1
Results Exposure was performed with average dose of 0.09 pC. When exposure was performed without dose control, the dot size spreads out more from the average value as the built-in dose variance increases. In contrast, dots exposed with dose control present a constant distribution variance. To improve the performance of dose control: Use different accelerating voltage and aperture to achieve higher SNR. Improve the quality of the mirror in the LCS to achieve better light yield. Implement the algorithm at higher rate. The ADC on XtremeDSP motherboard can sample at speed up to 105 MHz. Real-time Dose Control for Electron Beam Lithography UNIVERSITY OF KENTUCKY Yugu Yang, Stephen A. Maloney and J. Todd Hastings Department of Electrical and Computer Engineering, University of Kentucky, KY40506, USA Acknowledgements This work is supported by the National Science Foundation under Grant No. 0601351. Facilities and technical assistance for this work were provided by the University of Kentucky Center for Nanoscale Science and Engineering (CeNSE). We acknowledge Brian Wajdyk and Chuck May (CeNSE) for their valuable technical assistance. The FPGA design tools were donated by the Xilinx University Program. References [1] G. P Patsis, et al., Microelectronic Engineering 87, 1575-1578, 2010. [2] P. Kruit, et al., Journal of Vacuum Science & Technology B 24, 2931-2935, 2006. [3] N. Rau, et al., Journal of Vacuum Science & Technology B 16, 3784-3788, 1998. [4] H. I. Smith, Journal of Vacuum Science & Technology B 4, 148-153, 1986. [5] S. H. Lee et al., Photomask and NGL Mask Technology XVII, (SPIE, 2010) , pp. 77480J. [6] M. A. McCord and A. D. Brodie, US Patent 7091486 (2006); M. Mankos, et al., US Patent 6555830 (2003); A. Yamada and Y. Oae, US Patent 5449915 (1995); Y. Fujikura, US Patent 4937458 (1990). [7] J. S. Kim, et al., Journal of Nuclear Science and Technology, 515-517, Jun 2008. [8] T. Rahman, et al., Journal of Vacuum Science & Technology B 25, 655-660, Mar-Apr 2007. [9] K. Wilder, et al., Journal of Vacuum Science & Technology B 17, 3256-3261, Nov-Dec 1999. [10] L. R. Baylor, et al., Journal of Vacuum Science & Technology B 20, 2646-2650, Nov-Dec 2002. [11] C. S. A. Durisety, et al., Analog Integrated Circuits and Signal Processing 48, 143-150, 2006. [12] Xilinx XtremeDSP Development Kit-IV Reference Guide, NT 107-0272 – Issue 3. [13] Nallatech DIMEscript User Guide, NT 107-0113 – Issue 3. Introduction Fundamental limitation in Electron Beam lithography (EBL): Random Poisson distribution of electron arrival and resist interaction events [1-4], combined with system imperfections [5], limits the critical dimension (CD) control, line-edge roughness and throughput of EBL. Illustration of feedback system for real-time dose control. The substrate to be patterned is coated with a scintillating layer that produces an optical signal. The signal is detected and processed to determine when each pixel has received sufficient dose so that the control system can stop the exposure. Traditional way of determining exposure dose: •E-beam current is measured before exposure and shot exposure time is calculated to yield desired dose. •Statistical properties of e- beam are not taken into consideration. •Other methods have been implemented that use intervening apertures [6-7] or emitter control [8-11]. New approach: Real-time dose control for every exposed feature •A scintillating layer in resist stack emits several photons for each primary electron. The photons are collected and converted to electrical signal to estimate the dose that has been received by each pixel. •E-beam is blanked once sufficient dose is achieved. •Ultimately, individual electrons could be counted to overcome shot noise limit. Light Detection and Collection Light Sensing •Hamamatsu R4700U photomultiplier tube (PMT) offers gain up to with peak sensitivity through most of the visible spectrum. •Hamamatsu C4900 high voltage power supply provides 0 to -1250 V to power the PMT. •Low voltage electrical feed-through to EBL vacuum chamber Light Collection System (LCS) •Material: UV-curable photopolymer resin •Ellipsoidal hole formed in rectangular solid. •Light source (scintillator) is located at one focus and light sensor at the other. •Aluminum (Al) is thermally evaporated on the inside surface of cavity to increase the reflectivity. •LCS increases the light captured by PMT by 12 times. •Shielded with copper foil and aluminum to prevent charging 6 10 3 Side (a) of LCS is covered with copper foil. Type (a) sample is placed underneath the hole where E-beam focuses through. Pro/E light collection system model Side (a) Side (b) LCS and power supply on sample holder with metal shield. Side (b) of LCS is covered with copper foil. Type (b) sample sits on top of the mirror. Power supply is shielded with a sheet of Al. FPGA Implementation PMT-current integration algorithm: Start accumulating the electrical signal sampled by ADC once the exposure is initiated. A new beam-blanking signal from FPGA is generated to turn off the beam once the desired dose is achieved, or keep the beam on if otherwise. Scintillat or: • Composed of primary scintillator (p-terphenyl), secondary scintillator (POPOP), and polyvinyl toluene (PVT) polymer. Spin- coated from chlorobenzene solvent (C 6 H 5 Cl). 420-nm emission wavelength Samples Scintillator emits light in an upwards direction or through the edges of the film. LCS sits above the sample. Type (a) Type (b) Light emitted from scintillator is reflected downwards by Al film. Glass substrate allows the light to reach to the LCS underneath the sample. Scintillator (~450nm) SiO 2 (~40nm, sputtering) PMMA (~60nm) Silicon Substrate Glass Substrate Scintillator (~450nm) Al (~60nm, E-beam evaporation) PMMA (~60nm) Glass Substrate The change in nominal dose and the dose variance affects neither the average dot size, nor the standard deviation, when exposure is performed with dose control. 0 10 20 30 40 2 4 6 8 10 12 14 16 18 Std.ofExposure D ose(fC ) Std.ofDotDiameter(nm ) Std.ofD otD iam etervs.Std.ofExposure dose w ith D ose C ontrol nom inaldose 0.22pC nom inaldose 0.25pC nom inaldose 0.28pC Ave. Dot Dia. = 226.9nm Ave. Dot Dia. = 227.6nm Ave. Dot Dia. = 227.2nm Dots exposed at lower or higher average dose without feedback shows the same trend in std. of size as shown in the plot on the left when built-in dose variance increases. Average dot size increases as the base dose increases. 0 5 10 15 20 25 30 2 4 6 8 10 12 14 16 18 Std.ofExposure D ose (fC ) Std.ofDotDiam eter(nm ) Std.ofD otD iam etervs.Std.ofExposure D ose w ithoutD ose C ontrol 0.07pC 0.09pC 0.11pC Ave. Dot Dia. = 214.5nm Ave. Dot Dia. = 225.8nm Ave. Dot Dia. = 233.8nm Hardware: XtremeDSP Development Kit-IV motherboard is used for implementation. Signal from scintillator is sampled by ADC on the motherboard at a rate of 40 MS/s [12]. The algorithm is translated into DSP system in System Generator software and then converted into binary file to be downloaded to Virtex-4 user FPGA. Virtex-II clock FPGA synchronizes ADC with user FPGA [12]. DIMEScript language provides access to FPGA from host computer such as data transfer (“dose threshold” in the experiment reported here) and downloading binary design file to hardware [13]. 0 5 10 15 20 25 30 35 40 2 4 6 8 10 12 14 Std.ofExposure D ose (fC ) Std.ofD otSize (nm ) Standard D eviation ofD otD iam etervs.Standard D eviation ofExposure D ose w ith dose control w ithoutdose control 200 205 210 215 220 225 230 235 240 245 0 2 4 6 8 10 12 14 16 DotDiam eter(nm ) N um berperBin w ithoutdose control w ith dose control Dose Var. = ~24 fC 215 220 225 230 235 240 245 0 2 4 6 8 10 12 14 16 18 20 DotDiam eter(nm ) N um berperBin w ithoutdose control w ith dose control Dose Var. = 0 fC 215 220 225 230 235 240 0 2 4 6 8 10 12 DotDiameter(nm ) N um berperBin w ithoutdose control w ith dose control Dose Var. = ~10 fC 0 0.01 0.02 0.03 0.04 0.05 -0.5 0 0.5 1 1.5 2 2.5 3 Tim e (s) V o lta ge (V ) Beam B lank S ignalfrom R A ITH vs.Tim e 0 0.01 0.02 0.03 0.04 0.05 -0.1 0 0.1 0.2 0.3 0.4 Tim e (s) (a) V olta ge (V ) PM T S ignalvs.Tim e 0 0.01 0.02 0.03 0.04 0.05 -0.5 0 0.5 1 1.5 2 2.5 3 Beam B lank S ignalfrom R A ITH vs.Tim e Tim e (s) V o lta ge (V ) 0 0.01 0.02 0.03 0.04 0.05 -0.1 0 0.1 0.2 0.3 0.4 PM T S ignalvs.Tim e Tim e (s) (b) V o lta g e (V ) Comparison of EBL system beam blanker signal (upper traces, “BB_RAITH” in the schematics above) with scintillator signal from the PMT (lower traces). A voltage of 0 indicates that the beam would be on in the absence of feedback control. (a) The dwell time was set to 10 ms (nominal dose 2095 fC) and the feedback system achieved the required dose (629 fC) by terminating the exposure early. (b) The dwell time was set to 1.5 ms (nominal dose 314 fC) and the feedback system extended the dwell times to achieve the desired dose. Hardware testing Setup (a) was used to test the hardware functionality Accelerating voltage: 10 keV; Working distance: ~24mm Dose Control Experiment Exposure settings Setup (b) was used to write the patterns Accelerating voltage: 10 keV; Working distance: ~5mm Dot settling time was set to 1 ms to eliminate the modulated dwell time affecting the exposure of the next dot. Dot arrays in RAITH software • 7x7 dots in each field are separated by 0.9 m to avoid damaging the scintillator under neighboring dots. • Same base dose is applied to all the fields in the same row and different dose variations are introduced intentionally to each column. • Base-dose factor increases vertically (bottom to top) from 50% to 200%. • Dose variance increases Schematic of the experimental setup. The signal from PMT is routed through a transimpedance amplifier and ADC and finally processed by a Virtex-4 FPGA. The beam-blanking control signal from the EBL system (original connection shown with “X”) now provides the control signal for the logic, and a new signal from the FPGA controls the beam blanker. Transimpedance amplifier converts the negative current from PMT to 0-1V positive voltage. Setup (a) : sample (a) below mirror Setup (b) : sample (b) above mirror Schematics of elliptical mirror with video images captured in EBL chamber Determining the “Dose Threshold” Signal from PMT is captured by oscilloscope when scanning the sample near the focus of the elliptical mirror before performing exposure. Calculate the average voltage for the duration of desired dwell time (dt). Dose threshold = (# of samples sampled by ADC in dt) * PMT V PMT V -100 0 100 200 300 400 500 600 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Tim e (us) Voltage (V) Signalfrom Scintillatoron the Incidence ofElectrons vs.Tim e dt V V PMT 58 . 0 for dt Simulation of Future Algorithms Control algorithms 1)Pulse counting : Count current pulse associated with each electron arrival until proper exposure achieved, then stop writing the pixel. 2)Pulse analysis : Divide each current pulse by the average signal value produced by a single electron arriving. Once appropriate number of electrons have arrived, stop exposure. Investigation of the effects of system parameters on the performance of the control algorithms. (a) Normalized standard deviation of electrons required to expose one pixel as a function of photons produced by scintillator on the incidence of electron. (b) Normalized standard deviation of electrons required to expose one pixel as a function of sampling speed. Photons produced per electron is set to 5. 2 3 4 5 6 7 8 9 10 6 8 10 12 14 16 18 20 22 24 26 S tandard D eviation ofE lectrons/P ixel(% )vs.P hotons P erE lectron Photons P erE lectron S tandard D eviation ofE lectrons/P ixel(% ) No C ontrol C ontrol1 (P ulse C ounting) C ontrol2 (P ulse A nalysis) 0 0.5 1 1.5 2 2.5 3 3.5 4 x 10 -8 6 8 10 12 14 16 18 20 22 24 S tandard D eviation ofE lectrons/P ixel(% )vs.S econds P erS am ple S econds P erS am ple S tandard D eviation ofE lectrons/P ixel(% ) No C ontrol C ontrol1 (P ulse C ounting) C ontrol2 (P ulse A nalysis) (a ) (b ) Simulation Results Both algorithms offer improvement in standard deviation when the ratio of photons to electrons is larger than 5 for pulse analysis, and 3 for pulse counting respectively. Pulse counting algorithm depends highly on sampling speed, and provides significant improvement on the performance at sampling speed of 100 MS/s and higher. Conclusions The work reported here serves as a proof of concept for the novel approach of real-time feedback control of dose during electron-beam lithography without any modification of the patterning tool. The experimental results demonstrate that single-pixel exposures with intentionally varied dose are well controlled to yield constant variance and average value of feature size. Implementation on FPGA provides the flexibility to switch from one algorithm to the other, or combine both for future investigation. Schematic of signal routing among modules on the board and the host computer SEM images of dots exposed with average dose of 0.09 pC. (a) Dose control was incorporated during exposure to achieve the desired dose level. The dwell time was set to 600 s (nominal dose 0.28 pC, variance 0.04 pC) in the RAITH lithography software to avoid the effect of e-beam motion on the pattern shape. (b) Feedback control was not used during exposure. The variance of dose is 0.028 pC. ( a ) (b ) m m XtremeDSP Development Kit-IV 40MS/s ADC BB_RAITH Virtex-4 User FPGA ( XC4VSX35-10FF668) Spartan-II Interface FPGA PCI to User FPGA Interface Core data ctrl Host Computer User Interfa ce Softwar e Dose Threshold ctrl Dose Control Logics BB_FPGA Virtex-II Clock FPGA (XC2V80- 4CS144) cloc k Feedback clock Sig. from Scintillat or m m

Results Exposure was performed with average dose of 0.09 pC. When exposure was performed without dose control, the dot size spreads out more from the average

Embed Size (px)

Citation preview

Page 1: Results Exposure was performed with average dose of 0.09 pC. When exposure was performed without dose control, the dot size spreads out more from the average

Results

Exposure was performed with average dose of 0.09 pC. When exposure was performed without dose control, the dot size spreads out more from the average value as the built-in dose variance increases. In contrast, dots exposed with dose control present a constant distribution variance.

To improve the performance of dose control:

• Use different accelerating voltage and aperture to achieve higher SNR.

• Improve the quality of the mirror in the LCS to achieve better light yield.

• Implement the algorithm at higher rate. The ADC on XtremeDSP motherboard can sample at speed up to 105 MHz.

Real-time Dose Control for Electron Beam LithographyUNIVERSITY OF KENTUCKY

Yugu Yang, Stephen A. Maloney and J. Todd Hastings

Department of Electrical and Computer Engineering, University of Kentucky, KY40506, USA

AcknowledgementsThis work is supported by the National Science Foundation under Grant No. 0601351. Facilities and technical assistance for this work were provided by the University of Kentucky Center for Nanoscale Science and Engineering (CeNSE). We acknowledge Brian Wajdyk and Chuck May (CeNSE) for their valuable technical assistance. The FPGA design tools were donated by the Xilinx University Program.

References[1] G. P Patsis, et al., Microelectronic Engineering 87, 1575-1578, 2010.

[2] P. Kruit, et al., Journal of Vacuum Science & Technology B 24, 2931-2935, 2006.

[3] N. Rau, et al., Journal of Vacuum Science & Technology B 16, 3784-3788, 1998.

[4] H. I. Smith, Journal of Vacuum Science & Technology B 4, 148-153, 1986.

[5] S. H. Lee et al., Photomask and NGL Mask Technology XVII, (SPIE, 2010), pp. 77480J.

[6] M. A. McCord and A. D. Brodie, US Patent 7091486 (2006); M. Mankos, et al., US Patent 6555830 (2003); A. Yamada and Y. Oae, US Patent 5449915

(1995); Y. Fujikura, US Patent 4937458 (1990).

[7] J. S. Kim, et al., Journal of Nuclear Science and Technology, 515-517, Jun 2008.

[8] T. Rahman, et al., Journal of Vacuum Science & Technology B 25, 655-660, Mar-Apr 2007.[9] K. Wilder, et al., Journal of Vacuum Science & Technology B 17, 3256-3261, Nov-Dec 1999.[10] L. R. Baylor, et al., Journal of Vacuum Science & Technology B 20, 2646-2650, Nov-Dec 2002.[11] C. S. A. Durisety, et al., Analog Integrated Circuits and Signal Processing 48, 143-150, 2006.[12] Xilinx XtremeDSP Development Kit-IV Reference Guide, NT 107-0272 – Issue 3.

[13] Nallatech DIMEscript User Guide, NT 107-0113 – Issue 3.

IntroductionFundamental limitation in Electron Beam lithography (EBL): Random Poisson distribution of electron arrival and resist interaction events [1-4], combined with system imperfections [5], limits the critical dimension (CD) control, line-edge roughness and throughput of EBL.

Illustration of feedback system for real-time dose control. The substrate to be patterned is coated with a scintillating layer that produces an optical signal. The signal is detected and processed to determine when each pixel has received sufficient dose so that the control system can stop the exposure.

Traditional way of determining exposure dose:

•E-beam current is measured before exposure and shot exposure time is calculated to yield desired dose.

•Statistical properties of e-beam are not taken into consideration.

•Other methods have been implemented that use intervening apertures [6-7] or emitter control [8-11].

New approach: Real-time dose control for every exposed feature

•A scintillating layer in resist stack emits several photons for each primary electron. The photons are collected and converted to electrical signal to estimate the dose that has been received by each pixel.

•E-beam is blanked once sufficient dose is achieved.

•Ultimately, individual electrons could be counted to overcome shot noise limit.

Light Detection and CollectionLight Sensing

•Hamamatsu R4700U photomultiplier tube (PMT) offers gain up to with peak sensitivity through most of the visible spectrum.

•Hamamatsu C4900 high voltage power supply provides 0 to -1250 V to power the PMT.

•Low voltage electrical feed-through to EBL vacuum chamber

Light Collection System (LCS)

•Material: UV-curable photopolymer resin

•Ellipsoidal hole formed in rectangular solid.

•Light source (scintillator) is located at one focus and light sensor at the other.

•Aluminum (Al) is thermally evaporated on the inside surface of cavity to increase the reflectivity.

•LCS increases the light captured by PMT by 12 times.

•Shielded with copper foil and aluminum to prevent charging

6103

Side (a) of LCS is covered with copper foil. Type (a) sample is placed underneath the hole where E-beam focuses through.

Pro/E light collection system model

Side (a) Side (b)

• LCS and power supply on sample holder with metal shield. • Side (b) of LCS is covered with copper foil. Type (b) sample sits on top of the mirror.• Power supply is shielded with a sheet of Al.

FPGA Implementation

PMT-current integration algorithm:

• Start accumulating the electrical signal sampled by ADC once the exposure is initiated.

• A new beam-blanking signal from FPGA is generated to turn off the beam once the desired dose is achieved, or keep the beam on if otherwise.

Scintillator: • Composed of primary scintillator (p-terphenyl), secondary scintillator (POPOP), and polyvinyl toluene (PVT) polymer. Spin-coated from chlorobenzene solvent (C6H5Cl).

• 420-nm emission wavelength

Samples

Scintillator emits light in an upwards direction or through the edges of the film. LCS sits above the sample.

Type (a) Type (b)Light emitted from scintillator is reflected downwards by Al film. Glass

substrate allows the light to reach to the LCS underneath the sample.

Scintillator (~450nm)

SiO2 (~40nm, sputtering)

PMMA (~60nm)

Silicon Substrate Glass Substrate

Scintillator (~450nm)

Al (~60nm, E-beam evaporation)

PMMA (~60nm)

Glass Substrate

The change in nominal dose and the dose variance affects neither the average dot size, nor the standard deviation, when exposure is performed with dose control.

0 10 20 30 402

4

6

8

10

12

14

16

18

Std. of Exposure Dose(fC)

Std

. of

Dot

Dia

met

er (

nm)

Std. of Dot Diameter vs. Std. of Exposure dose with Dose Control

nominal dose 0.22pC nominal dose 0.25pCnominal dose 0.28pC

Ave. Dot Dia. = 226.9nm

Ave. Dot Dia. = 227.6nm

Ave. Dot Dia. = 227.2nm

Dots exposed at lower or higher average dose without feedback shows the same trend in std. of size as shown in the plot on the left when built-in dose variance increases. Average dot size increases as the base dose increases.

0 5 10 15 20 25 302

4

6

8

10

12

14

16

18

Std. of Exposure Dose (fC)

Std

. of

Dot

Dia

met

er (

nm)

Std. of Dot Diameter vs. Std. of Exposure Dose without Dose Control

0.07pC0.09pC0.11pC

Ave. Dot Dia. = 214.5nm

Ave. Dot Dia. = 225.8nm

Ave. Dot Dia. = 233.8nm

Hardware:

• XtremeDSP Development Kit-IV

motherboard is used for implementation.

• Signal from scintillator is sampled by ADC on the motherboard at a rate of 40 MS/s [12].

• The algorithm is translated into DSP system in System Generator software and then converted into binary file to be downloaded to Virtex-4 user FPGA. Virtex-II clock FPGA synchronizes ADC with user FPGA [12].

• DIMEScript language provides access to FPGA from host computer such as data transfer (“dose threshold” in the experiment reported here) and downloading binary design file to hardware [13].

0 5 10 15 20 25 30 35 402

4

6

8

10

12

14

Std. of Exposure Dose (fC)

Std

. of

Dot

Siz

e (n

m)

Standard Deviation of Dot Diameter vs. Standard Deviation of Exposure Dose

with dose controlwithout dose control

200 205 210 215 220 225 230 235 240 2450

2

4

6

8

10

12

14

16

Dot Diameter (nm)

Num

ber

per

Bin

without dose controlwith dose control

Dose Var. = ~24 fC

215 220 225 230 235 240 2450

2

4

6

8

10

12

14

16

18

20

Dot Diameter (nm)

Nu

mb

er

pe

r B

in

without dose controlwith dose control

Dose Var. = 0 fC

215 220 225 230 235 2400

2

4

6

8

10

12

Dot Diameter (nm)

Nu

mb

er

pe

r B

in

without dose controlwith dose control

Dose Var. = ~10 fC

0 0.01 0.02 0.03 0.04 0.05-0.5

00.5

11.5

22.5

3

Time (s)

Vo

lta

ge

(V

)

Beam Blank Signal from RAITH vs. Time

0 0.01 0.02 0.03 0.04 0.05-0.1

0

0.1

0.2

0.3

0.4

Time (s)(a)

Vo

lta

ge

(V

)

PMT Signal vs. Time

0 0.01 0.02 0.03 0.04 0.05-0.5

00.5

11.5

22.5

3Beam Blank Signal from RAITH vs. Time

Time (s)

Vo

lta

ge

(V

)0 0.01 0.02 0.03 0.04 0.05

-0.1

0

0.1

0.2

0.3

0.4PMT Signal vs. Time

Time (s)(b)

Vo

lta

ge

(V

)

Comparison of EBL system beam blanker signal (upper traces, “BB_RAITH” in the schematics above) with scintillator signal from the PMT (lower traces). A voltage of 0 indicates that the beam would be on in the absence of feedback control. (a) The dwell time was set to 10 ms (nominal dose 2095 fC) and the feedback system achieved the required dose (629 fC) by terminating the exposure early. (b) The dwell time was set to 1.5 ms (nominal dose 314 fC) and the feedback system extended the dwell times to achieve the desired dose.

Hardware testing • Setup (a) was used to test the hardware functionality

• Accelerating voltage: 10 keV; Working distance: ~24mm

Dose Control Experiment

Exposure settings

• Setup (b) was used to write the patterns

• Accelerating voltage: 10 keV; Working distance: ~5mm

• Dot settling time was set to 1 ms to eliminate the modulated dwell time affecting the exposure of the next dot.

Dot arrays in RAITH software• 7x7 dots in each field are separated by 0.9 m to avoid damaging the scintillator under neighboring dots.• Same base dose is applied to all the fields in the same row and different dose variations are introduced intentionally to each column.• Base-dose factor increases vertically (bottom to top) from 50% to 200%.• Dose variance increases horizontally (left to right) from 0 to 29% of base dose.

Schematic of the experimental setup. The signal from PMT is routed through a transimpedance amplifier and ADC and finally processed by a Virtex-4 FPGA. The beam-blanking control signal from the EBL system (original connection shown with “X”) now provides the control signal for the logic, and a new signal from the FPGA controls the beam blanker. Transimpedance amplifier converts the negative current from PMT to 0-1V positive voltage.

Setup (a): sample (a) below mirror

Setup (b): sample (b) above mirror

Schematics of elliptical mirror with video images captured in EBL chamber

Determining the “Dose Threshold”

• Signal from PMT is captured by oscilloscope when scanning the sample near the focus of the elliptical mirror before performing exposure.

• Calculate the average voltage for the duration of desired dwell time (dt).

• Dose threshold = (# of samples sampled by ADC in dt) *

PMTV

PMTV -100 0 100 200 300 400 500 600-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Time (us)

Vol

tage

(V

)

Signal from Scintillator on the Incidence of Electrons vs. Time

dt

VVPMT 58.0 for dt

Simulation of Future AlgorithmsControl algorithms

1)Pulse counting: Count current pulse associated with each electron arrival until proper exposure achieved, then stop writing the pixel.

2)Pulse analysis: Divide each current pulse by the average signal value produced by a single electron arriving. Once appropriate number of electrons have arrived, stop exposure.

Investigation of the effects of system parameters on the performance of the control algorithms. (a) Normalized standard deviation of electrons required to expose one pixel as a function of photons produced by scintillator on the incidence of electron. (b) Normalized standard deviation of electrons required to expose one pixel as a function of sampling speed. Photons produced per electron is set to 5.

2 3 4 5 6 7 8 9 106

8

10

12

14

16

18

20

22

24

26Standard Deviation of Electrons/Pixel (%) vs. Photons Per Electron

Photons Per Electron

Sta

ndar

d D

evia

tion

of E

lect

rons

/Pix

el (

%)

No Control

Control 1 (Pulse Counting)Control 2 (Pulse Analysis)

0 0.5 1 1.5 2 2.5 3 3.5 4

x 10-8

6

8

10

12

14

16

18

20

22

24Standard Deviation of Electrons/Pixel (%) vs.Seconds Per Sample

Seconds Per Sample

Sta

ndar

d D

evia

tion

of E

lect

rons

/Pix

el (

%)

No Control

Control 1 (Pulse Counting)Control 2 (Pulse Analysis)

(a) (b)

Simulation Results

• Both algorithms offer improvement in

standard deviation when the ratio of photons

to electrons is larger than 5 for pulse analysis, and 3 for pulse counting respectively.

• Pulse counting algorithm depends highly on sampling speed, and provides significant improvement on the performance at sampling speed of 100 MS/s and higher.

ConclusionsThe work reported here serves as a proof of concept for the novel approach of real-time feedback control of dose during electron-beam lithography without any modification of the patterning tool. The experimental results demonstrate that single-pixel exposures with intentionally varied dose are well controlled to yield constant variance and average value of feature size. Implementation on FPGA provides the flexibility to switch from one algorithm to the other, or combine both for future investigation.

Schematic of signal routing among modules on the board and the host computer

SEM images of dots exposed with average dose of 0.09 pC. (a) Dose control was incorporated during exposure to achieve the desired dose level. The dwell time was set to 600 s (nominal dose 0.28 pC, variance 0.04 pC) in the RAITH lithography software to avoid the effect of e-beam motion on the pattern shape. (b) Feedback control was not used during exposure. The variance of dose is 0.028 pC.

(a) (b)m m

XtremeDSP Development Kit-IV

40MS/sADC

BB_RAITH

Virtex-4 User FPGA( XC4VSX35-10FF668)

Spartan-II Interface FPGA

PCI to User FPGAInterface Core

data ctrl

HostComputer

UserInterfaceSoftware

DoseThreshold

ctrl

DoseControlLogics

BB_FPGA

Virtex-II Clock FPGA (XC2V80-

4CS144)

clock

Feedback clock

Sig. fromScintillator

m m