|
Why multiple columns?
-
Higher throughput: E-beam
systems today can take more than 20 hours to write a
single mask and more than 30 hours to direct write a
prototype on a wafer. As minimum feature sizes
become smaller, single-column e-beam systems are
struggling to maintain even their current low
throughputs. By combining multiple independent
e-beam columns into a single array, throughput is
increased sufficiently to make Multibeam e-beam
tools viable for production.
-
Lower cost per wafer: Much
of the expense in a modern e-beam lithography system
is in the wafer stage, data path, vacuum chamber,
wafer-handling mechanism and other “overhead”. By
combining multiple columns in one system, Multibeam
is effectively offering multiple individually
controlled e-beam lithography systems for about the
price and cleanroom footprint of only one. In the
fab, this translates directly into a lower cost per
wafer.
What is
the MBXÔ
Engine current density capability?
-
MBXÔ Engine e-beams are produced by hot Schottky sources
(one per beam) and then limited through apertures
and focused using patented optics, resulting in
>500A/cm2 shaped
beams or >2000A/cm2
Gaussian beams at the substrate (wafer or
photomask). This is a large advantage over other
e-beam systems that are limited to much lower
current densities (typically 10 to 200A/cm2).
-
Combined with Multibeam’s writing strategy and
distributed column architecture, this high current
density results in short flash times, quickly
patterning resist with minimal heating of the
substrate.
How fast
is the Multibeam MBXÔ
Engine?
-
Since Multibeam achieves high current density and
high edge acuity in each shaped beam using its
patented optics, each pattern can be exposed in a
matter of tens of nanoseconds. This, combined with
the high-speed electronics controlling Multibeam’s
all-electrostatic column, enables the MBXÔ
Engine to write each pattern sequentially with cycle
times in the tens of nanoseconds.
-
Throughput depends on application. A Multibeam MEBICÔ
has a throughput of 5 wafers per hour which is
compatible with process flows in low-volume ASIC
foundries.
What
about other multiple e-beam systems?
-
Multibeam’s expertise is in developing e-beam
columns with small footprints. Each Multibeam
column, about 350 mm tall, only has a 9cm2 footprint.
-
Traditional e-beam column designs use magnetic
instead of electrostatic electron optics, which
makes it much more difficult to achieve a small
column footprint. We are aware of attempts to pack
4-columns, even 16-columns, into one assembly to
increase throughput.
-
A
different multiple e-beam approach involves
splitting an electron beam originating from a single
source into thousands of beamlets. At first glance,
this would appear to increase throughput. However,
splitting the beam will result in lower current
density. More figures can be written simultaneously,
but each figure takes longer to write. Although
throughput is limited, this approach may be useful
for prototyping.
How does Multibeam solve field stitching?
-
At
the start of each run, the MEBICä automatically calibrates each beam’s
size and shape while also
calibrating the distance between each beam using
interferometry.
-
The
high-precision wafer stage combined with
high-resolution targets and a stable stage map,
allows nanometer placement precision of each e-beam
at the substrate.
-
Multibeam’s proprietary datapath controls each
column independently and simultaneously, enabling
high placement precision of each pattern with
respect to the substrate on the fly, allowing the
MEBICÔ
to stitch fields seamlessly and achieve high overlay
accuracy.
What about drift?
-
Drift can occur due to thermal expansion of the
wafer, stage, or column. Because drift is a slow
process where position error accumulates over time,
a fast system such as Multibeam’s MEBICÔ has an inherent advantage over systems that take hours
to write a wafer. Nonetheless, to minimize drift,
the MEBICÔ actively controls temperature within the MBXÔ
Engine, wafer stage, and wafer.
What
about space charge defocusing?
-
Multibeam benefits from a design that has zero beam
crossovers. This minimizes the space charge problem
that defocuses high current density beams in other
systems.
How is
substrate heating controlled?
-
The MBXÔ
Engine’s distributed column architecture, with each
e-beam at least 30mm away from any other e-beam,
means that heat caused by electron bombardment at
the substrate is spread across the wafer.
Localized heating is dispersed through thermal
conduction within the wafer.
-
The
MEBICÔ actively controls the wafer temperature using a cooled
electrostatic chuck.
What
about Line Edge/Width Roughness?
Line Edge Roughness
or Line Width Roughness (LER or LWR) is of concern for
critical dimensions at 65nm and below. Multibeam employs
a patented electron optics technology that results in
high edge acuity and reduced feature roughness. By
utilizing this technology, Multibeam can print sharper
features with straighter edges.
One example was presented by Multibeam in a paper on LER
at SPIE 2008
What is the
smallest feature Multibeam can write?
-
Electron beams have virtually unlimited resolution.
Ultimate resolution is limited by a combination of
the beam size, beam scattering in the resist, back
scattering, shot noise, resist characteristics, and
wafer processing.
-
Multibeam achieves high beam
edge acuity through a patented electron optics
design. Higher edge acuity translates into a sharp
e-beam that can write small high-resolution
features. See
this paper from SPIE 2008 for more information.
-
Since resolution is largely determined by resist and
scattering characteristics, Multibeam benefits from
the 3keV to 50keV beam energy capability of the MBXÔ
Engine. 50keV e-beam systems are standard within the
industry, and knowledge of appropriate resist
thicknesses, resist sensitivities, and processing
techniques to minimize resist related resolution
issues are well known.
Figure 1: Square beams
shaped using Multibeam’s patented optics have 5x sharper
edges than Gaussian beams. This allows Multibeam to
write patterns with high resolution and high contrast.
How
does Multibeam mitigate the surface charging effect?
-
Any
charge that accumulates on a surface will slightly
bend incoming electron beams.
The surface charge generated while writing one pattern
may cause an offset when writing a nearby pattern.
This effect is negligible except at the most
advanced technology nodes.
-
As
an e-beam scans a resist-coated wafer, most
electrons will pass through the resist, into the
wafer, and move to ground, but some electrons will
remain on the surface of the resist causing a
negative surface charge, and other electrons may
knock electrons out of the resist causing a positive
surface charge.
-
Multibeam’s 50keV beam is resistant to bending near
the substrate, minimizing this effect.
How
does Multibeam mitigate the proximity effect?
-
During e-beam lithography some electrons inevitably
scatter within the resist or substrate, dispersing
up to tens of microns from the primary beam location
and partially exposing resist – this is called
“proximity effect”.
To
correct for this, most e-beam systems implement
proximity effect correction (PEC).
-
Multibeam accomplishes PEC by adjusting flash time.
By controlling flash time the MEBICÔ
corrects for additional resist doses arising from
proximity effects.
Figure 2:
No proximity effect: Standard dose is used. Edges of
feature are determined by where TOTAL dose is greater
than Exposure dose.
Figure 3: Proximity effect from global (purple line) and
local (green line) backscattering. The exposure time is
reduced to adjust dose (blue line is adjusted to yellow
line). Edges of feature are determined by where TOTAL
dose is greater than Exposure dose. Feature position can
be affected by local proximity effect, but this is
corrected by adjusting the beam position.
What is
Multibeam’s writing strategy?
-
The beam is blanked while a subfield
deflector vector scans the beam in X-Y to the
location of the feature to be written within a
subfield, and then the beam is unblanked to write
the feature. After the feature has been written
(typically in tens of nanoseconds), the beam is
blanked again, and this process repeats until all
features within the subfield have been written.
-
After each subfield is written, a mainfield
deflector positions the beam at the center of the
neighboring subfield.
-
Each column independently writes pattern data within
each subfield while the stage moves continuously in
X and steps in Y using a standard write-on-the-fly
strategy.
-
Proximity and resist heating effects are corrected
using variable exposure time. Beam position is
adjusted on-the-fly to correct for stage error and
beam distortion (calibrated automatically using high
resolution targets, interferometry, and stage map)
How
reliable is the Multibeam MEBICÔ?
-
The
MEBICÔ
is
designed for reliability and consistency. The
system has only a few mechanical moving parts,
including the low-speed wafer stage. The system
automatically calibrates at the start of each run to
insure consist patterning throughout its life.
-
Multibeam’s Schottky Thermal Field Emitter (TFE)
electron sources wear out (exhausting a reservoir
containing zirconium-oxide) extremely slowly and are
replaced annually during scheduled maintenance.
|
|
