|Imprinting has been shown to replicates the mold to within 1-2 nm, resolves 2 nm line, and 1 nm
features. The challenge in imprinting is to do this on less than perfect substrates with devices,
and then separate the mold from the imprint material.
The imperfections in the substrates come in a range of heights and length scales, for more
details go to Substrates. The length scales can be summarized as;
Descriptor Length scale Height
Across wafer 50 - 200 mm up to 20 um
Across field 10 - 25 mm 100 nm – 1 um
Nanotopography 2-3 mm 50 – 100 nm
Roughness < 1 um < 1 nm
Devices 100 nm – 200 um 10 nm - 2 um
When a reflective wafer (such as Si) and a mold are both chucked to an optical flat, and then
imprinted, the variation in thickness > 10 nm shows up as changes in thin film interference
colors and can be detected by the naked eye. So all of the height changes above, except for
roughness will be seen as multicolor imprints.
The traditional contact printer had the same problem when trying to minimize diffraction effects
between the mask and wafer. The solution was to mount the wafer on a gimbal that can roll to
get the wafer parallel to the mask and then pull vacuum between the mask and wafer, and then
release the vacuum holding the back of the wafer. This forced the thin wafer (0.5 mm thick) to
conform to the mask (3-6 mm thick). This allowed “hard contact” printing with minimal diffraction
effects, however, the moment the backside vacuum was released, the system lost a positive grip
on the wafer and alignment quality was reduced.
It appears that all the high pressure imprint tools use a version of this strategy. The Nanonex
system uses “Air Cushion Press“, presumably a compressed air pocket, to provide uniform
compression behind both the mold and the substrate (Li 2004). This has been shown to
conform over nanotopography (Tam 2004). EVG bonder mounts the mold on a flexure that
allows the mold to accommodate across wafer variation (Vratsov 2003) .
MII (Watts 2005) and KIMM (Kim 2005) have published results using a thin hard template on a
vacuum chucked wafer, showing that the mold can be induced to conform to the surface of a
wafer and remove across wafer, across field and nanotopography. The result is single color
imprints while keeping a positive grip on the wafer which should enable alignment. PDMS
molds have also been used to provide conformality.
Transfer printing relies on molecular contact, and so must conform to surface roughness as
well. This has been achieved using a soft rubber (PDMS) mold.
Step and Repeat tools from MII uses rigid mold 6 mm thick, a field size of 10-25 mm, and relies
on a passive flexure to level the mold to the wafer taking out the across wafer variation . The
imprints shown by MII (REf) on Double Side Polished (DSP) wafers do not show evidence of
across field color variation. KIMM have shown that if you use a small number of drops with a low
viscosity imprint fluid, then the imprint shows the drop pattern (Sim 2005). This implies that
although the wafer is being vacuum clamped to a chuck, the wafer is being lifted off from the
chuck and deflected by the imprint forces. The same effect is probably the reason that there is
no evidence of within field variation in MII imprints. Imprints on Single Side Polished (SSP)
wafers do show nanotopography.
AIST has used a soft pad under the imprint field to provide conformality (Hiroshima 2005).
A simple model can be used to predict conformality (Watts 2005). Conformality is visualized as
the doming of a edge supported disk. The diameter and dome height are given by the length
scale and height of the non uniformity on the wafer, the thickness and properties of the disk are
given by the mold or wafer - whichever is the conforming element. The uniform pressure on the
disk is the imprint pressure.
The analytical solution for the deformation of a disk is :
The graph on the right shows that 1 atmosphere of pressure will allow a wafer to conform to a
mold with 20 um of non flatness over a 100 mm length scale. Nanotopography will be
conformed by a 0.5 mm thick wafer and 1/10 of an atmosphere. Conformality scales with the
cube of the thickness of the conforming element.
The presence of devices with height greater than 25% of the minimum feature size requires
some form of planarization. The most popular is spin coating which relies on surface tension to
pull the liquid film flat during coating.
Limitations to spin on planarization
There are two limitations to spin on planarization; first there is always solvent trapped in the film
so there is shrinkage after softbake that can be reduced by using the highest possible solids
solutions. Secondly, there is a length scale over which surface tension fails to planarization. In
the extreme example it is well known that single color layers (<10 nm variation) can be spun on
virtually any wafer showing that no planarization has occurred. Experimentally it has been shown
that spin on planarization looses its effectiveness for features > 20 um. Models have been
published that predict …..
Even for small features, there can still be a problem at the edge of a large array of patterns, the
coated film thickness will change for the edge features because of the resist film will change
over outside unpatterned area as shown on the right . The degree to which this is a problem is a
function of feature size relative to height change, and imprint wall angle. These problems with
spin on planarization have led to the popularity of CMP in device manufacture. Even CMP uses of
a “soft pad” is used to provide conformality over long length scales (MEMC).
If device performance will allow it, dummy features can be used to fill in large features, and
around the edge of areas of small features and allow spin on planarization to be a universal
solution. For example 20 um holes perforating any large raised pad area should be sufficient to
allow spin on planarization to work.
Imprint planarization must be used for features from 20 um 200 um. However conformality is
required for non flatness over length scales > 1 mm.
Step and repeat imprint with a unpatterned mold, will provide planarization < 10 mm, and
conformality > 25 mm.
Brewer Science have published data suggesting that they have a solution to local planarization
and global conformality.
Reverse tone imprint planarization, reported by MII as SFIL-R (MII ???) , uses a variation on a
bilayer process. A organic imprint material is used as the patterned imprint. This is followed by
a planarizing spin on post imprint step using a silicon containing polymer. This planarizing
coating fills in the imprint pattern. The spun on layer is then etched back to reveal the organic
material. The oxygen plasma is then used to etch the organic imprint. This process works well
for patterning fine lines over SSP wafers (LeBrake 2005).
The patterned imprint both patterns and planarizes the wafer. The limitation occurs in the post
imprint spin on step which planarizes the imprint pattern. Large features will not be planarized or
resolved after etch. Also there may be line width variation at edge features depending on imprint
wall angle. Dummy features will probably solve the spin on planarizing problem.
For the most general case of large features in both device and imprint, imprint planarization
either pre or post patterned imprint is required.
After molding the wafer and mold are in perfect contact with the imprint material as “filler”. The
surfaces and material is designed and treated so as to have good adhesion between imprint
and wafer, and low adhesion between imprint and mold (Shaefer 2002). Even so, as the layer
gets very thin relative the area in contact the force becomes very large. The only way to reduce
separation force is to make either the wafer or mold bend so that the layers are separated by
peeling. At the same time sideways motion must be eliminated to avoid shearing features.
In “Deformation and failure modes of adhesively bonded elastic layers”, Crosby (Crosby 2004)
describes 3 modes of adhesion failure for an elastic layer between two rigid surfaces; either
crack propagation at the interface, crack propagation in the elastic layer, or bulk material failure.
These modes are affected by force separating the surfaces, the aspect ratio of radius of the
adhesive area to thickness of the elastic layer, and adhesion between layers. For force applied
normal to the interface and good adhesion, as the aspect ratio gets large (> 1E2, crack
propagation at the surface is replaced by bulk failure. Successful separation in imprint must
occur by crack propagation at the surface, and the aspect ratio for thin imprint layers is very high
> 1E9. Even with release agents, this implies that some from of non normal force such as
peeling is required to get acceptable separation.
MII has reported that S&R imprint tools with a 25 x 25 mm field in a hard rigid template support
automated separation (Shumaker 2005). In addition, a 100 mm imprint with a hard flexible thin
fused silica mold can also be separated automatically (Watts 2005). Brewer Science has also
reported separating up to 200 mm wafers, using a thin flexible polymer film (ref ). EVG have
reported separating a 200 mm wafer by apparently bending the mold (Islan 2002). Soft rubber
PDMS molds also allow large wafers to be separated (Plachetka 2005)
For more on the imprint solutions try Imprint Molds, Materials and Tools
Otherwise use the back arrow or go the Imprint Overview or use the tool bars.