Scanned Mask Imaging Solid State Laser Tool for Cost Effective Flip

Scanned Mask Imaging Solid State Laser Tool for Cost Effective Flip
JLMN-Journal of Laser Micro/Nanoengineering Vol. 10, No. 1, 2015
Technical Communication
Scanned Mask Imaging Solid State Laser Tool for Cost
Effective Flip Chip – Chip Scale Package Manufacture
David T. E. Myles*1,2, Munya Ziyenge*1 , Jonathan D. Shephard *2 and David C. Milne*1
*1
M-Solv Ltd, Oxonian Park, Langford Locks, Kidlington, Oxford, UK, OX5 1FP
E-mail: Dave.Myles@M-Solv.com
*2
Institute of Photonics and Quantum Sciences Heriot-Watt University Edinburgh, UK, EH14 4AS
The IC packaging industry is now being driven by mobile devices, a market where cost and size
are key. Going to finer line widths and spacings allows a reduction in the number of layers making
up a multilayer chip package, giving a reduction in the cost and height profile of the device, as well
as improved signal latency. Embedding conductors within a dielectric film makes it possible to plate
to the required thickness without lateral growth of the traces. Using an ablative laser process to do
this avoids the financial and environmental costs of lithographic processes. A method for the 3D
structuring of dielectric films with pads, traces and vias with a resolution down to 2µm is described.
The set up uses a frequency tripled solid state laser to raster scan a binary mask, which is subsequently imaged onto a substrate with a maximum image field of 20x20mm. This offers performance
and cost advantages over alternative methods.
DOI: 10.2961/jlmn.2015.01.0019
Keywords: Mask imaging, ablative, microstructuring, micromachining, 355nm, chip packaging,
embedded, solid state laser.
resulting in capture pads which are twice as large as the
vias they capture. By micromachining the vias in the same
process step as the capture pads, the registration of the two
features can be as accurate as ±1µm. Having nearly landless microvias gives significantly more routing space per
layer, reducing the number of layers required which reduces both cost and signal latency.
Ablating features into a dielectric and subsequently
plating within these features to embed the conductor improves adhesion of small traces and reduces the chance of
copper bridging in the plating process, improving yield.
The dielectric also confines the plating to the feature,
meaning the trace can be plated to the required conductor
thickness without any lateral growth of the conductor.
Given the multitude of advantages presented by ablative laser processes in patterning dielectrics for advanced
packaging applications, this paper briefly discusses some of
the existing technologies for fine scale ablation of dielectrics. M-Solv’s patented Scanned Mask Imaging (SMI)
technology [6] is then introduced as a novel alternative,
which combines the benefits of multiple competing technologies.
1. Introduction
The semiconductor industry continues to meet the targets set by International Technology Roadmap for Semiconductors (ITRS) [1]. The targets follow Moore’s law,
which predicts that the number of transistors on a chip will
double every 2 years. With feature sizes down to 22 nm on
commercially available Integrated Circuits (ICs), the feature density on these semiconductor devices is rapidly increasing. For a given circuit design, the pitch of the input
and output (I/O) connection pads must therefore decrease
resulting in a higher I/O density. On top of this, Shannon’s
law of circuit complexity [2] predicts that circuit architecture complexity increases at a faster rate than the transistor
number predicted by Moore’s law. In combination with
Rent’s Rule [3], which describes a power law relationship
between the number of I/Os on a circuit and the circuit
complexity, these devices require increasingly higher density interconnects [4]. There is a need for increasingly more
complex interconnections with more input/outputs at finer
pitches with improved signal integrity and reduced signal
latency, all at reduced costs. This has driven research into
alternative manufacturing technologies that might bridge
the interconnection gap between sub-micron scale ICs and
millimetre scale Printed Circuit Boards (PCBs).
For economic reasons, the IC packaging industry has
preferred to incrementally advance photolithographic PCB
manufacturing technologies to date, and these are now
reaching their limits for cost effective IC package manufacture [5]. The vias which link the layers in the multilayer
packages are laser drilled prior to the multistep, lithographic processes used to form conductive pathways on the surface of substrates. As well as being a multistep process, this
has the additional disadvantage of requiring accurate registration of the features in the two process steps, typically
2. Current ablative laser technologies
Two ablative laser processes are the most obvious contenders to be used to structure dielectrics for advanced
package applications: excimer laser mask projection tools
and frequency tripled solid state laser direct write tools [7].
Both systems use light in the ultraviolet region which improves imaging resolution in the former, and the minimum
focal spot size in the latter. The high energy photons from
both lasers are strongly absorbed in the electron absorption
bands of many polymers, resulting in absorption lengths
typically shorter than a few microns. This reduces the
106
JLMN-Journal of Laser Micro/Nanoengineering Vol. 10, No. 1, 2015
fluence required for ablation, and limits the size of the heat
affected zone.
smaller than the feature size. These systems are therefore
not usually suited to high volume production in advanced
packaging applications where high densities and complex
pattern designs with large ground plane regions are a requisite.
2.1 Excimer laser processing
Excimer lasers offer the highest average power in the
ultraviolet region. Since throughput of a micromachining
tool scales approximately with average power reaching the
substrate, excimer laser systems can offer high throughput.
In an industrial environment, KrF and XeCl lasers emitting
radiation at 248nm and 308nm respectively offer the best
value in terms of cost per watt. However, the cost of ownership of excimer laser systems remains high compared to
solid state lasers due to the short lifetime of electrical components and optics, and the infrastructure needed to handle
the gases used.
Excimer lasers produce a highly multimode beam ideal
for imaging. In micromachining applications, average powers of a few hundred watts are achievable with repetition
rates of a few hundred hertz, giving pulse energies up to
around one joule. A typical system is comprised of beam
shaping and homogenizing optics used to illuminate a mask
which is subsequently imaged onto the substrate by a projection lens. This approach can achieve feature resolutions
down to a few microns, with good depth uniformity and
depth control. The etch rate of the many polymer films at 1
J cm-2 is of the order of 1µm per shot. Lowering the fluence
enables better depth control, but typically results in more
tapered side walls in the ablation. Excimer systems have a
process time independent of the pattern complexity, which
makes them ideal for the high density routing layers required in the next generation of advanced chip packages.
3. Scanned mask imaging solid state laser optics
The SMI process uses a frequency tripled, Q-switched,
multimode, solid state laser. Commercial models can have
average powers of up to 180W in a single cavity, with typical pulse lengths of tens to hundreds of nanoseconds. Manufacturers optimize the cavity for use at a particular repetition rate, often around 10kHz. This leads to pulse energies
in the millijoule regime, so clearly to obtain the same
fluence as an excimer laser process, a much smaller spot is
required. M-Solv’s solution is to raster scan a smaller beam
with a flat top profile across a binary mask, which is subsequently imaged onto the substrate. This gives imaging resolution and throughput similar to that of an excimer system,
with the much lower cost of ownership of a solid state laser
tool.
Flat top
plane
Homogeniser
Scanner
Singlet
Singlet
Telecentric
F-theta Lens
Mask
2.2 Frequency tripled solid state laser processing
Frequency tripled solid state lasers offer a lower cost
solution. Comparatively, they have much lower maintenance costs and a lower purchase cost per watt of laser
power. They can have comparatively good beam quality,
with M2 values close to 1, making them ideal for focusing
to a small spot. Such systems are typically comprised of a
pulsed or quasi-CW laser, beam expander, galvanometer
scan head to deflect the beam across the substrate and an ftheta scan lens such that the beam remains in focus across
the field of the scanner.
In this approach the pattern is defined by a CAD/CAM
file offering more flexibility than mask projection systems,
and fast on the fly changes to circuit design. Also, it does
not suffer from printed in defects: errors defined by a defect in the mask which are repeated in every device. In an
ablative process, the feature resolution is defined largely by
the size of the focal spot, the laser-material interaction and
the beam speed across the substrate. The best feature resolution achievable is therefore dependent on a number of
factors but is of the order of 10µm line width and space [8].
Flexibility in circuit design makes these systems ideal
for low volume prototyping [9], however they are limited
by process times dependent on pattern complexity, design
rules to keep beam velocity constant across the substrate
and some limitations in the features they can process.
Complicated control systems can be used to overcome
some of these issues, however it remains challenging and
time inefficient to ablate large features to a uniform depth
by overlapping scribes made by a focal spot size much
Beam profile at mask
Projection
lens
Image/substrate
Fig. 1 System schematic of M-Solv’s scanned mask imaging
technology.
A schematic of the optical system used for SMI can be
seen in Fig. 1. The laser beam is expanded using a
Keplerian telescope prior to the homogenizer. The beam is
then passed through the homogenizing element, which
forms a square, flat top plane at the beam waist of a planoconvex singlet. The homogenous plane formed is then imaged onto the mask using an infinity imaging system consisting of a second singlet and an f-theta scan lens, which is
mounted to the scanner. The divergence of the beam after
the beam waist at the mask is defined by the homogenizer,
the first singlet and the lenses in the infinity imaging system.
Care must be taken to design the system such that the
aperture of the scanner and the entrance pupil of the f-theta
lens are not overfilled, but that the divergence at the mask
is high enough to obtain the desired numerical aperture on
the object side of the projection lens to maximize the lens
resolution.
107
JLMN-Journal of Laser Micro/Nanoengineering Vol. 10, No. 1, 2015
The spot size at the mask should be such that the
fluence is below the damage threshold of the binary mask.
The results presented in the following section were obtained using a chrome on quartz mask, and the damage
threshold was assumed to be similar to the excimer laser
damage threshold: 100 mJ cm-2 [10]. The magnification of
the projection lens should be selected to give the desired
fluence at the substrate. A demagnification of 3.5 was chosen for the current lens, offering a good compromise between high resolution and fluence at the substrate, whilst
limiting the mask size. The mask is then imaged onto the
substrate by the projection lens.
By using a telecentric f-theta lens, light is incident approximately normal to the mask, limiting the size of the
first lens element required in the projection lens. An advantage of using a galvanometer scan head to deflect the
beam across the mask is that there is complete flexibility in
how the mask is raster scanned. This allows some control
over the time between consecutive shots on a given part of
the substrate, which can affect the thermal loading of the
substrate and the size of heat affected zones.
Illuminating the mask with a flat top, homogenous
beam offers several advantages over the approximately
Gaussian multimode beam profile from the laser. It lowers
the peak fluence of the beam, reducing the risk of damaging the mask, as well as reducing the percentage of the
beam with a fluence below the ablation threshold of the
substrate. It also reduces the coherence length of the beam
which is beneficial for imaging as it avoids the appearance
of diffraction fringes at the edge of ablated features in the
substrate [11]. The beam profile at the mask can be seen in
fig. 1.
By mounting the mask on a stage, it is possible to precisely overlay two images at the substrate to create 3D
structures in the material. This enables the drilling of
microvias in the same process step as the patterning of the
routing layer, facilitating landless pads. The accuracy of the
registration of the two images is dependent on the mask
alignment to the mask stage, the accuracy of the stage itself
and the magnification of the imaging system, with registrations better than ±5µm at the substrate easily achievable.
and the etch rates determined to be 0.50±0.01 µm/shot and
0.78±0.04 µm/shot for Kapton® and Ultimax™ respectively.
Fig. 2 Graph showing the etch rate of Kapton® and Ultimax™
with a fluence of 0.97±0.04 J cm-2 in the SMI system.
Fig. 3 Optical microscope image of the word “M-Solv”
ablated into Kapton® using the SMI process. The minimum line width is 4±1µm.
Fig. 3 shows an optical microscope (OM) image of the
word “M-Solv” ablated into Kapton® where the widths of
the lines making up the characters are 4±1µm. The mask
was raster scanned with a mark speed and pitch such that
each area received 25 shots, which gave an ablated depth of
10±2 µm.
Fig. 4 shows an OM image and SEM micrograph of a
resolution chart ablated into Kapton® using the 25 shots
per area mask scan, demonstrating the resolution of the
SMI system. The chart includes a 1 µm feature at the substrate which is not fully resolved by the projection lens.
The first set of 3 lines visible in the figure is 2 µm lines
separated by 2 µm at the edge of the lenses resolution limit.
This resolution is similar to that obtained with commercially available ablative excimer laser systems. These lines are
followed by sets of 3 lines with line widths and spacings of
3 µm, 4 µm, 5 µm, 10 µm and 15 µm.
The 3D structures in fig. 5 were micromachined to two
different depths by overlaying the image of two areas on
the mask. The image shows an 80 µm via machined down
to the copper clad laminate, and a capture pad and 16 µm
traces machined only part way through the Ultimax™ dielectric film. The features in the image are well registered
4. Scanned mask imaging results
In these preliminary studies, the laser ablation of
Kapton® (a polyimide film by DuPont™), and Ultimax™
(a dry film by Taiyo Ink) on a copper clad laminate, were
investigated. The Ultimax™ material is designed specifically as a build up material for flip chip applications. The
spot size at the mask was measured using a CCD beam
profiler. The power was then measured using a thermopile
power meter at a range of frequencies and the frequency
tuned to obtain a fluence of 79±3 mJ cm-2 at the mask.
With a lens demagnification of 3.5, this gives a fluence of
0.97±0.04 J cm-2 at the substrate. Increasing numbers of
shots were fired at different sites on each substrate to determine the etch rate at the specified fluence.
The depth of the ablated crater was measured using a
white light interferometer. The error in the depth measurements of the Ultimax™ substrate are significantly greater
due to an approximately 2µm peak to trough periodic variation in the height of the film propagated from the glass fibre weave in the laminate. The results are shown in fig. 2,
108
JLMN-Journal of Laser Micro/Nanoengineering Vol. 10, No. 1, 2015
with each other, highlighting the possibility of shrinking
the capture pad to create more routing space in a layer.
strated in commercially available thin films currently used
in the electronics industry. Excellent registration of vias to
pads was achieved, facilitating near landless pads and creating additional routing space. This SMI technology represents a highly cost effective solution to the demanding requirements of the advanced packaging industry with estimated cost of ownership less than half that of an excimer
laser based system.
Fig. 5 SEM micrograph of an 80 µm via machined through
the 20 µm Ultimax™ film to the copper clad laminate beneath. The image also shows the capture pad, 16µm signal
traces and a large uniform ground plane machined 8 µm deep
in the film and well registered to the via.
References
[1] "International Technology Roadmap for Semiconductors" (2012), retrieved http://www.public.itrs.net/.
[2] C. E. Shannon: Bell System Technical Journal, 28,
(1949) 59.
[3] E. F. Rent: IBM Journal of Research and Development, 49, (2005) 777.
[4] E. Beyne: Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers,
Detectors and Associated Equipment, 509, (2003) 191.
[5] R. Huemoeller, S. Rusli, S. Chiang, T. Y. Chen, D.
Baron, L. Brandt and B. Roelfs: “Unveiling the Next
Generation in Substrate Technology” white paper by
Amkor, Unimicron and Atotech (2007) retrieved
http://www.amkor.com/go/packaging/substratetechnology
[6] D. Milne, P. Rumsby and D. Myles: International
Patent WO 2014/068274 A1 (2014).
[7] A. Dietzel, J. van den Brand, J. Vanfleteren, W.
Christiaens, E. Bosman and J. De Baets: “Ultra-Thin
Chip Technology and Applications” ed. By J.
Burghartz (Springer, New York, 2011) p141.
[8] K. C. Yung and B. Zhang: Applied Physics A, 101
(2010) 385
[9] M. R. Nowak, A. J. Antończak, P. E. Kozioł, and K.
M. Abramski: Opto-Electronics Review, 21 (2013)
320.
[10] P. Rumsby, E. Harvey, D. Thomas and N. Rizvi:
Microelectronic Packaging and Laser Processing, 3184
(1997) 176.
[11] D. Bäuerle: “Laser Processing and Chemistry 4th
ed.” (Springer-Verlag, Berlin, 2011) p.239.
Fig. 4 OM photograph and SEM micrograph of resolution
chart ablated into Kapton® using the SMI optical system
with minimum line widths and spacings of 2 µm.
5. Summary
Embedding conductors within a dielectric confines the
conductor when plating, enabling finer feature resolutions.
Ablative laser processes can be used to structure dielectrics,
reducing the number of manufacturing steps compared to
photolithographic processes. They also allow vias to be
machined in the same process step as the routing layer,
significantly reducing the minimum size of the capture
pads increasing the space available for signal routing. This,
in combination with smaller signal pathways, reduces the
number of routing layers required in an advanced chip
package, which reduces the cost and height profile of the
package whilst improving the signal latency and integrity.
Frequency tripled, direct write, ablative solid state laser
systems have process times dependent on pattern complexity and therefore are not ideally suited to high volume manufacture of dense patterns comprised of complex and varied
features. Excimer laser systems can achieve a high
throughput at the required feature resolutions, however
have a higher cost of ownership than solid state laser systems.
The SMI optical system was introduced as an alternative to the above technologies offering a high throughput,
high resolution, solid state laser mask imaging system. The
results of the imaging system were presented and feature
sizes down to 2 µm line width and spacing were demon-
109
(Received: June 18, 2014, Accepted: December 5, 2014)
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising