Focused-ion-beam fabrication of nanoplasmonic devices

Transcription

1 Focused-ion-beam fabrication of nanoplasmonic devices H. J. Lezec Center for Nanoscale Science and Technology, NIST, Gaithersburg MD, USA..

2 Outline 1) Plasmon-induced negative refraction at visible frequencies 2) FIB dual-beam systems at NIST 3) The need for FIB automation

3 The need for general-purpose, user-friendly automation in the laboratory-environment. Automation is currently supported in very limited ways: Scripting/Macros a) Require text programming b) No graphical programming. c) No offline job setup and simulation. d) Limited programming architecture. e) Limited functionality (some machine functions not accessible). f) Limited machine-vision capabilites. g) No metrology capabilities (apart from eyeballing measurements) h) Milling not intimately linked to measurements. Prepackaged solutions(such as TEM lamella preparation) a) Lack flexibilty b) Are rarely robust (in part because they don t fully exploit machine vision/ metrology) c) Don t meet endless variety of situations encountered in the lab.

4 What makes automation a necessity Machines are vastly underutilized (could be milling day and night; they are mostly idle). Many prototyping applications require patterning hundreds of devices with small dimensional variations: impossible to do by hand without making mistakes. Ensures accuracy. Ensures repeatability. Self-documenting. Time is money. Repetitive tasks are boring.

5 What makes automation timely The FIB is a very robust tool (can maintain focus/stig for days). Can be managed in a very systematic manner. Tight beams at high current levels are now available: can take advantage of high throughput. Stages are adequate (don t need perfect landing accuracy if pattern recognition is involved) Pattern recognition/metrology is extremely fast and robust, and better than the human eye (sub pixel measurement accuracy)

6 Other Enabling options Live imaging in milling box (for facilitating recipe development) Same digital scan generator for imaging and writing (essential for feature placement accuracy).

7 What is still lacking A self calibrating system using in-situ calibration standards (focus, astigmatism, beam diameter and ellipticity, beam current, deflection gains, aperture erosion ). Necessary for tool matching! Providing quantitative results of those calibrations to user at all times to establish confidence levels. Beam is in focus, Measured beam diameter is 35nm x scan gain is within 99% of target. User should not have to defocus beam to check that it is in focus. Eliminate knob twirling. Real-time working-distance determination at point of beam impact would allow deterministic focusing on non-flat samples. Especially lacking: a user-friendly, visual approach to automation.

8 A possible approach: Smart shapes for robust and easily programmed automated metrology and milling recipes

9 FEI IC3D : Scalable smart-shapes enable complex automatic metrology sequences

10 Requested metrology layout

11 Load image from product database

12 Drag composite edge-find tool ( Outline ) from product stencil Blue boxes: patmax outputs. Grey boxes: horizontal-offset tool.

13 Run new image. 4 patmax outputs each snap into place to identify principal edges of structure

14 Drag edge-characterization tool from product stencil

15 Composite tool is snapped onto patmax outputs. Scaling of calipers occurs automatically as result of scaling of parent shape.

16 Tool is now in place.

17 Duplicate tool.

18 and snap it to other side of structure.

19 Drag line with extensions off of FEI metrology stencil

20 Attach to output connection points of horizontal edge-fine tools

21 Extend lines

22 Define intersection point on left

23 and on right.

24 Pull 2 angle center tools off of metrology stencil

25 Clip into place on lines and intersection points

26 Another tool is introduced from stencil library and clipped to image (referenced to first tools).

27 That was trivial (and fun) to program Now feed images to interface for automatic, robust smart metrology

28 Process image 1

29 image 2

30 image 3

31 image 4

32 image 5

33 and finally image 6.

34 All measurements are recorded in database. Measurement sequence is saved, including images. Perfect for off-line browsing, debugging, recipe refinement,

35 Conceptual proposal for programming arbitrary, fully automatic milling sequences based on smart shapes Lets extend the smart-shape concept to include milling, image acquisition, stage moves. As usual, all these are intimately tied to measurement. Here we are simply replacing the human eye The following is just a thought-experiment sequence of some of the kinds of things we would like to see the software do

36 Wafer Map with ordered stage sites tagged with saved FIB images or cartoons 1 2

37 1 Set FIB current. Align and autofocus on crosses: creates deterministic focus map for flat sample

38 Goto site 1 Set FIB focus. Grab image

39 Find, center and zoom tool

40 Find

41 Center, zoom, and grab image

42 Find upper edge of trench tool

43 Find

44 Attached box polishes only edge of trench. Real-time sidewall measurement: mill while d>50nm. d d=205nm

45 d=146nm

46 d=98nm

47 Stop polishing sidewall d=49.5nm (sub-pixel accuracy)

48 Increase beam current; image; find edges; offset by fixed amounts mill boxes (green) and circle find tool (red) Mill until circles appears with diameter φ>100nm ϕ2 ϕ1

49 Mill first box

50

51

52 Circle detected φ1=21nm

53 φ1=76nm

54 Stop milling first box: diameter tolerance achieved φ1=101nm

55 Mill second box

56

57

58 φ2=23nm Circle detected

59 φ2=77nm

60 φ2=101.5nm Stop milling second box

61 Center of each circle determined

62 Pt deposit box clips on

63 Machine Deposits Pt strap

64 Change beam current, grab image, invoke reference-image-tagged pattern recognition tool

65 Find Pattern. Center of box determines reference point for next stage move

66 Change magnification. Move stage in x and y (based on box angle)

67 Arrive at Site 2

68 Find, center and zoom tool

69 Find

70 Center and Zoom etc

71 All steps would be recorded All acquired images and metrology data stored in datase for further analysis Could browse through performed sequence offline, step by step (storyboard)

72 Starting with a monolithic metallo-dielectric multi-layer

73 We need to make hundred s of these, rapidly, accurately and repeatably

74 Robust, flexible, user-friendly machine-vision-based automation is the answer

75 Conclusion FIB fabrication of plasmonic metamaterial with negative index of refraction in the visible Applications issues: the need for automation