incident beam. Using the photodiode to detect the distribution of the back scattered electrons can be used as a sensitive measure of the electron beam position. Traditional Moiré patterning for IC wafer alignment achieves sub-micrometer accuracy which is insufficient for direct write ebeam lithography applications. Typically, one desires at least nanometer resolution for achieving lithography half pitch in the range of ten nanometers. An intuitive approach was to use a quadrature detector to measure the distribution of the back scattered electrons. This distribution caused imbalance in the detector currents and thus gave the electron beam deflection. And as described in the introduction, the back scattered electron signal was very noisy and yielded very poor measurement results. The resolution was in tens of micrometer range. Our attempts to deduce statistical properties from the time sequence had limited effect due to the inherent random shifting characteristic of the rebounding electrons and the drifting property of the beam. During our study, we noticed that the instantaneous electron image obtained from the back scattered electron is itself consistent with very rich positioning information. It would be possible to deduce position information from the electron image. Furthermore, this measurement on the relative position between the electron source and the substrate is the crucial information for the e-beam lithography process and can be used for direct servo feedback. All other positioning measurements such as the laser interferometry are non-collocated and require extreme precision and servo rigidity to achieve the same accuracy. This section then explains the procedures to derive the positioning information from the back scattering electron images.
sample and chuck. For the drift measurement study, an image was obtained every 2.5 min for the duration of 1 h. 2.2. Image-procession algorithm The image processing procedure follows the phase correlation approach [10] in which two images taken from the same position at different time instances are compared and the difference between the two frames was measured directly from their phases. First, the 2-D Fourier transform was applied to two frames:
Ga ¼ Ffg a g
The cross-power spectrum was then calculated by taking the complex conjugate of the second result, and normalizing this product as follows:
U¼
Beam drift can be detected by comparing the frames in a series of scanning-electron images using a convenient positioning mark, in this case a preprocessed silicon tip as shown in Fig. 1. The silicon tip had a pyramidal shape, as shown in the secondary electron images in Fig. 2. Before drift measurements, the sample with photoresist patterns was placed in the vacuum chuck for at least 5 hours so as to eliminate thermal interactions between the
Ga Gb jGa Gb j
ð2Þ
Finally, the inverse Fourier transform applied to seek the peak point corresponding to the difference in the spatial domain:
/ ¼ F 1 ½U ðDx; DyÞ ¼ maxf/g
ð3Þ
x;y
Fig. 3(a) and 1(b) show two sets of e-beam images of a preprocessed positioning reference mark on the substrate. This leads to the beam drift from the detection of the reference mark notable of the sample, while assumed being stationary. The offset between the images computed are then shown in Fig. 4. But beam drift and the difference between the two frames are in the opposite direction as shown in Fig. 4. For the x15000 magnification, there are 1280 horizontal pixels, the corresponding distance of 7935 nm, so a pixel is equivalent to 6.199 nm. 2.3. Responsivity For characterizing the magnification of the photodiode, we define the responsivity of a silicon photodiode as:
R¼ 2.1. Drift detection
ð1Þ
Gb ¼ Ffg b g
Iph Io ðE=eÞ
ð4Þ
where Iph is the measured photodiode current, Io is the incident electron-beam current, E is the electron landing energy, and e is the unit charge of the electron. 3. Experiment setup This section describes the measurement setup for the study. 3.1. Detector configuration
4 µm
Cr
0.1 µ m
2.6 µ m Si Fig. 1. Profile of the silicon tip.
A silicon photodiode is placed in a scanning-electron microscope (SEM). Four pico-ammeters were used to detect the quadrant currents. The central hole of the silicon photodiode had a diameter of 0.5 mm. The each quadrant sensing area was 11mm2. The responsivity, beam drift, and electron image could be obtained by computation. The entire system based on the silicon photodiode is shown in Fig. 5. A screen with a 60 V bias voltage was positioned directly in front of the silicon photodiode, as shown in Fig. 6(a). Notice that a negative bias was applied on the screen. Instead of the common use of positively biased screen to help trapping more secondary electrons for enhancing the images, we have used a negatively biased screen to filter out the secondary electrons. This was done in an effort to improve the positioning
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
139
Fig. 2. Secondary-electron image of the silicon tip.
Fig. 3. (a) Original frame. (b) Translated frame. The difference between these two frames were (3, 2) pixels.
The detector circuit was mounted in the linear motion feedthrough as shown in Fig. 7. This allows the photodiode detector move to the position just in front of the object lens by the distance of 1 mm, as shown in Fig. 8. The working distance between the photodiode detector and the sample was the normal working distance minus 3 mm. Fig. 9 shows the four quadrant photodiode detectors in the experiment. The back scattered electron forms a distribution over these quadrants. When the difference of the quadrature currents are used the accuracy and the displacement scope would be limited by the sizes of the back scattered electrons on the detectors. The smaller spots would result in a more-linear displacement scope. The image processing algorithm proposed in this paper, however, takes only the sum of all the detected currents for back scattered electron imaging. The position where the incident electron beam strikes the sample can then be analyzed from this image. Fig. 4. The resulting beam drift detected.
characteristics of the back scattered electron signals. Different voltages were tested; and as discussed before, the quad photodiode sensor signals were always too noisy and random to be used directly for positioning. The detector circuit shown in Fig. 6(b) and (c) was combined with the silicon photodiode and the printed circuit board by welding and wire bonding. The printed circuit board has 1 mm thickness, double-sided silver plated. The screen was meshed by 0.3 mmØ stainless wire and installed by 1 mm height above the photodiode detector.
3.2. Measurement procedure The pico-ammeters can convert the four-quadrant electron currents to the four channel analog voltage signals. By the operational amplifiers circuit for analog signal processing, the combined signal is set at ± 5 V range that is acceptable to the auxiliary input of the built-in imaging current board of the SEM. The back scatteredelectron image can be obtained by switching to the auxiliary input on the SEM for scanning. Notice that the strength of the back scattered electron is related to the atomic number. Collecting the four-quadrant currents with electron-beam scanning allows the composition image of the sample to be obtained. When scanning
140
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
Fig. 5. Architecture of the system hardware.
Fig. 7. The linear motion feed-through.
where A, B, C, and D are the currents in the four quadrants shown in Fig. 9. If necessary, the edge of the sample can be emphasized by subtracting the upper- and lower-quadrant currents or the left- and right-quadrant currents:
ðA þ BÞ ðC þ DÞ ðA þ DÞ ðB þ CÞ
Fig. 6. (a) The silicon photodiode detector. (b) The front view of the detector circuit. (c) The back view of the detector circuit.
the sample, the four quadrant currents can be added together to form an image according to
AþBþCþD
ð5Þ
ð6Þ
The experiments in this paper were carried out on an SEM; however, the underlining concept is developed for multiple ebeam direct write systems. The designs of these systems always involve tens of thousands of electron beams operating in parallel and on the same wafer. It is possible to dedicate a small number of the beams in convenient locations for servo feedback purpose. These beams can be outside of the circuit area or where the beams do not affect the circuit layout. We have checked with the semiconductor manufacturer to confirm that pre-patterned wafer for manufacturing is feasible. The authors are also developing the technique to integrate the sensors into the MEMS process for the electron optics. Such imaging can be taken parallel to the e-beam writing process and thus allow in situ feedback for the servo system. Notice that the proposed method provides relative position between the electron source and the wafer for correcting the effect
141
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
Fig. 8. The photodiode detector was positioned in front of the object lens of the SEM.
Fig. 9. The error-measurement theory of four-quadrant photodiodes [11].
of beam drifting over time. Until now there has been very little report on in situ detection of beam drift. The proposed method was not intended to provide information for scanning e-beam positioning. In the following subsections we explain the experimental setup first because the positioning calculation depends on the setup.
-4
2
x 10
1nA
1.9
500pA
4.1. Responsivity The responsivity of the photodiode was first tested to provide an assessment to the magnitude of the detected current. The detectors were exposed to the back scattered-electron from the interaction between the incident beam and the sample. The measurements were carried out for electron landing energy varying from 5 K to 20KeV. The working distance was fixed to 3 mm and the probe current was selected to be 500 pA and 1 nA. During these tests, the output current from the photodiode ranged from 0.3 to 4 nA. The resultant responsivity of the setup for probe currents of 500 pA and 1 nA are shown in Fig. 10. These results are in consistent with other measurements such as the results from [7]. The dark current, measured with the electron source off were 0.2 pA. This is enough to eliminate any possible measurement bias due to the beam blanking. 4.2. Electron-beam drift The beam drift experiments were carried out for probe current set at 500 pA and 1 nA. The settings are widely used in
RESPONSIVILITY (nA / W)
1.8
4. Experimental results
1.7 1.6 1.5 1.4 1.3 1.2 1.1 0
5
10 15 ELECTRON ENERGY (KeV)
20
25
Fig. 10. Responsivity of the photodiode for a working distance of 2 mm.
electron-beam exposures. The acceleration voltages were 5 and 10 kV. The working distance was defined as the distance between the silicon-photodiode detector and the sample. For the measurements performed in this study, the working distance was fixed to 3 mm. Data from the silicon-photodiode detector were plotted in Fig. 11. Similar procedures could be carried out for secondary
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
electron images, and those from the secondary-electron detector that was built into the SEM were plotted in Fig. 12. From these figures, one sees that the two figures are very similar. Both show larger amount of drift for higher acceleration voltages and for larger probe currents. We believe this is a clear hint on the thermal build up, higher voltages and currents resulted in more residual energy in the substrate and thus larger thermal deformation. From this point of view, it is also reasonable to believe that thermal effect on the substrate is the dominating effect on the misalignment between the electron source and the substrate. It is worth noting that the experiment was conducted in a scanning electron microscope whose electron gun and the substrate stages both carry very large thermal inertia. In a multiple beam system, the electron source and optics would be very small and placed very close to the substrate. The thermal effect could be further enhanced. The quantitative measures of the results are computed in Tables 1 and 2 for the images from the photo detector and from the secondary-electron detector, respectively. And the results for the two detectors are compared in Table 3. From Table 1, the back scattered electron images indicates more than 4 nm/min of drifts in both the X and the Y component of the motion axes. The higher acceleration voltage and the high probe current both resulted in larger drift speed. The secondary-electron detector in Table 2 provides the similar measurement. Table 3 listed the differences in the measurements. The secondary-electron detector gives a smaller drift measurement in general. The drift velocity ranges from 6.3 nm/min for 5 keV acceleration voltage and 500 pA probe current to 8.4 nm/min at 10 keV acceleration voltage and 1000 pA probe current. For e-beam lithography targeting 15 nm or less half pitch, this would be a very serious source of error for compensation.
0
-100
RELATIVE DRIFT (nm)
142
-200
-300 Y 5kV 500pA Y 10kV 500pA Y 5kV 1nA
-400
Y 10kV 1nA X 5kV 500pA
-500
X 10kV 500pA X 5kV 1nA X 10kV 1nA
-600
0
10
20
30 TIME (min)
40
50
60
Fig. 12. Electron-beam drift over time relative to the tip (for the secondary-electron detector).
Table 1 Beam drift parameters (for the silicon-photodiode detector). Probe current (pA)
500
Acceleration voltage (kV)
5
10
1000 5
10
X slope (nm/min) Y slope (nm/min)
4.699 4.262
5.445 5.119
5.916 5.382
6.157 5.682
Table 2 Beam drift parameters (for the secondary-electron detector).
4.3. Electron-beam images Comparing the scanning results obtained for the back scatteredelectron images with those for the secondary-electron detector at x100000 magnification revealed that the images sensed by the silicon photodiode showed much less noticeable Z-contrasts as shown in Fig. 13(a) and (b). The difference in contrast basically resulted from the limited input range provided by the auxiliary input of the SEM. But there are techniques available to further enhance the image contrast. Subtracting the left and right quadrant signals clearly revealed the edges of the sample Fig. 13(c), and these were
0
Probe current (pA)
500
Acceleration voltage (kV)
5
10
1000 5
10
X slope (nm/min) Y slope (nm/min)
4.371 4.188
5.964 5.355
5.964 5.758
6.460 5.943
Table 3 Comparison of parameters for the silicon-photodiode detector relative to the secondary-electron detector. Probe current (pA)
500
Acceleration coltage (kV)
5
10
1000 5
10
X slope (nm/min) Y slope (nm/min)
7.51% 1.76%
–8.70% –4.41%
–0.80% –6.53%
–4.70% –4.39%
RELATIVE DRIFT (nm)
-100
even more distinct when subtracting the upper and lower quadrant signals Fig. 13(d). Different subtraction also leads to enhancement in different portion of the image. Future study will look into whether the images from different subtraction can provide different meaning to the measurements.
-200
-300 Y 5kV 500pA Y 10kV 500pA Y 5kV 1nA
-400
5. Conclusions
Y 10kV 1nA X 5kV 500pA
-500
X 10kV 500pA X 5kV 1nA X 10kV 1nA
-600
0
10
20
30 TIME (min)
40
50
60
Fig. 11. Electron-beam drift over time relative to the tip (for the silicon-photodiode detector).
The usefulness of a silicon-photodiode detector in detecting electron drift has been demonstrated by comparing the images obtained by a silicon photodiode and formed from secondary electrons. Due to its low profile and manufacturing convenience, a silicon-photodiode detector is an ideal candidate for a multi-beam sensor array. The image technique, although more computational intensive, is necessary for achieving nanometer measurement resolution under the current setup.
Y.-H. Kuo et al. / Microelectronic Engineering 103 (2013) 137–143
143
Fig. 13. Comparison of images of Au particles obtained with a secondary-electron detector operating at a 10-kV acceleration voltage (a), and a silicon-photodiode detector operated at 10 kV with A + B + C + D quadrant processing (b), (A + B) – (C + D) quadrant processing (c), and (A + D)–(B + C) quadrant processing (d).
Acknowledgements This study was financed by the National Science Council of Taiwan under project No. NSC98-2622-E-002-003-CC1. References [1] M.J. van Bruggen, B. van Someren, P. Kruit, J. Vac. Sci. Technol. B 23 (6) (Nov/ Dec 2005). [2] M. Ogasawara, K. Ohtoshi, K. Sugihara, Jpn. J. Appl. Phys. 34 (1995) 6655–6657. [3] A. Ando, H. Sunaoshi, S. Sato, S. Magoshi, K. Hattori, M. Suenaga, H. Wada, H. Housai, S. Hashimoto, K. Sugihara, Jpn. J. Appl. Phys. 35 (1996) 6426–6428.
[4] J.G. Goodberlet, J.T. Hastings, H.I. Smith, J. Vac. Sci.Technol. B: Microelectron. Nanometer Struct. 19 (6) (Nov. 2001) 2499–2503. [5] J.T. Hastings, F. Zhang, H.I. Smith, J. Vac. Sci. Technol. B 21 (6) (Nov/Dec 2003). [6] International Radiation Detector, Inc. Available: http://www.ird-inc.com/ axuv.html. [7] C.S. Silver, J.P. Spallas, L.P. Muray, J. Vac. Sci. Technol. B: Microelectron. Nanometer Struct. 24 (6) (Nov. 2006) 2951–2955. [8] Y.-H. Kuo, C.-J. Wu, J.-Y. Yen, S.-Y. Chen, K.-Y. Tsai, Y.-Y. Chen, Nucl. Instrum. Methods Phys. Res. A 645 (1) (July 2011) 84–89. [9] Yi.-Hung Kuo, Cheng-Ju Wu, Jia.-Yush Yen, Sheng-Yung Chen, Kuen-Yu Tsai, Yung-Yaw Chen, Nucl. Instrum. Methods in Phys. Res. A 645 (2011) 84–89. [10] S. Ertürk, IEEE Trans. Consumer Electron. 49 (4) (Nov. 2003) 1320–1325. [11] L.P. Muray, C.S. Silver, J.P. Spallas, J. Vac. Sci. Technol. B: Microelectron. Nanometer Struct. 24 (6) (Nov. 2006) 2945–2950.