High-density Wafer Bumping with Solder Paste
HIGHEST BUMP HEIGHT-TO-PITCH RATIO
BY RICHARD LATHROP
Wafer bumping using solder paste was 6.5% of a $3.25 million (200-mm wafer equivalents) bumped wafer market in 2001, and is expected to grow to 16.5% of a $23 million market by 20061. The majority of this projected market is in memory, ASIC, and display applications. Current pitches are limited to 150 µm (5.9 mils) and greater, although some recent sub-100-µm experimental work has been published2. This article presents the highest bump height-to-pitch ratio possible with today’s solder paste and stencil technologies. Bump heights are quantified for numerous aperture area/pitch combinations via non-contact measurement methods for pitches in the 150- to 250-µm range, using full array designs.
Figure 1. Test coupon.
Apertures for full array applications are either round or square, and in many cases rectangular, for perimeter array designs. Adequate bump heights are achieved by overprinting the under bump metallization (UBM). Stencil fab technology is predominantly electroformed nickel in the 1- to 3-mil-thickness range. Electroformed nickel stencils are “grown” or electroplated onto a mandrel with apertures defined with photo-imaged plating resist. After extraction from the mandrel, this pure nickel stencil is mounted on a screen frame via a mesh that places the stencil foil in tension. This stencil technology offers the best aperture dimensional accuracy. The low surface energy and smooth aperture wall surface finish offered by this technology are required for good paste release from the apertures during printing. For this reason, electroformed (E-FAB) stencils were exclusively used in this study. Final bump formation is achieved by reflow in a convection oven in low O2 ppm nitrogen. Post-reflow flux residues are removed by cleaning with chemistries matched to the flux formulation. Most paste formulations for wafer bumping are water soluble to facilitate flux removal with simple aqueous cleaners. Solder powders are typically Type 5 (15 to 25 µm) and Type 6 (5 to 15 µm), available in variety of alloys to suit the application. Wafer bumping with printed solder paste can use standard SMT printing tools, with a few additional process elements not typically common in standard SMT printing. Literature on this subject stresses the need for polymer squeegees instead of the metal squeegee typical in today’s SMT process.3 Research on this subject suggests the requirement for a print gap or “off contact” printing for very high aperture density wafers.4 To establish the maximum bump height to bump pitch ratio (H/P), a test coupon was developed (Figure 1).
Coupon Development and Design
The coupon used in this study to determine the maximum H/P ratio was the second generation of wafer bumping coupon designs. The first design explored the overprinting of UBMs for 150-, 250- and 350-µm pitches5. Due to etch process limitations, the 150-µm-pitch grids were bridged. To correct this with the new coupon design, the thickness of the copper was reduced from 0.50 oz. to 0.25 oz. copper. The thickness measurement in Figure 2 illustrates the UBM thickness reduction for the new coupon. UBM designed diameter was 100 µm for the 250-, 225- and 200-µm-pitch arrays. For the 175- and 150-µm-pitch arrays, the pad diameter was reduced to 88 µm and 75 µm, respectively.
Figure 2. 3.25-oz. Cu on new coupon.
As with the first-generation coupon, lessons were learned with the first-generation stencil. Aperture wall failure was caused by insufficient webbing width (45 µm). This failure occurred during the plating mandrel extraction phase of the manufacturing process. This first study tested three stencil thicknesses (38, 51, and 64 µm), and both round and square apertures.6 Although this failure was noticed to some degree in all three thicknesses, it was the most pronounced with the thickest stencil and isolated to the largest square apertures. From this experience, a guideline was generated that the stencil webbing (distance between apertures) must be ≥1.2 × stencil thickness for square apertures and 0.8 × for round apertures.
Since the goal of this study was to push various solder paste formulations to yield the highest H/P ratio, stencil apertures had to be sized to the maximum. This meant that to get the largest apertures, the stencil webbing needed to be pushed to the minimum possible to get a usable stencil. A 50-µm-thick stencil was designed that was intended to push the solder paste to bridging (which occurs when two printed deposits are too close to each other), and to print release failures when the print area ratio (PAR) is too low. PAR is the area under the aperture divided by the aperture wall area. The aperture design objectives were:
- Test both square and round apertures.
- Largest aperture will have minimum stencil web width.
- Smallest aperture will be half the UBM pitch.
The stencil design essentially yielded 50 aperture design experiments per coupon with 20,000 bumps on each coupon. The top row is 250-µm pitch, the second is 225-µm pitch, the third is 200-µm pitch, the fourth is 175-µm pitch, and the bottom row is 150-µm pitch. Each grid is a 20 × 20 array of bumps. Table 1 shows the aperture design details, as well as PAR calculations. PAR figures are color-coded: yellow (0.6 to 0.66); orange (0.5 to 0.6); red (0.4 to 0.5); magenta (0.3 to 0.4); and green is any PAR over 0.66. All dimensions in the table are in microns.
Table 1. Stencil aperture design matrix.
Of equal challenge to forming bumps that are smaller than the head of a pin, is measuring them accurately. One of the difficulties encountered in the first study was measuring the bump height. With this laser triangulation technology, a point of light emitted from a laser diode passes through an objective, reflects off the surface to be measured, and a portion of this reflection excites a detector array yielding an index of height. If the portion of the reflection is occluded, no height data is returned from the sensor. The surface must also be sufficiently reflective. For PCB laminate, this can be problematic in that the material has considerable spectral absorption properties. On the other hand, if the material is too reflective, as in a polished wafer, then all light is reflected back to the emitter and again no height data is returned from the sensor. For these reasons, a new measurement technology was required for this new study. For this new study, a confocal point sensor was used to capture bump heights. The advantage of the confocal principle is its ability to measure surface structures independent of the reflectivity of surface materials and is very important for wafer bump structures because it does not produce artifacts at sharp edges like holes or interrupted areas that are common with laser triangulation. A laser diode illuminates a small pinhole that is focused with a high precision movable objective lens onto the surface of the specimen. The detector gets a signal only when the illuminated spot of the surface is exactly in focus. The signal processing software assesses the signal intensity and the position of the lens that is equivalent to the relative Z coordinate of the surface point measured. If the illuminated surface point is out of focus (Figure 3), then there is no signal and no surface point will be detected. The design of the sensor requires high-precision optics, optical alignment, and accurate fast movement of the objective lens without additional optical aberration. Table 2 shows a summary of the sensor performance.
Figure 3. Confocal point sensor.
Table 2. Sensor comparisons.
With the coupon and stencil aperture designs established and the measurement method selected, the test matrix must be identified. Focusing on determining the maximum H/P ratio for the F510 water-soluble wafer bumping flux formulation, two main formulation variables surfaced as potentially having an effect. Both powder size (Types 5 and 6) and alloy (Sn/Pb and Sn/Ag/Cu) were studied. Two stencils were used (45 and 53 µm), with the majority of the data measured on the 53-µm stencil. There were a total of 120,000 bumps measured, comprising 300 individual experiments (each a grid of 400 bumps).
Four coupons were printed for each paste-stencil combination. Coupons were contact printed at 25 mm/s, with a medium durometer polymer squeegee. All coupons were reflowed in nitrogen after a 45-min. purge. The final O2 level was 19 ppm. The reason for this low O2 requirement stems from the extremely high surface area of Types 5 and 6 solder powders, as shown in Figure 4. Standard Sn63 and SAC profiles were replicated on a new coupon. Profiles were a ramp-to-spike format. All coupons were cleaned with an in-line cleaner in DI water at 140°F/80 psi. Cross section specimen clips were attached in a way to prevent the top mesh belt in the cleaner from marring the bumps during the cleaning cycle. One coupon of each of the paste-stencil combinations was sent to a company* for measurements using their CF4 confocal measurement system. Bump height data and statistics, plus coplanarity data for each grid measured, were determined.
Figure 4. Surface area comparisons.
Determining H/P Ratio
Several factors must be considered when selecting the maximum H/P ratio. The calculation involved taking the average bump height and dividing by the pitch. Since the stencil design pushed the aperture size to the maximum, it also pushed the paste printing and slump properties to their limits. This resulted in shorting on some of the largest aperture designs. Since it was assumed that the maximum H/P ratio is defect-free and manufacturable, both coplanarity of each grid and data distribution were considered. Coplanarity was the largest bump minus the smallest bump. Data distribution was viewed in a box-whisker format. Figure 5 shows a bump height normal distribution and the box-whisker plot. When all three data attributes were plotted (Figure 6), the selection of the maximum usable H/P ratio became obvious. The red line is the H/P ratio, the blue line is coplanarity, and the green boxes are box-whisker plots. The highest H/P ratio that demonstrated a tight box-whisker plot, minimal outliers, and a low coplanarity relative to the other data in the pitch series was selected and highlighted (cyan circle) for each paste-stencil-pitch combination.
Figure 5. Distribution: a box-whisker comparison.
Figure 6. Data analysis plot.
Table 3 lists the results of examining 30 combination plots. The size of the aperture is in µm, the area is in µm2, and the shape of the apertures are listed.
Table 3. Maximum bump height-to-pitch ratios.
Further analysis and grouping of data has revealed three main trends:
- The finer Type 6 powder yielded a higher average H/P ratio. This was a result of higher packing density, which led to more solder deposited per µm3 of printed solder volume. Figure 7 shows that this trend is independent of solder alloy.
- Round apertures produced the majority of the maximum H/P ratios.
- Somewhat unexpected was a trend plotted in Figure 8. As pitch increases, so does the H/P ratio. Since the denominator of the ratio is pitch, it was anticipated that the H/P ratio should be a straight line when plotted against pitch. One theory is that since the webbing dimension is independent of pitch, a larger percentage of the available printing real estate is consumed by the webbing.
Figure 7. Type 6 powder yields higher bumps.
Figure 8. Pitch effect on H/P ratio.
- Prismark Partners LLC, The Electronics Industry Report, (New York 2003), pp 139-144
- Jackson, G.J. et al, ‘Differences in the Sub-Processes on Ultra Fine Pitch Stencil Printing due to Type-6 and Type-7 Pb-free Solder Pastes used for Flip Chip,” Electronics Components and Technology Conference, (2003), pp. 536-543.
- Johnson, A., ”Tutorial: How to Select the Best Stencil for SMT and Advanced IV Package Printing,” Chip Scale Review, (April 2003), pp. 51-57.
- Lathrop, R.R., “Optimizing Solder Paste printing For Wafer Bumping,” APEX West Conference, (February 2004).
- Coleman, W.E., “Stencil Design and Performance for Flip Chip/Wafer Bumping,” APEX West Conference, (February 2004).3
- Lathrop, R.R., “Solder Paste Printing and Stencil Design Considerations for Wafer Bumping,” SEMICON West Conference, (July 2004).
*NanoFocus AG, in Oberhausen Germany.
RICHARD LATHROP, Surface Mount Materials technical services manager, may be contacted at Heraeus Inc., Circuit Materials Div., 24 Union Hill Road, W. Conshohocken, PA 19428; (610) 825-6050; e-mail: firstname.lastname@example.org.