sign in sign up
<<
Home , Photolithography

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-28, NO. 11, NOVEMBER 1981 1375 Projection Photolithography-Liftoff Techniques for Production of 0.2-pm Metal Patterns

MARK D. FEUER AND DANIEL E. PROBER, MEMBER, IEEE

GLASS FILTER Abstract-A technique whichallows the useof projection photo- \ yPHOTORESIST with the liftoff process, for fabrication of sub- micrometer metal patterns, is described. Through-the-substrate (back- \II a projection) exposure of the photoresist produces the undercut profiles necessary forliftoff processing. Metal lines andsuperconducting microbridges of 0.2-pm width have been fabricated with this technique. I Experimental details and process limits are discussed. II OBJECTIVE

IMMERSION OIL ECENT DEVELOPMENTS inmicrolithography have SUBSTRATE - (d) R made possible the production ofa variety of devices with PHOTORESIST %- I-l submicrometer (submicron) dimensions [ 11 , [2] , offering the (a) advantages of higherspeed and packing density.For many Fig. 1. Schematicdiagram of back-projectionand metal-liftoff pro- Josephson-effect devices in particular, submicron dimensions cedure. (a) Exposure system, employing a Zeiss optical microscope. Image of the mask is projected through the substrate, which is shown are essential for achieving optimalperformance over awide in sideview. (b) Schematic contours of constant exposure intensity. range of operatingconditions [3] . Forsubmicron pattern (c) After photoresist development and metallization. (d) After liftoff. transfer, liftoff processing [ 11 generally has better resolution thanwet-chemical etching. Liftoff processing may also be after results achieved withthe back-projection technique preferred over the alternatives of chemical and etching are presented. for films which are difficult to etch, where the etching process Thetechnique we have developed, through-the-substrate can cause chemical or physical damage to the patterned film, exposure, involves projecting the image of a mask through or where resist masking for the etching process will lead to the back of a transparent substrate onto the photoresist,which contamination problems for subsequent use of the patterned is onthe front side. Optical absorption in thephotoresist film. leads to an undercut exposure profile, whchi.s preserved after This paperreports an opticalprojection technique which development. This exposure method is shown in Fig. 1. achieves undercut photoresist edge profiles necessary for the The exposure system is like that of Palmer and Decker [5] . liftoff process ’ [l] . With the projection technique we have A Zeiss Photomicroscope with a type 11-C epi-illuminator is developed, excellent liftoff results and yield are obtained even used. The microscope is adjusted for Koehler illumination, for dimensions

0018-9383/81/1100-1375$00.75 0 1981 IEEE 1376 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-28, NO. 11, NOVEMBER 1981

Photoresist processing is similar tothe manufacturer’s recommendations.* The basic steps are photoresistexposure and development, metal film evaporation, and metal film lift- off (see Fig. 1). A pre-exposure bakein air at 90°C for 15 min is used; no post-exposurebake is used. Development is for 30 s in stirred AZ developer* diluted 1 : 1 in deionized water. Iflarge-area patterning of the photoresist layer is required, contact exposure anddevelopment is carried outprior to projection exposure. Alignment of the microscope projection exposure to 90 percent. To achieve successful liftoff,particularly for the smaller structures, vertical orundercut photoresist edge profiles are essential. This is achieved with the back-projection technique, at least for the resist thicknesses >0.25 pm used. Absorption by 0.3 pm of unexposed photoresist is -25 percent [6],and this causes the fullyexposed region to be widest where the first enters the photoresist. Schematic contours of con- stant exposure intensity are shown in Fig. l(b). (Interference effects are here neglected.) If thedevelopment process had infinite contrast, these intensity contours would be produced in the developed photoresist pattern. In practice, even though thedevelopment process has only moderatecontrast [6], vertical or undercut edge profiles are obtained. Electron micrographs of photoresist edge profiles for three photoresist samples, prior to metallization, are shown in Fig. 2. The insetshows the pattern projected; the gap between the resulting photoresist “fingers” is 0.3 to 0.4 pm. Fig. 2(a)-(c) are side views of each photoresist pattern, viewed at 85” from u thesubstrate normal. In Fig. 2(a), thephotoresist film is Fig. 2. Photoresist edge profiles. The scanning-electronmicrographs 0.25 p thick,and showsvertical walls. Such athickness were taken at glancing incidence (i.e., nearly parallel to the substrate). (a) Through-the-substrate exposure of thin, 0.25-pmphotoresist and profile can be used for liftoff of thin metal films, <750 yields roughly vertical edge profiles. (b) Through-the-substrate expo- thick. In Fig. 2(b), the photoresist film is thicker, 0.5 pm sure of 0.5-pm photoresist yields clearly undercut edge profiles. (c) thick. This profile shows dramaticundercutting, and some Conventional top-surface exposureof thin, 0.25-pmphotoresist yields sloping edge profiles, which would not be suitable for liftoff process- (unexplained) raggedness. Such a profile is typicalfor this ing. 0.1-pm size scale is shown. Inset: Mask pattern. The photoresist thickne~s.~Finally, Fig. 2(c) showsa conventionalfront structures remainwhere mask is dark.The size of thegap at the (top)exposure of a thinphotoresist film. As expected,no center of the mask is 9 pm; this would be projected to 0.2 pm in the absence of effects. undercutting results, and liftoff is unreliable with such front exposures even for thin metalfilms. CS5-58, forexposure. The blue exposurefilter transmitsat Diffractioneffects set the lowerlimit onpattern dimen- around 400 nm.Exposure times are typically sions. For a lens with a NA ,the minimum 40 s for a thin (0.3-pm) layer of thepositive photoresist used, resolvable feature hasdimensions given approximately by Shipley AZ1350B.2Standard microscope cover glasses, type [71 PI No. 13, which are 0.17mm thick, are used as substrates. h (Most objectives designed for use with cover glasses are de- dmin= - 2NA signed forthis thickness. With oil-immersionlenses, the thickness is less critical.) The correct stage height for pattern 31n the discussion of resist edge profiles we have not included inter- reproduction is determined by scanning electron microscope ference effects, which are often of importance in front exposures on inspection of a test series of patterns exposed at various stage reflecting substrates. As seen in Fig. 2, interference effects appear to heights. For the 1OOX lens, the correct height is within 0.5 be small. This is due to the broad-band radiationused for exposure, the p smaller index mismatch between photoresist and air as compared with of the focus point for the mask image, as viewed through the case of reflecting substrates, and the large numerical aperture of the the microscope. lens; see Pierre Parrens and Paul Tigreat, in Microcircuit Engineering, H. Ahmed and W. C. Nixon, Eds. Cambridge: Cambridge Univ. Press, 2Shipley Co., Newton, MA 02162. 1980, pp. 181-198. FEUER AND PROBER: PROJECTION PHOTOLITHOGRAPHY-LIFTOFF TECHNIQUES with h theoptical . Thus lenses with large nu- merical aperture are required for best resolution; the highest values of NA are obtained with oil-immersion lenses. Unfor- tunately,the depth of field, given approximatelyby Ax = h/2(NA)* [7], [8] is smallest with such lenses, so that correct focusing is critical. While the ultimate resolution achieved in thephotoresist pattern depends on the development condi- tionsand the exposure-development characteristics of the particularphotoresist, (1) stillsets an approximate lower limit on device sizes which can be produced. Our experience shows that for isolated features, 0.2-pm linewidths are achiev- able with care; much smaller linewidths are rarely obtained. For dense patterns, a resolution approaching 0.2~may be achievable, but only withvery thin resist layers. Theextremely high resolutionof the combined back- exposure-liftofftechnique is demonstratedby the electron micrographs in Fig. 3. Fig. 3(a) shows a 0.25-pm-wide chrome line on a glass substrate. The film is 550 A thick. The edge roughness of G200 A is typical.Superconducting micro- bridge devices 0.2and 1 pm wide are shownin Fig. 3(b) and (c). (A microbridge is a narrowconstriction in a ;it is the thin-filmanalog of a point-contactstructure [3] .) Forsuch two-dimensional patterns, the mask is not precisely replicated because of diffractioneffects. Sharp corners of amask pattern are rounded significantly inthe photoresistpattern when photoresist pattern dimensions are G0.5 pm. Grating patterns with 0.2-pm electrode width also have been produced [7] byoptical-projection exposure and chemical etching. However, the edges obtained appear to be rougher than those in Fig. 3. Theprocedures described here are simple and reliable and allow excellent resolution, limited only by diffraction in the high-qualitymicroscope . Certain features are critical, however, and we wish tonote them explicitly. First, as should be obvious, careful alignment of the microscope and objective is essential for achieving best resolution. Even with careful alignment, best resolution is obtained only in approxi- H lym mately the central third of the field of view. Planachromat Fig. 3. Metal patterns produced withthrough-the-substrate exposure. lenses would be desirable for patterns which fill the field of Light region is the metal film; dark region is the substrate. Viewed with SEM at normal incidence. (a) Chrome line, 0.25 pm wide, on view. Thesecond point is that cover glass substrates are a cover-glass substrate; 550-8 film thickness; (b) Pb-In alloy micro- optimal. While patterns have been projected successfully bridge, 0.2 pm wide, on a cover-glass substrate; 250-8 film thickness. through 0.12-mm thick sapphire substrates, the higher index In (a) and (b), thephotoresist was exposed with lOOX, oil-immersion lens. (c) Pb-alloy microbridge, 1 pm wide, on a sapphire substrate; of refraction of sapphire (n = 1.8) leads to spherical aberra- -500-8 film thickness. Polishing marks in this sapphire substrate are tion. The resulting images are visibly fuzzier, and the ultimate visible. The photoresist was exposed with a 40X lens. In(b), the resolution is somewhat poorer, about 0.3 pm. (Fig. 3(c) does hazy lumps within -700 8 of the main film (the bright region) are due to migration or scattering under the overhanging photoresist (Fig. show a quite satisfactory 1-pm microbridge pattern on a sap- l(c)) of a thin (50-8) Pb-oxide undercoat layer, deposited and oxi- phire substrate.) A third important feature is that the expo- dized at room temperature to promote adhesion of the 250-8 thick sure time is critical, and must be accurate to approximately Pb-alloy film. The Pb-alloy film was then deposited at 77 K to avoid migration and reduce grain size. The masks used for producing the C5 percent to produce the smallest devices. Due to variations devices of Fig. 3(b) and (c) had the shape of the metal pattern of in processing, the optimum exposure time can vary up to-10 Fig. 3(c). percentfrom substrate to substrate. We thereforeexpose a series of devices on each substrate, over a range of exposure forsuperconducting Pb-alloy films, special proceduresmust times. While the exposure time affects device size, clean lift- be employed [4]. off of the undesired film is obtained for the whole range of The back-projection technique described in this paper serves exposuretimes. A fihal feature of note is thatthe use of a specific setof processing requirements:the use ofthin, ultrasonic agitation for liftoff requires excellentadhesion of transparentsubstrates, with single-layer metallization. These the metal film. For chrome films this is not a problem, but requirements apply for certain Josephsondevices and electron- 1378 TRANSACTIONSIEEE ON ELECTRONDEVICES, VOL. ED-28, NO. 11, NOVEMBER 1981 transport experiments. On thick or opaque substrates, under- [4] M. D. Feuerand D. E. Prober,“Properties of high-resistance cutphotoresist profiles maybe produced with other, com- superconducting microbridges based on lead alloy films,” IEEE plementarytechniques, such as multilayer resists [9] or Trans. Magn., vol. MAG-15, pp.578-581, 1979; M. D. Feuer, chemical treatment of the top layer of the photoresist to slow D. E. Prober, and J. W. Cogdell, “Fabrication of submicron Josephson microbridges using opticalprojection lithography development [ 101 . Also, for a limited number of simple struc- and liftoff technjques,” in AIP Con$ PYOC.,vol. 44, pp.317- tures, even smaller linewidths, G300 A, are possible if pattern 321, 1978; and M. D. Feuer, Ph.D. dissertation, Yale University, edges 1980; available from University ,Microfilms, AnnArbor, MI. can be used to define feature dimensions [ll] . The The application of the liftoff process in the production of Jo- back-projection technique is,however, ideally suited for the sephson tunneling junctions is described in [9] and [lo]. production of flexible submicron masks for conformal replica- [5] D. W. Palmer and S. K. Decker, “Microscopic circuitfabrica- tion on refractory superconductingfilms,” Rev. Sci. Instrum., tion [I]. Thusthe metal patterns produced with the back- vol. 44,pp. 1621-1624, 1973. Agood introductionto micro- projectiontechnique may be utilized as masksin amuch scope use is given by F. K. Mollring,Microscopy From The Very wider variety of applications. Beginning. New York: Carl Zeiss, 1978. [6] F. H. Dill, W. P. Hornberger, P. S. Hauge, and J. M. Shaw, “Char- acterization of positive photoresist,” IEEE Trans. Electron ACKNOWLEDGMENT Devices, vol. ED-22, pp. 445-452, 1975. Theauthors wish to thank Dr. A. Pooley of the Peabody [7] H.A.M. van den Berg and J.J.M. Ruigrok, “Theory and practice of image formation by the photoprojection methodof submicron Museum at Yale for assistance withthe scanning electron patterns,”Appl. Phys., vol. 16, pp. 279-287, 1978. micrographs,and P. Santhanamand C. Teng for excellent [8] J. H. Bruning, “Optical imaging for ,” J. Vac. technical assistance. The use of the clean room facilities Sci. Technol., vol. 17,pp. 1147-1155, 1980; D.A. Doane, at Yale is also acknowledged. “Opticallithography in the 1-bm limit,” SolidStateTechnol., vol. 23, no. 8, pp.103-114, 1980; W. G. Oldhamand A. R. Neureuther,“Projection lithography with high numericalaper- REFERENCES ture optics,” Solid-state Technol., vol. 24, no. 5, ~~pp. 106-111, H. I. Smith,“Fabrication techniques for surface-acoustic-wave 1981. andthin-film optical devices,” Proc. IEEE, vol. 62, pp. 1361- L. N. Dunkelberger,“Stencil techniquefor the preparation of 1387,1974. thin-film Josephson devices,” J. Vac. Sci. Technol., vol. 15, pp. A. N. Broers and T.H.P.Chang, in Microcircuit Engineering, 88-90, 1978; L. Jackel, E.Hu, R.Howard, L. Fetter, and D. H. Ahmed and W. C. Nixon, Eds. Cambridge: Cambridge Univ. Tennant, “Submicron tunneljunctions,” IEEE Trans. Magn., Press, 1980, ch. 1. vol. MAG-17, pp. 295-298, 1981; R. E. Howard, E. L. Hu, and M. Tinkham, “Junctions-Types, properties, and limitations,” in L. D. Jackel, “Multilevel resists for lithography under 100 nm,” Proc. AIP Confi, vol. 44, pp. 269-279,1978. this issue, pp. 1378-1381.

L.M. Hatzakis. B. J. Cavanello. and J. M. Shaw. “Single-steo ootical lift-off process,” IBM J. Res.Develop., vol. 24,lpp. 452-460, 41n additionto the techniquesdescribed in [9], a number of other1980. multilayer resist techniquesfor near UV-opticalexposure have been [ll] D. E. Prober, M. D. Feuer,and N. Giordano,“Fabrication of developed; see, forexample, Bruning [8]. However, these othertech- 300-8 metal lines with substrate-step techniques,”Appl. Phys. niquesrequire either a vapor-deposited resist or to etch Lett., vol. 37,pp. 94-96, 1980; M. D. Feuer and D. E. Prober, through an intermediate layer, and are thus more complex than those “Step-edge fabrication of ultrasmallJosephson microbridges,” of 191. Appl. Phys. Lett., vol. 36, pp. 226-228,1980.

Multilevel Resist for Lithography Below 100 nm

RICHARD E. HOWARD, MEMBER, IEEE, EVELYN L. HU, AND LAWRENCE D. JACKEL

Abstract-Features as small as 25 nm have been made with electron- I. INTRODUCTION beam lithography using multilevel resists on thick silicon substrates. Liftoff patterning of metal lines and reactive ion etching of silicon have URTHER INCREASE in complexity of integrated circuits demonstratedthe possibility of making device structures withlateral Fdepends in large parton continuing improvement in the dimensions belowdimensions 100 nm. lithographic patterningTechniques process. having higher resolutionthan optical exposure, such as e-beam or X-ray Manuscript received April 2, 1981; revised June 22, 1981. lithography, do not necessarily yield finer patterns. Each com- The authors are with Bell Laboratories, Holmdel, NJ 07733. ponent of the patterningprocess is important in achieving high

0018-9383/81/1100-1378$00.75 0 1981 IEEE