Microelectronic Engineering 23 (1994) 465-468 Elsevier 465 Application of electron beam lithography to pattern sub-micron features in a beam forming grating W. Parkes, S. Thorns and C.D.W. Wilkinson Department of Electronics and Electrical Engineering, Glasgow University, Glasgow G12 8LT, United Kingdom Techniques am described for the fabrication of both reflective and transparent beam dividers. The grating design uses a mean pitch of 960nm and groove width of 480nm with aspect ratios of 0.4 and 2.4. Electron beam lithography is exploited in order to obtain 25nm accuracy in positioning of the features. Limitations on the required groove depth accuracy of better than 5% are discussed and an interferometric method is described. 1. INTRODUCTION There is a considerable interest in optical data processing to increase the effective speed of computing machines by exploiting the massive potential for parallel switching [l]. A crucial element in optical circuits is the beam divider which must produce with minimal losses from a collimated input beam an array of output beamlets which are uniform in intensity to within about 10%. A number of methods meet these criteria [2,3], and it is now possible to make arrays at efficiencies of 70%~80%. To extend the performance further, an alternative approach is a diffraction grating with a variable period near to the wavelength of the incident light. Theory predicts that by correct choice of parameters the Bragg efficiency should closely approach 100% [4]. These gratings can be transparent or reflective depending on the type of construction. The small feature size of 470 nm and positional accuracy of 25 nm cannot be achieved by visible and near-UV lithography. The practical alternatives are deep UV photolithography or electron beam lithography. We use a method of fabrication based on the latter to define the sub-micron features. Gratings are made for reflection by two methods: simple metal liftoff and etching followed by a metal overcoat. Transmission mode elements are etched in quartz. While the surface feature dimensions are defined by the patterning, accuracy in the vertical direction is controlled by pre-calibration or by monitoring the deposition or etching process. 2. ELECTRON BEAM CONSIDERATIONS 2.1. Electron beam resist The pattern is written onto the sample by means of the positive electron beam resist polymethylmethacrylate (PMMA) [5] which is available in a variety of molecular weights and concentrations in solution. For lifting off a metal overlayer the initial resist profile must be undercut and the resist thickness should be several times that of the metal film to be deposited. A double layer coating of two different PMMA’s is used, the slower resist on top having a mean molecular weight of 350,000 (‘Elvacite’) and a faster one of m.w. 85,ooO (‘BDH’) underneath. To obtain a 200nm metal overcoat, a PMMA combination of 4% Elvacite dissolved in xylene and 12% BDH in chlorobenzene is selected, which gives after baking a thickness of about 900nm. In the case of a 30nm metal mask, the PMMA concentrations are 2.5% Elvacite and 8% BDH, giving about 300nm total. When masking a metal film a single PMMA layer suffices. The lower resist layer is baked for 30 mins. and the top layer overnight. 2.2 Design software The grating pattern is designed in a standard software format as a 8OOpmx 8OOpm unit. This can be multipied into a larger array by stitching to form a grating of sufficient size for laser testing. The file is fractionated in a Computer Aided Transcription System (CATS) program for reading by the electron beam writer. 2.3 Electron beam pattern generation A Leica Cambridge EBPGS-HR electron beam pattern generator writes the pattern into the PMMA using a gaussian beam shape. The resolution is fixed by the minimum feature size and its required positional accuracy. For a minimum ridge/groove width of 450nm and an allowed position error of 50nm, we select a beam writer resolution of 25nm. For liftoff the spot size must be kept small and 40nm is used as a compromise between positional 0167-9317/94/.$07.00 0 1994 - Elsevier Science B.V. All rights reserved.