Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Magnetron sputtering for the production of EUV mask blanks

Patrick Kearney, Tat Ngai, Anil Karumuri, Jung Yum, Hojune Lee,
David Gilmer, Tuan Vo*, Frank Goodwin
SEMATECH; Albany, NY, 12203, USA
*CNSE SUNY, Albany, NY, 12203, USA

1. ABSTRACT
Ion Beam Deposition (IBD) has been the primary technique used to deposit EUV mask blanks since 1995 when it was
discovered it could produce multilayers with few defects. Since that time the IBD technique has been extensively studied
and improved and is finally approaching usable defectivities. But in the intervening years, the defectivity of magnetron
sputtering has been greatly improved. This paper evaluates the suitability of a modern magnetron tool to produce EUV
mask blanks and the ability to support HVM production. In particular we show that the reflectivity and uniformity of these
tools are superior to current generation IBD tools, and that the magnetron tools can produce EUV films with defect
densities comparable to recent best IBD tool performance. Magnetron tools also offer many advantages in
manufacturability and tool throughput; however, challenges remain, including transitioning the magnetron tools from the
wafer to mask formats. While work continues on quantifying the capability of magnetron sputtering to meet the mask
blank demands of the industry, for the most part the remaining challenges do not require any fundamental improvements
to existing technology. Based on the recent results and the data presented in this paper there is a clear indication that
magnetron deposition should be considered for the future of EUV mask blank production.
Keywords: EUV mask blank deposition, magnetron, defectivity, manufacturability

2. INTRODUCTION
Magnetron deposition has been the standard method of EUV multilayer production for optics since the early 1990s. [1,2]
At that time magnetron demonstrated >50% reflectivity for Mo/Si multilayers, and people started considering using
EUV multilayers in lithography. The magnetron systems developed in the early 1990s were designed to give high
reflectivity and controllable uniformity, both of which are important for coating EUV optics. These systems were not,
however, designed with low defectivity in mind. Measurements of the defectivity of state of the art multilayers
deposited in the early 1990s showed defectivites of 10,000-100,000 defects per square centimeter at a detection
threshold of 250nm. As EUV masks require low defectivity an effort was launched at Lawrence Livermore National
Laboratory to drastically reduce EUV mask defectivity. The output of this effort was an Ion Beam Deposition (IBD)
tool that could produce EUV MLs with <0.1 defects per square cm@130nm defect size.[3]

Since Ion beam sputtering reported low defectivity, the vast majority of low defect ML deposition has been performed
with IBD using Veeco Nexus deposition tools. However, in the 20 years since IBD was selected, other deposition
techniques have advanced considerably. Re-evaluating the capability of alternate deposition techniques may reveal some
missed opportunities for mask blank improvement.

We have evaluated the capability of several deposition techniques to produce low defect, highly reflecting EUV mask
blanks:
• Evaporation- Evaporation can produce highly reflecting multilayers, but an assessment of the available tools for
purchase and their defectivitiy showed a lack of purchase ready tools.
• Magnetron- Several suitable, loadlocked, clean, multi target magnetron tools are for sale and magnetron itself is
known to be capable of producing highly reflecting multilayers.
• ALD/CVD- There is considerable interest in depositing Mo/Si multilayers with CVD or ALD or one of their
energy enhanced relatives (plasma enhanced, electron enhanced, etc.). At this time, these tools cannot deposit
Mo and or Si at low enough temperatures and high enough purity to enable their use for EUV mask blanks. For
example Si deposition is possible, but only at elevated temperatures. Pure Mo/Si multilayers react at
temperatures above ~200C and their reflectivity is greatly reduced. That said, much work is underway to reduce

Extreme Ultraviolet (EUV) Lithography VI, edited by Obert R. Wood II, Eric M. Panning,
Proc. of SPIE Vol. 9422, 94220H · © 2015 SPIE · CCC code: 0277-786X/15/$18
doi: 10.1117/12.2087773

Proc. of SPIE Vol. 9422 94220H-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/14/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


the deposition temperatures with these techniques and these tools may in a few years be capable of competing
for mask deposition. Currently, however, there are no ALD or CVD tools commercially available that can
produce high reflectance Mo/Si multilayers.

Based on our analysis of the alternate deposition techniques, it is clear that either IBD or Magnetron sputtering are the
best way to produce EUV mask multilayers. With IBD already explored by a major effort at SEMATECH, and efforts at
the major mask blank suppliers, it was decided to pursue magnetron deposition for EUV mask blanks in this work. As
magnetron is used extensively for high reflectance and high uniformity control work, we expect few problems in those
areas. The big question is can modern magnetron tools produce EUV mask multilayers with defectivity similar to IBD.

Historically it was assumed that EUV mask blanks could tolerate no defects. Recently the industry has planned on using
mitigation techniques to deal with a few defects per mask. (Pattern shift, pattern repair, etc.) Current thinking is that
masks with no large defects (> 80-100nm) and single digit small defects (<80nm) can be tolerated.

3. CANON ANELVA TOOL FOR EUV MASK BLANKS


SEMATECH was able to secure access to a Canon Anelva Multi target module installed at the CNSE SUNY campus in
Albany, NY. (College of Nanoscale Science and Engineering State University of New York) The tool coats 300mm
wafers, and has 4 targets in the deposition chamber, so handling was not required between each of the 80 layers that
make up an EUV mask multilayer. The Canon Anelva tool is well regarded in the industry as a low defect deposition
tool capable of depositing thin layers. Using the Canon Anelva tool allowed us to use the existing wafer infrastructure at
CNSE to inexpensively determine the defectivity of multilayers produced by the tool. It was not, however, possible to
modify the tool to handle EUV mask blanks, so an assessment of mask blank handling or deposition could not be made
at this time.

The defectivity of the Canon Anelva EUV MLs is shown in Figure 1. The defect measurements were made with a KLA
SP3 at 60nm sensitivity, and the number of defects was reduced by a factor of 4 to give the number of defects that would
be expected in the smaller area of an EUV mask blank quality area of 13.2x13.2 cm. The vertical lines in the plot
represent when changes were made to the tool (target changes and/or shield changes). After target and/or shield changes
the defectivity was elevated, and returned to a low level after several depositions were performed. There are two
instances where defectivitiy dropped below 10 defects per quality area for a significant period of time. Magnetron
deposition is capable of producing mask blanks with single digit numbers of defects. This defect level is similar to the
best defect level demonstrated by IBD as of mid-2014. Since that time IBD has seen a reduction in demonstrated
defectivity down to approximately 1 defect per mask blank at 50nm defect sensitivity. So, while the defectivity of an
“off the shelf” Canon Anelva tool compares well to that of relatively recent IBD results, IBD currently has the best
proven defectivity.

4. DISCUSSION
This work has demonstrated magnetron deposited multilayers that have defect levels close to what is required for EUV
mask blanks. This defect level is similar to what IBD had achieved in mid 2014, but is still several times the ultimate
performance demonstrated with IBD in late 2014 of ~1 defect per blank. Magnetron performance depositing on and
handling mask blanks must also be demonstrated.
Magnetron deposition does have several advantages over IBD that make pursuit of this technical direction warranted.
The first advantage is reflectivity. IBD MLs have typical reflectivities of 64% and maximum reflectivities of ~66%.
Magnetron deposition routinely achieves over 68% reflectivity. This difference in reflectivity would lead to
approximately 9% more stepper throughput in use. The second advantage of magnetron over IBD is general
manufacturability. IBD has very non-uniform target utilization which leads to rapid changes in rate and uniformity with
time compared to the Canon Anelva tool with its full surface erosion of its targets (See Figure 2). In our experience with
both tool sets, we find the center wavelength and uniformity specifications of EUV masks are more easily maintained on
the Canon Anelva tool than the IBD tools. Tool maintenance is also ~1/5 the time with magnetron than the 13%
downtime demonstrated for an IBD preventive maintenance.

Proc. of SPIE Vol. 9422 94220H-2

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/14/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


Equivalent defects in mask 13.2cm QA (CNSE Anelva tool)
10000 - I I
tAnelva 60nm Adders in QA
-IBD 2013 best median adders
-IBD 2014 best median adders

1000
I I

'
1

:y
100 I
w
,

I
0
J
06 . an
(.4")
C , , . J, .

! Al 1
J 4lI
.. `- 1 li.. I,
.0 . , .

10
1
,
z

Figure 1 Canon Anelva EUV ML defectivity on 300mm wafers.

Figure 2 Comparison of IBD (Left) and Canon Anelva (Right) silicon target at end of life. The IBD target has an eroded
elliptical pit in the center nearly the thickness of the target (~10 mm) while the peak to valley height range on the Canon
Anelva target is <2mm.

Proc. of SPIE Vol. 9422 94220H-3

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/14/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


5. CONCLUSION
While IBD is the process of record for EUV mask blank deposition, magnetron deposition with the Canon Anelva tool
may be a better long term solution for deposition of EUV mask blanks. With no optimization for the mask blank
problem, the Canon Anelva tool has produced EUV multilayers with defectivity approaching the best produced by IBD
after 20 years of tool optimization for the EUV mask problem. Magnetron also achieves higher reflectivity than IBD
and that should lead to higher stepper throughput in production. Finally the magnetron tool has better maintenance
times, and produce films with more stable wavelength and uniformity which should allow higher optical quality yields
during production.

REFERENCES
1

[1] Windt,. D, Waskiewicz, W, “Multilayer facilities required for extreme-ultraviolet lithography,” J. Vac. Sci. Technol.
B 12(6), (1994).
[2] Stearns, D., Rosen, R. and Vernon, S. “Multilayer mirror technology for soft-x-ray projection lithography,” Applied
Optics, Vol. 32, Issue 34, (1993).
[3] Kearney, P, Moore, C, Tan, S, Vernon, S, “Mask blanks for extreme ultraviolet lithography: Ion beam sputter
deposition of low defect density Mo/Si multilayers,” J. Vac. Sci. Technol. B 15, 2452, (1997).

Proc. of SPIE Vol. 9422 94220H-4

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/14/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

You might also like