RIKEN Review No. 43 (January, 2002): Focused on 2nd International Symposium on Laser Precision Microfabrication (LPM2001)

High-speed microvia formation with UVsolid state lasers

Corey Dunsky,∗ Hisashi Matsumoto, and Glenn Simenson

Advanced Packaging Products Division, Electro Scientific Industries, USA

Laser drilling has emerged in the last five years as the most widely accepted method of creating microviasin high-density electronic interconnect and chip packaging devices. Most commercially available laser drillingtools are currently based on one of two laser types: far-IR CO2 lasers and UV solid state lasers at 355 nm.While CO2 lasers are recognized for their high average power and drilling throughput, UV lasers are knownfor high precision material removal and their ability to drill the smallest vias, with diameters down to about25–30�m now achievable in production. This paper presents an historical overview of techniques for drillingmicrovias with UV solid state lasers. Blind and through via formation by percussion drilling, trepanning,spiralling, and image projection with a shaped beam are discussed. Advantages and range of applicabilityof each technique are summarized. Drivers of throughput scaling over the last five years are outlined andrepresentative current-generation performance is presented.

Electronic packaging and via drilling

In recent years, as the trend toward smaller, lighter, and morepowerful electronic products has continued, electronic circuitdesigners have been driven toward higher degrees of integra-tion. The number of devices in semiconductor integratedcircuits has continued its exponential growth, the numberof chips per product has increased steadily, and the num-ber of input/output signals per chip has grown significantly.In turn, the density of signal traces in electronic packagingschemes has been forced to keep pace. Indeed, the abilityto reliably and inexpensively manufacture high-density elec-tronic packaging has in many cases emerged as the bottleneckfor continued improvements in speed, functionality, and sizeof electronic products.

The term “packaging” refers to two broad classes of devices:integrated-circuit chip packages and high-density printedwiring boards. The former refers to the devices used to sup-port semiconductor chips, isolate them from the ambient, andprovide the means of routing signals on and off the silicon.Printed wiring boards (PWBs) refer to a broad class of de-vices which mechanically support the chip packages, routesignals between them and other, discrete electronic devices,and provide interface to the “outside world.”

Both classes of electronic packaging devices are constructedof multiple interleaved layers of conductive (metal) and di-electric (typically polymer) materials. Increasing the circuitdensity in the packages has proceeded by both increasing thelayer count and by decreasing the physical dimensions of cir-cuit features such as trace width, spacing between traces, andthe like. An important element of multilayer packaging archi-tecture is the layer-to-layer interconnect structures, which areformed by holes or vias drilled through the individual layersbefore lamination, or through the entire multilayer assemblyafter lamination. After drilling, the via walls are plated witha metal, typically copper, thereby forming the conductive Z-axis pathway.

∗ e-mail address: dunskyc@esi.com

Since the invention of PWBs in the 1950s, vias have beendrilled with high-speed mechanical drills. The drilling ma-chines have become increasingly sophisticated, fast, and ac-curate over the years and today the technology is relativelymature. Until recently, mechanical drilling has been a cost-effective solution as via diameters have shrunk. However, inthe last five years, as the drive toward higher circuit densityhas accelerated, there has been strong motivation to decreasevia diameter below the sizes attainable with mechanical drills.This is true because at each via site, each layer’s circuitrymust have a capture pad, larger in diameter than the viahole, the purpose of which is to provide physical connectionbetween the via and the circuit traces on each layer. Thecapture pad must be large enough to contain the via andaccommodate any errors in layer-to-layer alignment whichinevitably occur during the lamination process. It happensthat the size of the capture pad greatly affects the densityat which circuit traces can be routed in each layer: smallerpads mean more traces can occupy a given area. Since theareal circuit density in PWBs and chip packages is stronglydependent upon the capture pad size, and the capture padsize is largely driven by via diameter,1) there is a compellingdesire to shrink via diameters.

PWB via sizes have been in the range 250 to 400µm for sev-eral decades. Though early work in laser via formation wascarried out in the early 1990s,2–4) it was only with the ad-vent of High-Density Interconnect (HDI) circuit design rulesin the mid 1990’s that significant market demand developedfor much smaller vias. Mechanical drilling machines todayare capable of reliably drilling vias down to about 150µm di-ameter, and work is progressing toward 100µm. However,operational costs climb dramatically below about 200µm,and for so-called microvia drilling, where diameters are below150µm, the need developed for new technology.

In the late 1990’s, three technologies were competing for even-tual acceptance as the microvia drilling method of choice:plasma etching; use of photo-imageable dielectric materials;and lasers. Of the three, the first two have the advantage offorming all vias in each layer simultaneously, while the laser

80

Fig. 1. Evolution of microvia production technology.

process drills one via at a time. Despite this, by decade’s endlaser drilling had become established as the winner, with theother two methods relegated to niche applications (Fig. 1).

Laser via drilling

Today, laser microvia drilling is in worldwide volume produc-tion for chip packages and PWBs, with increasing fractionsof all packaging devices forecast to contain laser-drilled viaseach year throughout the next decade. Figure 2 shows thegrowth of the worldwide market for via drilling equipment.Though the laser technology has been widely implementedand has matured in the last five years, it is still young andthe rate of innovation remains high. In 1995, Electro Scien-tific Industries (Portland Oregon, USA) helped pioneer thelaser drilling market with a machine based on UV solid statelaser technology that was developed in house.5) At the time,two other companies were offering laser drills based on CO2

lasers.6) Early adopters quickly begin investigating the tech-nology7–9) and widespread acceptance followed within two tothree years. Today, approximately twenty companies world-wide are offering commercial laser drilling systems and themarket for this equipment continues to grow at a rate ex-ceeding the projection shown in Fig. 2.

Both UV solid states lasers and CO2 lasers continue to beused for via drilling. The former are primarily third-harmonicNd-doped designs (YAG and Vanadate) at 355 nm, while thelatter are a mix of RF-excited and TEA designs tuned to emitat wavelengths between 9.2 and 9.6µm. While CO2 laserdrills are recognized for their high average power and highdrilling throughput, UV lasers are known for high-precisionmaterial removal and their ability to drill the smallest vias,with diameters down to about 25–30µm now achieved in pro-duction.

Today, typical microvia diameters in HDI PWB designs arein the range 75–150µm, while higher-density chip packagesfeature vias in the 25–100µm range. As packages and devicesincrease in complexity, via sizes are expected to shrink in bothclasses of packages. Due to the long wavelength and shortdepth of focus associated with CO2 lasers, machines basedon this laser technology cannot drill vias as small as thoseachievable with UV wavelengths. Hence, it is recognized inthe industry that the high drilling throughputs that makeCO2 laser drills attractive today will give way to the small-hole capability of UV lasers over the long term. Further,recent increases in UV laser power yield drilling throughputs

Fig. 2. Growth of the via drilling equipment market.

competitive with those achieved with CO2 lasers. In the re-mainder of this paper, therefore, we concentrate on the UVlaser drilling technology.

UV laser technology

As areal via densities continue to increase, in order to main-tain and improve the cost-effectiveness of sophisticated pack-aging designs laser drilling speeds will have to increase apace.In 1995, typical UV drilling throughputs were 5–20 vias persecond. Today, throughputs achieved in similar materials are200–250 vias per second, and changes in the PWB manu-facturing sequence which eliminate the need to laser drillthrough copper layers permit throughputs of 450–600 vias/sto be realized. Numerous papers during the intervening yearsdocument the steady throughput increases and reliability en-hancements achieved in UV laser drills.10–13) These advanceshave been driven by improvements in laser power, beam po-sitioners, and optical design, and further improvements areon the way.

Figure 3 presents the historical progression of UV laser powerover the last five years. ESI’s 1996 arc-lamp-pumped third-harmonic YAG laser delivered well below 1 Watt to the worksurface. In contrast, new 355 nm lasers now commerciallyavailable generate up to 8 Watts, and power levels of 15 to20W are on the horizon.14) The Figure shows the advent ofUV Diode-Pumped Solid State (DPSS) lasers for via drillingin 1998. This marked the beginning of truly mass-production-qualified UV laser drills, equipment that does not requirefrequent tuning by skilled laser personnel.

Figure 3 also indicates that as third-harmonic laser designshave been scaled to higher powers, Nd:Vanadate has becomea desirable laser crystal in addition to Nd:YAG. The reasonis that as diode pump power has increased and cavity de-signs have been scaled up, Vanadate has offered a means ofobtaining a given pulse energy at significantly higher pulserepetition frequency (PRF). This provides the basis for scal-ing up drilling throughput by delivering a given number ofpulses at a particular pulse energy in a shorter time, a tech-nique we refer to as “PRF throughput scaling.” In 1996, forexample, a pulse energy of 150µJ could be delivered to thework surface at a PRF of about 3 kHz. By late 2001, thesame energy will be delivered at about 34 kHz. For 50µJ,the corresponding 1996 PRF was about 8 kHz whereas state-of-the art lasers in next-generation laser drills will deliver thisenergy to the work at over 70 kHz. The eight-to-tenfold in-

81

Fig. 3. Historical progress of UV solid state laser power for viadrilling.

crease in PRF at which a given pulse energy can be deliveredresults in a proportionate decrease in the drilling time pervia.

Beam positioner technology

The advances in laser power discussed above have necessi-tated corresponding increases in the speed of laser beam po-sitioners. Since the processing speed is determined by boththe time spent drilling each via and the time required to movebetween vias, both times must be driven down in order toincrease throughput. All commercial laser drills use a com-bination of two technologies to achieve fast, accurate beampositioning over a large area. XY tables or a pair of single-axis stages using linear motors provide the ability to movethe beam over standard-sized 500× 600mm PWB panels oreven larger areas. The stages, equipped with closed-loop po-sition sensing, move over this area at relatively low speedsand with accuracy on the order of 5µm. Stage speeds arelimited by maximum permissible accelerations on the orderof 1 g.

To move the beam from via to via at high frequency, an or-thogonal pair of galvanometer-driven mirrors is used to directthe beam to locations within a square area ranging from 10to 50mm on a side. A telecentric focusing lens maintains nor-mal beam incidence of the laser spot throughout the squarescan field. The galvo or scanner mirrors are also governedby closed-loop control electronics and maximum galvo mirroraccelerations are currently on the order of 300 g. Physicallimitations make it very difficult to push actuation frequencyabove about 1 kHz and mirror designs and materials are care-fully selected to maximize rotational speed and minimize po-sitioning errors. Via-to-via move time is a function of movelength, with short moves accomplished in a millisecond orless while moves of a few millimeters may take several ms.Between 1996 and 2000, minimum via-to-via move times inadvanced laser drills decreased from 4ms to 0.6ms.

Galvo actuation frequency is constrained by galvo mirror ma-terials, available high-speed error correction algorithms, andthe basic physics of precision high-speed motion. Hence, it isexpected that as UV laser power continues to increase, beam-

Fig. 4. Split-axis beam positioner architecture.

positioner limitations will emerge as the main rate-limitingfactor.

Figure 4 shows the combined stage+galvo scheme embod-ied in the split-axis design used in ESI’s laser drills. In thisconfiguration, the workpiece is placed on a vacuum chuckmounted on a linear stage moving along the Y axis. Abovethis, an X stage carries the “head” comprised of galvos, fo-cusing lens, and viewing cameras used for locating fiducialmarks on the workpiece panels. The lens and cameras aremounted on a Z-axis stage which permits focusing of boththe cameras and the laser beam. The Z-axis location of thetelecentric lens is also used as a process parameter for somedrilling processes, as discussed below. Other commercial laserdrill designs utilize fixed-path-length optics with a stationarylens and the workpiece fixtured to an XY table.

Beam motion is typically carried out by scanning the beamover the accessible scan field with the galvos, then using thestages to step the workpiece to an adjacent scan field and re-peating the process. Alternatively, the motion of the stagesand scanners can be coordinated so that both are moving con-tinuously. In this approach, stage movement is compensatedin real time by the galvos to maintain beam dwell at a viasite during drilling.15) This has a speed advantage in manycircumstances because the large mass of the stages or XY ta-ble does not have to be accelerated and decelerated after eachscan-field’s contents of vias are drilled. Details of the operat-ing principles and characteristics of this so-called compoundbeam positioner system have been presented in previous pub-lications.16)

Via formation processes

Beam positioner performance also comes into play whenprocessing certain multilayer constructions, namely those inwhich the top Cu layer must be drilled through. In this case,the laser beam must be focused tightly enough to achievesufficient fluence at the work for effective metal removal. At355 nm the laser/material interaction is purely thermal formetals and the material must be driven rapidly through themelt phase into the vapor phase prior to expulsion from thesurface by gasdynamic effects. If excess metal is heated onlyinto the liquid phase, a rim of undesirable re-solidified splashor slag is formed at the via periphery. When greater than a

82

Fig. 5. Percussion-drilled through-via array in 50-�m polyimide.

few microns high, this rim creates unacceptable conditions indownstream manufacturing processes such as sidewall platingand layer-to-layer lamination.

Due to these requirements, focused beam diameters at thework are generally in the range 10–20µm (1/e2), which cre-ates ablated spot sizes in the range 25–35µm. This permitsthe smallest vias in production today to be formed by per-cussion drilling, as shown in Fig. 5. For larger vias, fluencelimitations render percussion drilling infeasible; instead, thefocused laser spot must be moved within the via to sweep outthe desired area.

TrepanningThe most straightforward technique for forming larger viasinvolves moving the laser beam in a circular trajectory(Fig. 6). In the trepanning process, the tightly-focused laserspot cuts around the circumference of the via, allowing, inprinciple, any via size to be programmed. Typical trepannedvias, both blind and through holes, are shown in Fig. 7. Min-imum trepan time, typically achieved for vias smaller than75µm, has decreased from 6.0ms to 0.9ms in the last fouryears.

For blind vias, due to the beam trajectory shown in Fig. 6,when the programmed via size exceeds about 75µm an islandof un-ablated material can be left in the center of the hole.Therefore, for larger via sizes, another approach is required.

SpirallingFor larger via diameters, the intra-via motion necessarily be-comes more complex. As Fig. 8 shows, one solution is tomove the beam outward from the via center in a spiral mo-tion. This allows truly any via size to be created, althoughthe drilling throughput decreases with increasing via diam-eter. Programmable parameters here are the distance fromthe center at which the beam is initially applied; the numberof revolutions in the spiral (or equivalently, the radial pitchbetween spiral arms); the outer diameter at which the laseris turned off, and the number times the whole sequence is re-peated. With this scheme, complicated micron-scale motionscan be developed for tuning the material removal to partic-ular applications (material constructions and via sizes) andthe output characteristics of particular UV lasers. Resultsare similar to those shown in Fig. 7 for the trepan process.

In many PWB manufacturing situations, the laser drills a

Fig. 6. Beam motion in laser trepanning.

Fig. 7. Trepanned vias.

Fig. 8. Beam motion in laser spiralling.

multilayer construction in which an opening in the top cop-per layer has previously been formed by other means, suchas chemical etching. Here, the etched opening can be usedas a conformal mask and spiral parameters can be tuned toallow control over the slope of the via sidewall. The via inFig. 9 (a) was formed with a spiral that overfilled the etchedarea, whereas the via in Fig. 9 (b) was formed by underfill-ing the etched area. Figure 9 (c) shows a stepped-diametervia formed two layers down by using the copper planes asconformal masks.

Two-step processThe UV laser’s ability to effectively ablate copper makes it anattractive solution for many microvia drilling applications. Iteliminates the need for pre-etched openings in the top metallayer, and more importantly, can provide significant relieffrom layer-to-layer mis-registration occurring during the lam-ination process. If fiducial targets are precisely located on thebottom copper layer before drilling, machine coordinates can

83

Fig. 9. Spiralled vias. (a) and (b): taper control with pre-etched Cuopening as conformal mask. (c): Two-level via.

be set up to compensate for material distortions of the sec-ond layer. Then the vias can be drilled through the top metallayer and the intervening dielectric in locations keyed to theactual sites of the second-layer capture pads, with materialdistortions factored in. By not forcing via locations to thesites of the pre-etched layer-one openings, the ability to drillthe copper thereby relaxes the requirement for tight align-ment between metal layers one and two. This has becomeparticularly important in the past two years, as capture padsizes have shrunk to the point that layer-to-layer dimensionaloffsets can rival the size of the pads themselves.

The ability to drill copper, of course, means that measuresmust be taken to avoid laser damage to the bottom metal.This is accomplished by carrying out the via formation in twosteps. In the first step, the laser beam waist is placed at thework surface, achieving the high fluence required to removelayer 1 copper. In a second step, the beam waist is moved amillimeter or two away from the work surface, producing alow-fluence condition that is sufficient to ablate the organicdielectric but below the copper ablation threshold. Crosssections of vias after the first and second step are shown inFig. 10. Typically, step 1 is carried out with a spiral or trepanpattern while step 2 is conducted using either a trepan orpercussion drilling at much higher speed.

In order to maximize process speed, all vias on each largepanel are drilled at first-step conditions, then the focusinglens is moved and the vias are re-drilled with second-step con-ditions. State-of-the-art beam-positioner accuracy of ±20µmover a 500 × 600mm area is sufficient to ensure that beamplacement is aligned between steps 1 and 2. This two-pass ap-proach eliminates slow actuation of the focusing lens at eachvia site. However, it dictates that the inter-via moves be re-peated twice on each panel. As per-panel via counts reachinto the hundreds of thousands, total via-to-via move timecan be many minutes. An alternative to the two-pass tech-nique that would reduce total processing time is to changethe fluence between the high copper-cutting level and thelower dielectric-removing level at each via site. This elimi-nates one of the two sets of inter-via moves. The conceptcan be realized by either decreasing pulse energy or increas-ing spot size, or both. The former approach, which can beachieved by increasing laser PRF, suffers from lack of flex-

Fig. 10. Two-step process.

ibility and increases the step-two drilling time. The latterapproach requires very fast changes in the optical configura-tion. Hardware toward this end is currently in development.

Percussion drilling with aperture image projection

Despite the flexibility offered by the two-step spiral andtrepan processes described above, percussion drilling is moredesirable in many instances, particularly as advances in UVlaser power have enabled faster delivery of the energy requiredto drill each via. This is true because the extremely preciseand fast intra-via movement of the focused laser spot taxesbeam positioning technology, imposing difficult engineeringtradeoffs when attempting to drill faster. Percussion drillingeliminates the need for precise beam motion within the via.The high-speed beam positioner is thus only called upon toprovide via-to-via moves, eliminating the small-radius curvedpathways and attendant high accelerations associated withinside-the-via motions.

Percussion drilling with the focused Gaussian laser beam,however, is subject to a number of drawbacks for volume pro-duction implementation. Laser-to-laser variations in beamsize, roundness, astigmatism, and spatial mode content leadto variations of the ablated spot size and shape among laserdrills. On the production floor, this can result in the needto tune process parameters for each drill, increasing the levelof engineering involvement in routine operations. Further, asvia sizes continue to shrink, the sensitivity of the final viageometry to variations in beam characteristics becomes moresevere.

The excimer-based laser micromachining community has longimplemented image-projection techniques that mitigate theseproblems; the approach has been incorporated into solid-stateUV laser drills within the last three years. The beam is passedthrough a round aperture and an image of the aperture isprojected onto the work surface. Because the roundness ofthe aperture can be tightly controlled and the imaging opticscreate a crisp, faithful image of the aperture, the laser spotshape is excellent and system-to-system repeatability is de-coupled from laser-to-laser beam variation.

As an additional advantage, image projection reduces thenumber of process parameters to be optimized. Process de-velopment is therefore simpler, consisting of finding the bestcombinations of only two or three parameters. One simplyfinds the laser PRF that gives the best drilling fluence and thenumber of pulses required to open the via bottom diameterto the desired size.

A disadvantage of the imaging configuration is its more lim-ited applicability. Since the fluence in the imaged spot is sig-

84

Fig. 11. Schematic of imaged-aperture optics train. Beam shaperand imaging optics are mounted on a removable module. Beamshaper may be removed for clipped Gaussian imaging.

nificantly lower than the peak fluence in the focused Gaussianbeam, with today’s lasers the imaged beam can drill throughcopper for only the smallest vias. However, the techniqueproduces high-quality blind and through vias in a wide vari-ety of multilayer constructions without an overlying copperlayer.17)

Clipped Gaussian imagingTwo versions of projection imaging for via drilling have beendeveloped, as shown schematically in Fig. 11. In the simplerapproach, the UV laser’s Gaussian irradiance profile is used.In this case, the aperture masks the low-energy “wings” of thebeam, permitting only the central, high-intensity portions topass. A tradeoff between quality and speed emerges in thisso-called Clipped Gaussian configuration. This arrangementcan be set up so that varying fractions of the Gaussian beamare blocked by the aperture. If the Gaussian profile is highlyclipped so that only a small central portion is allowed to passthrough the aperture, then the irradiance profile imaged ontothe work surface will be more nearly uniform. This results inbetter control of the layer 2 copper conditions for blind vias,but that comes at the expense of rejecting a large fractionof the energy at the aperture mask. Decreased drilling speedresults.

If, on the other hand, a large fraction of the beam energyis permitted to pass through the aperture, then higher aver-age fluence is delivered to the work. However, the differencebetween the irradiance at the spot center, Ic, and the spotedges, Ie, will be large. At high transmission levels, the differ-ence between Ic and Ie leads to more difficulty in controllingdamage to the layer 2 copper: the center of the imaged spotcan exceed the fluence at which copper melts while the spotedges remain relatively cool.

The advantage to high aperture transmission though, isthat throughput increases with increasing aperture transmis-sion. Hence, there is a hard tradeoff between via qualityand drilling speed when using the clipped Gaussian imagingconfiguration. Increasing the transmission level to enhancethroughput results in a high differential in fluence betweenspot center and edge, either melting the layer 2 copper in thevia center or leaving the spot periphery starved for fluence.Reducing the pulse energy to the point where no copper melt-ing occurs in the center significantly decreases throughput.

Shaped beam imagingTo overcome this tradeoff, in an enhancement to the imaging

Fig. 12. Shaped beam irradiance profiles at the aperture plane.

technique, the laser beam’s natural Gaussian irradiance pro-file is transformed to a near-uniform “tophat” profile at theplane of the aperture, using beam shaping optics. Figure 12shows the irradiance distribution for the so-called ShapedBeam, both with and without the aperture.

By flattening the profile, the processing tradeoffs indicatedabove can be avoided. A high fraction of the beam energycan be passed through the aperture and delivered to the worksurface without a large difference in fluence between the cen-ter and edges of the imaged spot. This shaped, imaged beamremoves dielectric material uniformly across the via area. Itthereby permits the drilling process to be maintained at highfluence— and hence high speed— without creating undesir-able inner layer copper damage at the center of the imagedspot.

Further, by increasing the fluence at the periphery of the im-aged spot, the Shaped Beam clears the dielectric materialfrom the via edges with fewer pulses, significantly increas-ing the drilling throughput over that of the imaged ClippedGaussian beam, at the same time as it improves via qual-ity. Figure 13 compares throughputs achieved in 1999 withboth imaging techniques as well as the trepanning technique.Throughputs have increased significantly since then, but thefigure reflects the relative speeds of the two imaging processesas a function laser PRF. The two-step focused raw beam pro-cesses, in contrast, have increased in speed at a slower ratethan the imaging processes since 1999, since they are oftenlimited by the beam positioner rather than the available laserpower.

The curves in Fig. 13 exhibit maxima due to the competingeffects of PRF and pulse energy. As PRF increases, pulseenergy decreases, so more pulses are required to drill the via.

However, the pulses are delivered at a higher rate. This de-pendence of drilling speed on laser PRF is characteristic: aslaser power has increased in the last several years, the max-ima in the curves have shifted to higher PRFs, but are alwaysobserved.

With the uniform-fluence imaged spot, the layer 2 copper con-dition can be precisely controlled. Figure 14 shows a sequenceof vias drilled at varying laser PRFs and hence work-surfacefluences. The degree of layer 2 copper damage can be variedby simply changing the laser pulse repetition rate. Maximumthroughput occurs at a PRF (13 kHz) yielding moderate re-flow of the bottom copper. At PRFs producing little or no

85

Fig. 13. Throughput curves for two imaging techniques. The shapedbeam is about 25% faster than the clipped Gaussian.

re-flow (17–20 kHz), the throughput drops from the maxi-mum by 10–15%. Process engineers can thus optimize theprocess results for speed or bottom copper quality by choos-ing laser PRF; in any case the compromise between the two isnot nearly as severe as it is with the clipped Gaussian imagingtechnique (Fig. 13).

Figure 15 shows vias drilled in mid 2000 with the shaped,imaged beam. The vanadate laser used in this benchmarktesting had a specified maximum power of 4.5 Watts. For a62µm via, throughput is higher than that shown in Fig. 14,even though the resin is about 1.6x thicker. The latter weregenerated about 12 months earlier with a 3 Watt YAG laser.

Figure 16 shows further drilling performance data obtainedwith the 4.5W Vanadate laser. Here, we plot drilling time inmilliseconds as a function of laser PRF and the curves exhibitminima for the reasons discussed above. Note that highestspeeds (minimum drill times) occur with this more powerfullaser at significantly higher PRFs than the data shown inFig. 14. As the via size decreases, Fig. 16 shows that the drilltime decreases and the minimum shifts to higher PRFs. Thisoccurs because optimal results, both in terms of throughputand via quality, occur within a relatively narrow range ofwork-surface fluences. Therefore, as the size of the imagedspot shrinks, the pulse energy needed for fastest results alsodecreases and can be delivered at higher PRF.

Figure 17 shows recently obtained via-geometry statistics.These 50-µm vias were drilled in less than 0.8ms with the lat-est generation laser as of August, 2001 (compare Fig. 16). Toassemble this data set, 100 vias were sampled from each of sixpanels drilled every four hours during a continuous 24-hourrun. Each panel had 267,840 vias drilled over its 300×300mmarea; the 600-via data set was therefore sampled from over1.6 million vias. Measurements were carried out with an op-tical profilometer. This automated instrument fits an ellipseto the via top and bottom edges with sub-micron accuracyand 100-nm repeatability. Computed via diameters are theaverage of the ellipse major and minor axes and roundness isthe ratio of the two. Six-sigma variation falls less than 4µmfrom the mean, and typical roundness is greater than 95%.More extensive sampling of the 267,840 vias on one of the sixpanels indicated similar levels of process consistency.

Fig. 14. Dependence of throughput and bottom Cu condition onlaser PRF. Application is a 62�m via in 45�m epoxy resin.

Conclusion

Within the last five years, drilling of microvias in advancedelectronic packaging devices has become a widely commer-cialized laser precision microfabrication process. Steady ad-vances in drilling throughput, via quality, and process robust-ness have been driven by the advent of reliable, diode-pumped

86

Fig. 15. Vias drilled with 4.5W Vanadate laser and shaped/imagedbeam. Throughputs assume a 1.0-ms via-to-via move time.

Fig. 16. Shaped-beam drilling speed data from mid 2000. In com-parison, current-generation (mid-2001) lasers (Fig. 1) can drill50�m vias in under 1.0 ms.

solid state UV lasers, advances in beam positioner technol-ogy, and new optical designs. An example of the latter is theshaped, imaged beam, which offers superior via quality anddrilling throughput for substrates without outer layer copper.

In this paper, we have presented an overview of UV laserdrilling hardware and processing methods and have describedtwo variants of the image projection technique and the ben-efits of using beam shaping in laser via drilling. With theshaped-beam imaging configuration, tradeoffs between pro-cess speed and quality are less severe and throughput is about25% higher than achieved without the beam shaping optics.Maintaining the edges of the imaged spot at high fluence en-ables the dielectric material to be cleared from the via edgeswith fewer laser pulses. This is the key factor in the abilityof the Shaped Beam to increase drilling throughput over theClipped Gaussian.

The near future promises continued advances in equipmentand techniques. Process throughput scale-up will continue

Fig. 17. Via size and roundness statistics obtained during a 24-hrprocess stability test.

through higher laser power, low-level parallelism in the formof multi-headed machines, and to a lesser extent, faster beampositioning.

Other key areas of improvement will include enhanced au-tomation for system configuration and workpiece handling;increased process diagnostics and real-time control; and, ascapture pads continue to shrink, improved via placement ac-curacy, with position errors approaching ten parts per millionrelative to workpiece dimensions.

References1) S. Raman and A. F. Schreiner: Proc. IPC Printed Circuits

Expo Tech Conf., Long Beach, USA, 1998-10 (1998).2) A. Kestenbaum et al.: IEEE Transactions CHMT, 13 82

(1990).3) K. Arai et al.: Proc. PC World Conf. (1992).4) T. Tessier and G. Chandler: IEEE Transactions CHMT 16

43 (1993).5) M. Owen et al.: Proc. PC World Conf. (1996).6) B. Gu: Proc. ICALEO (1995), p. 351.7) J. Tourne: Proc. PC World Conf. (1996).8) D. Powell and M. Weinhold: Proc. PC World Conf. (1996).9) D. B. Noddin: Proc. ECTC (1998), p. 822.

10) C. A. Pico et al.: Proc. FLEXCONTM (1995), p. 152.

87

11) A. Cable: Proc. FLEXCONTM (1996).12) A. Cable: Proc. IPC Works ’97, High Density Interconnects

Conf., Oct. 5-9 (1997).13) S. Raman and A. Cable: Proc. IPC Works ’98, High Density

Interconnects Conf., Nov. (1998).14) N. Savage: OE Magazine 1, 46 (2001).

15) D. Cutler, B. Pailthorpe, and M. Unrath: Multi-Stage ToolPositioning System (1998).

16) A. Cable: Proc. SemiCon Singapore, May 4-6 (1999).17) C. Dunsky, H. Matsumoto, and B. Larson: Proc. IPC Printed

Circuits Expo Tech. Conf., San Diego, USA, April 4-6 (2000).

88