OPTIMIZATION OF DIELECTRICS SURFACE PREPARATION FOR VACUUM COATING Dr. Boris Statnikov Introduction Modern MICRO and NANO technologies in ultra- and high-frequency electronics are widely focused on application of highly effective organic dielectrics (Polyimide, BCB, Teflon, etc.) instead ceramics that essentially reduces the sizes, weight and cost of element base, and also opens essentially new technical opportunities. One of the most complicated technological problems of this new direction is a vacuum coating of organic surfaces by metals, as a rule in Sputtering machine, for Thin Film Technology application. The main problematic aspect is providing of the conductor sufficient adhesion to dielectric surfaces after deposition and Pattern Electroplating of conductive layers. The specified problem could be solved by preliminary (pre-vacuum) Dielectric Surface treatment (Grinding - 1 st step, Chemical Etch - 2 nd step, Low Temperature Plasma Etch - 3 rd step) and following Vacuum Processing: Outgassing, including heating, baking and cooling (4 th step), then Ion or RF etching (5 th step) before metal coating together with application of special methods of cooling during vacuum metal deposition in the Sputter machine. Each of the steps has effect theoretically but it is not known what a rate put in adhesion an each single action. On other hand all steps together take very long time and decrease throughput of production that is critically for High Volume Manufacture (HVM). The influence of PI surface preparation for Cu/Ti adhesion The experiment goal was to research a contribution of each above named process step to adhesion level, and also to obtain minimal combination of the steps, which will assure the Polyimide-Copper (PI-Cu) adhesion requirement Peel Force of 10 N/cm. After pre-vacuum treatment all tested panels were dried during 6 hrs and cooled down to 25 o C, and then were passed to Sputter machine directly from Vacuum Oven. The tested steps variations are given in Table 1. The adhesion was valuated with standard Peel Stress method, and also regarding the Patterning success in 5 points of wafers with conductors and spaces of 30-50 micron. Table 1. Conditions and results of the experiments Combination Grinding abrasive Type of surface preparation Quality criterions of KMnO 4 O 2 Ion Patterning Adhesion, N/cm factors #400 #600 #800 etch Plasma Outgas etch success, % Average Max Min 1 v v v v v v 80 11.28 13.12 9.21 2 v v v v v 90 7.42 9.97 5.17 3 v v v v v 80 11.11 12.62 9.97 4 v v v v 80 9.96 11.61 7.96 5 v v v 50 2.29 3.66 1.64 6 v v v v N/A 8.82 9.34 8.33 7 v v v v N/A 8.68 9.34 8.08 8 v v v v 60 10.63 12.23 8.58 2-99
Pre-vacuum surface treatment: The adhesion level of metal coating to dielectric is function of surface absolute and structure roughness formed during pre-vacuum treatment. The first step of surface roughness formation is a grinding after the dielectric lamination (pressing and polymerization). The abrasive defines the surface roughness, which shall be increased with abrasive grin sizes (the some is decreasing of mesh #, see Table 1). Our measurements showed that the grinding finish with abrasive # 400 versus # 800 decreased the average surface roughness R a from 1400 o A up to 5000 o A and adhesion level on 20% (see Table 1, combinations #6 vs. #8). In the same time adhesion for 8 th combination was less than for 1 st combination, where samples were grinded with finishing abrasive #800 but then treated additionally by Oxygen plasma. It is known that grinding creates a spire type of roughness (see Fig. 1, a) with simple surface. The chemical etching by KMnO 4 or Plasma develops the surface and opens a lateral caverns in dielectric resin. A filling of the caverns by sputtered-deposited metal (see Fig. 1, b) prevents a peeling of the coating and increases the adhesion level. a b c - metal coating - organic dielectric Figure 1. Cross-section (schema) of metal coating on dielectric: a sputtered-deposited metal coating on only grinded surface, b sputtered-deposited metal coating on grinded and etched surface, c electroplated coating on metallic seed-layer. All pre-vacuum steps of PI surface preparation make contributions for PI-Cu adhesion but least effective is the Plasma etching (see Table 1, combinations #3 vs. #1). The etching by permanganate increases the adhesion about 35% (see 2-100
combinations #2 vs. #1 & 3), therefore, its application is preferable if there is not some technological contra-indication. It has to be noted that rough grinding isn t preferable because creates difficulties for following Photolithography and increase number of shorts in Electronic Pattern. Herewith (see Table 1, combinations #3 vs. #8), the Adhesion is less because reduction of the PI-Cu contact surface with little number of big teeth regarding surface with many small teeth, especially after chemical etching of dielectrics. Vacuum surface treatment for coating: More expensive parts of the process are vacuum pretreatment steps and Cu/Ti coating. Herewith, the basic stage of Metal Deposition takes a less time than Outgassing and Ion etching, and therefore is a need to short or exclude those periods. With direct transfer of the panels from Vacuum Oven to Sputter machine there is possibility to exclude Outgassing without substantial damage for PI-Cu adhesion (see Table 1, combination #4 regarding #1 & 3). It is very important because Outgassing (including heating, baking and cooling) exceeds duration of the metal coating itself. Excluding of Outgassing reinforces the way for HVM on base of in-line Sputtering machines. Ion Etching before metal coating actualizes two functions: 1 - surface cleaning by sputtering of passive thin layer, 2 - chemical activation of the surface by destruction of the organic radicals for creation of a new metal-organic bonding. The data of the Table 1 show that it is not possibly to exclude Ion Etching from technology because the PI-Cu adhesion will be very low and even near to zero level (see combination #5). This situation leads also to bad condition of completed Electronic Pattern because many shorts and opens of interconnections. Dynamics of Cu/Ti on PI adhesion stabilization For majority applications the thin Cu/Ti coating on Polyimide or Teflon is only seed-layer for following Copper Pattern Electroplating. Evidently that filling of hollows between teeth and their caverns (see Fig. 1, c) will be increase Adhesion of metal to dielectric, but was found that many technological properties, obtained by adhesion level, had bad indications if substrates were passed to next operation immediately after Electroplating (EP). Therefore there was a need to research dynamics of the adhesion stabilization. Test method: In order to research adhesion stabilization dynamics the Polyimide- Glass substrates were coated by Ti/Cu seed layer together in Sputter, and then part of them were passed through EP after 2 hours, and others after 20 hours. At once after plating up to 35 microns of Copper, all substrates were prepared for the Standard Adhesion Test and checked during 96 hours (see Table 2). 2-101
Table 2. Dynamics of Cu on PI adhesion stabilization Time after EP, Average adhesion, Maximal adhesion, hr N / cm. N / cm. RPP 2 hr after Sputter 3.25 2.67 3.53 18.75 7.31 7.7 27.25 7.59 8.58 46.75 7.79 8.58 67.25 8.05 9.08 93.25 8.28 9.66 EP 20 hr after Sputter 2.5 2.91 3.15 8.5 3.94 4.8 28 7.51 8.58 48.5 7.68 9.21 74.5 8.01 9.09 Results: Adhesion stabilization is long time process and has two specific periods (see Fig. 2). During of 20-30 hours the Adhesion has high intensive rise and reach about 90% of Average and Maximal complete levels. Following there is period of Adhesion asymptotic low intensive rise, which isn t completed even during 93 hours after Electroplating. Fifure 2. Dynamics of Cu on PI adhesion stabilization Adhesion force, N per cm. 12 10 8 6 4 2 0 0 20 40 60 80 100 Time, hours Average Maximal Herewith the time interval between Sputter and EP don t influence on adhesion parameters that is seen from Table 2 and Fig. 2. Such process of Adhesion change in the time can have following explanation. Filling of hollows and caverns on dielectric surface by Copper during Electroplating begins in acute-angled points, evolves as dendrite rise and completes by their collision. Because the dendrite rise outstrips diffusion bonding of Copper a porous 2-102
and stressed structure is created. The relaxation time is needed in order to decrease the stress and to fill porous on means of metal molecular diffusion. This conclusion is confirmed by the got experimental data that is seen from interpolation and correlation analyzes (see Fig. 3). Figure 3. Interpolation of adhesion dynamics Adhesion, N per cm. 12 10 8 6 4 2 y = 1.8889Ln(x) + 1.5122 R 2 = 0.946 y = 1.6846Ln(x) + 1.2255 R 2 = 0.9131 0 0 20 40 60 80 100 Time, hours Log. (Maximal) Log. (Average) Conclusion Reaching of the needed level of Copper adhesion to dielectric substrate in manufacture of High Density Interconnection for ultra- and high-frequency electronics requires application of the Chemical Etching of substrate before vacuum treatment for surface development and Ion Etching in vacuum processing for molecular activation of coated dielectric. 2-103