BACKGROUND OF THE INVENTION In flip-chip bonding, a first component such as an integrated circuit (IC) chip is mounted on and connected to another component such as an interconnection component such as ceramic interconnection substrate or printed circuit board (PCB). A plurality (e. g. , an array) of solder balls (also called"solder bumps") are formed on a face of at least one of the components (e. g., the IC chip), and the"bumped"component is brought into a face-to-face relationship with the other component. The two components are then heated (such as in a furnace) to reflow (heat, then allow to cool) the solder bumps, thereby making electrical connections between respective terminals of the two components. A number of techniques for ball-bumping various substrates are proposed in the prior art.

In a basic"stenciling technique", a stencil (mask) having apertures (cells, openings) therein is placed over the substrate to be bumped, with the apertures overlying corresponding pads of the substrate. As the stencil is held in place, an amount of solder paste is dispensed onto the stencil, and a screening blade (sometimes called a"doctor blade") is moved across the stencil surface in a manner to force solder paste into the stencil apertures. The stencil is then removed, which leaves behind solder paste on the pads, and the substrate is heated to reflow the solder to form solder bumps on the substrate.

US Patent No. 5,539, 153 (Schweibert et al.;"HEWLETT PACKARD") discloses a method of bumping substrates by"contained paste deposition". A non-wettable metal mask (stencil) is disposed on a substrate such that a plurality of apertures in the mask align with a plurality of pads on the substrate. The apertures are filled with solder paste in a manner comparable to

that which was described hereinabove with respect to the stenciling technique. The solder paste is then reflowed with the mask in place. After reflow, the mask is removed.

US Patent No. 5,492, 266 ("IBM-1") discloses a process for forming solder on select contacts of a printed circuit board (PCB), and is generally similar to the aforementioned Hewlett Packard Patent. A non-wettable stencil having openings is positioned on the board, the openings are filled with solder paste and, with the stencil fixedly positioned on the board, the solder paste retained by the stencil pattern is reflowed to selectively form on the underlying contacts of the printed circuit board.

US Patent No. 5,658, 827 (Aulicino;"IBM-2") discloses a method for forming solder balls on a substrate. The solder balls are formed by squeegeeing solder paste through apertures in a fixture into contact with pads on a substrate, and heating the fixture, paste and substrate to reflow the solder paste into solder balls that attach to the pads and are detached from the fixture. After cooling, the fixture is separated from the substrate. In an embodiment of the method, the fixture and substrate are inverted, and another surface mount electrical component is placed on the opposite surface of the substrate prior to heating the substrate.

The aforementioned HEWLETT PACKARD, IBM-1 and IBM-2 patents all describe printing solder paste through a mask or stencil onto a substrate, and reflowing the solder paste with the stencil in place on the substrate. In each case, the cells formed by the stencil apertures/openings are open on one side (the side of the stencil opposite the side in contact with the substrate). This is referred to herein as"open cell".

USP 5,988, 487 ("MACKAY-1") discloses a"captured cell"solder printing and reflow technique. (Also referred to herein as"closed cell") A substrate (e. g. , wafer) is placed upon a base heater stage 20 having a flat top surface 21 upon which a substrate. A screening stencil is laid over the surface of the substrate and solder paste material is deposited into the stencil's apertures with a screening blade. The stencil is placed in such a manner that each of its apertures is positioned over a substrate pad, upon which a solder bump is to be formed. Next, a flat pressure plate is laid over the exposed top surface of the stencil, which creates a fully enclosed, or"captured"cell of solder paste within each stencil aperture. (This is in contrast to the captured, yet open cell technique of HEWLETT PACKARD. ) Then, with the stencil and

plate remaining in place on top of the substrate, the substrate is heated to a temperature sufficient to reflow the solder paste material. After reflow, the substrate is cooled, and the pressure plate and stencil are thereafter removed, leaving solder bumps on the substrate. The use of the pressure plate ensures the proper formation of the solder bumps at high densities of solder bumps (i. e. , high densities corresponding to small solder bump sizes and small pitch distances between solder bumps).

The captured cell technique of MACKAY-1 works, but has some limitations. The heat required to bring the mask to temperature and reflow the solder must be conducted through the substrate being soldered, which can cause damage to substrates, particularly to substrates which are printed circuit boards. Also, any gap at the mask-to-substrate interface allows solder to leak out during solder paste printing (filling the mask), and when the solder is reflowed it can be ejected onto the substrate due to escaping gas. The method is best suited to substrates with very flat faces (low topology).

The captured cell technique was improved upon in USP 6,293, 456 ("MACKAY-2"), which teaches heating the mask directly, rather than through the substrate. The heater stage may be used, without a pressure plate, to close the cells. The heater stage may be preheated before being brought into contact with the assembly of the mask and the substrate. Also, it is disclosed that the assembly of the mask and the substrate can be inverted, so that the gap (if any) between the mask and substrate is"up"rather than"down", particularly during reflow.

(See next paragraph. ) This works well when there is a large gap between the mask and the substrate, which may be due to irregular substrate topology. MACKAY-2 also teaches the "interference"method of ball attach. This generally means that the mask opening and mask thickness is designed to provide a solder past volume that will result in a reflowed ball diameter that exceeds the mask thickness, thereby causing the liquid solder ball to expand partially out of the mask, and contact and wet itself to the metal pad on the substrate (e. g., wafer). The expanding ball can bridge a gap between the mask and the substrate.

In Figure 4 of US Patent No. 5,658, 827 ("IBM-2"), the substrate 11 is shown with the pads on the top of the substrate, in what is referred to as an"inverted"position (column 6, lines 45-48). In Figures 5 and 6 of Aulicino et al. (USP 5,658, 827), the substrate 11 is shown in what is referred to as an"upright"position (column 6, lines 48-51), wherein the pads are on the bottom

of the substrate 11. In contrast thereto, in MACKAY-2, and generally throughout the present patent application, the orientation with the pads below the substrate, is referred to as"inverted", and the orientation with the pads on top of the substrate is referred to as"un-inverted"or"non- inverted". When appropriate, reference may be made to"pads up"or"pads down"to describe the orientation of the substrate being ball bumped.

The captured cell technique was further improved upon in USSN 09/962, 007 ("MACKAY- 3"), filed 9/24/01, which discloses BALL BUMPING SUBSTRATES, PARTICULARLY WAFERS. A mask (stencil) having cells (openings) is disposed on a surface of a heater stage, and is then filled (printed) with solder paste. (This is referred to as off-line printing of the mask. ) Then a substrate is assembled to the opposite side of the mask. Then the solder paste is reflowed. This may be done partially inverted, or partially inverted. Then the mask is separated from the substrate, either before or after cooling. Solder balls are thus formed on the substrate, which may be a semiconductor wafer. Methods for printing the mask with solder paste is described. (The MACKAY-3 application was filed 9/24/2001 and is priority to the present international (PCT) patent application.) Additional attention is directed to the following US patents: 5,079, 835 (Lam); 6,051, 273 (Dalal); 5,950, 908 (Fujino, et al. ) ; 5,539, 153 (Schweibert et al.;"Hewlett Packard"); 5,366, 760 (Fujii, et al. ) ; and 6,008, 071 (Karasawa, et al.).

BRIEF DISCLOSURE (SUMMARY) OF THE INVENTION It is an object of the invention to provide an improved process for forming solder balls on electronic components. It is another object of the present invention to provide a technique for ball-bumping a substrate so that the resulting solder balls have a clean, oxide-free surface, thereby improving wetting when the ball-bumped substrate is joined (soldered) to an interconnection substrate. It is another object of the present invention to accomplish the foregoing objects in a minimum number of process steps. It is another object of the present invention to accomplish the foregoing objects in a process which requires a minimal amount of manufacturing time. It is another object of the present invention to provide a technique for accomplishing the foregoing objects inexpensively. It is another object of the present invention to provide a technique for accomplishing the foregoing objects in a manner suited to high- volume production.

Generally, an electronic component (e. g. , wafer) substrate is processed ("ball bumped") to form a plurality of solder balls on a corresponding plurality of pads on the substrate. A mask (stencil) having a plurality of openings (cells) is disposed on the surface of the a heater stage and is printed (filled with solder paste). Then, the assembly of mask and heater stage is shuttled over to a substrate having pads (e. g. , a wafer) which is in a chuck. The filled openings of the mask are aligned over the corresponding plurality of pads on the substrate.

The mask is held in intimate contact with the heater stage and with the wafer. The cells are therefore"closed"or captured. Then the heater stage is heated to reflow the solder paste and form solder balls. The mask may be removed from the wafer (or vice versa) while still molten.

Reflow may also be performed in a partially-inverted orientation.

The process of the present invention is capable of achieving high densities of small solder balls, and is readily scalable to lower densities of large solder balls. The process proceeds relatively quickly, with low capital expenditure equipment, and without hazardous chemicals. The present invention provides a fast, low-cost, robust, non-capital-intensive method and apparatus for forming arrays of solder bumps at moderate to high densities on electronic components, including 150 ktm area arrays, 200 um area arrays, and 250 am area arrays, forming solder balls at 0.5 mm pitch and at 0.8 mm pitch.

The invention is more particularly set forth in the claims appended hereto, as follows.

A ball bumping machine comprises a machine base, a chuck for holding a substrate, a heater stage, and a frame for holding a mask, and is characterized by the chuck is disposed in a chuck base on one side of the machine base; the heater stage is disposed in a heater stage base on an opposite side of the machine base; an elongate shuttle mechanism is pivotally attached at one end to the machine base at a point between the. chuck base and the heater stage base, the shuttle mechanism having an opposite end; the frame is held in a mask carrier which is attached to the opposite end of the shuttle mechanism; and means are provided for controlling the position of the shuttle mechanism so that the shuttle mechanism can shuttle the mask carrier between the heater stage and the chuck.

Means are provided for selectively holding the chuck base to the machine base; means are provided for selectively holding the heater stage base to the machine base; means are provided for selectively holding the mask carrier to the heater stage base; and means are provided for selectively holding the mask carrier to the chuck base.

A chuck assembly for holding a semiconductor wafer substrate in intimate contact with the mask comprises a rigid, generally planar chuck base, characterized by a central recess extending into the stage from a top surface thereof, said recess sized and shaped to receive a generally planar, flexible diaphragm, said diaphragm extending across the bottom of the recess.

The method comprises ball bumping pads of a substrate in a chuck using a mask having cells filled with solder paste, and a heater stage, characterized by securing the mask to the heater stage; filling the cells of the mask; securing the mask/heater stage to the chuck; shuttling the mask/heater stage/chuck to a desired orientation; and heating up the heater stage to reflow the solder paste, thereby forming solder balls. According to a feature of the invention, while the solder balls are still molten, the wafer is separated from the from the mask.

A method of filling cells of a mask with a viscous material, such as solder paste, is also described, and comprises disposing a quantity of the viscous material on a surface of the mask; bringing a print blade to a distance of a few mils from the surface of the mask; contacting the mask with a cleaning blade; advancing the print blade across the surface of the mask to fill the cells; and advancing the cleaning blade across the surface of the mask, following behind the print blade.

Other objects, features and advantages of the invention will become apparent in light of the following description thereof.

BRIEF DESCRIPTION OF THE DRAWINGS Reference will be made in detail to preferred embodiments of the invention, examples of which are illustrated in the accompanying drawings. The drawings are intended to be illustrative, not limiting. Although the invention will be described in the context of these preferred embodiments, it should be understood that it is not intended to limit the of the invention to these particular embodiments.

Certain elements in selected ones of the drawings may be illustrated not-to-scale, for illustrative clarity. Often, similar elements throughout the drawings may be referred to by similar references numerals. For example, the element 199 in a figure (or embodiment) may be similar in many respects to the element 299 in an other figure (or embodiment). Such a relationship, if any, between similar elements in different figures or embodiments will become apparent throughout the specification, including, if applicable, in the claims and abstract. In some cases, similar elements may be referred to with similar numbers in a single drawing. For example, a plurality of elements 199 may be referred to as 199a, 199b, 199c, etc. The cross-sectional views, if any, presented herein may be in the form of"slices", or"near-sighted"cross-sectional views, omitting certain background lines which would otherwise be visible in a true cross- sectional view, for illustrative clarity.

The structure, operation, and advantages of the present preferred embodiment of the invention will become further apparent upon consideration of the following description taken in conjunction with the accompanying drawings.

Figure 1 is an exploded cross-sectional view of a method and apparatus for forming solder balls on substrates.

Figure 1A is an enlarged (magnified) view of the substrate (102) shown in Figure 1, after completion of ball bumping.

Figure 1B is an exploded cross-sectional view of an alternate embodiment of a method and apparatus for forming solder balls on substrates.

Figure 2A is a side cross-sectional view of another technique for forming solder balls on a surface of a substrate.

Figure 2B is a side cross-sectional view of another technique for forming solder balls on a surface of a substrate.

Figure 3A is a side cross-sectional view of an alternate embodiment of a technique for ball- bumping a substrate.

Figure 3B is a top plan view of a mask (stencil) used in the technique of Figure 3A.

Figure 3C is a top plan view of an alternate embodiment of a mask (stencil) used in the technique of Figure 3A.

Figure 3D is a top plan view of another alternate embodiment of a mask (stencil) used in the technique of Figure 3A.

Figure 3E is a side cross-sectional view of a further step in the technique for ball-bumping a substrate.

Figure 3F is a side cross-sectional view of a further step in the technique for ball-bumping a substrate.

Figure 3G is a side cross-sectional view of a further step in the technique for ball-bumping a substrate.

Figure 3H is a side cross-sectional view of a ball-bumped substrate which has been formed.

Figure 31 is a schematic illustration of a top plan view of a ball in a cell of a mask, such as the mask of Figure 3B.

Figure 3J is a schematic illustration of a top plan view of a ball in a cell of a mask, such as the mask of Figure 3C.

Figure 4 is a schematic diagram of a machine for ball bumping substrates.

Figures 4A-4B are schematic diagrams of a process flow for ball bumping substrates, using the machine of Figure 4.

Figure 4C is a schematic diagram of an alternate embodiment of a process flow for ball bumping substrates, using the machine of Figure 4.

Figure 4D is a schematic diagram of an alternate embodiment of a process flow for ball bumping substrates, using the machine of Figure 4.

Figure 4E is a partial cross-sectional view of a substrate being bumped according to the technique of Figure 4D.

Figure 5A is a side cross-sectional view illustrating a"composite"mask.

Figure 5B is a side cross-sectional view illustrating a"bridge the gap"feature.

Figure 5C is an exploded side cross-sectional view illustrating a"stacking masks"feature.

Figure 6A is a top plan view of a mask mounting technique of the prior art.

Figure 6B is a cross-sectional view taken on a line 6B-6B through Figure 6A.

Figure 6C is a cross-sectional view of the mask of Figure 6A.

Figure 7A is a top plan view of a mask mounting technique.

Figure 7B is an exploded side cross-sectional view taken on a line 7B-7B through Figure 7A.

Figure 8A is an exploded side cross-sectional view of an alternate embodiment of a mask mounting technique.

Figure 8B is a partially exploded side cross-sectional view of a technique for capturing the cells of the mask illustrated in Figure 8A.

Figure 9 is an exploded side cross-sectional view of a chuck assembly for holding a substrate which is a semiconductor wafer.

Figure 9A is a magnified cross-sectional view of a component of the chuck assembly of Figure 9.

Figure 10 is a schematic side view of a ball bumping machine of the present invention.

Figure 10A is a schematic side view of a heater stage of the present invention.

Figure 10B is a schematic cross-sectional view of a chuck assembly of the present invention.

Figure 10C is a schematic cross-sectional view of a wafer being ball bumped, partially inverted, according to the invention.

Figure 10D is a plan view of a mask, and its mounting arrangement, according to the invention.

Figures 11A-11C are schematic diagrams of a process flow for ball bumping substrates, using the machine of Figures 4 or 10, according to the invention.

Figure 12 is a schematic side view of a technique for applying solder paste to cells in a mask, according to the invention.

DETAILED DESCRIPTION OF THE INVENTION Figure 1 illustrates a technique 100 for forming solder balls on a surface of a substrate 102, such as is set forth in MACKAY-1.

The substrate 102 has number of pads 104 on its top (as viewed) surface. The pads 104 are typically arranged in an array, having a pitch (center-to-center spacing from one another). The substrate 102 is disposed atop a heater stage 106.

A mask (stencil) 110 is provided. The mask 110 is a thin planar sheet of relatively stiff material, such as molybdenum, having a plurality of openings (cells) 112, each corresponding to a pad 104 whereupon it is desired to form a solder ball on the substrate 102.

The mask 110 is placed on the top (as viewed) surface of the substrate 102 with the cells 112 aligned over the pads 104. The cells 112 in the mask 110 are filled with solder material 114.

This is done in any suitable manner such as by smearing solder material on the top (as viewed) surface of the mask 110 and squeegee-ing the solder material 114 into the cells 112 of the mask 110.

A typical solder paste contains particles of lead/tin solder, in a matrix of flux, with the following proportions : 80% (by weight) solid material (e. g. , particles of lead/tin solder), and 20% (by weight) flux (including volatiles). In terms of relative volume percentages, the same typical solder paste may contain approximately 55% (by volume) of solid material (metal) and 45% (by volume) of flux. As discussed in greater detail hereinbelow, it is preferred that a"solder material"be used in lieu of regular solder paste.

The cells 112 in the mask 110 may be filled with solder material prior to placing the mask 110 on the top surface of the substrate, in which case the solder-material-filled cells 112 would be aligned over the pads 104.

A pressure plate 120 is disposed onto the top (as viewed) surface of the mask 110. This holds the mask 110 down onto the substrate 102, and the substrate 102 down onto the heater stage 106. This also closes off the cells 112-hence, the terminology"captured cell".

The heater stage 106 is heated up, typically gradually, to a temperature sufficient to cause the solder material in the cells 112 to melt (reflow). When the solder material melts, the individual solder particles will merge (flow) together and, due to surface tension, will try to form (and, typically, will form) a sphere.

When the solder material re-solidifies, it assumes a general spherical or hemispherical shape.

The mask 110 is then removed from the substrate 102.

Figure 1A is an enlarged (magnified) view of the substrate 102 shown in Figure 1, after completion of ball bumping. Herein it can be observed that the solder balls 130 are generally spherical, have a diameter"D"and have a height"H".

MACKAY-1 describes exemplary substrate heating programs (profiles, recipes) in terms of temperature as a function of time.

A drawback of the technique 100 is that no provision is made for"outgassing"of volatiles when the solder material is reflowed. Another drawback of the technique 100 is that heat is directed through the substrate 102.

Figure 1B illustrates an alternate technique 150 (compare 100) for forming solder balls on a surface of a substrate 152 (compare 102).

The substrate 152 has number of pads 154 (compare 104) disposed on its top (as viewed) surface. The substrate 152 is disposed atop a chuck (base) 158, rather than atop a heater stage (106).

A mask (stencil) 160 (compare 110) having cells 162 (compare 112) filled with solder material 164 (compare 114) is disposed on the surface of the substrate 152 with the cells 162 aligned with the pads 154. The cells 162 may be pre-filled or filled with the mask 160 atop the substrate 152.

A pressure plate 170 (compare 120) is disposed onto the top (as viewed) surface of the mask 160. This holds the mask 160 down onto the substrate 152, and the substrate 152 down onto the chuck base 158. This also closes off the tops of the cells 162.

A heater stage 156 (compare 106) is disposed onto the top (as viewed) surface of the pressure plate 170. The heater stage 106 is heated up, typically gradually, to a temperature sufficient to cause the solder material in the cells 162 to reflow. When the mask 160 is removed, solder balls such as those (130) shown in Figure 1A will be present on the pads 154.

A drawback of the technique 150 is that no provision is made for"outgassing"of volatiles when the solder material is reflowed. However, in contrast to the technique 100, the technique 150 directs heat through the pressure plate 170 rather than through the substrate 152.

Figure 2A illustrates a technique 200 for forming solder balls on a surface of a substrate 202.

The substrate 202 has a top surface 202a and a bottom surface 202b.

In this example, forming solder balls on an external surface of substrate (or board) which is a BGA substrate (board) is discussed as exemplary of forming solder balls on (ball-bumping) a substrate. It should, however, be understood that the techniques described herein have applicability to ball bumping other substrates, such as semiconductor wafers.

A typical BGA substrate 202 has a plurality of contact pads 204 on its surface, each of which measures 35 mils across. In the typical case of round contact pads, each pad would be 35 mils in diameter. These contact pads 204 are typically spaced 50 mils (center-to-center) apart from one another. Often, the pad-surface 202a of the substrate is covered by thin (e. g. , 2 mil) layer of insulating material 206, such as a polymer, which has openings 208 aligned with (centered over) the pads 204. The insulating material 206 has a top surface 206a.

The openings 208 in the insulating material 206 are typically somewhat smaller in size (area) than the pads 204-for example, each opening measuring only 30 mils across. Evidently then, the top surface 202a of the BGA substrate 202 will be quite irregular, exhibiting peaks where the insulating material 206 overlaps the pads 204 and valleys between the pads 204.

A mask (stencil) 210 is provided. The mask 210 is a thin (e. g. , 30 mils thick) planar sheet of relatively stiff material, such as molybdenum, having a plurality of openings (cells) 212, each corresponding to a pad 204 whereupon it is desired to form a solder ball on the substrate 202. A typical cross-dimension for a cell 212 is 40 mils across.

In a first step of forming solder balls on (ball bumping) the substrate 202, the mask 210 is placed on the top surface 202a of the BGA substrate 202 with the cells 212 aligned over the pads 204, more particularly, over the openings 208 in the layer of insulating material 206. As illustrated, due to the size (diameter) of the cells 212, and the irregular surface 206a of the insulating material 206, there will be gaps 214 between the mask 210 and the insulating material 206. A typical dimension for the gap is 1-2 mils. As will be evident, these gaps 214 have benefits and disadvantages.

In a next step of forming solder balls on the substrate 202, the cells 212 in the mask 210 are filled with solder material 220 which is shown as a number of various-size spheres : (The middle cell 212 in the figure is shown without solder material 220, for illustrative clarity. ) This is done in any suitable manner such as by smearing solder past on the top surface 210a of the mask 210 and squeegee-ing the solder material 220 into the cells 212 of the mask 210. The cells in the mask may be filled with solder material prior to placing the mask 210 on the top surface 202a of the BGA substrate 202 with the (filled) cells 212 aligned over the pads 204.

In a next step of ball bumping the substrate 202, a heater stage (platen) 230 is disposed onto the top surface 210a of the mask 210, and the substrate 202, mask 210 and heater stage 230 are held together with clamps (not shown), in the orientation shown in the figure-namely, with the heater stage 230 on top of the mask 210, and with the mask 210 on top of the substrate 202. A pressure (contact) plate (not shown, compare 170) may be disposed on the top surface 210a of the mask 210, between the heater stage 230 and the mask 210.

In a next step of forming solder balls on the substrate 202, the heater stage 230 is heated up, typically gradually, to a temperature sufficient to cause the solder material 220 to melt within the cells 212. When the solder material 220 melts, the individual solder particles will merge (flow) together and, due to surface tension, will try to form (and, typically, will form) a sphere.

During reflow heating, small-sized solder particles within the solder material can"leak"out of the gap 214. This is not desirable. On the other hand, the gap 214 allows volatile material to "outgas".

After reflowing the solder material 206, the heater stage 230 is either removed immediately, so that the solder can cool down, or is kept in place and allowed to cool down until the solder has re-solidified as solder balls. As described in greater detail hereinbelow, often, as the solder material cools off, it will try to form a ball which has a larger diameter than the cell. This results in (i) there being an interference fit between the resulting solder ball and the sidewalls of the cell and (ii) a deformed solder ball. Regarding the latter, it is known to reflow the resulting deformed solder balls after removing the mask in order to cause them to assume a more spherical shape.

The forming of solder balls (240) on a substrate (202) is suitably carried out in the orientation illustrated in Figure 2A-namely, the mask (214) is disposed on top of the substrate (202) and the heater stage (230) is disposed on top of the mask (214). Alternate embodiments, where reflow heating is carried out with the mask/substrate assembly inverted, or partially inverted, are described hereinbelow.

An inherent"side-effect"of the described technique 200 is that the flux material in the solder material (106) will liquefy and may run down onto the top surface 202a of the substrate 202 or, in the case of there being an insulating layer 206, onto the top surface 206a of the insulating layer 206. In that the ball-bumped BGA substrate (or ball-bumped semiconductor package assembly) may be"warehoused"for months, prior to being mounted to an interconnection substrate, it is known that it should be cleaned of flux (de-fluxed) soon after the solder balls have been formed on the pads (204). Furthermore, whatever flux was present in the solder material (220) will largely have been dissipated (run-off and cleaned off) in the process the flux ran off (and cleaned) off the solder balls, resulting in that they will need to be re-fluxed prior to assembling to the interconnection substrate. Typically, the flux component of solder material will lose its viscosity and start running at a much lower temperature than the melting point of the solid particulate (solder) component of the solder material.

Figure 2B illustrates another prior art technique 250 (compare 200) for forming solder balls on a surface of a substrate 252 (compare 202) -more specifically on contact pads 254 (compare 204) of a substrate 252. The substrate 252 has a top surface 252a (compare 202a) and a bottom surface 252b (compare 202b), contact pads 254 (compare 204) disposed on its top surface 252a, and a thin layer of insulating material 256 (compare 206) which has openings 258 (compare 208) aligned with (centered over) the pads 254. The insulating material 256 has a top surface 256a (compare 206a).

A mask 260 (compare 210) has a plurality of cells 262 (compare 212). In this example, the cross-dimension of a cell 262 is smaller than in the previous example (for example only 25 mils across). Due to this smaller cross-dimension, a gap (compare 214) is not formed between the mask 260 and the insulating material 256, and the mask 260 is essentially"sealed"to the substrate 252. This has the advantage that small solder balls and flux material will not"leak out" (through the gap) onto the surface of the substrate 252 (except in the case that the mask is

held off of the surface of the substrate by a defect or by contamination). However, the lack of a gap also means that volatiles have no place to escape (vent, "outgas"). Thus, the rate at which the temperature of the solder material 270 is elevated becomes critical. More particularly, if the solder material is heated too fast, the volatiles will try to escape the cell (262) in a"violent" manner, often tending to lift the mask 260 off of the substrate 252. This is not desirable.

As in the previous example, in a first step of forming solder balls on the substrate 252, the mask 260 is placed on the top surface 252aof the BGA substrate 252 with the cells 262 aligned over the pads 254, more particularly, over the openings 258 in the layer of insulating material 256.

As in the previous example, in a next step of forming solder balls on the substrate 252, the cells 262 in the mask 260 are filled with solder material 270 (compare 220) which is shown as a number of various-size spheres. (The middle cell 262 in the figure is shown without solder material 220, for illustrative clarity. ) The cells 262 in the mask 260 may be filled with solder material prior to placing the mask 260 on the top surface 252a of the BGA substrate 252 with the (filled) cells 262 aligned over the pads 254.

As in the previous example, in a next step of forming solder balls on the substrate 252, a heater stage (platen) 280 (compare 230) is disposed onto the top surface 260a of the mask 260, and the substrate 252, mask 260 and heater stage 280 are held together with clamps (not shown), in the orientation shown in the figure-namely, with the heater stage 280 on top of the mask 260, and with the mask 260 on top of the substrate 252. A pressure (contact) plate (not shown, compare 170) may be disposed on the top surface 260a of the mask 260, between the heater stage 280 and the mask 260.

As in the previous example, in a next step of forming solder balls on the substrate 252, the heater stage 280 is heated up (gradually, as noted hereinabove), to a temperature sufficient to cause the solder material 270 to melt within the cells 262. When the solder material 270 melts, the individual solder particles will merge (flow) together and, due to surface tension, will try to form (and, typically, will form) a sphere.

As in the previous example, after reflowing the solder material 270, the heater stage 280 is either removed immediately, so that the solder can cool down, or is kept in place and allowed to cool down until the solder has re-solidified as solder balls.

As described in greater detail hereinbelow, often, as the solder material cools off, it will try to form a ball which has a larger diameter than the cell. This results in (i) there being an interference fit between the resulting solder ball and the sidewalls of the cell and (ii) a deformed solder ball. Regarding the latter, it is known to reflow the resulting deformed solder balls after removing the mask in order to cause them to assume a more spherical shape.

As in the previous example, the forming of solder balls on a substrate (252) is typically carried out in the orientation illustrated in Figure 2B-namely, the mask (260) is disposed on top of the substrate (252) and the heater stage (280) is disposed on top of the mask (260). Alternate embodiments, where reflow heating is carried out with the mask/substrate assembly inverted, or partially inverted, are described hereinbelow.

A benefit of the techniques 200 and 250 shown in Figures 2A and 2B is that the mask and the solder material contained within the cells of the mask are heated essentially directly, rather than through the substrate as was the case with the technique 100 shown in Figure 1. Also, as shown in Figure 2A, a gap 214 allows for outgassing, which permits faster reflow times.

Figure 3A illustrates a technique 300 (compare 100,200, 250) for ball bumping a substrate 302 (compare 102,202, 252) -more specifically on contact pads 304 (compare 104,204, 254) of a substrate 302. The substrate 302 is any electronic substrate, including a semiconductor wafer or a BGA board. The substrate 302 has a top surface 302a (compare 102a, 202a, 252a) and a bottom surface 302b (compare 102b, 202b, 252b). A plurality of contact pads 304 (compare 104,204, 254) are disposed on the top surface 302a of the substrate 302, and are covered by a thin layer 306 (compare 206,256) of insulating material, such as a polymer, (or, in the case of the substrate 302 being a semiconductor wafer, a passivation layer) which has openings 308 (compare 108,208, 258) aligned with (centered over) the pads 304. The insulating material 106 has a top surface 306a (compare 106a, 206a, 256a). The top surface 302a of the substrate 302 has an irregular topology, exhibiting peaks where the insulating material 306 overlaps the pads 304 and valleys between the pads 304.

A mask (stencil) 310 (compare 110, 210,260), which is suitably a thin planar sheet of relatively stiff material, such as molybdenum, has a plurality of cells 312 (compare 112,162, 212,262), each corresponding to and aligned with a pad 304 whereupon it is desired to form a solder ball on the substrate 302. The cells 312 in the mask 310 may be round (circular), as illustrated by the array of cells 312b in Figure 3B. Preferably, however, the cells are not round (circular). For example, as illustrated by the array of cells 312c in Figure 3C, the cells 312c may be square. In this manner, for a given spacing, e. g. , 10 mils between the peripheries of adjacent cells 312c (in other words the size of the"web"in the mask between adjacent cells 312c), each individual cell 312c can have a larger area, hence a larger volume for a given thickness mask, than a round cell (312b). Alternatively, as illustrated by the array of cells 312d in Figure 3D, the cells 312d may have a trapezoidal shape, and be arranged in alternating orientations. As in the example of square cells (See Figure 3C), in this manner, for a given spacing, e. g. , 10 mils between the peripheries of adjacent cells 312d (in other words the size of the"web"in the mask between adjacent cells 312d), each individual cell 312d can have a larger area, hence a larger volume for a given thickness mask, than a round cell (312c). All other things being equal, the volume of a trapezoidal cell (312d) can be greater than that of a square cell (312c) which, in turn in greater than that of a round cell (312a). Non-round cells (e. g. , 312c and 312d) in a mask (e. g. , 310) for forming solder balls on a surface of a substrate are viable alternatives to round cells. It should be noted that Figures 3B, 3C and 3D are not drawn to the same scale as Figure 3A.

Returning to Figure 3A, in a first step of forming solder balls on the substrate 302, the mask 310 is placed on the top surface 302a of the substrate 302 with the cells 312 (preferably the cells 302c or 302d) aligned over the pads 304. Evidently, the irregular surface 306a of the insulating material 306 will result in there being gaps 314 (compare 114) between the mask 310 and the insulating material 306. These gaps 314 can perform a beneficial purpose of allowing volatiles to vent (outgas).

The mask 310 is held in any suitable manner either in direct face-to-face contact with the substrate 302, or ever so slightly spaced therefrom.

Then, the cells 312 are filled with solder material 320. (The middle cell 312 in the figure is shown without solder material 320, for illustrative clarity. ) The cells 320 of the mask 310 may be filled with solder material either when the mask is in face-to-face contact with the substrate

302, or"off-line" (prior to bringing the mask into face-to-face contact, or near contact, with the substrate. At this point in the process, the technique of the present invention deviates significantly from the techniques (100,200, 250) described hereinabove.

Figure 3E illustrates a next step in the process, wherein the assembly of the mask 310 and the substrate 302, with solder material 320 loaded into the cells 312 of the mask 310 is inverted, so that the substrate 302 is physically atop above the mask 310, as is illustrated in the figure. In this"upside-down"orientation, the solder material 320 will not fall out of the cells 312 in the mask 310, because it is"sticky", being a combination of solid particles and relatively viscous (at room temperature) flux material. The solder material 320 has the general consistency of toothpaste. It should be noted that in this figure (Figure 3E) the middle cell 312 is shown filled with solder material 320. Alternatively, a pressure (or"contact") plate may be placed against the mask, as described with respect to other embodiments.

As illustrated, this upside-down assembly of the mask 310 and the substrate 302, with solder material 320 loaded into the cells 312 of the mask 310 is brought into contact with a heater stage 330 (compare 130,230) which is either brought up to or which has been pre-heated to a temperature which is greater than the melting point of the solid particles in the solder material 320.

It is generally preferred that the solder material is gradually rather than abruptly reflowed. For example, by bringing its temperature up to less than its melt point to allow it to"condition"prior to causing it to reflow. Any suitable heat profile can be used.

For example, "63/37"lead/tin solder has a melting temperature of approximately 183°C (Centigrade). In which case, the heater stage 330 may be preheated to 140°-150°C for conditioning the solder material, then brought up to a temperature of at least 215°C, preferably to a temperature which is 20°C-40°C higher than the melting temperature of the solid particles of the solder material (i. e. , the heater stage 330 is preferably heated to approximately 220°C- 225°C for reflowing the aforementioned 63/37 solder material).

The upside-down assembly of the mask 310 and the substrate 302, with solder material 320 loaded into the cells 312 of the mask 310 is held in contact with a heater stage 330 for a sufficient period of time"t"for the solid particles in the solder material 320 to melt, and preferably not much longer. Given the dynamics of the overall system, this period of time"t"is preferably determined empirically. However, since the heater stage 330 was already preheated, and since the solder material 320 and the solder mask 310 are both fairly good conductors of heat, and based on experimental trials of the technique described herein, it is contemplated that, for most anticipated microelectronic applications, a period of time"t"of 5-20 seconds will be sufficient time for the solder material 320 to liquefy. However, in the case of a board (substrate) having heatsinks, for example a thick copper heatsink, the time"t"required to form the solder balls on the substrate may more than 20 seconds, for example 30 seconds.

Figure 3F illustrates a next step of the process wherein, after the solid particles in the solder material 320 have liquefied, the heater stage 330 is removed from being in further contact with the upside-down assembly of the mask 310 and the substrate 302. This can be done either by lifting the upside-down assembly of the mask 310 and the substrate 302, or by lowering the heater stage 330. The liquefied solder particles of the solder material 320 will begin to cool off and coalesce into one solid mass, typically generally in the form of a sphere.

Figures 3G and 3H illustrate the solder balls 340 that are formed by the process described hereinabove. In Figure 3G, the mask 310 is still in place. In Figure 3H, the mask has been removed, and the ball bumped substrate has been re-flipped over.

While Figure 3G illustrates an"ideal"situation where the resulting solder balls are perfectly centered within their respective cells, the real world tends not conform so neatly to perfection.

As illustrated in the schematic illustration of Figure 3I, a solder ball 342 which is slightly off- center in a round cell 344 (compare 312b) will exhibit an arcuate area of contact with the sidewall of the cell. In contrast thereto, as illustrated in the schematic illustration of Figure 3J, a solder ball 346 which is slightly off-center in a square cell 348 (compare 312c) will exhibit only minimal (e. g. , point) contact with the sidewall of the cell. The cumulative effect of a number of solder balls misaligned with the mask openings (cells) and being in contact with the mask can have an adverse undesirable effect on subsequent separation of the mask from the substrate.

A benefit of this"inverted"technique is that, due to the influence of gravity (i. e. , the earth's pull on objects towards the center of the earth), flux material within the solder material 320, which also has been liquefied, will run down the surface of the solid mass, rather than up to the surface of the substrate 302. This is in marked contrast to the previous examples wherein it was observed that the tendency was for the liquefied flux to run down onto the substrate (102,202, 252). This has some important beneficial results, including: the substrate (board) 302 does not need to be cleaned; 'the resulting solder balls 340 are"pre-fluxed" ; and 'the resulting solder balls 340 have a clean, oxide-free surface for better (subsequent) soldering.

Another benefit is that the resulting solder balls 340 will have a height (diameter) which is greater than the thickness of the mask 310. Generally, large solder balls 340 having approximately a 1: 1 aspect ratio (height: width) are readily formed on pads of substrates using the techniques described herein. As a result, the molten solder ball can join itself to the substrate without there needing to be any direct contact between the mask and the (pads of the) substrate. Also, the mask can be removed while the solder is still molten, thereby greatly facilitating mask/substrate separation.

Figure 4 illustrates major components of a"bumping"machine 400 for ball bumping substrates, both in the manner described hereinabove as well as using alternate techniques. The machine 400 comprises a stable platform 402.

The machine 400 comprises a chuck 404 which is disposed on the platform 402, for holding a substrate 406. (The substrate 406 is not a component of the machine 400.) The machine 400 comprises a mask holder 408 for holding a mask (not shown), and which is mounted in an articulated manner to the platform 402 so that it can be moved from a one position to another position.

The machine 400 comprises a pressure plate holder, such as a simple framework, for holding a pressure plate 410 (compare 120), and which is preferably mounted in an articulated manner to

the platform 402 so that it can be moved from a one position to another position. In use, it is preferred that the pressure plate be held in intimate contact with the surface of the mask opposite the substrate during reflow of the solder material in the mask.

A heat source 412 is provided for reflowing solder material in the mask, and which is preferably mounted in an articulated manner to the platform 402 so that it can be moved from a one position to another position. The heat source 412 may be a heater stage, or may be a radiant (e. g. , infrared) heat panel, such as may be obtained from Watlow Electric Mfg. Co. , St. Louis, MO, USA.

A print station 414, which may be a flat, non-wettable surface, is optionally provided, for off- wafer filling of the cells of the mask with solder material, as mentioned hereinabove.

One having ordinary skill in the art to which the invention most nearly pertains will understand how to implement the machine 400, for performing the various techniques described herein, in light of the descriptions set forth herein.

Inverted Reflow, Inverted Cooling Figures 4A-4B illustrate a technique 420 for ball bumping substrates. In this technique, the pressure plate is positioned above the heat source, at a location on the machine platform, as illustrated. The mask is positioned on the substrate, which is positioned on the chuck, at another location on the machine platform. With the mask positioned on the substrate, the mask cells may be filled with solder material. Next, the assembly of the chuck/wafer/mask are shuttled into position, upside down, on the pressure plate. The heat source is turned on, and the solder material in the mask melts. Then the heat source is shut off to allow the solder material to cool and coalesce into solder balls. Finally, the mask is separated from the substrate and the substrate is separated from the chuck.

It should be noted that in this, as well as in certain other embodiments described herein, that heat must pass through the pressure plate to melt the solder material within the mask. In the case of using a heat source which is an infrared-type heat source, a quartz pressure plate may be used. Otherwise, the pressure plate may be molybdenum, stainless steel, or the like.

The mask cells may be pre-filled with solder material, such as by positioning the mask on a print station surface (414, described hereinabove), or by utilizing the pressure plate as a print station (in which case, the heat source should not be"on").

The heat source may have a flat surface so that it can perform the function of the pressure plate, without an additional component.

Inverted Reflow, Un-Inverted Cooling Figure 4C illustrates a technique 440 for ball bumping substrates. This technique proceeds in the manner of the technique 420 described hereinabove, up to the point of melting the solder material with the substrate inverted, as illustrated in Figure 4B. Then, rather than allowing the solder material to cool in this orientation, the assembly of the chuck/wafer/mask are repositioned away from the heat source so that the wafer is"right side up" (un-inverted), and the solder material is allowed to cool. Finally, the mask is separated from the substrate and the substrate is separated from the chuck. The heat source may"follow"the assembly of chuck/wafer/mask when it is repositioned, in which case it would be switched"off"to allow the solder material to cool.

Partially-Inverted Reflow And Cooling As mentioned hereinabove with respect to the technique 300, an advantage of reflowing the solder material in the inverted position, as described by the techniques 420 and 440 is that outgassing may occur in gaps (e. g. , 314) between the mask and the substrate, thereby permitting relatively rapid heating (melting) of the solder material. However, it is possible that oxides may become trapped in the interface between the solder material and the substrate pad when reflowing in the inverted orientation.

Figures 4D and 4E illustrate an alternate technique 460 for ball bumping substrates. In this technique, rather than inverting the substrate (from 180° to 0°) to reflow the solder material, the substrate is positioned at an angle between 90° (on its side) and 0° (inverted), such as at 45° from inverted, as illustrated. (This also includes orientations for the substrate which are beyond inverted, such as-45°.) As illustrated, the substrate has been rotated 135° from being face (pads) up to being partially face down.

As best viewed in Figure 4E, a mask 462 (compare 310) having openings (cells) 464 (compare 312) extending from a one surface to an opposite surface thereof and filled with solder material 466, has its one surface disposed against a surface of a substrate 468 having pads 470. A pressure plate 472 is disposed in intimate contact against the opposite surface of the mask 462.

A middle one of the cells 464 is illustrated without solder material 466, for illustrative clarity, so that the gap 474 (compare 314) can clearly be seen. The assembly of substrate 468, mask 462 and pressure plate 472 are oriented as shown, and it can be seen that the gap 474 is at the highest point of the cell. This facilitates outgassing of volatiles during reflow. The chuck and the heat source are omitted from the view of Figure 4E, for illustrative clarity.

This technique proceeds in the manner of the techniques 420 and 440 described hereinabove, up to the point of securing the solder-laden mask to the substrate and mounting the pressure plate to the assembly. Then, the assembly is positioned as shown, partially inverted, so that a corner of each cell is the highest point in the cell (see the corner at the gap 474). Reflow is performed in this position, using the heat source (not shown). Finally, the mask is separated from the substrate and the substrate is separated from the chuck.

Alternatively, rather than allowing the solder material to cool in the partially-inverted orientation, the assembly of the chuck/wafer/mask are repositioned away from the heat source so that the wafer is"right side up" (un-inverted, 180°), and the solder material is allowed to cool. The heat source may"follow"the assembly of chuck/wafer/mask when it is repositioned, in which case it would be switched"off'to allow the solder material to cool.

Composite Mask And Pressure Plate The benefit of using a pressure plate to capture the solder material in the cells of the mask has been discussed hereinabove. It is generally preferred that the pressure plate be intimately held against the mask so that there are no gaps for leakage, particularly when reflowing inverted or partially inverted. This function can be implemented by a composite mask performing the functions of a mask and a pressure (contact) plate are formed as an integral unit, thereby assuring no leakage between the two.

Figure 5A illustrates an embodiment of a composite mask 500. The composite mask 500 is a rigid planar structure having two portions-a mask portion 510 comparable (e. g. ) to the mask

110 described hereinabove, and a pressure plate portion 520 comparable to the pressure plate 120 described hereinabove. A plurality of cells 512 (compare 112) extend from a one surface of the composite mask 500, through the mask portion 510, to the pressure plate portion 520. These "blind hole"type openings 512 are filled with solder material 514 (compare 114) in the manner described hereinabove.

The composite mask 500 is suitably formed of a sheet of metal, such as molybdenum, which is etched to have cells 512 extending into a surface thereof (but not all the way through the sheet). Alternatively, the composite mask 500 can be formed from a sheet of metal comprising the pressure plate portion 520, a surface of which is masked, patterned, and plated up to form the mask portion 510 (with cells 512).

Alternatively, a composite-type mask can be formed from a discrete mask welded or otherwise intimately joined (including adhered) to a discrete pressure plate.

Bridging A Gap An interesting feature/capability of the technique is illustrated in Figures 5A and 5B, but is not limited to the use of a composite mask. The composite mask 500 is illustrated disposed beneath a substrate which is in an inverted position, for example the substrate 302 from Figure 3A (see also Figure 3E). Note that no part of the substrate 302 actually is in contact with the composite mask 500-rather, that there is a small gap 524 between the opposing faces of the substrate and the mask.

As best viewed in Figure 5B, when the solder material 514 reflows and forms a ball, the ball has a diameter (height) which is greater than the thickness of the mask (in this illustrative case, greater than the thickness of the mask portion 510 of composite mask 500), so it sticks out of the mask, "bridges"the gap 524, and wets itself to the pad 304 on the substrate 302. The solder ball does this while it is in a liquid state, at which point the mask can easily be separated from the substrate, thereafter allowing the solder ball to cool off (solidify).

Stacked Masks Figure 5C illustrates a mask stack 550 comprising a first or"Liftoff"mask 552 (compare 110) having a plurality of cells 554 (compare 112) and a second or"volume control"mask 556

having a plurality of cells 558. For example, the mask 552 is 4 mil thick, and the mask 556 is 3 mil thick. The cells 558 are tapered, as illustrated, to provide reduced hole volume control The orientation of the mask stack 550 as it would be employed for ball bumping a substrate is illustrated by the substrate 560 having pads 562 and a pressure plate 564.

The mask stack 500 is beneficial in applications where particularly tall (high aspect ratio) solder balls (columns) are desired to be formed on a substrate, tending to overcome inherent limitations in the aspect ratio of holes that can be formed in masks. The two (or more) masks may be removed one at a time after solder ball formation to reduce liftoff stress.

There have thus been described, with respect to Figures 5A, 5B and 5C a number of mask "variations", including a composite mask, a mask which is spaced from the substrate being bumped, and a mask stack. Other mask variations may occur to one having ordinary skill in the art to which the present invention most nearly pertains, in light of the teachings set forth herein.

Mask Mounting Techniques Figures 6A and 6B illustrate, in top plan and side cross-sectional views, respectively, a conventional (prior art) mask setup 600. A generally rectangular (typically metal, such as molybdenum) mask 602, having a plurality of openings 606 for screening solder material as described hereinabove, is disposed in a frame (or mask mount) 604, and is secured (fixed) by at least two opposite, and in some cases by all four of its edges 602a, 602b, 602c and 602d to the mask mount (or frame) 604. Figure 6B shows the mask 602 in its cold (not heated) state, secured to and stretched between opposite sides of the frame 604. Upon application of heat to the mask 602, it expands, and having essentially no place to go, buckles or warps, as illustrated in the side cross-sectional view of Figure 6C. Such warpage (e. g. , in the Z-axis) reduces the pad-to-pad distance in one axis only.

Assembling The Mask To The Substrate Figures 7A and 7B illustrate, in top plan and side cross-sectional exploded views, respectively, a mask setup 700. A generally rectangular mask 702 has one edge 702a fixed, such as with allen cap screws (not shown), to a mask mount 704. The other, opposite edge 702c of the mask is disposed in a printer frame holder 708 which permits the mask to expand (allows for

movement in the Y-axis, due to elongation) during the heat cycle, without buckling. The edges 702b and 702d are not clamped.

Preferably, the mask 702 should not be allowed to move freely in both the X and Y directions, else high warpage and feature misalignment may occur. Pins 710 and 712 which are in the form of"diamond points"are mounted on the printer frame holder 708 and extend through corresponding elongate slots 720 and 722, respectively, which are cut or etched in the mask to permit the mask to expand in one direction (the Y-direction) only.

A plurality of elongate rail pins 730 are disposed on the bottom surface of the mask 702 and held thereto by any suitable means such as flush mount (recessed head) screws 732. The pins 730 extend in a"normal"direction to the plane of the mask, and are generally cylindrical, having a stepped portion of reduced diameter which will allow the pins 730 to be"captured"by a corresponding element of the work holder, as described hereinbelow. The elongate pins 730 are disposed outside an area having a pattern of holes 706 for applying solder material (not shown) onto the surface of a substrate 734.

The mask 702 and substrate 734 may be"assembled"for ball bumping in the following exemplary manner. The substrate is held in a workholder having"rails"736. Preferably, there are two rails 736, spaced apart from one another in the X-axis, and both running parallel to the Y-axis. The rails 736 have holes 738 for receiving the pins 730 (shown in phantom). A portion of the rails, or a separate element 740 associated with the rails, are movable, and have slots 742, and may include a cam-surface (e. g. , a tapered landing) for capturing the distal ends of the elongate pins 730, thereby forming a secure mortise (the pins) and tenon (the slots in the rails) type of connection between the rails and the pins.

In use, the mask 702 is brought down onto the workholder, with the holes 706 in the mask aligned with pads on the substrate 734, the rail pins 730 are captured by the element 740, thereby assembling the component (s) to the mask. The holes 706 are filled with solder material, and the assembly of the mask and substrate (component) are flipped over, in the manner described hereinabove, and placed against a heater stage, the heater stage being in contact with the mask, the solder material is reflowed, the heater stage is dropped, the molten solder material cools off, and solder balls are formed on the component.

The rails 736 running along the bottom of the mask 702 aid in maintaining the mask flat during the solder reflow heating process. The rails 736 are suitably at least two piece units-the piece contacting the mask is preferably made of a material selected for its low thermal conductivity (e. g. , 12 BTU/hr/sq. ft. ) such as Maycor (tm) ceramic. Preferably, the screws 732 mounting the pins 730 to the mask are also made of a material having low thermal conductivity (e. g. , 12 BTU/hr/sq. ft. ) such as Maycor (tm) ceramic.

Figure 8A illustrates an alternate embodiment of a an"assembly"800 of a mask 802 (compare 110,702) and a substrate 804 (compare 102,734). The substrate 804 is supported by a workholder (stage, chuck) 806, which is on the machine platform 810. Vacuum pedestals 812 and 814 extend upward (towards the mask) from the workholder 806. The mask 802 is shown having a mask mount 816 (compare 704) fixing a one edge thereof. When the mask 802 is brought down onto the substrate 804, a vacuum is drawn through the vacuum pedestals 812 and 814 to hold the mask 802 intimately against the substrate 804.

Figure 8B illustrates how a pressure plate 820 (compare 410) may be added to the"assembly" (800) of mask 802 and substrate 804. The assembly 800 is inverted and disposed onto the pressure plate 820. The pressure plate may simply be a stainless steel plate which is held by pedestals 822 extending upwards (towards the pressure plate) from the machine platform 810.

A heater stage 824 (compare 412) is disposed underneath the pressure plate. If desired, the pressure plate 820 may be secured to the assembly 800, in intimate contact with the mask 802 using magnets, vacuum chucks and the like.

It is within the scope of the invention that any combination of gizmos, gadgets, and the like (cam surfaces, vacuum chucks, magnets, electromagnets) can advantageously be utilized to hold the mask to the substrate and to hold the pressure plate to the mask.

Biased Chuck As mentioned above, a mask is placed substantially into face-to-face contact with a substrate being bumped. When the assembly of the mask and the substrate are moved (re-positioned), such as to an inverted or semi-inverted position, the mask may separate somewhat from the substrate, allowing solder material to enter gaps between the mask and the substrate. Also,

during reflow, the mask may warp or buckle, also allowing solder material to enter gaps between the mask and the substrate. A biased chuck assembly is provided for maintaining an intimate face-to-face contact between a mask and a substrate being bumped.

Figure 9 illustrates a biased chuck assembly 900 for holding a substrate 902 such as, but not limited to, a semiconductor wafer in positive contact with a mask 904. In a manner such as described hereinabove, two opposite edges of the mask 904 may be retained by rails 906 and 908, so that the mask 904 can be tensioned (stretched).

Semiconductor wafers are relatively brittle, but are known to have a certain degree of flexibility. For purposes of practicing this invention, the degree of flexibility possessed by a semiconductor wafer is sufficient to allow the semiconductor wafer 902 to deflect when urged against the mask 904 so as to maintain substantially intimate contact between the surface of the mask 904 and the surface of the semiconductor wafer (substrate) 902.

The substrate 902 is urged against the mask 904 in the following manner. A rigid, generally planar chuck base 910 has a central recess (cavity) 912 extending into the chuck base 910 from a top (as viewed) surface thereof. The recess 912 is sized and shaped to receive a generally planar, flexible diaphragm 914. The diaphragm 914 extends across the bottom of the recess 912, and is secured to the chuck base 910 such as with a bead 916 of a suitable adhesive 916 disposed about the periphery of the diaphragm 914. An inlet tube 920 extends from exterior the chuck base 910 to within the cavity 912, underneath the diaphragm 914. In this manner, when a gas such as nitrogen is introduced at a positive pressure into the inlet tube 920, the diaphragm 914 is caused to deflect upwards (as viewed), urging anything disposed atop the diaphragm 914 (in this case, the wafer 902) upwards (in this case, against the mask 904). The diaphragm 914 is suitably a 0.125 inch thick sheet of silicon rubber material. The peripheral edge of the diaphragm 914 is preferably"contained"by the sidewall of the cavity 912, as illustrated.

Preferably, a permeable substrate 928, such as a 100 mil thick powdered metal plate, is disposed beneath the diaphragm 914, between the diaphragm 914 and the bottom surface of the cavity 912. When a suction is applied to the inlet tube 920, the permeable substrate 928 will prevent the diaphragm 914 from closing off the opening.

A second central recess (cavity) 922, coaxial with and larger (wider, of greater diameter) than the recess 912 extends into the chuck base 910 from the top surface thereof, and is sized and shaped to receive a generally planar, flexible manifold element 930.

As best viewed in Figure 9A, the manifold element 930 has a top surface 932 and a bottom surface 934. A plurality of grooves 936 extend, such as criss-cross style (2 parallel sets of intersecting grooves), across the top (as viewed) surface of the manifold element 930. An opening 938 extends from the top surface 932 of the manifold element 930 (of from a bottom of one of the grooves 936) through to the bottom surface 934 of the manifold element 930. The opening 938 is aligned with an inlet orifice 940 in the chuck base 910.

As best viewed in Figure 9, the manifold element 930 extends across the recess 922, and may be secured to the top (as viewed) surface of the diaphragm 914. (Alternatively, the manifold element 930 may be formed integrally with the diaphragm. ) In this manner, when a vacuum is pulled on the inlet tube 940, a substrate 902 sitting atop the manifold element 930 is held in intimate contact with the manifold element 930. The manifold element 930 is suitably a 5 mil thick sheet of a film material such as kapton (tm).

In use, a wafer 902 is loaded onto the chuck assembly 900. The wafer 902 is disposed atop the manifold element 930. The mask 904, which may previously have had solder material introduced into its cells (apertures), is disposed against (including nearly against) the surface of the wafer. A positive pressure is introduced into the inlet tube 920, and the assembly of mask and wafer can be manipulated (e. g. , inverted, partially-inverted) for reflowing the solder material, as discussed hereinabove. Intimate contact is assured between the mask and the substrate by the positive pressure at the inlet tube 920. After the solder material has been reflowed, preferably after the solder balls have formed on the substrate, a negative pressure (vacuum) is applied to both of the inlet tubes 920 and 940 to hold the wafer 902 firmly to the chuck assembly 900 so that the mask 904 may be lifted off of (released from) the wafer 902.

An additional advantage of the chuck assembly 900 is that the wafer 902 is disposed upon a non-metallic film 930 which, in turn, is disposed upon a non-metallic membrane 914, both of which (930 and 914) serve as thermal barriers to isolate the thermal mass of the chuck base 910

from the wafer 902. Inasmuch as it is generally preferred to keep the thermal mass"seen"by the heater element to a minimum so that the solder material in the mask may efficiently be reflowed, this serves to reduce the effective thermal mass of the chuck assembly.

Examples Of Solder Materials And Mask Dimensions A suitable solder material for use with the present invention comprises"63/37"lead/tin solder having a melting temperature of approximately 183°C (Centigrade), and has relatively large particle sizes. Large solder particles are less likely to leak out of any gap (e. g. , 314) between the mask and the substrate being bumped. The following chart lists a number of exemplary dimensions and relationships between: D, the diameter of the desired resulting solder ball; W, the cross-dimension of the cell in the mask; T, the thickness of the mask; d, the particle size (e. g. , diameter); #, the approximate number of particles in a cell; and %, the final percentage of metal, by volume, in the cell. D (mils) W (mils) T (mils) d (mils) # % 4 6 3 1.5 18 31 5 7-8 4 2 15 28 10 12-13 8 4 37 42 20 25 15 5 63 44 Notes: 1. The pitch of the pads on the substrate being bumped is typically twice the diameter (D) of the resulting solder ball.

2. The size of a pad on the substrate being bumped is typically approximately equal to the diameter (D) of the resulting solder ball.

3. The final percentage (%) metal is determined without compression of solder material in the cell.

From the chart presented above, it is evident that: The cross-dimension (W) of a mask cell is always greater than the thickness (T) of the mask.

The solder material filling the cells in the mask preferably comprises solder particles which of a size (d) which is relatively"huge"in comparison to the cell cross dimension (W) or diameter (D) of the resulting solder ball. As is evident from the chart presented above, the dimension"d"

is at least approximately 20% of the dimension"W". And, the dimension"d"is at least approximately 25% of the diameter"D"of the resulting solder ball.

The solder material may comprises solder particles of a size (d) which is at least 10% of either the cross-dimension (W) of the mask cell or the diameter (D) of the resulting solder ball, including at least 20% of the cross-dimension (W) of the mask cell or which is at least 25% of the diameter (D) of the resulting solder ball. As compared to mask thickness (T), the smallest particle diameter (d) should be at least 40% of the mask thickness, including at least 50%.

An advantage of using"huge"solder particles in the solder material is that the particles will be less likely to"leak out"of any gap (e. g. , 314) between the mask and the substrate. A typical dimension for a gap between a mask and a substrate being ball-bumped, due to non-planarities in the substrate, may be on the order of 1-2 mils.

Another advantage of using"huge"solder particles is volume control, and increasing the percentage of solid material in each cell of the mask, so as to maximize resulting solder ball size. Using a typical solder paste, which is a homogeneous suspension of metal powder in a flux vehicle, the percent solid material is limited by the solder paste composition. In contrast thereto, huge particles, when forced into the cell, will displace flux, and may also compact (deform). In this manner, a surprisingly large volume percentage can be achieved.

It should also be understood that the solder particles in the solder material used to fill the cells in the masks are not necessarily spherical, in which case they would have a width or cross- dimension rather than a"diameter".

In the context of there being gaps between a mask carrying the solder material and a surface of the substrate being ball-bumped, the solid particles preferably exhibit a minimum diameter which is larger than the largest gap between the mask and the substrate.

A suitable solder material contains particles of lead/tin solder, in a matrix of flux, with the following proportions: 80% (by weight) solid material (e. g. , particles of lead/tin solder), and 20% (by weight) flux (including volatiles). In terms of relative volume percentages, the same

typical solder material may contain approximately 55% (by volume) of solid material (metal) and 45% (by volume) of flux.

A suitable solder material for use in being applied to a substrate and reflowed to form solder balls on the substrate has the following composition and characteristics: a plurality of solid particles of solder material suspended in a flux-material; the solid particles having diameters in the range of from approximately 1.5 mils to approximately 5.0 mils.

Preferably, the average size of the solder particles is such that they number (#) in the range of a few dozen to a few hundred solder particles filling each cell of the mask.

A principal advantage of the techniqe (s) described hereinabove is that since the solder ball has a diameter which exceeds the thickness of the mask and sticks out when reflowed, it can join itself to the (pads of the) substrate without there having been any contact between the mask and the (pads of the) substrate. Also, the mask can be removed while the solder is still molten, thereby greatly facilitating mask/substrate separation.

Although a solder material comprising solder particles and flux is described, the solder material may be dry, such as fluorine-treated, or using a forming or reducing gas.

The mask may be coated with a polymer such as photo-imageable polyimide or silicone rubber.

This will protect the substrate against damage if the mask is in contact with the substrate. The coating, if sufficiently thick, can also serve as a conforma mask mating to irregular surfaces, and improve the volume of solder per cell, and help release the substrate.

Many of the features discussed hereinabove can be"mixed and matched"with one another.

Other features are generally incompatible with one another-for example, it might be inapposite to have a biased chuck as in Figure 9 along with a bridging the gap embodiment as in Figure 5A. One having ordinary skill in the art to which the invention most nearly pertains will understand which features work well with (are compatible with) one another and which do not.

Developments, Improvements, Insights, Clarifications

and Further Embodiments Since the time of filing MACKAY-2 various insights have been gained, and developments either made or conceived of. Nevertheless, many of the features described hereinabove remain completely valid and effective. In the comments and embodiments that follow, there is substantial overlap with various portions of the disclosure of the parent application, and there are some new concepts. Unless specified otherwise, any matter that follows should not automatically be presumed to be new matter. In many cases, the descriptions that follow merely represent slight improvements or clarifications to what was already disclosed.

In the main hereinafter, substrates which are semiconductor wafers ("wafers") are discussed, but the invention is not limited to wafer substrates.

Captured Cell One of the distinguishing, and rather critical features of the invention over many of the prior art approaches is that the present invention uses"captured cell"technology. As described above, the cells can be closed by a pressure plate by the heater stage itself or by using a mask with blind holes. In the embodiments described hereinbelow, the cells of the mask are typically closed off by the heater stage itself, without a separate pressure plate.

Characteristics of the Mask The mask should have low thermal expansion, with holes which are etched rather than drilled.

This is applicable to masks that have cells which are either through holes, or which are blind holes.

Holding the mask tightly It was previously believed (see description of Figures 7A and 7B, above) that permitting the mask to expand freely, in one axis, would be the best way to alleviate problems associated with warpage (warping). A more preferred system for mounting the mask has been developed. The mask is, for example, a molybdenum sheet with holes. The mask is preferably mounted to a stainless steel (SS) mesh (screen) which is pre-tensioned on a disposable frame. The mask and SS mesh are glued together. Then the SS mesh is cut away from the center of the mask (this applies tension to the mask), where the holes (cells) are. One edge of the mask is directly attached. The opposite edge has approximately one inch (2.54 cm) of SS mesh between the

mask and the frame. This allows the mask to expand, and the SS mesh takes up the slack. This also allows the frame to change temperature without affecting mask tension. If all four sides of the mask were directly mounted to the frame, as the frame cooled, the mask could buckle (or "oil can").

Reducing forces In the embodiments described hereinbelow, printing (filling the mask cells) is mainly done"off- line", without the mask first being on the wafer. This is important in that it reduces the force required on the wafer from print blade forces, and reduces cell volume variations, as described in greater detail hereinbelow.

Capturing the cells The cells are closed by the heater stage itself. Magnets are disposed about the periphery of (located outside of) the mask frame and the heater stage to hold the heater stage to the mask frame. Also, after filling the cells with solder paste, the opposite side of the mask is closed by the wafer (as an example of a substrate being ball bumped). Magnets are disposed about the periphery of (located outside of) the chuck assembly to hold the chuck assembly to the mask frame. The mask frame slides into the carriage with land areas for magnets to contact. The carriage moves the mask and frame assembly to the chuck, etc. In this manner the force required on the wafer to maintain the captured cell can substantially be reduced, which has been found to be beneficial.

An Exemplary Machine and Process Flow Figure 10 illustrates an exemplary ball bumping machine 1000 having a base 1002, a chuck 1004 on the left side for holding a wafer 1006 and a heater stage 1008 on the right side. A mask 1010 is held in a frame 1012. The chuck 1004 is disposed in chuck base 1014. The heater stage 1008 is disposed in a heater stage base 1016.

An elongate shuttle (carriage) mechanism 1018 is pivotally attached to the base 1002 at a point "P"between the chuck 1004 and the heater stage 1008. The frame 1012 is held in a mask carrier 1020 which is attached to the opposite (free) end of the shuttle mechanism 1018. A motor 1021 controls the position of the shuttle mechanism 1018. The shuttle mechanism 1018 can shuttle the mask 1010 (i. e. , the carrier 1020) between the heater stage 1008 on the right side

(as shown) and the chuck 1004 on the left side. The shuttle mechanism 1016 pivots about the point"P". Cameras (not shown) are used to make alignments, for example of the mask 1010 to the wafer 1006.

A set of holddown magnets 1022, which preferably are electromagnets, selectively hold the chuck base 1014 to the machine base 1002. Similarly, a set of holddown magnets 1024, which preferably are electromagnets, selectively hold the heater stage base 1016 to the machine base 1002. The carrier 1020 is ferrous, or has ferrous"lands". A set of lift magnets 1026, which preferably are electromagnets, selectively hold the carrier 1020 to the heater stage base 1016.

Similarly, a set of lift magnets 1028, which preferably are electromagnets, selectively hold the carrier 1020 to the chuck base 1014.

In this manner, the mask can be brought down onto the heater stage, the magnets 1026 turned on, the magnets 1024 turned off, and the heater stage can be lifted by the shuttle mechanism 1016. In other words, when the mask is shuttled, it can take the heater stage with it. Similarly, the mask can be brought down onto the chuck, the magnets 1028 turned on, the magnets 1022 turned off, and the chuck can be lifted by the shuttle mechanism 1016.

A more detailed example of mask mounting is shown in Figure 10D where one can see the mask 1010 glued (mounted with an adhesive 1011) to a SS mesh 1013 in a frame 1012, as described hereinabove, and an area 1032 of cells 1034, as described hereinbelow.

Figure 10B illustrates an embodiment of a chuck assembly, according to the present invention in somewhat more detail than it was illustrated in Figure 10. This is similar to the chuck of Figure 9, but without some elements and with the addition of other elements. But the basic idea is the same-namely, to hold the wafer and be able to introduce pressure to flex it into intimate contact with (in this example) the printed mask.

What was shown as chuck assembly 1014 in Figure 10 can be seen to comprise an inner chuck base 1054 and an outer chuck base 1056. The outer chuck base 1056 sits atop the machine base 1002. The lift magnets 1058 (compare 1028) are located in the outer chuck base 1056. A wafer 1006 is shown disposed above everything, merely for illustrative perspective.

The inner chuck base 1054 is mounted on a set of legs 1062 within the outer chuck base 1056, and the legs allow the inner chuck 1054, hence the wafer 1006, to be raised or lowered by a stepper motor or other suitable actuator (not shown), as discussed above. An air cylinder 1064 provides pressure for flexing the wafer, as described hereinabove.

A vacuum line 1066 extends through various (three shown) insulating layers 1068 (three shown) to a manifold element 1070, for holding the wafer. The manifold element is suitably mica ceramic.

Figures 11A-11D illustrate an exemplary process flow, as described hereinafter. Various alignment steps are omitted from the description, as they will be well understood by the person of ordinary skill in the art to which the invention most nearly pertains.

In a first process step (Figure 11A), the mask 1010 is disposed on the heater stage 1008, and is secured (assembled) to the heater stage by turning on the lift magnets 1026. This is before the cells of the mask are filled with solder paste, and before the heater stage is heated up.

As best viewed in Figure 10A, the mask 1010 has an area 1032 (typically centrally located on the mask) that has cells 1034 extending completely therethrough. A groove 1030 is formed in the top surface of the heater stage, and preferably extends entirely around an area corresponding to the area 1032 of cells 1034. The groove 1030 communicates with an orifice 1036 which extends to (beyond) an outer surface of the heater stage. When vacuum is applied to the orifice, the mask 1010 is held firmly onto the heater stage 1008. The groove 1030 is preferably at least one inch (2.5 cm) away from (outside of) the area 1032 of cells 1034. It is preferred not to have the vacuum groove too close to the area of the cells so as to avoid the vacuum applied thereto exerting a suction on the molten solder paste (including flux) when the heater stage is heated up (as described hereinbelow). Since the heater stage is functioning as the pressure plate in capturing (closing off) the cells, it is important to maintain intimate contact with a mask having cells which are through holes and the surface of the heater stage, and to substantially prevent the mask from warping. The vacuum groove 1030 achieves this purpose, while allowing for some expansion and/or contraction of the mask 1010 without buckling.

Generally, blind hole masks (e. g. , 500) are not preferred, it having been found that to manufacture a blind hole mask is difficult with respect to maintaining uniform hole depth (hence, cell volume), particularly when etching is the preferred hole-making process (in favor of drilling). The vacuum groove 1030 in the heater stage makes a through-hole mask behave like a blind hole mask, in the sense that leakage between the mask and the heater stage (in the role of closing off the cells) is substantially eliminated.

The magnets 1026"assemble"the heater stage to the mask carrier so that the heater stage can shuttled along with the mask. The vacuum holds the mask to the heater stage, thereby capturing the cells on one side of the mask. These two features have the following benefits: - keeps solder paste from leaking under through-hole type mask - holds mask to ensure uniform heating or outer mask area - holds mask during extraction, keeps mask from warping In the case of a mask with cells which extend through the mask (as illustrated, and as preferred), any leakage between the mask and the heater stage will adversely affect the subsequent ball formation. A pressure plate may optionally be disposed between the heater stage and the mask, but is not necessary. With a blind hole type mask, the holes would be disposed away from the surface of the heater stage (e. g. , the"pressure plate portion"520 of the blind hole mask 500 would be against the surface of the heater stage), and leakage between the mask and the heater stage would not be an issue, but it is nevertheless important to maintain intimate contact between the heater stage and the mask. The mask is relatively thin (e. g. , 0.003 inches = 3 mils), and is therefore prone to warping, particularly when heated and constrained by a frame. The heater stage is relatively thick, and (in relative terms) not prone to warping. It is important in any case to maintain intimate surface-to-surface contact between the heater stage and the mask during not only the mask printing step (discussed hereinbelow), but throughout the entire process of forming solder balls, to avoid mask warping. Maintaining mask flatness (i. e. , avoiding mask warping) is very important to successful ball formation and yield (e. g. , avoidance of voids).

Meanwhile, as shown in Figure 11A, the wafer 1006 is loaded onto the chuck 1004, which is movable in the vertical axis, and the chuck may be positioned slightly (e. g. , 0.005 inches, 5 mils) below"contact height" (see, e. g. , the dashed line in Figure 10). Contact height is the

height at which the mask will contact the wafer, when shuttled over to the left, and it is simply a good idea to leave a small clearance between the mask and the wafer so that the mask can be positioned onto the wafer without mechanical interference. This step can be before, during or after the first process step of securing the mask to the heater stage.

About the"clearance", which is comparable to the"gap"described hereinabove. The clearance dimension of 5 mils is about 5 times as great as the average size of a typical 1 mil diameter solder particle filling a mask cell. The typical mask cell has a cross-dimension (diameter, in the case of a cylindrical cell) of approximately 4-10 mils.

Next, the mask is printed-in other words, the cells 1034 of the mask 1010 are filled with solder paste (not shown, see, e. g. , Figure 1B). This can be done in any suitable manner, so long as the cells of the mask are substantially and uniformly filled, and that there is substantially no excess solder paste on the surface of the mask. An exemplary technique for printing the mask is described hereinbelow, with respect to Figure 12.

It is preferred to print"off-line"-in other words, not on the wafer. If printing on the wafer (as described in the parent application), it must be appreciated that the surface of the wafer is often not very flat, topographically speaking. And this topography can lead to variations in the effective overall volume of a cell being filled with solder paste. As a general proposition, any variations in the process, from cell-to-cell, are simply not desirable. Hence, printing on a known flat surface-i. e. , the surface of the heater stage-is preferred. Also, by printing"off- line", the wafer is spared from the sometimes excessive forces required to get a good print (effective cell filling).

The heater stage is, of course, at this point in the process, substantially at"room temperature" (not heated). Else, flux in the solder paste in the cells of the mask would start to vaporize, etc. This represents a departure from many of the processes generally described in the parent application, where it was described to be desirable to have the heater stage preheated, at all times.

Next, as shown in Figure 11B, the assembly of the printed mask and the heater stage ("mask/stage", as held together by magnets 1026) are shuttled over to the wafer 1006 which is sitting on the chuck 1004.

Then, the lift magnets 1028 are turned on firmly secure ("assemble") the mask carrier to the chuck. This ensures that the chuck and wafer pads will maintain alignment to the mask holes during transfer. Then chuck can then be shuttled to the 135-degree (90 + 45 degree) position for reflowing the solder paste. (The 135 degree position is shown in Figure 10C.) The heater stage lift magnets 1026 and the chuck lift magnets 1028 are phased (poled) oppositely so that they do not cancel out when everything (heater stage-mask-chuck) is assembled together.

Although this step of contacting the wafer to the mask is shown with the wafer in the non- inverted position, it is within the scope of the invention that the wafer and chuck could be shuttled over to the mask/stage, or that both the wafer/chuck and mask/stage could be shuttled to an intermediate position/orientation.

Next, the chuck is pressurized-for example to approximately 3 psi. As described hereinabove, this will ensure positive intimate contact of the wafer with the mask. (This will also take up the 0. 005" clearance, mentioned above. ) This intimate contact is beneficial because: - it contains the solder paste during heat up; - keeps the mask from warping; and - ensures contact of the wafer pads with the balls which will be formed in the mask.

Next, the heater stage is heated up, according to a desired profile (temperature schedule). For example, the heater stage is first heated to approximately 150 degrees (C), which will activate the flux.

With the flux activated, the assembly of chuck/wafer/mask/stage may be shuttled to t nearly inverted position, such as 135 degrees (Figures 10C, 11C; compare Figure 4D). This is

advantageous for wafers having irregular topography, but is not necessary for wafers having relatively flat top surfaces.

Next, the temperature is increased sufficiently to reflow the solder paste and permit balls to form. For conventional 63/37 Pb/Sn eutectic, this is at least about 183 degrees (C). The preferred temperature for the described process is 195-200 degrees.

As shown in Figure 10C, when the solder paste reflows, it forms balls 1040 (compare 340) on pads 1044 (compare 304) of the substrate (wafer) 1006 (compare 302). The balls extend (grow) out of the cells 1024 of the mask 1006 and wet themselves to the pads of the wafer.

This semi-inverted orientation causes solder paste to be forced (by gravity) into a 90 degree corner (bottom left, as viewed) in each cell of the mask, allowing venting at the opposite corner (top right, as viewed).

Finally, the wafer is extracted after some time (dwell) at maximum (solder reflow) temperature. The pressure (e. g. , 3 psi) is turned off at the chuck, and the wafer is slowly pulled away from the mask. This is advantageously done before the re-flowed solder material has solidified, thereby facilitating mask removal. However, caution should be exercised with respect to slowly separating the mask from the wafer (or vice versa) to that air currents and/or suction are not created. For example, a separation speed of about 2 inches per second has been found reasonable.

A chamber is optionally formed between the chuck and the mask holder so that the atmosphere can be controlled, e. g., NO2.

After mask removal, the heater stage can be shuttled back (with the mask holder) to its original position, awaiting the next cycle.

Then, the mask can be moved to a neutral position for removal or cleaning.

Observations The molten solder ball remains in contact with the mask edges inside the aperture (cell), depending on the amount of interference. For example, a 0.004 ball inside a 0.003 mask will

have 0.001 interference. The ball flat will be located at some distance from the aperture wall at any rotational position. That means if the pad were to be skewed to one side of the aperture and the ball on the other side, a"miss"could occur (no copper pad in contact with liquid solder). Therefore, good initial alignment is very important.

Normally, the balls are formed on the wafer with the wafer uninverted-with the pads atop the wafer (rather than below, or"inverted", as discussed in detail in the parent application). The 135 degree partially-inverted scenario (Figure 10C) appears to only be required with wafers having high topography (deviating from flat). Other angles, such as between 20 or 30 degrees and 60 degrees are believed to be beneficial for partially-inverted. With highly planar (low topography) wafers, however, little difference is observed between reflowing partially- inverted and non-inverted.

During reflow, solder paste (particles of solder in flux) first outgases some flux, then the solder balls begin to shrink into a slug (no interference is observed when solder is a slug).

Then a complete melting and complete surface tension equilibrium causes interference and the liquid solder wets to the solid copper pad. This is the reason for having little to no voids in the solder pad interface. In an experimental bumping situation, only 0.4% of the pads had voids, and the voids were less than 5% of pad diameter.

A significant benefit accrues to printing the mask without the wafer being present. Normal print pressure is on the order of 60 psi, and this is a lot of pressure to subject a wafer to. By avoiding this, the only pressure exerted on the wafer is the 3 psi used to flex the wafer into intimate contact with the mask prior to reflow.

Printing (Filling the Mask Cells) Generally, any technique may be employed for filling the mask cells with solder paste, and the mask may be printed off-line. It has been found that printing a mask normally takes multiple passes with a single squeegee to ensure complete cell filling, without"gouging". And, it is also important that the surface of the mask be clean (free of excess solder paste) afterwards.

Figure 12 illustrates a beneficial technique for filling the cells of a mask with solder paste. It should be understood that this technique is not limited to ball-bumping. Nor is it limited to

solder paste. The technique is useful for filling any mask or stencil with any viscous material.

The technique is well suited to filling a mask with solder paste having a viscosity in the range of 20K-300K cps.

According to the inventive technique 1200, first a blob (glob, quantity) of solder paste 1202 is disposed on the surface of the mask 1210 (compare 1010). The mask 1210 is on a support surface 1208 (compare 1008). The support surface 1208 may be a wafer, for printing in place on a wafer, if so desired.

The mask 1210 has a plurality of cells 1234 which may be arranged in an array. The cells may be round, square or the like. The mask has a thickness, typically 3 mils. The cells are preferably, but not necessarily, uniform in size, hence volume. For example, a square cell may have a cross-dimension of 6 mils.

A print blade 1220 is brought to a distance of a few mils (e. g. , 5-10 mils) from the surface of the mask. It is preferred that the blade not drag across the mask. The print blade is advanced in the direction of the arrow 1222. As the first blade advances, the cells 1234 become filled with solder paste. Because the print blade is spaced from the mask there will inevitably be an amount (film) of excess solder paste on the surface of the mask behind (to the left of) the blade on the surface of the mask. The residual film has a thickness essentially equal to the liftoff distance of the print blade. Since the print blade is not in contact with the mask, the contact pressure is essentially zero. This can be important when the mask is supported on a delicate electronic component (when not printing off-line). The print blade 1220 may be of stainless steel or soft rubber (60 shore A hardness) rubber. A good example of a squeegee blade is a rubber blade made of 90 durometer ULON.

The leading edge angle of the print blade can be adjusted to help"roll"the paste downwards onto the mask. And, its flat bottom surface helps force the solder paste into the cells with less pressure, because the cells are not sealed off by the blade 1220.

A cleaning blade 1230 is disposed so as to contact the mask, and advanced in the direction of the arrow 1222. In essence, the metal blade follows behind the squeegee blade and performs "clean up duty". By way of example, for printing a mask for a 6 or 8 inch wafer, a the metal

blade follows behind the squeegee blade by a distance of one inch (2.5 cm). The end portion of the metal blade contacting the mask describes approximately a 45-60 degree angle with the surface of the mask. The cleaning blade 1230 may be of metal, such as stainless steel. Or, have a metal edge. A good example of a metal blade is a Permalex blade by Transition Automation SPK-PLX-1. 5-9.

Two main issued with the cleaning blade are length and cleaning. If the blade length is too short (e. g. , < 1.5 inch), the blade may be too stiff, resulting in past being dragged out of the cell behind the blade. Blade contact angle does not seem to matter as much as stiffness. The cleaning blade should be cleaned after each print cycle.

The benefit of the two blade system is that the print blade 1220 fills the cells without needing to clean the surface of the mask, and the cleaning blade 1230 cleans the surface of the mask without gouging the already filled cells. The print blade exerts no direct pressure on the mask (and underlying component, if any). The cleaning blade exerts very little pressure on the mask. Any suitable mechanism can be used to control the movement of the blades across the surface of the mask. The present technique is beneficial in that it enables filling the cells of a mask in only one pass.

Further Observations.

Heating through the mask is preferred. Substrate heating is minimized, and in the case of a gap between the substrate and the mask, heating through the substrate is not viable.

The"interference"method is preferred. This means that the mask (cell) opening and mask thickness is designed to produce a solder paste volume that will result in a reflowed ball diameter that exceeds the mask thickness. This forces the liquid solder ball to contact and wet the metal pad on the substrate (wafer). This appears to be a key reason why the solder ball can be attached without voids in the bump.

Warpage of the mask is to be avoided. It has been found that the lower the TCE of the mask, the less pressure is required to capture the warpage. Pure molybdenum masks are generally preferred because of their availability, and the availability of etching services, as well as its low thermal expansion (TCE) and metal wear and tear behavior. Polyimide masks are also suitable.

The use of water-soluble solder pastes is preferred. Rosin-based systems leave hard to remove residues which can be especially difficult to remove if allowed to cool before cleaning. They are also very tacky and exert large separation forces on the mask during wafer separation."No clean"fluxes are completely incompatible with this solder deposition method. The water clean family of pastes yields very good printing, fluxing and separation characteristics, as well as easy cleanup.

The orientation for reflowing shown in Figure 10C can be varied ("programmable"), as suited to a particular application. Reflow can be done inverted, partially inverted, non-inverted.

Wafer extraction after the solder balls have solidified ("solid extraction"), as described hereinabove, proved to be somewhat problematic. Solidified solder balls can tend to stick to the knife edges of the mask cell openings, making removal difficult and stressful on the substrate (e. g. , silicon wafer). It was thought that releasing mask pressure while the solder was still molten would allow the mask to warp, thereby damaging the resulting solder bumps (balls).

However, it was determined that the liquid solder bumps, once wetted to the substrate (wafer) pads, were very resistant to damage. Preferably, separation of the mask from the substrate is performed at the maximum solder temperature ("liquid separation"), resulting in less wear and tear on the masks, as well as improving solder ball surface finish. It is generally preferred to remove the wafer from the mask before the mask is removed from the heater stage. This is possibly due to wetting between the mask and the heater stage maintaining the mask flat during extraction.

Off-line printing is preferred, particularly in the case of delicate substrates such as wafers. In any case, printing on the heater stage (or contact plate, or any suitable planar surface), rather than on the substrate, lends more certainty to the process, resulting in the best cell volume consistency. This also eliminates printing forces at the wafer surface. However, printing on the wafer may be preferred for wafers with deeply recessed pads.

Although the invention has been illustrated and described in detail in the drawings and foregoing description, the same is to be considered as illustrative and not restrictive in character - it being understood that only preferred embodiments have been shown and described, and that all changes and modifications are desired to be protected.